Evolving Approaches to Understanding Natural Hazards [1 ed.] 9781443883870, 9781443876094

Citation preview

Evolving Approaches to Understanding Natural Hazards

Evolving Approaches to Understanding Natural Hazards Edited by

Graham A. Tobin and Burrell E. Montz

Evolving Approaches to Understanding Natural Hazards Edited by Graham A. Tobin and Burrell E. Montz This book first published 2015 Cambridge Scholars Publishing Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2015 by Graham A. Tobin, Burrell E. Montz and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-7609-7 ISBN (13): 978-1-4438-7609-4

CONTENTS

List of Figures............................................................................................. xi List of Tables ............................................................................................. xv Acknowledgments .................................................................................. xviii Chapter One ................................................................................................. 1 Natural Hazards: Evolving Approaches Burrell E. Montz and Graham A. Tobin Part I: Risk and Vulnerability Introductory Comments ............................................................................... 8 Graham A. Tobin and Burrell E. Montz Chapter Two .............................................................................................. 11 The Magnitude, Frequency and Spatial Distribution of Tornadoes in Texas: An Assessment of Tornado Risk Kevin R. Mulligan, Lucia S. Barbato and Stephen Weinbeck Chapter Three ............................................................................................ 23 Disaster Warnings in San Marcos, Texas: Communicating in Ethnically Diverse Places Carolyn E. Springer and R. Denise Blanchard Boehm Chapter Four .............................................................................................. 36 Hazardousness of the Tampa Bay Region: Evaluating Physical Risk and Socio-Economic Vulnerability Burrell E. Montz and Graham A. Tobin Chapter Five .............................................................................................. 47 Classifying Heat Stress Events in the Central United States Erik H. Bowles

vi

Contents

Chapter Six ................................................................................................ 57 Vulnerability to Flooding in Columbia County, PA: The Role of Perception and Experience among the Elderly Jennifer Webb Chapter Seven............................................................................................ 68 Framing Flood Risk: Comparing Perceptions of the “100 Year Flood” and Two Alternate Descriptions Heather M. Bell Chapter Eight ............................................................................................. 82 International Students Vulnerability to Emergency Events: Does Tenure of Residence Make a Difference? Xueqin (Elaine) He Chapter Nine.............................................................................................. 96 Chronic Hazard: Weighing the Risk against the Effects of Emergency Evacuation from Popocatepetl, Mexico Graham A. Tobin, Linda M. Whiteford, Eric C. Jones and Arthur D. Murphy Chapter Ten ............................................................................................. 110 Assessing the Impact of Variable Weighting Strategies on the Spatial Pattern of Hazard Vulnerability Lyndsey E. Abel and James K. Lein Chapter Eleven ........................................................................................ 124 Assessing Vulnerability to Hurricanes in Harris County, Texas Chunling Liu and Gang Gong Chapter Twelve ....................................................................................... 137 Refining the Spatial Prediction of Frequency and Probability of Hurricane Strikes in South Carolina Shuang Kang Wu and Shou Lu Part II: Technology Introductory Comments ........................................................................... 154 Burrell E. Montz and Graham A. Tobin

Evolving Approaches to Understanding Natural Hazards

vii

Chapter Thirteen ...................................................................................... 157 Drought Assessment through Advanced Very High Resolution Radiometer Satellite Data Stephen J. Walsh Chapter Fourteen ..................................................................................... 169 Graphic Narratives for Emergency Mapping Mark Monmonier Chapter Fifteen ........................................................................................ 178 The Application of Remote Sensing and GIS for Mapping Flood Prone Agricultural Land Randall S. Pearson and Stephen A. Kay Chapter Sixteen ....................................................................................... 189 Mapping Landslide Susceptibility: A Travis County, Texas, Case Study David J. Wachal Chapter Seventeen ................................................................................... 204 The Hazardousness of Place: Risk from Multiple Natural Hazards Burrell E. Montz Chapter Eighteen ..................................................................................... 217 Monitoring Weather Hazards on Rural Roads Using Remote Sensing and GIS Richard P. Watson, Karl K. Benedict and Theresa R. Watson (Kuntz) Chapter Nineteen ..................................................................................... 230 The Use of GIS, Remote Sensing and Virtual Reality in Flood Hazard Modeling, Assessment and Visualization Shunfu Hu Chapter Twenty ....................................................................................... 240 Sinkhole Mapping in Pinellas County, Florida: Problems and Solutions Robert Brinkmann, Don L. Seale, Kelly Wilson and H. Len Vacher Chapter Twenty-One ............................................................................... 252 A GIS-Based Approach to Analyzing Warning Siren Networks: An Analysis of Riley and Wabaunsee Counties, Kansas Mitchel J. Stimers

viii

Contents

Chapter Twenty-Two............................................................................... 267 Flash Floods, Low Water Crossings and Swift Water Rescues in Texas Pamela S. Showalter and Yongmei Lu Chapter Twenty-Three............................................................................. 277 Using GIS to Examine Evacuation Need and Shelter Locations in Broward County, Florida Shivangi Prasad Chapter Twenty-Four .............................................................................. 290 Flood Damage Assessment Using Remote Sensing, GIS and GPS Technologies: A Case Study in Golestan Province, Northeastern Iran Parviz Zeaiean Firouzabadi, Saeed Saroei and Esmat Madaniyan Part III: Impacts and Mitigation Introductory Comments ........................................................................... 302 Graham A. Tobin and Burrell E. Montz Chapter Twenty-Five ............................................................................... 305 Land Use Policy and Flood Hazard Mitigation in the Development of Eastern New Orleans Merrill L. Johnson and Robert A. Sauder Chapter Twenty-Six................................................................................. 318 Earthquake Insurance Reform in New Zealand Iain Hay Chapter Twenty-Seven ............................................................................ 331 The Floodplain as a Housing Submarket: A Preliminary Analysis Burrell E. Montz Chapter Twenty-Eight ............................................................................. 339 Impacts of a Second “Once in a Lifetime” Flood on Property Values: Linda and Olivehurst, California Revisited Graham A. Tobin and Burrell E. Montz Chapter Twenty-Nine .............................................................................. 349 To Stay or Leave: Residents’ Evaluation of Hurricane Evacuation Warnings Kirsten Dow, Patrice Burns and Susan L. Cutter

Evolving Approaches to Understanding Natural Hazards

ix

Chapter Thirty ......................................................................................... 360 The 1986 Edmonton, Alberta Canada Tornado: Perception, Experience, and Response Then and Now R. Denise Blanchard-Boehm and M. Jeffrey Cook Chapter Thirty-One ................................................................................. 373 Chronic Hazards: Health Impacts Associated with On-Going Ash-Falls around Mt. Tungurahua in Ecuador Graham A. Tobin and Linda M. Whiteford Chapter Thirty-Two ................................................................................. 387 Predictors for Public Response to Tornado Warnings: The May 4, 2003 Tornadoes in Kansas, Missouri, and Tennessee Bimal Kanti Paul and Biqing Huang Chapter Thirty-Three ............................................................................... 396 The Role of Tourism in Post-Disaster Recovery: A Case Study of Patong Beach, Phuket Jacqueline Salmond Chapter Thirty-Four................................................................................. 411 Developing a County-Level Natural Hazards Mitigation Plan with GIS: Practices, Problems, and Challenges Yu Zhou Chapter Thirty-Five ................................................................................. 420 Coping with the Flood Hazard in the Nyando River Basin, Kenya Charles G. Manyara Chapter Thirty-Six ................................................................................... 431 The Locations of Temporary Shelters after Hurricane Katrina R. Stacy Brown, Jonathan C. Comer, and Thomas A. Wikle Chapter Thirty-Seven .............................................................................. 444 Managing Response and Recovery to Mississippi River Flooding: Applying Spatial Analysis in Memphis, Shelby County, Tennessee Brian Waldron, Arleen Hill and Bob Nations Jr.

x

Contents

Chapter Thirty-Eight ............................................................................... 458 Site, Situation, and Property Owner Decision- Making after the 2002 Guadalupe River Flood Elyse M. Zavar, Ronald R, Hagelman III and William M. Rugely II Contributors ............................................................................................. 471 Index ........................................................................................................ 484

LIST OF FIGURES

Figure 2-1. Magnitude and frequency of tornadoes in Texas ..................................14 Figure 2-2. Area struck by tornadoes by magnitude ................................................15 Figure 2-3. Spatial distribution of tornadoes in Texas .............................................16 Figure 2-4. Spatial distribution of strong and violent tornadoes..............................17 Figure 2-5. Population density of Texas counties....................................................19 Figure 2-6. Tornado density in Texas counties .......................................................19 Figure 2-7. Relative tornado risk in Texas ..............................................................20 Figure 3-1. Study area, San Marcos, Texas .............................................................25 Figure 3-2. Revised general risk communication model .........................................34 Figure 4-1 Geo-physical risk of the Tampa Bay region ..........................................41 Figure 4-2. Social vulnerability of the Tampa Bay region ......................................42 Figure 4-3. Hazardousness of the Tampa Bay region ..............................................43 Figure 5-1. Moisture gradient decreases from east to west across Nebraska and Kansas ........................................................................................................49 Figure 5-2. Geographic variation of heat stress events across the moisture gradient and by latitude ......................................................................53 Figure 6-1. Columbia County, Pennsylvania...........................................................60 Figure 8-1. International student population in the USA by academic level with selected academic year ...........................................................83 Figure 8-2. International students and emergency events ........................................88 Figure 9-1. Cascade of Impacts ...............................................................................97 Figure 9-2. Villages around Popocatépetl .............................................................100 Figure 9-3. Age distribution of respondents’ households .....................................102 Figure 10-1. Study area and regional setting .........................................................112 Figure 10-2. Storm-event impact of A) hurricane, B) flooding, C) total storm .....116 Figure 10-3. Surfaces depicting A) predicted vulnerability, B) 1-year actual storm impact, C) 2.5- year actual storm impact ..............................................117 Figure 10-4. Deviation from A) predicted vulnerability and 1-year storm impact, and B) predicted vulnerability and 2.5-year storm impact..................119 Figure 10-5. Vulnerability based on multiple weighting: A) age, race, gender, wealth; B) non-white, poor children; C) non-white, poor female, non-vehicle......................................................................................................120 Figure 11-1 Harris County, Texas .........................................................................128 Figure 11-2 Spatial distribution of geophysical vulnerability ...............................131 Figure 11-3 Social vulnerability by census block group based on population and structure....................................................................................................131 Figure 11-4. Social vulnerability by census block group based on differential access to resources ..........................................................................................132

xii

List of Figures

Figure 11-5. Social vulnerability by census block group based on population with special evacuation needs .............................................................................. 132 Figure 11-6. Social vulnerability by census block group based on all three evacuation needs .............................................................................................133 Figure 11-7. Social vulnerability by census block group based on geophysical vulnerability and all three evacuation dimensions ........................................133 Figure 12-1. Tracks of hurricanes that affected South Carolina, 1906-2005 .........145 Figure 12-2. Frequency of hurricanes that affected South Carolina, 1906-2005 ...146 Figure 12-3. Cumulative frequencies of hurricane strikes in South Carolina between 1906 and 2005...................................................................................147 Figure 12-4. Expected probabilities of hurricane strikes once per year in South Carolina ............................................................................................148 Figure 13-1. Location of Oklahoma cooperative weather stations ........................162 Figure 13-2. Oklahoma climatic divisions ............................................................163 Figure 13-3. Crop moisture index – June 22-28, 1980 ..........................................164 Figure 13-4. Crop moisture index – July 13-19, 1980 ...........................................164 Figure 13-5. Crop moisture index – July 20-26, 1980 ...........................................165 Figure 13-6. Crop moisture index – August 17-23, 1980 ......................................166 Figure 15-1. Study area .........................................................................................180 Figure 15-2. Fifteen consecutive day flood recurrence interval for the Yalobusha River at Whaley, Mississippi ..........................................................................184 Figure 15-3. Fifteen consecutive day flood recurrence interval for the Yazoo River at Greenwood, Mississippi ....................................................................185 Figure 16-1. Travis County, Texas ........................................................................193 Figure 16-2. Landslide susceptibility index...........................................................194 Figure 16-3. Travis County slope ..........................................................................195 Figure 16-4. Travis County geology .....................................................................196 Figure 16-5. Travis County faults and fault buffers ..............................................197 Figure 16-6. Travis County vegetation ..................................................................199 Figure 16-7. Travis County landslide susceptibility index ....................................199 Figure 17-1. Combined probability of occurrence.................................................213 Figure 17-2. Impact index map .............................................................................215 Figure 18-1. McKinley County, New Mexico showing WRRHAMS interface and road network.............................................................................................219 Figure 18-2. WRRHAMS client interface: 24-hour precipitation total data ..........222 Figure 18-3. WRRHAMS client interface with a close-up view of the flow accumulation raster and analysis basins ..........................................................224 Figure 18-4. WRRHAMS client interface with a close-up view of the flow accumulation raster and TIGER road network................................................225 Figure 18-5. Road/stream intersections ranked by probability of Erosion ............226 Figure 19-1. Location of the Piasa Creek watershed .............................................232 Figure 19-2. 3D representation of the Piasa Creek watershed: (A) 3D scene displayed in ArcView 3d scene Viewer; (B) DOQ superimposed on the 3D scene; (C) 100-year flood zone superimposed on the DOQ ...........................235 Figure 19-3. A close-up view of a portion of the Piasa Creek watershed showing locations of single-family houses inside the 100-year flood zone ...................236

Evolving Approaches to Understanding Natural Hazards

xiii

Figure 19-4. 3D representation of the same portion of the Piasa Creek watershed as shown in Figure 19-3 .................................................................................236 Figure 19-5. VRML 3D scene played back using Microsoft Internet Explorer .....237 Figure 20-1. Pinellas County, Florida ...................................................................243 Figure 20-2. An area of Pinellas County, Florida mapped using ALSM techniques .......................................................................................................246 Figure 20-3. Air-photographs showing changes in sinkhole landscape from 1926 to 2002 ...........................................................................................248 Figure 21-1. Riley and Wabaunsee Counties, Kansas ...........................................254 Figure 21-2. Population analysis of Riley County, Kansas ...................................255 Figure 21-3. Population analysis of Wabaunsee County, Kansas ..........................256 Figure 21-4 Potential warning siren locations for Wabaunsee County Kansas .....260 Figure 21-5. Potential siren locations for Lake Wabaunsee siren..........................261 Figure 21-6. Siren location analysis, Riley County, Kansas..................................263 Figure 21-7. Siren location analysis, City of Manhattan, Riley County, Kansas ...264 Figure 22-1. Study area: the flash flood alley consisting of 44 counties .............269 Figure 22-2. Swift water rescues in study area: 2007..........................................272 Figure 22-3. LWCs and SWRs with FEMA floodplain, Hays County, Texas .....272 Figure 22-4. LWCs, 2007 SWRs, FEMA Floodplain, East Hays County, Texas ............................................................................................................273 Figure 23-1. Distribution of evacuation need ........................................................283 Figure 23-2. High evacuation need clusters ..........................................................284 Figure 23-3. Distribution and direction trend of mobile homes and RV parks ......286 Figure 23-4. Potential shelter sites ........................................................................287 Figure 24-1. Location of Golestan province in Iran ..............................................292 Figure 24-2. ETM acquired in Oct. 2001, IRS 1D acquired in Oct. 2002 and photo mosaic of pictures taken from starting point of floods ...................293 Figure 24-3. Fuzzy change map between 2000 and 2001. Light tones show severe changes ................................................................................................294 Figure 24-4. Fuzzy change map between 2000 and 2002. Light tones show severe changes ................................................................................................294 Figure 24-5. Scatterplot of fuzzy change maps of 2001 and 2002.........................295 Figure 24-6. Fuzzy change map between 2001 and 2002. Dark tones show less change ......................................................................................................295 Figure 24-7. Areas of more than 50 percent changes; darker tones along river .......................................................................................................296 Figure 24-8. Flood boundary and land use/ land cover map in 2001 .....................296 Figure 24-9. Flood boundary and land use/ land cover map in 2002 .....................297 Figure 25-1. New Orleans and vicinity ................................................................308 Figure 25-2. Expansion of built-up area in eastern New Orleans, 1951 to 1980 ..309 Figure 25-3. Approximate flood zone after rain of May 3, 1978 ..........................314 Figure 28-1. Extent of the 1997 floods in Linda and Olivehurst, California ........341 Figure 31-1. Outpatient consultation rates: 1995-2001 ........................................379 Figure 31-2. Outpatient consultation rates: Acute respiratory infections ..............379 Figure 32-1. The study area ..................................................................................391 Figure 33-1. Location of Phuket, Thailand ........................................................397

xiv

List of Figures

Figure 33-2. Tourism promotion ..................................................................... 407 Figure 34-1. Location of five Ohio counties ........................................................412 Figure 34-2. Reference map: Streams and 100-year floodplain in Preble County, Ohio ..................................................................................................415 Figure 34-3. Impacts of the Maumee River 100-year flood on Grand Rapids, Wood County, Ohio .......................................................................................416 Figure 35-1. The Study Area ................................................................................... 422 Figure 35-2. Rainfall Patterns ..............................................................................423 Figure 36-1. Locations of formal and informal shelters in Louisiana ...................435 Figure 36-2. Locations and types of shelters in Louisiana ...................................436 Figure 36-3. Distances of formal and informal shelters from the Superdome ......438 Figure 37-1. Shelby County, Tennessee ...............................................................446 Figure 37-2 Mississippi River flood event ...........................................................449 Figure 37-3. The planning tool .............................................................................453 Figure 37-4. Example of public map released for impacted zip codes .................454 Figure 38-1. Study area within Comal County, Texas ..........................................460

LIST OF TABLES

Table 2-1. Weighting factors based upon tornado density.......................................18 Table 3-1. Ethnic composition of study area population in 2000 ............................25 Table 3-2. Variables significant at the 0.05 level or below .....................................30 Table 4-1. Return periods and probabilities of occurrence ......................................38 Table 4-2. Variables used in determining social vulnerability ................................39 Table 5-1. Previous heat stress classification method using heat index...................50 Table 5-2. Heat stress classification model .............................................................52 Table 5-3. Heat stress classification results .............................................................54 Table 6-1. Age and total perception score ...............................................................62 Table 6-2. Age and ever evacuated home due to flooding ......................................63 Table 6-3. Flood experience and perception............................................................64 Table 6-4. Evacuation and perception .....................................................................64 Table 6-5. Damage and perception ..........................................................................65 Table 7-1. Variation in possibility of occurring more than once per year: Cochran’s Q ............................................................................................................... 72 Table 7-2. Participant comments on multiple floods per year .................................72 Table 7-3. Participant comments on change in size over time.................................74 Table 7-4. Participant comments on concern ..........................................................75 Table 7-5. Variations in most concerning flood: Cochran’s Q ................................76 Table 7-6. Variations in least concerning flood: Cochran’s Q ................................76 Table 7-7. Variation in relative concern: Friedman test ..........................................78 Table 7-8. Variation in combined effectiveness: Friedman test ..............................78 Table 8-1. Length of residence vs. environmental familiarity .................................90 Table 8-2. Length of residence vs. knowledge of hurricanes ..................................90 Table 8-3. Length of residence vs. Hurricane Rita evacuation behavior .................91 Table 8-4. Length of residence vs. expected evacuations ........................................91 Table 8-5.The top 4 identified evacuation determinants .........................................92 Table 9-1. Summary demographic statistics by age .............................................102 Table 9-2. Number of times evacuated .................................................................104 Table 9-3. Evacuation by age ...............................................................................105 Table 9-4. Evacuation and occupation .................................................................106 Table 9-5. Evacuation and education ...................................................................107 Table 9-6. Evacuation by village area ..................................................................108 Table 10-1. Variables employed in the analysis ....................................................114 Table 11-1. Variables used to determine social vulnerability ................................129 Table 12-1. Hurricanes making landfall in South Carolina, 1906-2005 ................144 Table 13-1. Twelve step multiple regression model of CM and remotely sensed variables for Oklahoma regions ...........................................................166 Table 15-1. Flood recurrence intervals..................................................................182

xvi

List of Tables

Table 15-2. Stage to discharge conversion ............................................................183 Table 15-3. River stages from March 8, 1989 through March 10, 1989 ................185 Table 15-4. Images selected for this study ............................................................185 Table 16-1. Travis County slope, fault buffers, and land cover.............................196 Table 16-2. Travis County geology .......................................................................198 Table 17-1. Indices for determination of hazardousness .......................................207 Table 17-2. Return periods and probabilities of occurrence for different storm intensities ........................................................................................................209 Table 19-1. Project data ........................................................................................232 Table 22-1. Incidents by year .............................................................................270 Table 23-1 Variables used to determine evacuation need .....................................280 Table 23-2. Mean variable values for clusters and entire county ..........................285 Table 24-1. Area damage statistics ........................................................................297 Table 27-1. Variation in price as influenced by variables ....................................335 Table 27-2. Floodplain location and price ............................................................335 Table 27-3. Floodplain location and age of housing ............................................336 Table 27-4. Floodplain location and number of days on the market ....................337 Table 28-1. House sizes - 1986 and 1997 flood area ............................................344 Table 28-2. House list prices - 1986 and 1997 flood areas ...................................344 Table 28-3. House sold prices – 1986 and 1997 flood areas ................................344 Table 28-4. Median house sold prices as a function of flood area.........................345 Table 29-1. Survey responses of Myrtle Beach residents ......................................352 Table 29-2. Evacuation rates reported by Myrtle Beach respondents for hurricanes over the last three years............................................................353 Table 29-3. What convinced you to leave the area? ..............................................354 Table 29-4. Reasons for non-evacuation for hurricanes .......................................355 Table 29-5. What makes a source reliable? ..........................................................357 Table 30-1. Statistically significant study variables .............................................366 Table 30-2. Levels of experience: Respondents in the path as compared to those out of the path during the tornado ..................................................................367 Table 30-3. Levels of experience: Respondents in the path of the tornado as compared to those not in Edmonton at the time .........................................368 Table 30-4. Preparedness (response) toward future tornado occurrence: All respondents ..............................................................................................369 Table 31-1. Questionnaire sample and survey statistics .......................................377 Table 31-2. Reported exposure to ash ..................................................................380 Table 31-3. Reported frequency of exposure to ash .............................................380 Table 31-4. Perceived effect of ash on respondent and family .............................380 Table 31-5. Perceived risk of ash to family health ...............................................381 Table 31-6. Activities undertaken to counter effects of ash .................................381 Table 31-7. Reported illnesses of respondents over preceding six months ...........382 Table 32-1. Factors for public response to tornado warnings ...............................393 Table 33-1. Tourism data 1995-2003 ..................................................................399 Table 33-2. Phuket hotel occupancy rates 2004-2005 ........................................402 Table 34-1. Impacts of the 100-year flood on Preble County, Ohio .....................416

Evolving Approaches to Understanding Natural Hazards

xvii

Table 34-2. Impacts of the 100-year flood of the Maumee River on Wood County ............................................................................................417 Table 34-3. Community profile: Incorporated areas of Wood County affected by Maumee River flood .................................................................................417 Table 35-1. Number of people per household .......................................................425 Table 35-2. Roofing material ................................................................................425 Table 36-1. Correlation matrix for all shelters .....................................................441 Table 37-1. Select demographic characteristics of Shelby County .......................448 Table 37-2. Parcels inundated at specific predicted Mississippi River flood stages ..............................................................................................................448 Table 37-3. GIS data sets and sources ..................................................................450 Table 38-1. Building permits for single family homes from 2000-2005 for Comal County and Central Texas region ..................................................461 Table 38-2. Descriptive statistics of spatial characteristics ..................................464 Table 38-3. Land values .......................................................................................465 Table 38-4. Multinomial logistic regression results .............................................466

ACKNOWLEDGMENTS

The editors would like to thank all those who have participated in the Applied Geography Conferences over the years. The conference began in 1978 established thanks to the efforts of John Frazier and Bart Epstein who saw the need for applied research that focused on societal problems and their solutions. The conference has published 36 volumes (1978 and 2013) edited by a number of scholars: Tom Bell, Dawna Cerney, Bart Epstein, John Frazier, Ed Hillsman, John Harrington, Lisa Harrington, Ken Jones, Jay Lee, Dennis Lord, Burrell Montz, Henry Moon, Andy Schoolmaster, and Graham Tobin. In addition, we are grateful to the many reviewers who have spent time going through articles and of course the paper presenters, all of whom have added to the quality of the Applied Geography Conferences. Burrell E. Montz Graham A. Tobin

CHAPTER ONE NATURAL HAZARDS: EVOLVING APPROACHES BURRELL E. MONTZ AND GRAHAM A. TOBIN

Introduction The 21st century presents many challenges to the hazard manager; dynamic climatic conditions, with many possible outcomes, combined with seemingly unlimited population growth, rapid urbanization and changing socio-economic relationships are reshaping disaster impacts, community responses, and social safety mechanisms. Indeed, the many facets of human vulnerability are being constantly restructured with the ongoing interplay of the physical environment and social, economic and political forces (Montz and Tobin, 2012; Tobin and Montz, 2009). At the same time, reducing vulnerability and enhancing community resilience require policies aimed at mitigating the consequences of disasters as they affect different locations and different groups. As a result, sound scientifically-based research of an interdisciplinary nature is required to further our understanding of the forces at play, and to devise appropriate means to counter them (Montz and Tobin, 2012). There is a kaleidoscope of realities to confront and one set of strategies will not fit every geographic location, social, political, and economic context. Our models must be flexible to address such moving targets and at the same time provide an understanding of the interactions between people, natural hazards, and disaster events. These are issues that have been addressed for some time, and we have made progress – but there is much more to be done. It is in this context that we look at evolving approaches to natural hazards.

2

Chapter One

Evolving Traditions Research into natural hazards has had a long tradition beginning with a focus on physical processes and gradually evolving into a full-blown interdisciplinary agenda with the recognition of the interactions between the physical and human environments. It was Gilbert White’s pioneering work (see White, 1945; 1961; 1964; White et al., 1958) that demonstrated unambiguously the fundamental significance of the human element in determining disaster settings and outcomes. As a result, the hazard agenda now embraces initiatives ranging from the physical, to the socio-economic, to the political, and incorporates a broad array of methodological approaches and technological advances, and employing both quantitative and qualitative procedures. For much of the second half of the twentieth century, hazards research adopted two foci: 1) furthering knowledge of the physical and spatial dimensions of natural events and 2) understanding individual decisionmaking and responses to natural hazards and disasters (Montz and Tobin, 2012). Much of this research was applied, seeking to solve societal problems -- a laudable goal but for many academics it lacked a sound theoretical foundation (Alexander, 1997; Hewitt, 1997) and tended to overlook the systemic causes of disasters (Pidgeon et al., 2003). As the twentieth century progressed, these criticisms were addressed and increasingly research was directed at understanding differences in human vulnerability and community resilience (Cutter et al., 2003; Smith and Petley, 2009; Tobin and Montz, 1997). These concepts, firmly established by Blaikie et al. (1994) (see also Wisner et al., 2004), have now become the modus operandi for much hazard research. Montz et al. (2004) demonstrated that hazards research had evolved from looking at individual risks and one-off, extreme events, to understanding complex events that include political, social, and economic forces that impact the vulnerability of individuals and communities, to a range of hazards at different geographic scales and at various governmental levels. Disasters, therefore, are now understood to have systemic root causes embodied in socio-economic and political systems that place people, often marginalized groups, in hazardous locations, in unsafe structures, and with few resources available for them to respond with security to events. There has indeed been a paradigm shift, turning attention to more general principles of hazardousness, risk, vulnerability, and resilience. Part I of this book reflects this shift, with papers that illustrate the changing approaches to the study of risk and vulnerability.

Natural Hazards: Evolving Approaches

3

At the same time as concepts of hazards, vulnerability and risk were evolving, so too were new technologies coming to the fore which vastly increased the arsenal of tools available to the researcher and practitioner. The development and expansion of geographical information systems (GIS) have provided new ways to explore multiple aspects of hazards and disasters. Investigations of dynamic forces, whether they are geophysical, social, economic or political, can now be visually represented over time and space. In addition, improvements in statistical techniques that examine spatial patterns and correlations have brought a scientific rigor to many studies. Along with GIS, the refinement of remote sensing, such as LiDAR, now promote more sophisticated models of landforms and human activities. Furthermore, others have utilized web-based resources and sophisticated statistical modeling to further academic and practical advances in hazards research. In Part II of the book, the reader will find examples of articles that incorporate these technologies as they have been applied to hazards research over several decades. The combination of these new hazard concepts and frameworks and the incorporation of innovative technologies and analytical techniques has encouraged a more systematic approach to problem solving. Applied research has taken a prominent place in the academy as hazard researchers have sought to explain and ultimately reduce the impacts of disasters. Questions regarding appropriate mitigation practices that account for both the geophysical event and the human landscape have been addressed. For example, consider the nature of hazards, both those which have long been the subject of geographic research, such as floods, earthquakes, hurricanes, and tornadoes, and those which have more recently been recognized to be of increasing significance. These include slow-onset, pervasive hazards such as drought, chronic hazards associated with continued exposure to, for example, repeated ash-fall from relatively frequent volcanic eruptions, and changing hazardousness due to climate change. How do we mitigate these impacts? In a way, we have come full circle; White’s arguments are just as pertinent today as they were in the mid-twentieth century. The advantage we have now, however, is superior technology and a better grasp of those root causes and conditions that exacerbate or ameliorate disaster impacts, risk, vulnerability and resilience. In Part III of the book, therefore, we provide some articles to illustrate how impacts and mitigation have been examined by geographic scholars. The collection of papers included in this book offers the reader insight into the development of applied hazards research, as it has built on previous work, evolving technologies, improved understanding of the factors involved, and increased awareness of the needs of those who make

4

Chapter One

decisions with respect to hazard management. The interplay of factors contributing to disaster losses has become ever more complex, as too has our ability to incorporate those complexities into a coherent response strategy, as we strive to reduce losses and increase resilience. It is hoped that this volume provides not only an appreciation for the foundation that has been set (as illustrated by the examples here), but also inspires future researchers as they look to address these very pressing issues. Editors’ Note: Because the papers included in this volume cover more than three decades of hazards research, tools, techniques, and even office software have changed significantly. As a result, the figures included in the chapters are of the best quality possible. Tables could be recreated but figures could not, explaining the variable quality of figures found in this volume.

References Alexander, D. 1997. The Study of Natural Disasters, 1977-1997: Some Reflections on a Changing Field of Knowledge. Disasters 21, 284-304. Blaikie, P., T. Cannon, I. Davis, and B. Wisner. 1994. At Risk: Natural Hazards, People’s Vulnerability, and Disasters. London: Routledge. Cutter, S.L., B.J. Boruff, and W.L. Shirley. 2003. Social Vulnerability to Environmental Hazards. Social Science Quarterly 84, 242-261. Hewitt, K. 1997. Regions of Risk: A Geographical Introduction to Disasters. Harlow, UK: Longman. Montz, B.E. 2009. Emerging Issues and Challenges: Natural Hazards. Journal of Contemporary Water Research and Education 141, 1-4. Montz, B.E., J. Cross, S. Cutter. 2004. Hazards. In Geography in America at the Dawn of the 21st Century, G.L Gaile, and C.J. Wilmott, eds. Oxford: Oxford University Press. pp. 479-491. Montz, B.E. and G.A. Tobin. 2007. Three Decades of Applied Geography: Themes from the Applied Geography Conferences since 1978. Papers of the Applied Geography Conferences 30:1-9. Montz, B.E. and G.A. Tobin. 2011. Natural Hazards: An Evolving Tradition in Applied Geography. Journal of Applied Geography 31(1):1-4. Montz, B.E. and G.A. Tobin. 2012. Natural Hazards and Natural Disasters. In 21st Century Geography: A Reference Handbook, J.P. Stoltman ed. London: Sage Publications Volume 2, Chapter 46, pp. 509-518.

Natural Hazards: Evolving Approaches

5

Montz, B.E. and G.A. Tobin. 2013. Vulnerability, Risks, and Hazards. In Oxford Bibliographies in Geography, B. Warf ed. Oxford: Oxford University Press. Pidgeon, N., R. Kasperson, and P. Slovic 2003. The Social Amplification of Risk. Cambridge: Cambridge University Press. Smith, K. and D. Petley. 2009. Environmental Hazards: Assessing Risk and Reducing Disaster. 5th edition. London: Routledge. Tobin, G.A. and B.E. Montz. 1997. Natural Hazards: Explanation and Integration. New York: Guilford Press. Tobin, G.A. and B.E. Montz. 2004. Natural Hazards and Technology: Vulnerability, Risk and Community Response in Hazardous Environments. In Geography and Technology, S.D. Brunn, S.L. Cutter, and J.W. Harrington eds. Dordrecht, Netherlands: Kluwer Academic Publishers. Chapter 23, pp. 547-570. Tobin, G.A. and B.E. Montz. 2009. Environmental Hazards. In International Encyclopedia of Human Geography, R. Kitchin and N. Thrift eds. Oxford: Elsevier. Volume 3, pp. 521-527. White, G.F. 1945. Human Adjustments to Floods. Department of Geography Research Paper 29. Chicago: University of Chicago Press. White, G.F. ed. 1961. Papers on Flood Problems. Department of Geography Research Paper 70. Chicago: University of Chicago Press. White, G.F. 1964. Choice of Adjustments to Floods. Department of Geography Research Paper 93. Chicago: University of Chicago Press. White, G.F., W.C. Calef, J.W. Hudson, H.W. Mayer, J.R. Sheaffer, and D.J. Volk. 1958. Changes in Urban Occupance of Flood Plains in the United States. Department of Geography Research Paper No. 57. Chicago: University of Chicago Press. Wisner, B., P. Blaikie, T. Cannon, and I. Davis. 2004. At Risk: Natural Hazards, People’s Vulnerability and Disasters (2nd ed.). London: Routledge.

PART I RISK AND VULNERABILITY

INTRODUCTORY COMMENTS GRAHAM A. TOBIN AND BURRELL E. MONTZ

As hazards research has evolved, both risk and vulnerability have evolved into key themes that now dominate the literature. In this regard, much attention has been paid to establishing and further developing a sound conceptual foundation for understanding, differentiating, characterizing, and measuring risk and vulnerability. At the same time, recognition of the significance of addressing both risk and vulnerability, separately and in combination, in emergency management, hazard mitigation, and loss reduction has captured the attention of applied geographers. The collection of papers in this section illustrates the varying concerns, methodologies, and approaches through which both risk and vulnerability have been studied. The first two papers in this section, originally published in 2001 and 2002 respectively, set the stage by focusing on risk as created by the interplay of forces of the geophysical environment and human vulnerability as defined by the locations of populations. Mulligan and Barbato mapped frequency and spatial distribution of tornadoes in the state of Texas to evaluate patterns of risk while Springer and Blanchard-Boehm evaluated differing perceptions of risk to flooding in one Texas city, San Marcos. This article is centered on risk communication, arguing that only through understanding of the risk perceptions of different ethnic groups can the risk be successfully communicated. A year later, we incorporated both geo-physical risk and socio-economic vulnerability and found that, at least in the Tampa, Florida region, vulnerable populations do not live in the most hazardous locations – contrary to what had often been assumed. Together, the case studies in these three papers reflect important aspects of understanding risk and vulnerability: what is likely to happen where, to whom, and how might that influence emergency management. The next set of papers in this section was originally published in 2006 and 2007 and focuses on how we might characterize risk and how people understand the risk to which they are exposed. Erik Bowles makes the case for a classification scheme for heat stress, as a means of communicating the potential risk. He also undertook an analysis of the frequency and

Introductory Comments

9

probability of heat stress occurrences for the study area – again with an eye to communicating risk. This theme is followed later in Bell’s work that evaluates how different ways of describing or defining the 100-year flood influences perceptions of risk. Communicating risk is based on understanding both the nature of the hazard and the characteristics and perceptions of the populations who are at risk. Webb focused on the elderly in Columbia County, Pennsylvania, an area that has a long flood history. One would probably assume, then, that the elderly understand the risks to which they are exposed, given their longevity in the county – however, this proves not to be the case. These findings have important applications to emergency management and planning. In her contribution, He evaluates the risk perceptions and vulnerability of international students – an often overlooked segment of the population. Knowing that both local knowledge and social networks can be important to understanding one’s risk and preparation options, she presents and tests a model to explain the evacuation decisions of international students. The theme of local knowledge, social networks and evacuation decisions is continued in the research that Tobin and his colleagues undertook in Mexico, in this case with respect to risk from a chronic hazard, a volcano. Published between 2009 and 2012, the final set of papers in this section, all of which are GIS-based, addresses measurement questions in evaluating both risk and vulnerability, each based on the argument that differing emphases on one or another of the relevant variables contributing to risk or vulnerability will lead to different results. Together, these papers make the case that significant thought and care must be used in selecting and measuring variables. Abel and Lein evaluated the extent to which anticipated patterns of losses, as determined by vulnerability maps developed using varying weighting strategies, compared to actual losses following Hurricane Katrina in New Orleans. Similarly, using one county in Texas, Liu and Gong illustrate how focusing on different socioeconomic characteristics leads to very different spatial patterns of vulnerability. Finally, Wu and Lu sought to refine the spatial prediction of frequencies and probabilities of hurricane strikes in South Carolina in order to assist emergency managers, planners, and insurance companies with detailed spatial information that has direct application to their responsibilities. Their results clearly illustrate the spatial dimensions of risk to hurricanes. Taken together, the papers in this section cover spatial and perceptual aspects of risk and vulnerability to a range of hazards, with various applications to emergency management, hazard mitigation, and planning. The methods used, including GIS, surveys, and focus groups, and the case

10

Part I

study approach are common to hazards research as it has been undertaken for decades. The lessons learned from the papers presented here have influenced subsequent research as we continue to improve our understanding of risks and vulnerabilities as they play out in different places.

CHAPTER TWO THE MAGNITUDE, FREQUENCY, AND SPATIAL DISTRIBUTION OF TORNADOES IN TEXAS: AN ASSESSMENT OF TORNADO RISK KEVIN R. MULLIGAN, LUCIA S. BARBATO AND STEPHEN WEINBECK

Introduction Understanding the risk posed by tornadoes is an important component of emergency planning and management efforts. In any given year tornadoes have the potential to cause millions of dollars in property damage and, in some cases, a significant number of injuries or loss of life. To understand the risk posed by tornadoes, it is important to understand the frequency and spatial distribution of tornadoes of varying magnitudes. Moreover, any assessment of tornado risk must take into account the density and spatial distribution of the population. The purpose of this paper is to analyze the magnitude, frequency and spatial distribution of tornadoes in the State of Texas and then apply this knowledge to an assessment of tornado risk. In the first part of this paper, maps are generated to show the spatial distribution of tornadoes in the state. These data are then used to create a map showing the density of strong or violent tornadoes by county. The aim is to identify those regions of the state that have a higher probability of significant tornadoes. In the second part of this paper, the county tornado data are combined with county population data to create a spatial model of tornado risk. The purpose of this analysis is to create a composite variable for tornado risk that takes into account: 1) the historical frequency of strong or violent

Papers and Proceedings of the Applied Geography Conferences (2001) 24: 121129

12

Chapter Two

tornado events, and 2) the likelihood that a tornado will cause significant property damage, injuries, or loss of life, given the population density within a county.

Tornado Data Background In 1987, Fujita published an exceptional map showing the spatial distribution of tornadoes in the United States (Fujita, 1987). The map was based on a dataset containing the magnitudes and locations of 23,264 tornadoes observed between 1930 and 1978. Given such a large dataset, the map was exceptional insofar as each tornado event was mapped by hand and the map took three years to construct. After Fujita's tornado map was completed, the University of Chicago Tornado Tape became available (Fujita, 1987). This dataset included the magnitudes and locations of 31,054 tornadoes observed between 1916 and 1985. Subsequently, Fujita used this dataset to produce computer generated grid-point tornado maps. Although Fujita used these grid-point maps to analyze both the spatial and temporal distribution of tornadoes in the United States, by today's standards, these maps appear unrefined and are difficult to interpret. In addition to Fujita's classic work, Grazulis has published what is perhaps the most comprehensive description of tornadoes in the United States (Grazulis, 1990). In the first volume of Significant Tornadoes, Grazulis provides a comprehensive analysis describing the spatial distribution of tornadoes mapped by state. In the second volume, Grazulis provides an exceptional chronology of tornadoes. In this volume, the author describes the location, time of day, magnitude, path length and width, and the damage associated with every known significant tornado in the United States from 1880 to 1989.

The SPC Tornado Archive Although the books authored by Fujita and Grazulis provide an excellent foundation for understanding the spatial and temporal distribution of tornadoes in the United States, it is difficult to analyze the original tornado data because these data are not readily available in a digital format (Fujita, 1987; Grazulis 1990). To overcome this problem, digital tornado data were obtained from the NOAA Storm Prediction Center (SPC). A project was then undertaken to assemble a tornado

Tornadoes in Texas: An Assessment of Tornado Risk

13

database for the United States that can be easily imported into a geographic information system (Weinbeck et al., 2000). By compiling tornado data in a GIS format, those interested in tornadoes can readily view, query, and analyze the tornado data to suit their needs. In a GIS format, users can zoom into an area of particular interest, and they can create custom maps that are suitable for printing. More advanced GIS users can also integrate the tornado dataset with other GIS data layers to complete more sophisticated types of overlay analysis or build spatial models. The SPC tornado archive is a digital database containing the magnitudes and locations of tornadoes observed between 1950 and 1995. The tornado database was compiled from two sources. The data for 195059 were compiled from National Weather Service Office reports published as part of the Climatological Data: National Summary (US Weather Bureau, 1950-1959). The tornado data for 1959-95 were derived from reports published as part of the Storm Data publication TD3910 (National Climate Center, 1959-1995). Within the SPC database, tornadoes are located by either a single latitude and longitude coordinate pair (single point tornadoes), or by multiple coordinate pairs (long track tornadoes). In the case of long track tornadoes, two coordinate pairs can be combined to form a tornado track line segment, and one or more line segments can be combined to locate the path of the tornado. In addition to the coordinate information, the SPC database contains more than thirty attribute fields including the time, date and year of the tornado event, the magnitude of the tornado on the Fujita scale, the length of the tornado path and the width of the tornado path. Given the difficulty of estimating tornado path length, path width and magnitude from postevent damage surveys, there is almost certainly some interpretation in the reporting of tornado attributes. This is especially true in the case of tornadoes that occur in non-urban environments where tornado damage is not as evident. Nevertheless, the SPC tornado archive represents the most complete digital database available (National Climate Center, 1959-1995).

The Nature of Tornadoes in Texas Magnitude and Frequency To analyze the magnitude, frequency and spatial distribution of tornadoes in Texas, the SPC database was queried to select only those tornado events that occurred within the state borders. Figure 2-1 shows the frequency of different magnitude tornadoes recorded between 1950 and

14

Chapter Two

1995. For T Texas, there are a 5633 sing gle point andd long track tornadoes t within the S SPC databasee. Of these, 4339 4 (77 perccent) are classsified as relatively w weak tornadoees (F0-F1), 1247 (22 perccent) are classsified as strong tornaadoes (F2-F3)) and 47 (lesss than 1 perccent) are classsified as violent (F4 aand F5).

Magnitude and frequency f of torrnadoes in Texaas Figure 2-1. M

Althoughh the data in Figure 2-1 sh how a decreaase in the freq quency of higher magnnitude tornaddoes, it is imp portant to reccognize that both b path length and path width teend to increaase with F-scaale (Fujita, 1987). To analyze the pprobability off being struck by different m magnitude torn nadoes, it is therefore important to consider thee total area sttruck by each h class of tornado usinng the path length multip plied by the path width for each tornado. In ssome cases, thhe average patth length or avverage path width w for a given magnnitude tornadoo was substiituted if one of these vaalues was missing. In Texass, F2, F3 and F4 tornadoes affect larger areas, even th hough F0 and F1 tornaadoes are morre common (F Figure 2-2). Th This result suggests that the probabillity of being struck s by F2, F3 or F4 torrnadoes is greeater than the probabillity of being struck s by an F0 F or F1. F5 ttornadoes tend to have long path llengths and path widths, but the low w frequency of these tornadoes reesults in a muuch smaller total t area beinng struck. In terms of risk, the historical recordd in Texas sug ggests that F22, F3 and F4 tornadoes t are the mostt hazardous. When W the larg ge area struckk by these torrnadoes is combined w with their viollent and destrructive force, the element of risk is very high inn terms of propperty damage,, injury and looss of life.

Tornadoes in Texas: An Asssessment of Torrnado Risk

15

Area struck by toornadoes by maagnitude Figure 2-2. A

Spatial Disttribution Figure 22-3 shows thee distribution of all singlee point and lo ong track tornadoes (F F0-F5) recordeed between 19 950 and 1995 . Based upon this map, it becomes aapparent that the state can be b roughly diivided into tw wo regions with respectt to the occurrrence of torn nadoes. Alongg the Mexico border in the southweestern part of the t state, the frequency f of toornadoes appeears to be significantlyy lower whenn compared to o the northerrn, central and eastern parts of the state. Moreovver, a visual in nterpretation oof the data su uggests an almost randoom distributioon of tornadoees in the non-bborder region. To interppret these resuults correctly, it is importannt to keep in mind m that data concerrning the frequency of tornadoes caan be influeenced by population density. In thhe earlier yeaars before Dooppler radar (1950 to about 1990), it is likely thhat many tornaadoes in the soouthwestern part p of the state were nnot observed given the reelatively low population density d in much of thiis part of thee state. Neverrtheless, the ddifference bettween the border regioon and the resst of the statee is quite pronnounced, and d it seems likely that thhe difference is not solely an artifact off observation bias. The relatively loow frequencyy of tornadoes along the bborder region n is most likely the rresult of som me observatio on bias and the dry atm mospheric conditions thhat tend to dominate the cliimate in this aarid region. In termss of tornado risk, r it is also o important too note that F0 and F1 tornadoes do relatively liittle damage. In fact, a torrnado's rankin ng on the Fujita scale is based on a post-event daamage survey and, by defin nition, the damage is m minor in the case c of F0 and d F1 tornadoees. To evaluate the risk to life and pproperty, it is therefore best to examine the spatial disstribution of higher maagnitude tornaadoes (Grazulis, 1990).

16

Chapter Two

Figure 2-3. Sppatial distributiion of tornadoess in Texas

Figure 22-4 shows the spatial distrib bution of stronng (F2-F3) an nd violent (F4-F5) tornnadoes in the state. Althoug gh the overalll density of sttrong and violent tornaadoes is muchh lower, the spatial s patternn is very similar. Once again, alongg the Mexico border b region,, the frequencyy of strong an nd violent tornadoes iss quite low. In I the northerrn, central annd eastern parrts of the state, the freequency of strong or violentt tornadoes is significantly greater.

Historical Assessmen nt of Tornaado Risk Metho ods For a pparticular counnty, the risk of significannt damage caaused by tornadoes iss largely a funnction of two variables: 1) the county population density andd 2) the probaability of a strong s or viollent tornado occurring o within the ccounty. Simplly stated, the risk of propeerty damage, injury or death shouldd increase direectly with eith her an increasee in populatio on density or an increasse in the frequuency of tornaadoes. To analyyze tornado rissk, a spatial model m was connstructed that takes t into account bothh of these varriables. The po opulation dennsity for each county in the state w was mapped based b on 1999 census daata. These population

Tornadoes in Texas: An Asssessment of Torrnado Risk

17

densities weere then weighted based on the probaability of a strong s or violent tornado occurringg in the county. To create this weightin ng factor, the historicaal record of significant s tornadoes was uused. For each h county, the total num mber of recorrded strong an nd violent tornnadoes was divided by the county aarea to calcullate the tornad do density. Thhe tornado deensity for each countyy is expressedd here as the number n of siggnificant tornaadoes per 100 square m miles. The couunty tornado densities d weree then grouped d into ten classes usingg an equal inteerval classificcation.

nd violent tornaddoes Figure 2-4. Sppatial distributiion of strong an

Table 2--1 shows the weighting faactors based oon the county y tornado density. Althhough the usee of ten classees is somewhaat arbitrary, itt provides for a 10 perrcent increasee in the weigh hting factor too distinguish one class range from tthe next. To calculate c relatiive tornado rissk for those co ounties in the highest class, the coounty population densities were multipllied by a weighting ffactor of 1. To T calculate the t relative ttornado risk for those counties in the second highest classs, the popuulation densitties were multiplied bby a weightinng factor of 0.9. 0 For thosee counties in the third highest classs, population densities d weree multiplied byy a weighting g factor of 0.8, and so oon. The ratioonale for usinng this weighting scheme iis quite simplle. If two similar sizedd counties haave the same population deensity, and on ne county

Chapter Two

18

has historically had twice as many significant (F2-F5) tornadoes, then presumably the tornado risk for that county is twice as great. Once again, there should be a linear relationship between tornado density and risk of significant damage for counties with the same area and population density. Table 2-1. Weighting factors based upon tornado density Tornado Density Class 1 2 3 4 5 6 7 8 9 10

Class Range (Tornadoes/100mi2 0.000-0.343 0.343-0.686 0.686-1.029 1.029-1.372 1.372-1.715 1.715-2.058 2.058-2.401 2.401-2.744 2.744-3.087 3.087-3.430

Weighting Factor 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Counties in Class (N) 97 53 36 30 12 13 6 4 1 2

Mapping Population and Tornado Densities Figure 2-5 shows the population density of counties in Texas. The greatest population densities are obviously associated with the major urban centers in the state including Dallas, Fort Worth, Houston, Austin, San Antonio, El Paso, Lubbock and Amarillo. In addition, population densities are relatively high along the I-35 corridor connecting the Dallas-Fort Worth area to Austin and San Antonio. Given the higher population densities and the built environment of these urban centers, the potential for significant damage, injury, and loss of life is obviously greater when compared to more rural counties with lower population densities. Figure 2-6 shows the frequency of significant tornadoes by county in the state, with the number of significant tornadoes normalized by the county area. These are the same data shown in Figure 2-4 expressed as the density of strong or violent tornados per 100 square miles. For cartographic reasons, the ten classes described in Table 2-1 are shown as five categories in the legend of Figure 2-6. As might be expected, the counties along the Mexico border region tend to have lower tornado densities and those in the northern, central and eastern parts of the state tend to have higher tornado densities. For counties with a high population density in the southwestern part of the state, the lower tornado densities should reduce the overall tornado risk.

Tornadoes in Texas: An Asssessment of Torrnado Risk

unties Figure 2-5. Poopulation densiity of Texas cou

Figure 2-6. Tornado density in Texas countties

19

20

Chapter Two

Resu ults To assesss the relative risk of signifiicant tornado ddamage for co ounties in the state, a dimensionleess risk valu ue was derivved using the county population ddensity weightted by the torn nado density. For example, the 1999 population ddensity of Luubbock Countty was 251 ppeople per squ uare mile and the tornnado density iss 1.98 significcant tornadoess per 100 squaare miles. From Table 2-1, the tornnado density weighting w facttor is 0.6. Mu ultiplying the populatiion density byy the weightin ng factor yieldds a relative risk r value of 151. Figure 22-7 shows thee relative torn nado risk for all counties. Because county popuulation densitiies tend to varry over severaal orders of magnitude, m with many rrural counties having a low population deensity and a few f urban counties havving very a high h populatio on density, thee risk values also vary over severall orders of magnitude.

Relative tornadoo risk in Texas Figure 2-7. R

As mighht be expected, the relative tornado riskk map (Figurre 2-7) is similar to thhe populationn density map p (Figure 2-5)). The greatest risk of significant ddamage is asssociated with h the major urban countiees in the northern, central and easttern parts of th he state. Thesee major urban n counties tend to havve both a higgh population density and a high probaability of tornado occuurrence (exprressed as the tornado t densitty). For exam mple, both

Tornadoes in Texas: An Assessment of Tornado Risk

21

Dallas County and Tarrant County (the Dallas- Fort Worth area) have a very high tornado risk. These particular counties have both a very high population density and, historically, have seen a high frequency of strong or violent tornado events. Those major urban centers in the southwestern part of the state tend to have a lower tornado risk because the density of significant tornadoes in this part of the state tends to be lower. El Paso, for example, has a very high population density, but the tornado risk is an order of magnitude lower because the historical record suggests that the probability of significant tornadoes is very low. The lowest risk value (0.017) is associated with Loving County. In this case, the county population density is very low and there are no recorded F2-F5 tornadoes in the database.

Discussion This analysis has shown how relative risk of significant tornado damage varies spatially across the State of Texas based solely on the population density of counties and the historical tornado density. Another factor that might also play an important role in assessing tornado risk is the integrity of building construction. In particular, it is widely recognized that mobile homes are very susceptible to tornado damage and provide relatively little protection in the case of a strong or violent tornado. During the course of this study, mobile home density was considered as a variable in the spatial model. The results, however, tend to show that mobile home density is closely correlated with population density. In other words, the counties with the greatest number of mobile homes tend to be the counties with the greatest population densities. Nevertheless, this assessment of tornado risk might be significantly improved if mobile home density was incorporated as part of the risk factor for less populated rural counties.

Conclusions The purpose of this paper was to analyze the magnitude, frequency and spatial distribution of tornadoes in Texas. The tornado data used in this study were obtained from the NOAA Storm Prediction Center (SPC). This digital archive contains tornado data for the United States recorded between 1950 and 1995. Within the SPC dataset, tornado events are recorded as latitude and longitude coordinates and classified according to the Fujita scale. To examine the spatial distribution of tornadoes in Texas, these data were mapped and analyzed in a GIS environment. The results from this analysis suggest that the Mexico border region has a significantly

22

Chapter Two

lower occurrence of strong and violent tornadoes. In the northern, central and eastern parts of the state, the historical record shows a significantly higher frequency of potentially destructive tornadoes. To assess tornado risk in the state, a spatial model was constructed to incorporate both the population density of counties and the probability of significant tornadoes based on the historical record. The results from this analysis strongly suggest that the greatest tornado risk in the state is associated with the high population densities of the urban counties in the northern, central and eastern parts of the state. In the southwestern part of the state, counties with high population densities tend to have a lower risk given the relatively low frequency of tornado occurrence.

References Fujita, T.T. 1987. U.S Tornadoes, SMRP Research Paper Number 218. Chicago: The University of Chicago, Satellite and Mesometeorology Research Project. Grazulis, T.P. 1990. Significant Tornadoes 1880- 1991. Vol. 1 and 2. St. Johnsbury, Vermont: Environmental Films. National Climate Center, 1959-1995. Storm Data. National Atmospheric and Oceanic Administration, Environmental Data and Information Service, TD3910. U.S. Weather Bureau, 1950-1959. Climatological Data: National Summary, Vol. 1-10. Weinbeck, S., R. Peterson, A. Doggett and K.R. Mulligan. 2000. Using GIS to Generate Spatial Statistics for Tornado Occurrences. In 20th Conference on Severe Local Storms, Orlando, Florida. Boston: American Meteorological Society.

CHAPTER THREE DISASTER WARNINGS IN SAN MARCOS, TEXAS: COMMUNICATING IN ETHNICALLY DIVERSE PLACES CATHRYN E. SPRINGER AND R. DENISE BLANCHARD-BOEHM

Introduction In October of 1998, Central Texans experienced a severe flash flood event, in which 31 people drowned. For the San Marcos community, the “Great Flood of 1998” threatened many lives, caused more than $12 million in damages, and tested the city’s disaster preparedness and response. During the flood event, the town experienced chaos because of a lack of effective communication. Many individuals did not know how, or when, to respond to the flooding, and some residents were even unaware of the impending disaster. This lack of communication during the 1998 flood prompted this research. Risk communication efforts face a complexity of issues depending on the audience’s needs and characteristics. In an effort to reach everyone in the public, “emergency planners are tailoring programs to meet the needs of populations that are particularly vulnerable to disasters . . . and to take [into] account the varying cultural and linguistic requirements of the communities’ members” (Drabek, 1996). Special target audiences of risk communicators include groups such as the elderly, children, minorities, and tourists. San Marcos consists of several of these vulnerable populations, particularly minority and non-English speaking audiences. This study examines the influence of ethnicity on communicating urgent,

Papers and Proceedings of the Applied Geography Conferences (2002) 25: 334342

24

Chapter Three

short-term disaster warnings in San Marcos, Texas. This paper focuses on only the differences that were found between Anglo and Hispanic citizens in their risk communication behaviors. From these findings, disseminators of warning messages might consider ways to target and tailor vital information so that the receivers effectively internalize and engage in action to save their lives and properties.

Study Area and Background San Marcos is located in central Texas, along Interstate 35, approximately 30 miles south of Austin and 45 miles north of San Antonio. This area, known as the Austin-San Antonio Corridor, has a propensity for flash flooding and is experiencing tremendous population growth. Together, these factors increase the risks associated with flood events and pose a challenge to emergency and city planners. Two main streams flow through San Marcos, the Blanco and the San Marcos rivers. Flood flow for these rivers is produced from ephemeral streams in the central Texas region. The physical geography of the central Texas region causes it to be especially prone to flash flooding. The numerous creeks, limestone bedrock, steep slopes, and shallow, clay soils all contribute to large amounts of runoff. Furthermore, the climate provides ample opportunities for large rain events from mid-latitude cyclones that move across the continental United States bringing fronts southward, and from tropical cyclones that travel through the Gulf of Mexico and along the west coast of Mexico. In addition, water can easily become “backed up” from the confluence of the Blanco and San Marcos rivers, which is located just south and east of San Marcos. Figure 3-1 illustrates the location of the City of San Marcos, as well as the location of the San Marcos and Blanco rivers and their confluence in relation to the city. From 1990 to 2000, the San Marcos population grew 21 percent, Austin, 41 percent, and San Antonio, 22 percent. All three cities consist of two dominant ethnic groups: Hispanic and White (non-Hispanic or Anglo). In 2000, San Marcos had a population of 34,733 people; 36 percent were Hispanic and 55 percent were non-Hispanic Whites (US Census, 2000). Although both Austin and San Antonio consist of a substantial Hispanic population, the two cities differ in that the Hispanic population in Austin is 31 percent of the total population, whereas in San Antonio it comprises 59 percent of the total population. Table 3-1 displays the ethnic composition of these three cities.

Disaster Warnings in San Marcos, Texas

25

Figure 3-1. Study area, San Marcos, Texas (Source: Department of Geography, Texas State University-San Marcos) Table 3-1. Ethnic composition of study area population in 2000 Ethnic Group Non-Hispanic, White Non-Hispanic, Black Hispanic Other

Austin 53 10 31 6

San Marcos 55 5 36 4

San Antonio 32 7 59 2

Flood of October 1998 The “Great Flood of 1998” occurred during the weekend of October 17-19, 1998. Two hurricanes off the west coast of Mexico pumped moisture into the region, and a stalled cold front provided continuous uplift. These meteorological conditions produced the largest recorded rain event in San Marcos. The city received more than 25 inches of rain during the flood—15 inches fell within a 24 hour period (Earl and Wood, 2002) Weather forecasters had predicted a chance of rain but did not expect a large rain event for the third weekend of October 1998. Thus, the heavy rainfall event which produced record flooding surprised residents and city officials. Several factors contributed to the confusion that occurred during the flood. The rivers’ overflow divided the city into two parts, stranding some people away from their destinations. Many at home were without electricity. Emergency response crews and residents alike knew little about

26

Chapter Three

where to turn for lifesaving information, such as evacuation locations, what roads were closed, and what to expect of the rising river. Evacuated residents were transported to shelters in university buses, school buses, and dump trucks only to be transported again because of shelters next to rising creek waters being flooded. In addition, San Marcos had limited media resources as the city is situated outside of both Austin and San Antonio news coverage areas. Residents who did have power during the flood relied mainly on television and radio stations from San Antonio and Austin; however, these stations only broadcast short descriptions of flooding in San Marcos and did not offer information regarding street closures, shelters, or important phone numbers for people in San Marcos. This event indicated a risk communication weakness to the city’s emergency planners (O’Leary, 1999) As a result of this event, the city has discussed implementing a radio station for communication during disaster events such as the 1998 flood.

Risk Communication: Theory and Research The General Risk Communication Model (GRCM), developed by Sorensen and Mileti (1987) and revised by Blanchard (1992), illustrates the process and main factors involved in communicating risk: the actual message itself, those sending the message, the receivers of the message, and the responses to those messages. The model is tailored to risk communication by assigning specific criteria to each of these factors and by identifying key components in the process of risk communication. The four components of the GRCM include: 1) message/sender characteristics, 2) receiver characteristics, 3) receivers’ perceptions of risk, and 4) receivers’ response behavior. Currently, the model places ethnicity within the receivers’ characteristics component. The following studies, discussed in the context of the United States, consider the majority population to be Anglo or White and minority populations to be Hispanic, AfricanAmerican, or Asian.

Influence of Ethnicity on Risk Perception Prior to the 1980s, larger themes of study in hazards research subsumed issues of ethnicity, gender, and special populations. Investigation of these sub-populations has developed over the past 15 years; however, these studies present mixed conclusions. Several studies have suggested that minorities have lower levels of individual risk perception than Anglos. More specifically, these studies indicated that

Disaster Warnings in San Marcos, Texas

27

Mexican-Americans perceive dangers at lower levels than Anglos even when warnings have been heard (Perry and Green, 1982; Perry and Mushkatel, 1984). Contrarily, studies of the 1989 Loma Prieta earthquake in the San Francisco Bay area reported that Mexican-Americans had higher levels of individual risk perception (Blanchard, 1992; Fothergill et al., 1999). Other studies suggested that: 1) African-Americans appear more fatalistic than Anglos and Mexican-Americans, 2) Anglo males seem the least concerned about natural hazard risks, and 3) African-Americans tend to perceive a high risk of future damage to their homes (Blanchard, 1992; Fothergill et al. 1999). However, Ives and Furseth (1983), in their study of Charlotte, North Carolina, found no statistically significant differences in levels of risk perception between various ethnic groups. Researchers have attributed some of these conflicting findings to other variables that seem to interact with ethnicity. Perry and Green (1982) suggested that minorities appear to define danger differently than Anglos; thus contributing to varying levels of perceived danger. Furthermore, Sokolowska and Tuszka’s (1995) study of perception in lesser-developed nations indicated that minorities have a “real” perception of danger and are just as concerned about environmental risks as Anglos. In addition, researchers attributed the higher risk perception of Mexican-Americans in the 1989 Loma Prieta earthquake to their knowledge of, and experience with, the 1985 Mexico City earthquake (Blanchard, 1992). However, Drabek (1986) concluded that both minorities and Anglos have higher levels of perceptions of personal vulnerability when warned by a credible source. This agglomeration of findings supports the proposition that each sub-population has unique characteristics and that ethnic groups perceive their vulnerability to various types of individual risk in diverse ways.

Influence of Ethnicity on Message-Receiver Interactions In several studies, Perry and colleagues concluded that minorities have less faith in warnings than Anglos (Perry and Green, 1982; Perry and Mushkatel, 1984). Factors known to influence individuals’ faith in warnings are those of sources and channels. Message sources refer to the person or people responsible for message dissemination, whereas, message channels pertain to types of media through which disseminators transfer messages. However, a message source may sometimes act as a medium as well. Ethnic groups seem to have different perceptions as to what constitutes “credible sources.” From the literature, these generalizations include: 1) Anglos receive information from formal, English speaking sources, 2) Latin-Americans obtain risk information from informal sources

28

Chapter Three

such as friends and family, and 3) African-Americans and Hispanics are more likely to use social networks and relatives for risk communication (Drabek, 1986; Fothergill et al., 1999). Perry and Lindell (1991) also found that, with short-term warnings, both minorities and Anglos view authorities as the credible source. However, in a study of Hurricane Andrew, urban families, despite ethnicity, showed dependency on family relations for information during the disaster (Peacock et al., 1997). This finding supports claims that ethnicity interacts with other cultural and environmental factors. Faith in warning messages also seems to be influenced by the recipients’ confirmation of the message. From their study, Perry and Lindell (1991) found that Mexican-Americans tend to confirm messages more than Anglos and contact a higher number of confirmation sources than Anglos. Findings regarding faith in warnings may be a result of other influential variables, such as level of community involvement, socioeconomic conditions, forewarnings, event familiarity and perceptions of message source (Drabek, 1986; Perry and Green, 1982).

Influence of Ethnicity on Behavior towards Disasters The majority of literature that discusses ethnicity and disaster behavior addresses the response stage of an event. The sub-field of mitigation and preparation as it impacts ethnicity is limited in research, but some differences are still evident. Studies have indicated that Anglos: 1) make more structural changes, 2) are more likely to develop a plan, 3) have more opportunities for hazard education, and 4) obtain more adequate insurance than minorities (Fothergill et al., 1999; Peacock et al., 1997). Studies in the aftermath of Hurricane Andrew exposed insurance “redlining,” a practice by insurance companies where purchasing opportunities are provided only to certain ethnic groups. Thus minorities in the area, African-Americans and Mexican-Americans, were less likely to be insured by one of the top three companies in the nation. In addition, this study found that African-Americans were the least likely to store food and supplies, while Asians were the least likely to have a plan (Peacock et al., 1997). Hazards researchers concluded that because minorities have more limited financial resources than Anglos, they are unable to make structural changes and invest in adequate insurance. The most extensive portion of the literature in ethnicity and warning response addressed evacuation practices. Results from studies conflict regarding minority evacuation compliance. Perry and Lindell (1991) suggested that ethnicity does not affect evacuation compliance but does

Disaster Warnings in San Marcos, Texas

29

influence individual motivation and emergency decision-making processes. Their study concluded that evacuation compliance interacts with risk perception, planning, and family context. However, several studies indicated that minorities are less likely to evacuate than Anglos (Drabek, 1986; Fothergill et al., 1999; Perry and Green, 1982; Perry and Mushkatel, 1984). Further, these researchers demonstrated that during a disaster, Mexican-Americans tend to take protective actions instead of evacuating (Perry and Mushkatel, 1984). In reviewing the literature on ethnicity and response, it is important to keep in mind certain variables that appear to interact with ethnicity. Studies have suggested that minorities tend to have stronger family ties, have a lower economic status, and are less involved in the community than Anglos (Drabek, 1986; Perry and Mushkatel, 1984). These variables might aid in explaining disaster response differentials of various ethnic groups. As seen in this review, no overriding trends have emerged in the literature on risk communication and ethnicity.

Data and Methods Data were collected from San Marcos residents through telephone interviews. Using the 1999 San Marcos Telephone Directory, names were stratified according to ethnicity. Hispanic surnames were identified with a U.S. Census Bureau Hispanic surname list. Every fiftieth name was included in the sample. Before interviews began, individuals received a letter informing them of the study and forthcoming telephone interview. The survey measured variables through questions that asked respondents to rate certain items on a scale of 1 to 4. In these questions, “4” represented the most positive end of the spectrum, for example, “very effective,” and “1” marked the negative end such as “very ineffective.” Other questions asked for “yes” or “no” answers, which were assigned values of 1 and 0 respectively. Television and radio stations were ranked by their latitudinal location: Austin represented the northernmost stations, while San Antonio represented the southernmost stations. Respondents were also asked to identify their ethnicity as “Anglo,” “Hispanic,” “African-American,” or “Other.” The data were analyzed using the nonparametric Wilcoxon Rank Sum Test for Independent Samples which tests the null hypothesis that there is no statistically significant difference between the means of two independent samples. The survey instrument gathered data for 58 variables. These variables examined each of the four main components of the GRCM: perceptions of risk, receiver characteristics, message/sender characteristics, and response and preparation behavior.

Chapter Three

30

Results and Discussion A total of 60 residents participated in the telephone interview, with 48 percent of respondents identifying themselves as Hispanic and 52 percent as Anglo. Of the 58 variables tested, nine of them showed statistically significant differences between Anglo and Hispanic respondents. Table 32 displays the means and Z-scores for the nine variables that tested significant at the .05 level or below. Table 3-2. Variables significant at the 0.05 level or below Variable

Mean Hispanic

Mean Anglo

Z-Score

Scale 1-4

3.38

2.55

2.666**

Scale 1-4 0-no;1-yes 0-no;1-yes 1-Austin; 2both; 3-San Antonio

3.21 0.52 0.69 2.93

2.72 0.19 0.26 2.19

2.500** -2.324* -3.50*** 2.299*

0.58

-2.121*

3.62 0.06 0.23

-2.650* 2.449* 2.236*

Measure

Perception of Risk Likelihood of future October 1998-type flood Message-Sender Characteristics Effectiveness of radio Friends/family trustworthy Neighbors trustworthy Radio station location

Response/Preparation Behavior Knew if lived in floodplain 0-no;1-yes 0.41 Previous Exposure Distance from river Miles 2.29 Live in 100-year floodplain 0-no;1-yes 0.50 Live in 500-year floodplain 0-no;1-yes 0.70 Levels of significance as follows: *=.05, **=.01, ***=.001

Perception of Risk and Previous Exposure The risk perception component of the GRCM contained one statistically significant variable—respondents’ opinions toward the likelihood of a 1998-type flood occurring in San Marcos again within the next 10 years. Approximately 90 percent of Hispanics and 52 percent of Anglos in the study believed that the recurrence of a 1998-type flood was “somewhat likely” or “very likely.” The difference between the two means was 3.38 for Hispanic and 2.55 for Anglo, and was significant at the .01 level. Respondents’ proximity to the river was considered a possible

Disaster Warnings in San Marcos, Texas

31

explanation for discrepancies in risk perceptions between the two ethnic groups. The analysis indicated that Hispanics lived closer to the river than Anglos, and that more Hispanic respondents than Anglos lived within the 100-year or 500-year floodplain.

Message-Sender Characteristics Anglo and Hispanic respondents differed significantly for four of the message/sender characteristics variables: effectiveness of radio, location of radio stations, trust in neighbors, and trust in family and friends. Approximately 75 percent of Hispanic participants and 59 percent of Anglo participants viewed radio as a “somewhat effective” or “very effective” channel for receiving warning messages. The means of the two groups, which were 3.21 for Hispanic and 2.72 for Anglo, were significantly different at the .01 level. In addition, these respondents seemed to rely on emergency information mostly from radio, and radio stations located in different cities. About 55 percent of Hispanic respondents listened to stations from both Austin and San Antonio, whereas only 23 percent of Anglo respondents listened to stations from both cities. Approximately 45 percent of Anglo respondents preferred to listen to radio stations from only Austin, and just 17 percent of Hispanics listened to radio stations from Austin alone. Hispanic and Anglo respondents differed in their preference for radio station locations at the .01 level. This difference is attributed to the nature and composition of the two cities. San Antonio consists of a dominantly Hispanic and middle class population and has a cultural atmosphere that embraces the heritage of Hispanics from Mexico. In contrast, Austin, the capital of Texas and home to the University of Texas, thrives with a generally young and affluent population that pursues activities of the arts and environment. More Anglo than Hispanic participants judged neighbors and family and friends to be trustworthy, or credible, sources for disaster warning information. About 48 percent of Hispanic respondents and 81 percent of Anglo respondents felt that friends and family were trustworthy sources, and approximately 31 percent of Hispanic respondents and 74 percent of Anglos thought that neighbors were trustworthy sources. Differences found for trust in friends and family and trust in neighbors were significant at the .05 and .001 levels, respectively. These trends suggest that Hispanic participants depend more on formal sources, such as television and radio, than on informal sources, such as neighbors, friends, and family, whereas, Anglo participants rely on both formal and informal sources.

32

Chapter Three

Response and Preparation Behavior From the analysis of response and preparation behavior, the variables of knowledge of location in floodplain and distance from the river emerged as statistically significant. About 58 percent of Anglo participants and 41 percent of Hispanic participants exhibited knowledge of their location in relation to the floodplain. This difference was significant at the .05 level. Characteristics of the City of San Marcos may cause differences between the two ethnic groups in their preparation behavior. The floodplain boundaries of San Marcos, established by the Federal Emergency Management Agency (FEMA), have served as sources of heated debates between residents of San Marcos and emergency management experts and have changed periodically during the building of five retention dams in San Marcos. Furthermore, damage assessments of the 1998 Central Texas flood showed that a large percentage of damage claims occurred outside of these boundaries. These shifting floodplain boundaries and unpredictable advances of the river, joined with the evidence that the Hispanic respondents live closer to the river, suggest that the Hispanic population may be more familiar with the fluctuation in river levels and physical water boundaries but not with the predefined FEMA boundaries. Thus, it is speculated that the Hispanic population perceives no need for the knowledge of floodplain boundaries.

Conclusions Two main conclusions are drawn from this study. First, the study reinforces assertions that messages need to be tailored for a defined audience according to its characteristics and that community profiles should be used in defining these characteristics to facilitate communication operations and programs (Drabek, 1986; Nathe et al., 1999). Second, the role of ethnicity and its actual level of influence within risk communication depends upon its interaction with other variables pertaining to the cultural and physical attributes of a place and its inhabitants. In some contexts, ethnicity may have a tremendous impact on risk communication behavior, whereas, it may have no effect in other situations. For San Marcos, Texas, ethnicity does appear to affect risk communication behavior. Messages that are directed toward a particular target audience seem to be the most effective for ensuring successful risk communication (Blanchard, 1992; Drabek, 1996; Handmer and Penning-Rowsell, 1990; Nathe et al., 1999). Therefore, it is advised that developers of risk

Disaster Warnings in San Marcos, Texas

33

communication programs and materials consider attributes of the target audience. The City of San Marcos would benefit from the development of a more comprehensive community profile than what this study provides. Based on the results of this study, the community profile should take into account the ethnic diversity of San Marcos. The lack of communication during disaster events, such as the Flood of October 1998, could be mitigated by implementing disaster warning systems tailored specifically for the San Marcos community. Furthermore, it is recommended that the City of San Marcos educate its floodplain inhabitants of their locations and of important laws that pertain to people living in floodplains. From this study, the existing GRCM (Blanchard, 1992) is modified in the very first component—message creation—in which the message, or messages, is/are tailored toward particular target audiences that reflect attributes identified in a community profile. Figure 3-2 shows the revised model. The target audience component is placed, as its own entity, before the dissemination of the message and is called “Message Created for Target Audience of Defined Place.” Examples of characteristics to be included in a community profile are listed below the title of this component. The placement of this new component signifies the necessity of developing appropriate messages and properly disseminating these messages, given the context of communication. Ethnicity is one of the cultural attributes within the new target audience component that developers of risk communication efforts need to consider in a community profile. Other characteristics that should be considered in a community profile but are beyond the scope of this study address the political, cultural, and social conditions of a targeted audience (Blanchard, 1992; Fothergill et al., 1999; Handmer and Penning-Rowsell, 1990). This study recommends future research to analyze the influence and role of these audience characteristics with the GRCM. Characteristics to consider include social-networks, values, receivers’ political identification and beliefs, political infrastructures, organization of emergency management, economic status, and social-class identification. This study has shown that ethnicity plays a significant role in risk communication behaviors of residents in San Marcos, Texas. These findings suggest that other communities might consider ethnic diversity in developing and implementing risk communication efforts. By examining groups separately, researchers can identify important characteristics of a population that determine the structure and success of emergency management. By studying these attributes, emergency planners will be able to develop target audience profiles of communities and to establish effective risk communication programs.

34

Chapter Three

Revised general risk communiccation model (S Source: Adapteed from Figure 3-2. R Blanchard, R. D. 1992)

References Blanchard, R. D. 1992. Risk Commu unication andd Individual Response: R Impact oof the 1990 Revised Earrthquake Probbabilities for the San Franciscco Bay Areaa. Ph.D. Disssertation, Unniversity of Colorado, C Boulder.. Drabek, T.E E. 1986. Human Systems Responses R to D Disaster: An Inventory I of Sociollogical Findinngs. London: Springer-Verl S lag. —. 1996. Thhe Social Dim mensions of Disaster: D A FE EMA Higher Education E Course. Emmitsburg, MD: Federal Emergency M Management Agency. A Earl, R.A., and C.R. Woood, 2002. Upstream U Channges and Dow wnstream Effects oof the San Marcos M River of o Central Teexas. Texas Journal Jo of Science 554(1): 69-88. Fothergill, A A., E. Maestass, and J.D. Daarlington. 19999. Race, Ethn nicity and Disasterss in the Uniteed States: A Review R of thee Literature. Disasters D 23(2): 1556-173.

Disaster Warnings in San Marcos, Texas

35

Handmer, J., and E. Penning-Rowsell. 1990. Hazards and the Communication of Risk. Great Britain: Gower Technical. Ives, S.M., and O.J. Furseth. 1983. Immediate Response to Headwater Flooding in Charlotte, North Carolina. Environment and Behavior 15(4): 512-525. Nathe, S., P. Gori, M. Greene, E. Lemersal, and D. Mileti. 1999. Public Education for Earthquake Hazards. Natural Hazards Informer 23(2): 1-11. O’Leary, D. 1999. October 1998 Flood San Marcos, Texas. San Marcos, TX: City of San Marcos. Peacock, W., B.H. Morrow, and H. Gladwin, eds. 1997. Hurricane Andrew: Ethnicity, Gender and the Sociology of Disasters. New York: Routledge. Perry, R.W., and M.R. Green. 1982. The Role of Ethnicity in the Emergency Decision-Making Process. Sociological Inquiry 52(4): 306334. Perry, R.W., and M.K. Lindell. 1991. The Effects of Ethnicity on Evacuation Decision-Making. International Journal of Mass Emergencies 9(1): 47-68. Perry, R.W., and A.H. Mushkatel. 1984. Disaster Management: Warning Response and Community Relocation. Connecticut: Quorum Books. Sokolowska, J., and T. Tuszka. 1995. Perception and Acceptance of Technological and Environmental Risks: Why Are Poor Countries Less Concerned? Risk Analysis 15(6): 733-743. Sorensen, J.H., and D.S. Mileti. 1987. Public Response to Emergency Warnings. Reston, VA: United States Geological Survey, Department of the Interior. United States Census Bureau. 2000. Census 2000 Summary File 1. http://factfinder.census.gov.

CHAPTER FOUR HAZARDOUSNESS OF THE TAMPA BAY REGION: EVALUATING PHYSICAL RISK AND SOCIO-ECONOMIC VULNERABILITY BURRELL E. MONTZ AND GRAHAM A. TOBIN

Introduction Much attention in hazards research has recently centered on risk and vulnerability, including how it should be defined, how it varies from place to place, how to measure it, and how to portray it. While distinctions between geo-physical risk and social vulnerability have been made, most work generally recognizes that both are important. Indeed, there is an evergrowing body of research that addresses hazardousness and particularly vulnerability as the intersection of geo-physical conditions and social systems (Cutter et al., 2000; Dow, 1992; Liverman, 1990; Montz and Evans, 2001; Tobin and Montz, 2004). Still, questions remain about the extent to which each plays a role, particularly when considering the range of hazards to which populations and settlements are exposed at a specific location. Further, even within each category of risk and vulnerability, there is discussion about what are the dominant contributing factors. Thus, hazardousness and vulnerability are not well understood. An additional problem in addressing vulnerability is the extent to which it is under or overestimated (The H. John Heinz III Center, 2000). This can result from a lack of accurate data, from the way in which the available data are incorporated, or from some combination. The rapid growth in populations and development in many coastal areas has exacerbated this problem, and suggests a need for the continued revisiting

Papers of the Applied Geography Conferences, (2003) 26:380-388

Hazardousness of the Tampa Bay Region

37

of vulnerability analyses, both in terms of accuracy of the data and determining appropriate measures of vulnerability. The research presented here approaches these issues by evaluating spatial variations in both geo-physical risk and social vulnerability, at a specific location, and then combining them to assess the hazardousness of that location. In doing so, this project is aimed at answering three questions: 1. How does the spatial distribution of vulnerable populations and land uses compare to the spatial distribution of geo-physical risk? 2. To what extent does changing the measures of social vulnerability alter patterns of hazardousness? 3. How might the responses to the previous questions affect emergency management and planning approaches?

Study Area The coastal and near coastal areas of the Tampa (Florida) metropolitan region serve as the study area for this project. The region was chosen because it incorporates a range of hazard probabilities, including areas subject to storm surge, coastal flooding and other hazards associated with hurricanes, as well as areas with relatively less risk. The study region has a total population of 502,152, or 21% of the population of the Tampa Metropolitan Statistical Area (Binghamton University, 2002). Similarly, the region encompasses approximately 20% of the total number of housing units for the Tampa area. Tampa is subject to numerous natural hazards, including hurricanes, floods, lightning, tornadoes, and droughts. Only two of these, hurricanes and floods, are used in this analysis because of the spatial differentiation that exists throughout the region regarding levels of risk. However, a more comprehensive review of the hazardousness of Tampa to multiple hazards can be found in Montz (2000). The region has been threatened numerous times by hurricanes and tropical storms, with approximately 60 coming within 75 miles of Egmont Key, at the entrance to Tampa Bay since the beginning of the twentieth century (Hillsborough County Emergency Planning Operations, 1997). The most recent events include Hurricanes Elena (1985), Georges (1998), Gabrielle (2001) and Eduoard (2002). Risk zones for hurricanes and tropical storms were determined by using the National Hurricane Center Risk Analysis Program (HURISK), an application of which was centered on Egmont Key (Neumann, 1987). From this, it is possible to calculate

38

Chapter Four

recurrence intervals and probabilities of occurrence for different category storms (Table 4-1). Flooding is not as widespread a hazard as are hurricanes, but the small urban watersheds characteristic of the region are affected by both localized summer thunderstorms and frontal systems. Hillsborough County has been in the Regular Phase of the National Flood Insurance Program since 1980, with an update of the floodplain maps accomplished in 1992. These maps provide the means of mapping the extent of the flood hazard based on the 100-year flood area (0.01 probability) and the 100-500 year floodplain (0.002 probability). Coastal flooding areas are not included here because they are considered above, in the hurricane data. Table 4-1. Return periods and probabilities of occurrence Category Tropical Storm 1 2 3 4 5

Wind Speed (kts) 135

Return Period 14 years 19 years 65 years 160 years 400 years 500 years

Prob. of Occurrence 0.07 0.05 0.02 0.006 0.003 0.002

Methods Geo-physical Risk Index As mentioned above, only hurricanes and floods are used in this analysis of geo-physical risk because droughts, lightning and tornadoes can occur anywhere within the county and, thus, do not provide sufficient spatial differentiation to address the questions asked here. In addition, there are various technological hazards that exist in the study area, but they are not included for several reasons, including the difficulty in determining probabilities of occurrence and the dependence of those events on other events, like hurricanes. Probabilities of the occurrence of both hurricanes and floods were based on the information presented in the previous section. Total probability was derived by summing the two individual probabilities, and the result was mapped using ArcView.

Social Vulnerability Index There is a variety of ways that have been used to determine and portray social vulnerability. Some use weighting systems (Lowry et al., 1995;

Hazardousness of the Tampa Bay Region

39

Montz and Evans, 2001) that reflect the relative contributions of the variables under study, and others are based on indices (Cutter et al., 2000; Montz, 2000), with differing elements that contribute to them. Rather than develop yet another method, the procedure used by Cutter et al. (2000) was adopted for this project. In this approach, each variable is standardized based on the ratio of that variable in each census tract to the total for the region. The resulting index value equals the ratio divided by the maximum ratio value for each variable for each census tract. The index values for all variables for each census tract were summed to produce the social vulnerability index. Different variables are also used to define social vulnerability usually by representing specific characteristics of populations that are known to differentiate more vulnerable sectors of society from those less vulnerable. Again, this project adopted the characteristics used by Cutter et al. (2000), although the variables used differ somewhat. Both the characteristics and variables used are shown in Table 4-2. In all cases except median house value, the higher the ratio, the greater the level of vulnerability is believed to be. With respect to median house value, this relationship is generally the same, except those with higher value homes have more to lose in an absolute sense, even if they have more assets to cover the losses so vulnerability can be considered in two ways. Table 4-2. Variables used in determining social vulnerability Vulnerability Factor Population size Access to resources Level of wealth Structural vulnerability

Measure Total population Number of housing units Population under age 17 below the poverty level Population over age 65 below the poverty level Median house value Number of mobile homes

Clearly, additional and different variables could and probably should be used to measure vulnerability. The factors listed in Table 4-2 were selected for several practical reasons. First, they are similar, though not identical, to those used by Cutter et al (2000), and thus provide a different context for looking at similar measures. Second, they provide a range of considerations, including numbers at risk, differential susceptibilities to loss, and structural measures. Finally, they are readily available on the Internet for census tracts in Tampa (Binghamton University, 2002). This final reason also explains why census tracts are used as the scale of

40

Chapter Four

analysis. The size of census tracts may mask more local variations in risk, but it also facilitates analysis for the purposes of this project.

Determining Overall Hazardousness Overall hazardousness is defined as the product of the geo-physical risk index and the social vulnerability index. This approach, used by Cutter et al. (2000), gives equal weight to all variables. The importance and use of weights has been discussed elsewhere (Lowry et al., 1995; Montz and Evans 2001) with the argument being that some variables contribute more to a population’s hazardousness than do others. Yet, despite a great deal of work on social and economic vulnerability, and on hazardousness, there is little agreement on the relative contributions. Thus, in order to avoid the difficulties associated with determining relative contributions of various social and economic characteristics to vulnerability, each is assumed to contribute equally at this time.

Results Spatial Distribution of Geo-Physical Risk The risk map depicted in Figure 4-1 presents the results of combining the probabilities of storm surge and flood probabilities. Not surprisingly, the zones of highest risk are mostly along the coast because tropical storms and hurricanes of low intensity have relatively high probabilities of occurrence. It is important to note that the way risk is measured here does not directly address intensity of impact, but it is reflected in the results. For instance, a Category 5 hurricane, with an estimated annual probability of occurrence of 0.002, will have a greater inland impact than will a Tropical Storm. At the same time, that Category 5 hurricane will hit the coast harder than will a Tropical Storm, with its higher storm surge and stronger winds. Thus, the risk ranges used here reflect both the chance of an event and the intensity of impact.

Spatial Distribution of Social Vulnerability The map of social vulnerability (Figure 4-2) shows several census tracts in the high vulnerability category, only two of which are along the coast. Many of the most densely populated areas, around and north of the central business district of Tampa, are characterized by medium to low levels of social and economic vulnerability. Those census tracts in the

Hazardousness of the Tampa Bay Region

41

highest vulnerability category exhibit very different characteristics. In some cases, one variable goes a long way to explaining vulnerability and in others it is a mix of characteristics. For instance, of the two coastal census tracts classified as high vulnerability the most southern one, and the largest, is characterized by two variables: the highest median house value index, which, in this case, means the lowest median house value, at $16,000 and the highest mobile home index. In contrast, the other high vulnerability coastal census tract does not have the highest index for any single variable, but rather scores high on several indices, including number of houses, median value (with a rather high value at $88,000), and population, suggesting dense, somewhat high value development.

Figure 4-1 Geo-physical risk of the Tampa Bay region

The patterns in this map illustrate the great variability that exists in characteristics that help to define social and economic vulnerability, even

42

Chapter Four

over relatively small areas. The northern portion of the study area has several census tracts with high levels of vulnerability, but many more in the medium and low categories. Both within and between these categories, there are distinct variations in factors in neighboring tracts.

Figure 4-2. Social vulnerability of the Tampa Bay region

Hazardousness The combination of geo-physical risk and social and economic vulnerability leads to very different patterns from those depicted in the first two maps (Figure 4-3). There are only a few areas defined as highly hazardous, suggesting that vulnerable populations do not live in the most hazardousness locations. There are exceptions, however. Portions of two census tracts that were in the high social vulnerability category are, indeed, highly hazardous. As noted earlier, this is for different reasons. While both census tracts in which these high hazard areas are found are in regions of high geo-physical risk, owing to their coastal locations, the extreme

Hazardousness of the Tampa Bay Region

43

southwestern tract is highly vulnerable due to the relative preponderance of mobile homes and the related lower value of housing. In contrast, the other tract is wealthier but more densely settled and has more elderly people under the poverty level.

Figure 4-3. Hazardousness of the Tampa Bay region

The remaining portions of census tracts that were categorized as highly vulnerable, based on the social indices used, are now in the low hazardousness category. They are mostly located away from the coast and, with one or two exceptions, are not subject to riverine flooding. Yet, coastal vulnerability remains relatively high. The fact that vulnerable populations are not characteristic of these regions does not negate their innate hazardousness, as shown in the moderate or higher vulnerability categorizations for all coastal regions.

44

Chapter Four

According to Figure 4-3, floodplains, although more restricted in area, remain areas of high overall hazardousness. This, however, does not necessarily reflect the intensity of impact as floods in this region are neither fast moving nor quick onset, owing to the relatively flat topography. In addition, Hillsborough County and the City of Tampa have participated in the National Flood Insurance Program (NFIP) since 1982, so presumably floodplain development since then has been somewhat curtailed. Certainly, this map speaks to the need to continue to enforce the provisions of NFIP.

Conclusions and Recommendations Three maps were developed to provide an illustrative tool for evaluating patterns of hazardousness based on geo-physical risk and social vulnerability in the Tampa Metropolitan Region. It is clear that the coastal location of this area presents a very hazardous situation, and geo-physical risk is highest along these coastal zones. However, for the most part, vulnerable populations, at least as measured here, do not occupy the most hazardous locations. Indeed, portions of only a few census tracts are characterized as high overall vulnerability. This is not to say that the socially and economically vulnerable populations are not at risk, nor does it suggest that the high hazard regions can be discounted because vulnerable populations do not live there. What it does suggest is that emergency managers must take an in depth look at these patterns in order to develop appropriate plans because the commonly held belief that vulnerable populations reside in the most hazardous areas does not apply here. Thus, in response to the first research question posed in this project, there are important differences in the two patterns, suggesting that emergency managers in other areas need to undertake similar analyses in order to understand more fully the nature of local vulnerability. Clearly, context matters. This work also points up the important fact that it makes a difference what variables are used for analysis. The census tracts that came out as having high social and economic vulnerability are characterized by very different variables. This represents a strength of this type of research and of the methodology because it allows for different mixes of factors to be incorporated. Indeed, some census tracts that were characterized as high vulnerability had both high median housing values and high indices for population over age 64 that is under the poverty level. On the other hand, it may also misrepresent vulnerability in some cases. For instance, the coastal census tract in the most southwestern portion of the study area has

Hazardousness of the Tampa Bay Region

45

a lower value for housing and a high ratio of mobile homes. Obviously, these two characteristics are correlated, and the result of including both may overestimate the level of vulnerability. Finally, the results of this research have important implications for emergency management and planning. For obvious reasons, emergency managers concentrate much of their efforts on high hazard areas, like those in the moderate to high categories in Figure 4-1. No matter who lives in these areas, the risk is high and appropriate measures need to be in place prior to an event. On the other hand, other areas are at risk, too, but not necessarily simply because of their geo-physical risk. There is a need for emergency managers to evaluate in some detail how patterns of geophysical risk and social and economic vulnerability intersect, or do not intersect, in their region. This would entail, as an early step, some discussion of how vulnerable populations should be defined locally. As noted above, context matters, so emergency managers must give some consideration to the characteristics of the local population, both in conjunction with and separate from the hazardousness of the location in which they live.

References Binghamton University. 2002. U.S. Census 2000 Map Server. GIS Core Facility, Binghamton University, Binghamton, NY. http://censusmap.Binghamton.edu. Last accessed 18 June 2003. Cutter, S.L., J.T. Mitchell, and M.S. Scott. 2000. Revealing the Vulnerability of People and Places: A Case Study of Georgetown County, South Carolina. Annals of the Association of American Geographers 90 (4): 713-737. Dow, K. 1992. Exploring Differences in Our Common Future(s): The Meaning of Vulnerability to Global Environmental Change. Geoforum 23 (3): 417-43. Hillsborough County Emergency Planning Operations. 1997. Hillsborough County Evacuation Map. Tampa, FL: Hillsborough County Emergency Planning Operations. H. John Heinz III Center. 2000. The Hidden Costs of Coastal Hazards: Implications for Risk Assessment and Mitigation. Washington, D.C.: Island Press. Liverman, D.M. 1990. Vulnerability, Resilience, and the Collapse of Society. In Understanding Global Environmental Change: The Contributions of Risk Analysis and Management, R.E. Kasperson, K.

46

Chapter Four

Dow, D. Golding, and J.X. Kasperson eds. Worcester, MA: Clark University. Lowry, J.H., Jr., H.J. Miller, and G.F. Hepner. 1995. A GIS-based Sensitivity Analysis of Community Vulnerability to Hazardous Contaminants on the Mexico/US Border. Photogrammetric Engineering and Remote Sensing 61: 1347-1359. Montz, B.E. 2000. The Hazardousness of Place: Risk from Multiple Natural Hazards. Papers and Proceedings of the Applied Geography Conferences 23:331-339. Montz, B.E. and T.A. Evans. 2001. GIS and Social Vulnerability Analysis. In Coping with Flash Floods eds. E. Gruntfest and J. Handmer. Dordrecht, Netherlands: Kluwer Academic Publishers. pp. 37-48. Neumann, C.J. 1987. The National Hurricane Center Risk Analysis Program (HURISK). NOAA Technical Memorandum NWS NHC 38. Coral Gables, FL: National Hurricane Center. Tobin, G.A. and B.E. Montz. 2004. Natural Hazards and Technology: Vulnerability, Risk and Community Response in Hazardous Environments. In Geography and Technology, S.D. Brunn, S.L. Cutter, and J.W. Harrington eds. Dordrecht, Netherlands: Kluwer Academic Publishers. Chapter 23, pp. 547-570.

CHAPTER FIVE CLASSIFYING HEAT STRESS EVENTS IN THE CENTRAL UNITED STATES ERIK H. BOWLES

Introduction The central United States experiences a wide variation of climate from year to year, which has a significant influence on the livelihood of living organisms (Rosenberg et al., 1993). Heat stress events produce the potential for harm to people, and hazards researchers are examining phases of the oldest of the traditions within geography: the manner in which humans interact with their environment (Mitchell, 1989). There is a practical aspiration to inform decision makers about the specific environmental conditions that have harmful effects on humans, animals, and natural resources (Harrington and Bowles, 2002). Therefore, it has become necessary to advance our capability to measure the level of exposure people experience during extreme heat events. Geographical contrasts of heat stress is an issue because people’s acclimatization to their regional climate norms would then infer that similar heat events would affect people more severely in areas that do not commonly experience major heat events. This project aims to gain knowledge of an underrepresented climatic hazard. Extreme heat events cause more human mortality on an annual basis than almost all other weather-related hazards combined (Robinson, 2001). However, these events have still proven difficult to universally define with no rigorous definition yet in use. The American Meteorological Society defines a heat wave as “a period of abnormally and uncomfortably hot and usually humid weather” (AMS, 2000). There are several other attempts at adding detail to this definition, which has unfortunately made the problem more complex by not having a collective

Papers of the Applied Geography Conferences (2006) 29: 40-48

48

Chapter Five

set of parameters (Burroughs, 2003; Choi and Meentemeyer, 2002; Kalkstein et al., 1996; Robinson, 2001; and Ward, 1925). Events like the ones in Chicago in 1995 and Paris in 2003 have proven that heat stress affects tens of thousands of people at a time, even in economically developed locations (Kalkstein, 1995; Vandentorren et al., 2004). The Chicago event is known to have killed over 500 people, and some numbers for Western Europe from 2003 have surpassed 20,000 deaths. Smoyer et al. (2000) provide information that, not only are cycles of hot conditions to continue, but also to increase in intensity into the future. When addressing the problem of excessive heat, one must account for two basic components, temperature and moisture. It has long been known that the stress caused by heat is compounded by the amount of moisture in the air (Lally and Watson, 1960). Water, itself, is a natural energy storage substance, especially with respect to heat (Oke, 1987). As air temperatures rise, the potential vapor content rises, meaning even more energy can be stored in the air. The effect that humidity has on warm temperatures is similar to how cold temperatures are affected by wind to produce windchill (Steadman, 1979). As heat and humidity continue to rise, humans and animals become more uncomfortable, and the sultry surroundings can become hazardous to an organism’s health to the point of causing death (Kalkstein and Valimont, 1986; Steadman 1984; and Thom, 1959). Temperatures in western Kansas can exceed those in the eastern parts of the state, but the moisture in the east forces the heat events to be much more intense. Because of this environmental synergism, several indices have been suggested to characterize how the air actually feels to an organism (Hevener, 1959; Houghton and Yaglou 1923; Quayle and Doehring, 1981; Steadman, 1984; and Thom, 1956). Dodd (1965) used US Department of Defense data to illustrate where the moisture gradient exists for the United States using dew point averages (Figure 5-1). There is a basic southeast-tonorthwest decrease of atmospheric moisture content from the Gulf States to the northern Rockies, indicating that typical heat stress magnitudes would follow the same pattern. For the study area of Kansas and Nebraska, the gradient is practically east to west; thus, two data points were used to represent conditions across each state. The most commonly used index for measuring the levels of heat stress in the U.S. is the heat index (HI). Originally termed ‘apparent temperature’, the heat index quantifies heat stress values to more accurately depict the combined effect of temperature and humidity on the skin of humans (Steadman, 1979). This study used the temperature and humidity combination of the heat index to represent heat stress values on an hourly

C Classifying Heaat Stress Events in the Central U United States

49

basis and tw wo values froom a table pro ovided in Chrristopherson (2001) ( as parameters ffor a new classsification mo odel (Table 5--1). The inform mation in Table 5-1 iis an attempt at categoriziing the effeccts of high heeat index temperatures and their corresponding impacts on ppeople. Unforrtunately, the categoryy numerals increase i as th he conditionss decrease, and a these categories aare purely chharacterizing the heat inddex temperatu ures. One would believve that if condditions have reached r a ‘Cattegory 4,’ thee situation would have become worsse, as is the case c with otheer natural hazards. The general effects from the thhermal extrem mes indicate thhe problems associated a with prolonnged exposuree, but there is no parameeter denoting g what is ‘prolonged.’’ However, there t are soliid categoricall thresholds that t have been established here andd it is felt thatt these parameeters could prrovide the categories too use in the neew heat wave model.

Moisture gradiennt decreases fro om east to west across Nebrask ka and Figure 5-1. M Kansas (Sourrce: Adapted froom US Departm ment of Defensee 1965)

Tornado hazards havee been classified for decadees employing the t Fujita Scale from 0-5, and hurrricane hazard ds have been categorized from 1-5 using the Saaffir-Simpsonn scale. These classificationn methods hav ve helped the general ppublic understtand the magn nitude of the eevent that has occurred, is occurringg, or will occuur. Therefore, when designning a modell for heat waves, it waas believed relevant to follo ow the same m methodology. A model has been deeveloped to claassify heat waaves as a cateegory from 1--5, with 5 being the hiighest. It will then be impllemented for 4 cities in Kaansas and

Chapter Five

50

Nebraska spanning across the known moisture gradient from west to east in the years of 1980-2000. Table 5-1. Previous heat stress classification method using heat index Category I II

Heat Index/ Apparent Temperature 54°C (130°F) or higher 41°C - 54°C (105°F - 130°F)

III

32°C - 41°C (90°F - 105°F)

IV

27°C - 32°C (80°F - 90°F)

General Effect of Heat Index on People in High Risk Groups Heat/sunstroke highly likely with continued exposure Sunstroke, heat cramps, or heat exhaustion likely and heatstroke possible with prolonged exposure and/or physical activity Sunstroke, heat cramps, or heat exhaustion likely and heatstroke possible with prolonged exposure and/or physical activity Fatigue possible with prolonged exposure and/or physical activity

(Source: Adapted from Christopherson 2001)

Data and Methods The National Climatic Data Center (NCDC) is a useful source for obtaining long-term weather data for stations across the United States. Previous work with these data has provided assurance that the hourly information needed is available for the temporal period of 1980 to 2000. A goal of this study was to work with hourly data, so the Surface Airways dataset of the U.S. first-order and airport stations was acquired. Only temperature and humidity were extracted for this research and used to find hourly heat index (HI) values. The four stations used for this study were Topeka in northeast KS, Dodge City in southwest KS, North Platte in western NE, and Omaha in southeast NE. Their physical locations allow observations to be made in quadrants across the gradient. After the hourly temperature and humidity data were obtained in English units from the Surface Airways dataset, hourly values of HI are to be determined using equation [1] from Dixon (1997). Since the heat stress management categories for HI begin at 80° F (Christopherson, 2001), it was decided that all values greater than 70° F would be kept in the analysis spreadsheets. Maintaining the hourly HI values between 70° F and 80° F allows analysis of conditions prior to or just after a heat stress event and analysis of the degree of nighttime relief provided during the

Classifying Heat Stress Events in the Central United States

51

early morning hours. Any day that does not have at least one hour of HI greater than 70° F was eliminated from the data set. This procedure leaves out the usually cold periods of winter, early spring, and late fall. When measuring heat, simply stating that an hour exceeded a parameter is not enough information to express the magnitude of the heat stress. Heat index values that surpass the thresholds of 90° F and/or 105° F must be measured by how much the threshold is surpassed. If HI = 16.923 + 0.185212 * T + 5.37941 * H - 0.100254 * T * H + (0.941695 * 10-2) * T2 + (0.728898 * 10-2) * (H2) + (0.345372 * 10-3) * T2 * H - (0.814971 * 10-3) * T * H2 + (0.102102 * 10-4) * T2 * H2 - (0.38646 * 10-4) * T3 + (0.291583 * 10-4) * [1] H3 + (0.142721 * 10-5) * T3 * H + (0.197483 * 10-6) * T * T3 - (0.218429 * 10-7) * T3 * H2 + (0.843296 * 10-9) * T2 * H3 - (0.481975 * 10-10) * T3 * H3 + 0.5

T = temperature °F H = humidity %

Topeka registered a 3:00 pm HI of 109° F, that hour measures 19° F above the 90° F threshold, and 4° F above the 105° F threshold. The amount that an hour exceeds either threshold is determined and then the hours are summed to find a heat stress magnitude for each day. Parameters for event classification were based on the amount which a day exceeded the 90° F and 105° F thresholds. In the above example, results would be termed 19 hours above HI 90 and 4 hours above HI 105 for that hour. Classification of daily heat events was executed by applying the model in Table 5-2. If a daily magnitude exceeded 30 hours above HI 90, but had no HI 105 hours, the heat stress for the day would be considered Minor, or a Category 1. However, the minimum duration for an event to be classified is either three consecutive days of Category 1, or two consecutive days where one is a Category 1 and the other is a category between 1 and 5. A third parameter determining daily heat stress magnitudes is recovery, or the number of actual hours that the heat index was measured at or below 75° F. Recovery time attests to how intense an event is throughout its duration. For a day to reach conditions hazardous enough to warrant a Category 5 rating, more than 210 hours of HI 90 must occur with at least 35 hours of HI 105, as well as having zero actual hours that register a HI less than 75° F. When classifications are assigned for a string of consecutive days, this set of days is considered a heat stress event. Much like tracking changes in hurricane strength, daily heat stress classifications vary from day to day depending on local conditions. Also like hurricanes, no matter how long the heat event lasts, or how many different classifications are recorded during the event, the highest category recorded is the classification

Chapter Five

52

assigned to the event. Summary statistics were compiled to identify the number of times each category occurred during the time period, as well as the average probability that an event could occur each year. Table 5-2. Heat stress classification model Category 1. Minor 2. Moderate 3. Strong 4. Severe 5. Extreme

HI – 90

HI - 105

Recovery

> 30 > 80 > 120 > 160 > 210

>1 > 10 > 20 >35

” 10 ”6 ”2 0

Results and Discussion Results for the magnitude and frequency of heat events from station to station were primarily as expected (Figure 5-2). Topeka had the most events (113) with the most major events, Category 4 or Category 5, as well (18). North Platte experienced the least number of events (33) with zero major events over the 20-year period. Although Dodge City totaled more events (91) than Omaha (66), Omaha’s moisture availability is easily observed with the number of major events during this period (Table 5-3). Geographically, one can observe the difference in heat events across the east-west moisture gradient in Kansas and Nebraska. Category 1 events are the most dominant of the events that occur in North Platte and Dodge City. Dodge City in the southwest of the study area averages over 3 Category 1 events each year, whereas North Platte in the northwest averaged just over one Category 1 event per year. Comparatively, the two eastern cities also exhibit a north-south disparity in the frequency of Category 1 events. Topeka in the southeast experiences 2.2 Category 1 events on average, and Omaha averages less than one per year with a 0.8 percent chance of one occurring. A significant drop off in event frequency is observed with Category 2 events in the western cities. This abrupt reduction in heat stress occurrence, 67 percent chance annually in Dodge City and a mere 20 percent chance annually in North Platte, is a testament to the lack of the moisture needed to produce conditions beyond Minor on a regular basis. Eastward, Topeka has 1.4 Category 2 events per year, and Omaha sees almost 1 annually. Heat stress events of higher magnitude are extremely rare in the western areas of the study area. Together, both stations only experienced 4

Classifying Heat Stress Events in the Central United States

53

actual events between Category 3 and Category 5 over the 20-year period. North Platte never saw an event beyond a Category 3, but Dodge City did register a Category 5 heat stress event in 1980. The event of 1980 was widespread throughout the central U.S., and these results indicate that the highest order of heat stress can occur in the western Plains during the most extreme of warm season conditions. Topeka still experiences over 1 Category 3 events per year, and these events are relatively common in Omaha as well. North Platte

Omaha

80

80

70

70

60

60

50

50

40

40

30

30

20

20

10

10

0

1

2

3

4

5

0

1

2

Dodge City

80

70

70

60

60

50

50

40

40

30

30

20

20

10

10 1

2

3

4

5

Topeka

80

0

3

4

5

0

1

2

3

4

5

Figure 5-2. Geographic variation of heat stress events across the moisture gradient and by latitude

Frequency of the two major categories, 4 and 5, decreases for both Topeka and Omaha, but the probability of a major event occurring is still around 50 percent annually. Conditions that develop into a Category 5 heat

Chapter Five

54

stress hazard level are extremely rare regardless of location in this area. Adaptation to such high levels would therefore not be expected for people in any of the areas in this study, however the knowledge of their occurrence in 3 of 4 points in this region warrants a necessity to mitigate for future extreme heat stress hazards. Table 5-3. Heat stress classification results

Topeka Frequency Average/21 Omaha Frequency Average/20 Dodge City Frequency Average/20 North Platte Frequency Average/20

CAT1

CAT2

CAT3

CAT4

CAT5

TOT

44 2.20

28 1.40

22 1.10

13 0.65

5 0.25

112

16 0.80

19 0.95

16 0.80

9 0.45

6 0.30

66

75 3.57

14 0.67

1 0.05

0 0.00

1 0.05

91

27 1.35

4 0.20

2 0.10

0 0.00

0 0.00

33

Although a hurricane can reach a Category 5 status on the SaffirSimpson scale, it may reduce to a Category 3 by the time it makes landfall; either way, the damage is measured with the current strength of the storm. Losses from a heat stress event can be identified in a similar manner with this model as each day is assigned a value, and prior magnitudes of the event can be accounted for. The difference between heat stress events and the hurricane example is that the whole heat stress event is widely experienced whether a Category 1 or a Category 5. The Category 5 event at Dodge City in 1980 lasted 28 days, but only the third day of the event actually registered a 5 classification. Topeka experienced an event at the same time that was ten days shorter, but eight of the days (four of which were consecutive) registered a Category 5. The magnitudes of these events are notably different, and at present it is still difficult to confidently describe the concrete differences between locational impacts. Further analysis will include data from several points around the country, as well as an ecological collaboration to continue a progressive understanding of human capacity to adapt during extreme climatic events.

Classifying Heat Stress Events in the Central United States

55

Summary Implementation of the new heat stress classification model was designed to be objective as well as descriptive of extreme heat events. However, objectivity is difficult to preserve when geographic location, local climate, and adaptive capacity of the people involved are all variables to consider when determining the magnitude of heat events. Work in this study did exhibit positive results in the direction of heat stress classification in a manner that could inform people of the hazard levels during and after an event. Results in this research also provide the frequency and probability of specific heat stress occurrences for an area. Future work in this area would allow for better geographic knowledge of heat stress magnitudes, as well as a refinement of the techniques necessary to describe the effects certain magnitudes and frequencies of heat stress levels have on people.

References AMS. 2000. Glossary of Meteorology, 2nd ed. T. Glickman, ed. Boston, MA: American Meteorological Society. Burroughs, W. 2003. Climate into the 21st Century. New York: Cambridge University Press. Choi, J. and V. Meentemeyer. 2002. Climatology of Persistent Positive Temperature Anomalies for the Contiguous United States (1950-1995). Physical Geography 23(3): 175-195. Christopherson, R. 2001. Elemental Geosystems, 3rd Ed. Upper Saddle River, NJ: Prentice Hall. Dixon, R. 1997. A Heat Index Climatology for the Southern United States. National Weather Digest 22(1): 16-21. Dodd, A. 1965. Dew Point Distribution in the Contiguous United States. Monthly Weather Review 93(2):113-122. Harrington, J., and E. Bowles. 2002. Assessing the Impact of Heat and Humidity on Livestock: Development of an Hourly THI Climatology. Papers and Proceedings of the Applied Geography Conferences 25: 311-315. Hevener, O. 1959. All about hHumiture. Weatherwise. 12(2): 56-85. Houghton, F. and C. Yaglou. 1923. Determining Equal Comfort Lines. Journal of the American Society of Heating and Ventilating Engineers 29: 165-176. Kalkstein, L. 1995. Lessons from a Very Hot Summer. Lancet 346: 857859.

56

Chapter Five

Kalkstein, L. and K. Valimont. 1986. An Evaluation of Summer Discomfort in the United States using a Relative Climatological Index. Bulletin of the American Meteorological Society 67(7): 842-848. Kalkstein, L., P. Jamason, J. Greene, J. Libby, and L. Robinson. 1996. Health Watch/Warning System: Development and Application, Summer 1995. Bulletin of the American Meteorological Society 77(7): 1519-1528. Lally, V. and B. Watson. 1960. Humiture Revisited. Weatherwise. 254256. Mitchell, J. 1989. Hazards Research. In Geography in America, G. Gaile and C. Willmott eds. Columbus, OH: Merrill Publishing Company. pp. 410-424. Oke, T. 1987. Boundary Layer Climates, 2nd Ed. New York: Routledge. Quayle, R. and F. Doehring. 1981. Heat Stress: A Comparison of Indices. Weatherwise 34: 120-124. Robinson, P. 2001. On the Definition of a Heat Wave. Journal of Applied Meteorology 40: 762-775. Rosenberg, N., P. Crosson, K. Frederick, W. Easterling III, M. McKenney, M. Bowes, R. Sedjo, J. Darmstadter, L. Katz, and K. Lemon. 1993. The MINK Methodology: Background and Baseline. Climate Change 24: 7-22. Smoyer, K., L. Kalkstein, J.S. Greene, and H. Ye. 2000. The Impacts of Weather and Pollution on Human Mortality in Birmingham, Alabama and Philadelphia, Pennsylvania. International Journal of Climatology 20: 881-897. Steadman, R. 1979. The Assessment of Sultriness. Part I: A TemperatureHumidity Index based on Human Physiology and Clothing Science. Journal of Applied Meteorology 18: 861-873. —. 1984. The Universal Scale of Apparent Temperature. Journal of Climate and Applied Meteorology 23: 1674-1687. Thom, E. 1956. Measuring the Need for Air Conditioning. Air Conditioning, Heating, Ventilating 53(8): 65-70. —. 1959. The Discomfort Index. Weatherwise 57-60. Vandentorren, S., F. Suzan, S. Medina, M. Pascal, A. Maulpoix, J. Cohen, and M. Ledrans. 2004. Mortality in 13 French Cities during the August 2003 Heat Wave. American Journal of Public Health 94(9): 15181520. Ward, R. 1925. Climates of the United States. Boston, MA: Ginn and Company.

CHAPTER SIX VULNERABILITY TO FLOODING IN COLUMBIA COUNTY, PA: THE ROLE OF PERCEPTION AND EXPERIENCE AMONG THE ELDERLY JENNIFER J. WEBB

Introduction The U.S. Census estimated that 12.4 percent of the total population is comprised of persons aged 65 and older, or the elderly (U.S. Bureau of the Census, 1996). This is quite an increase from the 8 percent of elderly residents in the United States in 1990. It is also likely that a substantial number of people in this group require some kind of assistance due to limited mobility and/or self-care. The elderly has been identified as a very important subgroup within the population in recent years by numerous hazards researchers. The concept of vulnerability has also emerged over the past decade and essentially aims to identify the potential for loss. More specifically, social vulnerability seeks to identify specific groups within the general population that may be more vulnerable to hazards than others. Age and gender are important indicators of vulnerability so, of course, the elderly are among those social groups which have been identified. These social groups all have defining characteristics that may compromise or hinder their mobility in times of disaster, during the evacuation phase as well as in the recovery phase. This research focuses on the perceived vulnerability of the elderly. The research reported here is part of a larger study aimed at better understanding vulnerability and the role of intervening variables such as experience, perception, and behavior in the Columbia County, Pennsylvania

Papers of the Applied Geography Conferences (2006) 29: 168-176

58

Chapter Six

study area. Of particular concern here is the role of age and experience in influencing perceptions and actions among different groups of the elderly population.

Vulnerability of the Elderly Much hazards research suggests that a relationship exists between social and economic factors and increased risk or vulnerability. In other words, studies have concluded that there may be particular place characteristics that isolate and position certain groups of people in harm’s way. In Hurricane Andrew, age (the elderly or the very young), single parent households (especially those headed by females), those with physical disabilities, and residents of public housing were found to be most affected (Morrow, 1999). This study focuses on the elderly. A variety of factors influence hazard perception, particularly past experience with hazards (Burton et al., 1993, Smith, 2001). People who have had experiences with hazards in the past tend to act differently in terms of preparation and response in the wake of future hazards. Personal experience with hazards, particularly direct experience, can serve to amplify or attenuate risk (Kasperson et. al., 1988). How different factors serve to amplify or attenuate perceptions of risk are the subject of much research. For instance, some argue that the magnitude and frequency of a hazard also influences hazard perception (Theios et al., 1978). Psychologists, on the other hand, have supported the idea that risk explanation is rooted in the cognitive behavior of the individual, and this claim has been reinforced by anthropologists, who strongly believe that both social and cultural contexts help shape hazard perceptions (Kasperson et al., 1988). In other words, some believe that risks and hazards interact with psychological, social, and cultural processes “in ways that can heighten or attenuate public perceptions of risk and related risk behavior” (Kasperson et al., 1988, pp. 178-179). Mayhorn (2005) discussed the importance of focusing on the elderly in hazards research. First, compared to their younger counterparts, the elderly are more likely to underutilize aid following disasters. Second, the elderly tend to suffer more psychological effects in the long term. Finally, the elderly tend to take longer to recover economically or financially postdisaster. However, social support systems and networks have been found to be important in the aftermath of an event—whether they are family, friends, or spouses. Specifically, because the elderly may well be less likely to take measures to ensure disaster preparedness, strong social support networks are critical (Ellen, 2001).

Vulnerability to Flooding in Columbia County, PA

59

In order to increase disaster preparedness and response among the elderly, such factors as physical impairments, structural characteristics, accessibility, safety, and financial capabilities need to be considered (Eldar, 1992). This is true with those who are institutionalized or otherwise require additional care (Ellen, 2001). Indeed, all elderly are not the same, based on age categories and other factors. Phifer (1990) distinguished age categories of older adults in terms of their defining characteristics. He refers to older adults aged 55 to 64 as the “healthy, working age group”. Adults ages 65 and 74 are labeled as “young old,” and have usually just entered retirement. The elderly ages 75 to 84 are referred to as “aged”, indicating that they are well beyond retirement and some may experience mobility limitations. Finally, older adults ages 85 and above are referred to as the “oldest old.” This is the age group where many typically experience mobility and/or self-care limitations and decreases in visual and auditory acuity. The differences among the elderly suggest variations among them in terms of perceptions as well.

Methods Using a case study approach of Columbia County, Pennsylvania (see Figure 6-1), this research focuses on perception of the flood hazard among the elderly populations. Floods have devastated Columbia County in the past, with the most destructive one being a result of Hurricane Agnes in 1972. However in recent years, Columbia County has also experienced major flooding as a result of Hurricanes Ivan and Frances in 2004. This approach provides a means of examining hazard perception and attitudes that possibly can aid emergency managers with decision making as it relates to hazard response and mitigation. Most importantly, however, this study furthers our understanding of vulnerability on a local scale. In order to identify where elderly populations live for purposes of administering the survey, the Q3 flood data from FEMA (FEMA, 2006) were overlain with census block groups. The elderly living within the floodplains were categorized based on Phifer’s classification (Phifer, 1990). The sample was restricted to residents of the 100-year floodplain, but no attempt was made to isolate the different risk areas within this zone. The number of elderly people per block group in Columbia County was identified and mapped, providing the sample for administration of the survey. The sample was drawn from the 33 municipalities in Columbia County using a proportionate stratified random sampling design. The stratification of the sample was based on the address lists that were developed for each age group, totaling 1,694. The breakdown of each is as

60

Chapter Six

follows: 68 percent from ages 55 to 59, 16 percent from ages 60 to 64, 10 percent from ages 65 to 74, 4 percent from ages 75 to 84, and 2 percent from ages 85 and older. Because of the small numbers in the oldest two age groups, they were combined into one group for statistical analyses. For this research, the sample size was 200, divided among the five age groups. Due to financial and time limitations, the survey was administered both door-todoor and through phone interviews. This could affect the results, though no differences were noted in the responses of those surveyed each way.

Figure 6-1. Columbia County, Pennsylvania.

Vulnerability to Flooding in Columbia County, PA

61

The survey presents a series of questions that were divided into four parts: household attributes, flood experience, attitudes, and flood behavior and policy. Attitudes were measured using a variation of the Likert scale as a means of measuring frequency and likelihood. Typically, the measurements are on a five-point scale, ranging from positive to negative responses, however in this research, a three-point scale was used. Both frequency tables and the Chi-Square Statistical Analysis were used to analyze the data gathered from the questionnaire, and to draw conclusions regarding attitudes and experience with flooding.

Results Age and Perception Rather than present the individual results of each of the five perception questions, a summation, representing a total perception score, was measured to determine if an association is present regarding overall perception. More specifically, the perception score was derived by adding the ones, twos, and threes from the group of questions. The range of the total perception scores was from five (lowest) to fifteen (highest). In addition, the original five age categories were collapsed into three new categories, 55 to 64, 65 to 74, and 75 and up, both to obtain better responses and mediate any model violations. More specifically, this ensures that the expected frequency of observations in any cell is at least 5 to meet the requirements of Chi square tests. No statistical association was found between perception and age (X2= 1.775 and Significance= .777). Within all three age groups, only slightly more than 20 percent of respondents fell into the high perception category (Table 6-1). It is noteworthy too, that a higher percentage of those respondents who are 75 and older fell into the low perception category. This lack of difference leads to consideration of factors that might influence perceptions, particularly the impact of experience.

Age and Experience Given that age was not found to have a significant association with perception, it is important to consider the role of experience, particularly given that older populations are likely to have more experience. Experience was measured by three variables: evacuation, damages from flooding, and protective actions taken to reduce flooding. A significant association was found between age and respondents who ever evacuated

Chapter Six

62

their homes due to flooding (Table 6-2). More respondents said that they had never evacuated, regardless of age but the highest percentage is found within the age category of 55 to 64. In contrast, a larger proportion of the older elderly reported having evacuated. The results indicate that the proportion of respondents who have evacuated increases with age. This relationship between age and evacuation yielded a statistically significant association at the .05 level. Table 6-1. Age and total perception score Age 55-64

65-74

75 and older

Total

N % within age % within total Perception score N % within age % within total perception score N % within age % within total perception score N % within age % within total perception score

Total Perception Score Low Medium High 29 47 23 29.3 47.5 23.2

Total 99 100

46.0

50.5

20.9

49.5

12 27.9

22 51.2

9 20.9

43 100

19.0

23.7

20.5

21.5

22 37.9

24 41.4

12 20.7

58 100

34.9

25.8

27.3

29.0

63 31.5

93 46.5

44 22.0

200 100

100

100

100

100

None of the other relationships between age and flood experience was found to be significant. Thus, the assumption that those in older age groups are more likely to have experienced flooding is erroneous. The fact that the older elderly are more likely to have evacuated, however, is important and seems to contradict some of the literature. These results, combined with those regarding age and perception, suggest that age is not a defining factor, at least in the study area. This leads to consideration of the extent to which other relationships might exist, particularly that between experience and perception.

Vulnerability to Flooding in Columbia County, PA

63

Table 6-2. Age and ever evacuated home due to flooding Age 55-64

65-74

N % within age % within total Perception score

N % within age % within total perception score 75 and N older % within age % within total perception score Total N % within age % within total perception score X 2 = 6.557; Significance = .038

Ever Evacuated No Yes 65 34 65.7 34.3

Total 99 100

56.5

40.0

49.5

24 55.8

19 44.2

43 100

20.9

22.4

21.5

26 44.8

32 55.2

58 100

22.6

37.6

29.0

115 57.5

85 42.5

200 100

100

100

100

Experience and Flood Perception In order to evaluate relationships between experience and perceptions, each of the perception questions was tested with each of the experience variables. The questions on perception asked about views of the flood risk faced by the community, expectations of future flooding, concerns about future flooding, views of social support available, and whether or not flooding has led to consideration of moving. Space limitations make it impossible to present all of the contingency tables, so summaries of results are presented, with discussion centering on the most salient results. All results are available from the author. Flood experience was found to have an association with hazard perception in only one of the five independent tests, flood experience and the likelihood of one’s home being flooded in the near future (Table 6-3). The association is significant at the 0.05 level, but the resulting Cramer’s V of 0.257 suggests a weak relationship. The results indicate that more than half of the respondents, no matter how many floods they have experienced, view flooding as not likely, but those who have not experienced flooding or who have experienced one or two floods are proportionally more likely to hold this view. Thus, while it does not

64

Chapter Six

appear that perception increases with flood experience, lack of flood experience seems to be important. Table 6-3. Flood experience and perception Flood Experience and: Flood risk to community Future flooding Concern about flooding Social support Moving from area

Pearson's Chi-Square 4.033 25.143 3.775 7.832 10.509

Asymp. Sig. 0.672 0.000 0.707 0.251 0.105

Cramer's V. 0.257

The associations between evacuation and flood perception were all statistically significant at the 0.05 level, with the exception of one (flood risk to community) (Table 6-4). Over 58 percent of respondents who had evacuated reported that it was either somewhat likely or very likely that their homes will be flooded in the near future. In contrast, only about 31 percent of respondents who had never evacuated indicated that it was somewhat or very likely that their homes would be flooded in the future. Further, while most respondents have never evacuated due to flooding, those who have are more concerned about flooding, whether it be a little or a great deal. Evacuation is also associated with social support. A total of 67 respondents out of 85 (79 percent) who have evacuated in the past indicate they have some form of social support network, in contrast to the 65 percent who have never evacuated. Finally, a larger number of those who have evacuated in the past think about moving from the area because of the flood risk, compared to those who have never evacuated. However there is still a substantial proportion (71.8 percent) of evacuees who have never thought about moving. Table 6-4. Evacuation and perception Evacuation and: Flood risk to community Future flooding Concern about flooding Social support Moving from area

Pearson's Chi-Square

Asymp. Sig.

Cramer's V.

3.866 17.064 7.479 7.169 7.213

0.145 0.000 0.024 0.028 0.027

0.292 0.193 0.189 0.190

Vulnerability to Flooding in Columbia County, PA

65

The last measure of past experience with floods is damages that occur as a result of flooding. All relationships between damages and flood perception are statistically significant at the 0.05 level (Table 6-5). The individual Chi square results show that over 88 percent of respondents who have experienced damages from flooding tend to see flooding as somewhat of a risk or a high risk to the community, compared to 67 percent of respondents who have not experienced damages. These small differences and the Cramer’s V of 0.272 indicate a weak relationship. However, an association is seen such that perception of the flood risk to the community increases with damages. Table 6-5. Damage and perception Damages and: Flood risk to community Future flooding Concern about flooding Social support Moving from area

Pearson's Chi-Square

Asymp. Sig.

Cramer's V.

14.826 47.306 16.782 6.138 18.568

0.001 0.000 0.000 0.046 0.000

0.272 0.486 0.290 0.175 0.305

Not surprisingly, over 68 percent of respondents who have reported damages from flooding report that their homes are likely to be flooded in the near future. In contrast, approximately 78 percent of respondents who have not experienced damage do not believe that their homes will be flooded in the future. This is the strongest association among the perception questions, with a Cramer’s V of 0.486. As before, perception increases with damage experience. Seventy-eight percent of respondents who have experienced damages from flooding consider it to be at least somewhat of a risk in comparison to 58 percent of respondents with no damage experience. This is similar to the results above, but in this case, the risk is personalized. Here, too, perception of flood concern increases with damages experienced. More respondents who have experienced damages from flooding talk about flooding with family, friends, or neighbors. Almost two-thirds of those who report having experienced damage talk frequently about flooding compared to less than half of those in the other two groups. It is not surprising that there was a much larger proportion of persons who have experienced damages from flooding that think about moving more often (32.2 percent) than those who have not experienced damages (9.1 percent). Some 90 percent of those who have not experienced damages have also never talked about flooding. However, it is also the case that more than

66

Chapter Six

two-thirds of respondents with experience have never talked about moving. Despite this, it is important to note that thoughts about moving increase with damages.

Summary and Conclusions The findings seem to indicate that only slight differences exist among the elderly groups in terms of perceptions and/or experience. Interestingly, there was no statistically significant association between age and perception. More respondents in the oldest age group, 75 and older, fell into the low perception category. However, a statistical association was present between age and evacuation. In other words, older respondents were found to have evacuated more than their younger counterparts, which contradicts the literature. This may be because older adults consider their lives to be more important than personal possessions, as was conveyed in many personal interviews with respondents. Flood experience, in terms of the number of times one experienced a flood in their lifetime, evacuation, and damages resulting from flooding, also influenced the flood perceptions of respondents. This supports the literature in that people who have had experiences with events in the past tend to act differently in terms of preparation and response in the wake of future occurrences. Although the relationship between age and flood perception was not as straightforward as originally thought, the statistically significant findings are still quite interesting, and more research on those specific areas in the future may yield even more explanation for the associations found. Intervening factors such as the length of residence and damage seemed to play more of a role in perception than the age of respondents, which in turn, directly results from flood experience. Overall, there are significant relationships between experience and perception that may serve to overshadow any differences of the various age groups among the elderly. Because this research is a case study, it is difficult to generalize and relate these results on a larger scale to different areas. The results are really dependent on the mix of hazards that a particular area encompasses, in which case for more than one hazard, a multiple hazards approach may be taken which will significantly impact the overall place vulnerability. However, this does not undermine the importance of case studies in hazards research. This research in particular has offered more information that was required to understand and evaluate the vulnerability of Columbia County to flooding, with an emphasis on the elderly population.

Vulnerability to Flooding in Columbia County, PA

67

This study has illustrated the importance of experience in flood perceptions among the elderly, while actually undermining the role of age in influencing perceptions. Interestingly, perceptions of community risk and personal risk vary among respondents. More specifically, respondents view flooding as a risk to the community, however they do not personalize it. This may be because where they reside is not an area that gets flooded frequently, whereas their community at large faces more of a flood risk. The knowledge obtained from this study can be of importance to both emergency management and mitigation planning. Social vulnerability measures such as these should be considered in existing vulnerability assessments so that special needs populations can be targeted before an event occurs and officials and others can be prepared.

References Burton, I., R.W. Kates, and G.F. White. 1993. The Environment as Hazard, 2nd Ed. New York: The Guilford Press. Eldar, R. 1992. The Needs of Elderly Persons in Natural Disasters: Observations and Recommendations. Disasters 16 (4): 355-358. Ellen, E.F. 2001. The Elderly May Have Advantage in Natural Disasters. Psychiatric Times XVII (1). FEMA. 2006. Digital Q3 Data. http://www.fema.org/hazard/map/q3.shtm Last accessed 30 July 2006. Kasperson, R. E., O. Renn, P. Slovic, H.S. Brown, J. Emel, R. Goble, J.X. Kasperson, and S. Ratick. 1988. The Social Amplification of Risk: A Conceptual Framework. Risk Analysis 8 (2): 177-191. Mayhorn, C.B. 2005. Cognitive Aging and the Processing of Hazard Information and Disaster Warnings. Natural Hazards Review 6: 165169. Morrow, B.H. 1999. Identifying and Mapping Community Vulnerability. Disasters 23:1-18. Phifer, J.F. 1990. Psychological Distress and Somatic Symptoms after Natural Disaster: Differential Vulnerability among Older Adults. Psychology and Aging 5 (3): 412-420. Smith, K. 2001. Environmental Hazards: Assessing Risk and Reducing Disaster. London: Routledge. Theios, M., S. Dragula, and R. Graff. 1978. Hurricane Perception in New Orleans: A Comparison of Five Neighborhoods. Bulletin of the Illinois Geographical Society 20 (2): 22-32. U.S. Bureau of the Census. 1996. 65+ in the United States. Washington, DC: U.S. Government Printing Office.

CHAPTER SEVEN FRAMING FLOOD RISK: COMPARING PERCEPTIONS OF THE “100 YEAR FLOOD” AND TWO ALTERNATE DESCRIPTIONS HEATHER M. BELL

Introduction From 1980 through 2005, 67 weather related disasters in the United States exceeded one billion dollars each in direct damage (Ross and Lott, 2006). In over half of them, flooding was either the primary cause or a significant component of a compound disaster like a hurricane. According to Perry (2000), more property damage was caused by and more lives lost to flooding than any other disaster type in the twentieth century. Effective loss prevention and distribution begin well before a flood event. Structural flood mitigation first became the official responsibility of the federal government in 1936 (USWRC, 1971); the National Flood Insurance Act of 1968 emphasized non-structural methods and created the National Flood Insurance Program (NFIP). This Act and subsequent legislation established a flood with a one percent chance of occurring in any year as the foundation for both structural mitigation and non-structural management programs in the United States. Commonly known as the “100 year flood,” its parameters are neither completely accurate nor static. Over half of U.S. flood losses occur outside this “high risk” area, in the 500 year floodplain, or outside both mapped floodplains (Frech, 2005; Smith, 2000). The 100 year flood has been upheld as a useful policy criterion (GFWNFPF, 2004; NRC, 2000), but practitioners and researchers alike

Papers of the Applied Geography Conferences (2007) 30: 362-371

Framing Flood Risk

69

have questioned the effectiveness of hundred year flood terminology in public communication (Frech, 2005; GFWNFPF, 2004; Gruntfest et al, 2002; NRC, 2000). Its use may emphasize risk dichotomies and mask the irregularity and uncertainty associated with both the timing and consequences of flooding. New terms have been introduced to the public through education campaigns and other official communications, but testing has been limited (Bell and Tobin, 2007). It is not clear that they more successfully convey uncertainty or induce concern. This study compares the efficacy of three methods commonly used to describe the policy’s benchmark flood: the 100 year flood, a flood with a one percent chance of occurring in any year, and a flood with a 26 percent chance of occurring in 30 years. To be mathematically correct, the third description (based on cumulative probability) should read “at most a 26 percent chance of occurring in 30 years.” The phrase “at most” is generally not included in public communication, however, and was therefore not used in this research. Investigating how risk messages are interpreted may help to improve communication and create more realistic expectations on the parts of managers and communicators as well as “the public.”

Methods Field work was conducted in the Towns of Union and Vestal, New York. The communities lie in the south-central portion of the state on opposite banks of the Susquehanna River. Both experienced major flooding in June of 2006. Data were collected through a face to face survey administered to adult residents of single family homes located in official floodplains. Areal weighting of 2000 Census block group data indicated that approximately 15 percent of Vestal residents and 18 percent of unincorporated Union residents lived in one of the two mapped floodplains. In 2000, this population was 94 percent white and nonHispanic. Over 20 percent had a baccalaureate degree and more than three quarters owned their home. A higher proportion of floodplain residents in Vestal were over 65 and Vestal does not currently participate in the NFIP Community Rating System. The sample was spatially stratified by floodplain designation and Town. A total of 114 people participated; sixty lived in Union, 54 in Vestal, 50 in the 100 year floodplain and 64 in the 500 year floodplain. Floodplain status was determined by overlaying county parcel data with FEMA’s Q3 data in ArcGIS. Parcels with more than half their area in a particular floodplain designation were considered to be “in” that floodplain. Data collection began in late October, 2006. Though generally

70

Chapter Seven

removed from the study area’s designated floodplains, locally severe overland flow and tributary flooding occurred in November, 2006. In order to maintain reference consistency in the survey, areas impacted in November were not sampled on subsequent trips. The questionnaire was divided into five major themes: 1) flood experience and loss mitigation activities, 2) general perception of flood risk and cause, 3) flood information infrastructure, 4) perceptions associated with specific flood risk descriptions, and 5) basic demographic data. The study outlined in this paper made use of questions in the fourth theme. These questions were closed, but the face to face format allowed for the collection of qualitative data. Participants were given cards with the three descriptions on them and were instructed that they could also answer the questions with a combination, “All”, or “Don’t Know”. Relative effectiveness was judged through both understanding and persuasion. Understanding of flood related uncertainty over time and space was evaluated through two questions: 1. Which of these floods, if any, do you think could happen more than once in a year? 2. Do you think the size of any of the floods described on these cards could change over time? If yes, which? Relative concern was used to measure persuasion and was also assessed through two questions: 1. Which of the floods described on the cards concerns you most? 2. Which of the described floods concerns you the least? If more than one description was given as an answer to a question, each pertinent response was coded as one. If a participant said he or she didn’t know, all descriptions were coded as zero. “Don’t Know” responses were also recorded as a separate variable. A scale of relative concern was constructed for each description. If a description was perceived as the most concerning by an individual, it was assigned a positive one; if perceived as the least concerning, it was assigned a negative one. Descriptions eliciting no response were treated as neutral and assigned a zero. Points were then added together and adjusted to a range of zero to two. In addition, respondents were asked to rate their level of concern associated with the most and least concerning floods on a seven point scale that ranged from “Not Concerned at All” to “Completely Concerned”. Quantitative analyses consisted of standard nonparametric tests of variance. All statistical tests were conducted at a .05 level of significance in SPSS 15.0. The Holm method was used to adjust for multiple

Framing Flood Risk

71

comparisons. Participant commentary was not specifically asked for in this section of the survey, but interviewer observations and participant comments, if provided, were recorded for each question. The comments were written up and themes within and across questions were identified.

Results and Discussion Quantitative results are presented for the entire sample and each of the four spatial subsets in order to illustrate pattern consistency and identify possible areas of further investigation. Qualitative themes are summarized by question or question set. Commentary is meant to assist in the practical understanding of general quantitative trends and is not separated into spatial groupings.

Understanding of Flood Related Uncertainty When asked which of the described floods they thought could happen more than once per year, about nine percent of the whole sample said they didn’t know and approximately 45 percent said all of them could happen more than once. These figures become neutral in analyses of variance. Affirmative response percentages and the results of comparisons are included in Table 7-1. It should be noted that only two people said that all three descriptions referred to the same flood; six identified the one percent chance and the 100 year floods as equal. Description rankings were consistent across spatial subsets. The one percent chance description always had the highest response rate and the 100 year description the lowest. The omnibus test indicated significant general variation between response percentages in each spatial grouping. Post hoc testing to identify specific differences showed the 100 year description to perform significantly worse than both the one percent chance and 26 percent chance descriptions in all categories. Differentiation between the two probability based descriptions was less clear, showing significance only within the total sample. Unlike the ranking of descriptions, percentages varied. Every description had a higher response rate in the 100 year floodplain. This was not surprising, as 13 percent of those in the 500 year floodplain answered “Don’t Know” versus only four percent of those in the SFHA. The result may be a function of experience. Experience did come up in participant commentary on this question (Table 7-2). Remarks were associated with the one percent chance and 100 year descriptions as well as responses of “All” and made reference to concrete examples, general personal experience and general community

Chapter Seven

72

experience. If a specific description was chosen, the choice seemed to depend on what label a respondent had assigned to recent floods and if that label was consistent. This would suggest that the naming of experience, in addition to experience itself, may be important to individuals’ understanding and assessment of risk. Table 7-1. Variation in possibility of occurring more than once per year: Cochran’s Q

100 year 1% chance 26% chance

All (N=114) % Yes 46 78 67

Union (N=60) % Yes 40 78 67

Vestal (N=54) % Yes 52 78 67

100 Year (N=50) % Yes 54 84 76

500 Year (N=64) % Yes 39 73 59

Omnibus

43.14**

28.76**

Q Statistic 14.80**

18.10**

25.31**

Post Hoc Significance a 100 & 1% ** ** ** ** ** 100 & 26% ** ** * ** ** 1% & 26% * NS NS NS NS a McNemar’s test used in post hoc comparisons; significance adjusted using Holm method. * p>0.05 **p>0.01 Table 7-2. Participant comments on multiple floods per year Themes Experience

Outlook Likelihood Climate Change Difficulty

Representative Comments Our car almost got wiped out in June AND November The 1% HAS happened more than once a year I say they could happen more than once only because of personal experience – anything’s possible Anything can happen Just had the 100 year, won’t happen anytime soon The 1% and 26% could happen more than once. With climate change, you never know Whether they happen more than one time depends on the size It’s hard to tell which floods could happen more than once by the descriptions

In the third comment, “anything’s possible” hints at the potential interaction of experience and general outlook. This phrase or something

Framing Flood Risk

73

like it was usually accompanied by a response of “All,” and, like climate change, came up more often when discussing change in size over time. The linking of climate change only to the overtly probabilistic descriptions may demonstrate the strength of association between the 100 year description and ideas of a strict cycle. In the course of the survey, two people stated that the 26 percent chance flood occurred every seven years. A few said that the one percent chance flood happened every year. Many people made statements consistent with a belief that the “100 year flood” happened once every 100 years. Comments were made when discussing the possibility of multiple floods per year, relative concern, and when asked directly about likelihood of occurrence in the next year. About 72 percent of respondents chose the 100 year flood alone as the least likely to occur. Only eleven percent of participants chose the one percent chance flood, the description with the second highest response rate. These statements again show the potential relationship between experience, the naming of experience, and perception, and give further credence to critiques of the 100 year flood description. Another theme running through commentary on each of the questions was dissatisfaction with the descriptions. In some cases, it was a general sense of difficulty or confusion. Others pointed to a particular problem. The participant who said that, “Whether they happen more than one time depends on the size,” ultimately answered “Don’t Know” and indicated that a piece of information vital to making a differentiation (the physical size of a flood) was missing. While some difficulty with the questions themselves might be expected, this man may have articulated specifically a frustration others expressed more generally. A separate individual gave a somewhat exasperated response decrying the lack of size markers in the descriptions when asked about change in size over time (see Table 7-3). A much higher proportion of participants (24 percent) answered “Don’t Know” to this question and made comments about having trouble. This perhaps reflects the level of difficulty of the question as well as the perceived lack of pertinent information in the descriptions. It may also indicate that a number of people are unfamiliar with the kinds of conditions and data constraints that impact flooding and flood estimation. Those who believed at least one of the described floods could change often explained their reasoning, however. Most cited either climate change or human alteration of the environment as a factor in their response. Some linked the two. Participants relied on their own interpreted experience as well as the experiences of their social contacts to estimate change. One

74

Chapter Seven

person specifically mentioned “An Inconvenient Truth” as an influence. Informal communication appeared to have made a difference in some people’s decisions, as did a general expectation of change or lack of control. Table 7-3. Participant comments on change in size over time Themes Outlook Experience Climate Change Human Influence

Difficulty

Representative Comments Time can change anything Floods are getting bigger and bigger My neighbors say there are big time increases in flooding With global warming all of them could change I suspect they can change, especially due to human activity All of them could change when we’re fooling with Mother Nature But these descriptions don’t GIVE the size! This is confusing

None of the comments listed in Table 7-3 contain references to the descriptions. None of the recorded comments did either. Statistical tests showed no difference between any of the terms in any spatial category with regard to change in size over time. The largest difference was only four percentage points. In fact, 65 percent of the total sample said that all could change over time, 24 percent answered “Don’t Know,” and six percent believed none could change. That left potential variation among only six people. Similar patterns were found in the subsets, though all six who chose only one or two descriptions lived in the five hundred year floodplain. Relative experience may be a factor, but thought processes associated with this question appeared unrelated to the terms presented.

Relative Concern When asked about the most and least concerning of the floods described, about 43 percent said they were equally concerning. The comments in Table 7-4 indicate that a response of “All” could reflect either a general position that, as one participant put it, “Any flooding equals concern,” or an opposite contention that flooding isn’t really a problem for them and they don’t worry about it much. For these 43 percent, the descriptions were somewhat irrelevant; other situational and cognitive factors seemed to have more effect on response.

Framing Flood Risk

75

Table 7-4. Participant comments on concern Themes Outlook Mitigation Likelihood

Size Difficulty

Representative Comments Any flood is bad I’m not personally in danger It’s pretty safe if they do something I’ve insured the hell out of the house This (100 year) already happened It (100 year) makes you leery, but I won’t see it again in my lifetime If you have a disaster (100 year), you’re probably okay 30 years is a short time. Wow, that concerns me I’m least concerned about the 26% because it’s the smallest I’m most concerned that one will exceed the last. These aren’t very good indicators.

Like general outlook (perhaps associated with experience or location), actions to mitigate physical or financial impact influenced answer choice. Though individuals who made action related comments chose “All,” the actors involved and levels of concern differed. Comments involving “they” (rather than “I”) were attached to both relatively high and relatively low levels of concern; variation might be connected to the perceived likelihood of governments taking “proper” action. Trust and/or credibility may play a part. While responses of “All” perhaps had little to do with the descriptions, statistical results included in Tables 7-5 and 7-6 show significant variation in the remainder of the responses across all spatial categories. If a participant said that the descriptions were equally concerning, each was assigned a one in both questions, neutralizing the effect in analyses of variance. Response rate rankings were internally consistent for both most concerning and least concerning. In all subsets, more people chose the 100 year flood than either of the other two descriptions as the most concerning. The 26 percent description ranked second, but post hoc tests showed no significant difference between it and either the 100 year or one percent chance descriptions. The percentage range between rankings was most variable in the floodplain groupings. The one percent chance flood description had the lowest response rate for the most concerning flood and ranked highest in all subsets for the least concerning. One might expect to see a similar inversion with the 100 year description. Instead, the 26 percent chance description had the lowest response rate. In post hoc comparisons, however, only the differences between the 26 percent and the one percent descriptions were significant.

Chapter Seven

76

Table 7-5. Variations in most concerning flood: Cochran’s Q

100 year 1% chance 26% chance

All (N=114) % Yes 75 52 62

Union (N=60) % Yes 72 52 62

Vestal (N=54) % Yes 78 52 63

100 Year (N=50) % Yes 76 54 70

500 Year (N=64) % Yes 73 50 56

7.46*

9.05*

Q Statistic Omnibus

15.39**

6.00*

9.87**

Post Hoc Significance a 100 & 1% ** * ** * * 100 & 26% NS NS NS NS NS 1% & 26% NS NS NS NS NS a McNemar’s test used in post hoc comparisons; significance adjusted using Holm method. * p>0.05 **p>0.01. Table 7-6. Variations in least concerning flood: Cochran’s Q

100 year 1% chance 26% chance Omnibus

All (N=114) % Yes 61 73 50 15.86**

Union (N=60) % Yes 63 72 47 9.72**

Vestal (N=54) % Yes 57 74 54 Q Statistic 7.36*

100 Year (N=50) % Yes 64 76 54

500 Year (N=64) % Yes 58 70 47

7.28*

8.67*

Post Hoc Significance a 100 & 1% NS NS NS NS NS 100 & 26% NS NS NS NS NS 1% & 26% ** ** NS * * a McNemar’s test used in post hoc comparisons; significance adjusted using Holm method. * p>0.05 **p>0.01.

The comments in Table 7-4 hint at an explanation for the lack of logical consistency in the ranking of most concerning and least concerning descriptions. Participants did not relate the 100 year flood to concern in a consistent manner. Almost identical statements were accompanied by very different responses and ratings on the concern scale. Several justified their answers with variations of “It already happened.” One might rank it as least concerning; another might have said most concerning, but rated it a

Framing Flood Risk

77

one. Similar discrepancies emerged with more cautious individuals. The person who wouldn’t “see it again in my lifetime” was still “leery” and ranked the 100 year flood as most concerning, but rated it a four. The individual who equated the 100 year flood with disaster, but thought that “you’re probably okay” for the future if you experienced it, ranked the description as least concerning and also rated it a four. Elsewhere in the interview, she said, “Every time it rains, I worry. I think about it constantly.” This combination illustrates a sort of wishful thinking underlain by apprehension that was not uncommon. The above comments, along with the sheer proportion of statements dealing with the 100 year flood, indicate that this phrase elicits strong responses and has power. While a repeated measures ANOVA identified no significant differences in mean concern ratings, words mattered. Unfortunately, they mattered in different ways to different people. The 100 year description, perhaps because it is more frequently used and seemingly straightforward, appeared more prone to inconsistent interpretation. The results of the combined concern analyses presented in Table 7-7 also reflect irregularity. The one percent chance description had the lowest mean ranking in all spatial categories. Contrary to previous analyses, however, the relative rankings of the other two descriptions did not hold across subsets. Post hoc comparisons showed a pattern consistent enough to conclude that the one percent description was significantly less effective in inducing concern than the other two descriptions. There is not enough evidence in any grouping to suggest that the 100 year description is more persuasive than the 26 percent chance description, or vice versa. Comments in Table 7-4 indicate that participants came to contradictory conclusions regarding the 26 percent chance description, as they did for the 100 year description. Unlike the 100 year description, though, the 26 percent chance flood ranked highly because participants generally did not have strong reactions to it vis-à-vis concern.

Combined Measure of Effectiveness Because answers to the question about change in size over time appeared to have little to do with the descriptions themselves, the question was not included in either a scale of overall understanding or a scale of combined effectiveness. The concern scale ranged from zero to two. In order to equally weight understanding and persuasion components in a scale of overall effectiveness, responses to the question regarding more than one flood per year were multiplied by two and added to the concern

Chapter Seven

78

score. The combined scale ranged from zero to four. The results of a Friedman test are presented in Table 7-8. Table 7-7. Variation in relative concern: Friedman test

100 year 1% chance 26% chance

All (N=114) % Yes 2.12 1.77 2.11

Union (N=60) % Yes 2.07 1.78 2.15

Vestal (N=54) % Yes 2.18 1.75 2.07

100 Year (N=50) % Yes 2.11 1.75 2.14

500 Year (N=64) % Yes 2.13 1.78 2.09

Chi Square Statistic Omnibus

17.27**

7.82*

10.69**

9.71**

8.05*

Post Hoc Significance a 100 & 1% ** NS ** * * 100 & 26% ** NS ** NS NS 1% & 26% ** * * * * a Wilcoxon test used in post hoc comparisons; significance adjusted using Holm method. * p>0.05 **p>0.01. Table 7-8. Variation in combined effectiveness: Friedman test

100 year 1% chance 26% chance

All (N=114) % Yes 1.83 2.04 2.13

Union (N=60) % Yes 1.75 2.08 2.18

9.12**

9.48**

Vestal (N=54) % Yes 1.92 2.01 2.07

100 Year (N=50) % Yes 1.79 2.00 2.21

500 Year (N=64) % Yes 1.86 2.08 2.06

Chi Square Statistic Omnibus

1.22

8.17*

3.01

Post Hoc Significance a 100 & 1% NS NS NS NS NS 100 & 26% ** ** NS * NS 1% & 26% NS NS NS NS NS a Wilcoxon test used in post hoc comparisons; significance adjusted using Holm method. * p>0.05 **p>0.01.

Relative rankings were consistent in four of the five groupings, with the 26 percent chance description at the top and the 100 year description at the bottom. The scale’s calculation gave a slight advantage to descriptions that performed well in the understanding component; descriptions received

Framing Flood Risk

79

either a zero or a two, whereas in the concern portion, a score of zero, one, or two was possible. This may explain, in part, the one percent chance description’s relatively high ranking, given its poor performance in the concern section. However, percentage ranges between the 100 year and one percent descriptions were closer in the concern results than they were in the results for understanding. The 100 year description was also hurt by the split response with regard to concern. General variation between the descriptions was significant in only three of the groupings. In each case, the result was driven by the disparity between the 26 percent and 100 year descriptions. Post hoc comparisons showed no significant differences between either of these and the one percent description. While the rank order of descriptions was fairly consistent, the somewhat ambiguous results may point to a problem in combining the two conceptions of effectiveness.

Conclusions In order to promote a better understanding of flood related uncertainty, several organizations have recommended using probabilistic descriptions rather than the return period in public communication (Faber, 1996; Gruntfest et al., 2002; NRC, 2000). Previous research has shown that the one percent description may be associated with higher levels of understanding compared to the 100 year description (Bell and Tobin, 2007). In this study, both the one percent and 26 percent chance descriptions performed better than the 100 year description in terms of understanding. The NRC (2000, 2006) has conceptually linked better decision making and behavior to the understanding of uncertainty, but the results presented here suggest that understanding and persuasion may not necessarily be linked. Other research has led to similar conclusions (e.g., Beehler et al., 2001; Bell and Tobin, 2007; Sjoberg, 2000). Which description is best might depend on whether public understanding or persuasion is the goal; risk managers and communicators may have to decide which is ethically and/or practically more important in the current system of flood loss mitigation and distribution. The 26 percent chance description appears the most effective overall, given the shortcomings of the other two. It also received the highest response rate when participants were asked which of the floods described was most likely to occur within the next year. There are reasons to be cautious, however. The 26 percent description scored well, in part, because it did not stand out. This may be due to its unfamiliarity to most respondents. While lack of familiarity could be seen as a benefit, research

80

Chapter Seven

involving a similarly unfamiliar group in a different location showed strong negative reactions to the 26 percent description (Bell and Tobin, 2007). Comments in Table 7-4 touch on another potential difficulty for all the descriptions. In a setting that encourages heuristic processing, individuals react to different message components and come to contradictory conclusions. Future research to clarify relationships among perceived size, likelihood, concern, and specific descriptions would be useful. Many individuals admitted trouble attaching a physical size to the descriptions, however, and vocalized a general frustration with all of the terms. The effectiveness of each might be improved with the inclusion of damage estimates (NRC, 2000; Smith, 2000), early and continuing education (Frech, 2005), and visual reminders (Siegrist and Gutscher, 2006). Participatory research could result in additional relevant suggestions. Researchers can help practitioners by examining: 1) what people want to know about flooding, 2) how they talk about flooding, 3) what concerns them about flooding, and 4) who they talk to about flooding and who they want to talk to them. It is unlikely that the answers to these questions will be identical across physical or social space, but common themes might be identified. In this study, participant comments pointed out the importance of experience, the naming of experience, outlook, mitigative behavior, and information networks in forming understandings of and attitudes toward flood risk. Researchers must continue to tease out the complicated relationships between these factors and others identified in the literature in order to improve our understanding of not only perception and behavior, but the role of risk communication in creating resilient individuals and communities.

References Beehler, B.P., B.M. McGuinness, and J.E. Vena. 2001. Polluted Fish, Sources of Knowledge, and the Perception of Risk: Contextualizing African American Angler’s Sport Fishing Practices. Human Organization 60(3). Bell, H. and G.A. Tobin. 2007. Efficient and Effective? The 100-Year Flood in the Communication and Perception of Flood Risk. Environmental Hazards, 7(4):302-311. Faber, S., 1996. On Borrowed Land: Public Policies for Floodplains. Institute of Land Policy. Cambridge, MA: Lincoln. Frech, M. 2005. Flood Risk Outreach and the Public’s Need to Know. Journal of Contemporary Water Research and Education 130:61-69.

Framing Flood Risk

81

Gilbert F. White National Flood Policy Forum (GFWNFPF). 2004. Reducing Flood Losses: Is the 1% Chance Flood Standard Sufficient? Report of the 2004 Assembly of the Gilbert F. White National Flood Policy Forum, Washington, D.C. Gruntfest, E., K. Carsell and T. Plush. 2002. An Evaluation of the Boulder Creek Local Flood Warning System. Colorado Springs: University of Colorado at Colorado Springs. National Research Council (NRC). 2000. Risk Analysis and Uncertainty in Flood Damage Reduction Studies. Washington, D.C.: National Academy Press. National Research Council (NRC). 2006. Completing the Forecast: Characterizing and Communicating Uncertainty for Better Decisions Using Weather and Climate Forecasts. Washington, D.C.: National Academy Press. Perry, C.A. 2000. Significant Floods in the United States During the 20th Century-Measures a Century of Floods. USGS Fact Sheet 024–00. Ross, T. and N. Lott. 2006. Billion Dollar U.S. Weather Disasters, 1980 – 2005. Ashville, N.C.: National Climatic Data Center. Smith, D.I. 2000. Floodplain Management: Problems, Issues and Opportunities. In Floods: Volume I, D.J. Parker, ed. London: Routledge. Siegrist, M. and H. Gutscher. 2006. Flooding Risks: A Comparison of Lay People’s Perceptions and Expert’s Assessments in Switzerland. Risk Analysis 26(4):971-979. Sjöberg, L. 2000. Specifying Factors in Radiation Risk Perception. Scandinavian Journal of Psychology 41(2):169-174. United States Water Resources Council (USWRC). 1971. Regulation of Flood Hazard Areas to Reduce Flood Losses: Volume 1. Washington, D.C.: U.S. Government Printing Office.

CHAPTER EIGHT INTERNATIONAL STUDENTS’ VULNERABILITY TO EMERGENCY EVENTS: DOES TENURE OF RESIDENCE MAKE A DIFFERENCE? XUEQIN (ELAINE) HE

Introduction At present approximately 2 million university (college) students, worldwide, study outside of their home country, and the predicted international student population is expected to expand to 18 million by 2025 (Altbach, 2004). The U.S. had housed near 600,000 international students in 2004, which was approximately 25 percent of the world’s foreign students (Altbach, 2004). The population of international students in the U.S. increased dramatically over the past five decades (Figure 8-1). During the 19541955 academic year, there were 34,232 international students, which composed about 1.4 percent of all students enrolled in the U.S. Toward the end of the 1970s, international students accounted for 2.4 percent (286,343) of all students enrolled. By the 1999-2000 academic year, 514,723 (about 3.8 percent of the total student population) international students enrolled in American colleges and universities. During the 20022003 academic year, 586,323 international students, or 4.6 percent of total students in the U.S., studied at American colleges and universities (Altbach, 2004). The most recent data indicate that the size of the international student population in the U.S. has decreased since 20032004, but the absolute number is still significant. There are 564,766 international students in the U.S., which accounts for 3.9 percent of total student enrollment (Open Door, 2006).

Papers of the Applied Geography Conferences (2007) 30: 257-266

IInternational Students Vulneraability to Emerggency Events

83

The eveer-increasing size and th he unique ccharacteristicss of the international student population has motivaated studentt affairs professionalls, policy makers, m and scholars to study this specific population. These studiess have focused on two majjor themes. One O theme pays attentioon to the inceentive/motivattion of cross--border study (Chirkov et al., 20077; Peyton, 20005) while the t other foccuses on inteernational students’ miigrant adjustm ments, which include languuage confiden nce (Kim, 2006; Yangg et al., 2006)), socio-culturral adjustmennts (Kashima and Loh, 2006; Li annd Gasser, 2005; Major, 20 005), and psyychological ad djustment (Wang andd Mallinckroddt, 2006). However, H few w studies haave been conducted to examine internationaal students’ familiarity with w the biophysical environmentts of their host h cities, thheir level off hazards awareness, aand their emerrgency behaviior.

on in the USA A by academic level with Figure 8-1. International stuudent populatio selected academic year (Souurce: Open Doorrs, 2006)

This papper focuses onn the study of hazards and innternational education. e First, it ennumerates pottential explan nations for iinternational students’ potential vuulnerability to hazards. Then n, it presents a conceptual model to illustrate the connection between inteernational stuudents and em mergency events. Reseearch questionns are develop ped and tested based on the proposed model.

Inteernational Students and a Emergeency Eventts There arre a number off reasons to ex xpect that inteernational stud dents may be more vullnerable to ennvironmental hazards. h Mostt internationall students

84

Chapter Eight

are unfamiliar with the environments of host countries and cities. They are faced with a new culture and new biophysical environments, the absence of established social networks, and occasional problems with language. All these work together to make international students more susceptible to hazards than domestic students, and they may encounter more difficulties during emergency situations.

Language Barriers Research about international tourism has found that language difference creates barriers that are perceived as troublesome by tourists (Basals and Klenosky, 2001). Students from non-English speaking countries usually encounter language barriers when they come to the U.S. as international students without prior foreign study experiences (Kim, 2006; Yang et al., 2006). Language barriers may impede international students’ understanding of emergency messages about impending hazardous events. For example, Perry and Mushkatel (1986) have documented that non-English speaking groups were less likely to be aware of a warning. Since understanding the emergency information is the first step of emergency response (Fitzpatrick and Mileti, 1991), misunderstanding or incomplete understanding of warning messages will affect emergency response and evacuation decisions.

Cultural Differences Cultural norms, values, and beliefs can influence perceptions of risks and their consequences (Weber and Hsee, 1998). Studies found that nationality not only affected behavior but also how people are influenced by others’ actions (Markus and Kitayama, 1991; Pizam and Sussmann, 1995). Typically, Asians wish to assimilate with others, while Americans stress self-awareness, appreciate people’s differences, and emphasize the importance of independence, suggesting that Asians may have a greater desire to behave similarly to others. In the case of hurricane emergency response, they may be more influenced by suggestions from relatives or friends and actions of their neighbors than Americans.

Environmental Familiarity The importance of “knowing” the residential biophysical environment has been tested. An empirical study of the decision to evacuate from Hurricane Bret found that the choice was significantly related to self-

International Students Vulnerability to Emergency Events

85

reported awareness of the spatial distribution of risk (Zhang et al., 2004). A comparative study of beach safety between international and Australian students found that international students were more likely to engage in risky behavior and were less aware of beach safety practices than their Australian counterparts (Ballantyne et al., 2005). As international students are newcomers, they need time to get to know both the biophysical surroundings of their homes and their hosts.

Disaster Experiences International students’ disaster experiences in their home countries may affect their perceived risk and actual response in the U.S. If students know nothing about a hazard, they may either fear or be unaware of the approaching threat. If they have experienced a hazard, they may understand it and know how to respond to their perceived threats (Correia et al., 1998; Lindell and Whitney, 2000). If students have been continuously exposed to a hazard, they may habituate to it and may not believe the warning is important (Richardson et al., 1987).

Interrupted Social Networks Social networks established in their home countries are dismantled when international students leave home. Some international students are fortunate to have relatives or friends, but most of them are on their own. Social networks can provide support in times of emergency and crisis (Elliott and Pais, 2006). Time is required to build social networks anew and the rate at which it occurs varies from one person to the next.

Financial Burdens Disaster evacuation can be expensive. An empirical study found that the average Hurricane Bonnie evacuee spent nearly $470 living in a hotel, $144 staying in a public shelter, and $174 sheltering at the home of a friend or their family (Whitehead, 2003). About two thirds of international students are supported by their families (Open Doors, 2006). Discovery of the financial burden that evacuation has on their families often produces guilt (Yen and Stevens, 2004). The financial burden contributing to international students’ vulnerability to hazards can be further illustrated by the cost of hazard-proof facilities.

86

Chapter Eight

A Model of International Students and Emergency Events Based on the above discussions, a conceptual model was developed to illustrate the vulnerability of international students to emergency events (Figure 8-2). The model consists of three parts: the international student, an emergency situation, and the interaction between the first two components. The international student component includes potential difficulties that international students may encounter in host countries. The emergency situation is comprised of the information about and responses to the hazard (receiving warning information, perception of a threat, decisions about evacuation, and evacuation behavior). The outcome of the interaction of the problems encountered by international students during an emergency event is illustrated by the third part of this model. Length of residence of a student may change the potential difficulties that may increase the vulnerability of international students during emergency events. It is usually that a language barrier will diminish with increasing years in the host country, cultural differences will weaken, familiarity with biophysical environments will increase, new social networks will be established, and financial burdens may also be attenuated.

Research Questions Many problems encountered by international students gradually diminish with the increasing length of residence, regardless of whether students intend to assimilate or not. Does the negative impact caused by these challenges with respect to international students’ emergency response also diminish? To answer this general question, five specific research questions were developed as follows: 1. Do international students become more familiar with their biophysical environments as the length of residence increases? 2. Do international students become more knowledgeable about the hazards of their biophysical environments as the length of residence increases? 3. Was the likelihood of international students’ decision to evacuate during Hurricane Rita associated with their length of residence? 4. Was international students’ evacuation expectation associated with their length of residence? 5. Were the orders to evacuate, the evacuation decisions of neighbors or the cost of evacuation important influences that guided international students to decide to evacuate during a hurricane?

International Students Vulnerability to Emergency Events

87

Survey An online survey was designed to collect data from international students. The survey was comprised of questions from a variety of areas. Survey questions focused on the length of residence of the respondent, their environmental familiarity, their knowledge of hurricanes, their likely decision to evacuate in the future, determinants of their evacuation decisions, and their hurricane evacuation experiences. The question, “How long have you lived in the Houston area?” was used to measure respondents’ length of residence. Environmental familiarity was assessed by the following three questions, “Do you live in a floodplain?” “Do you live in a hurricane evacuation zone?” and “Are you familiar with the recommended evacuation routes for your residence?” Survey respondents were asked to identify the relative severity of “hurricane watch” versus “hurricane warning,” and between “mandatory evacuation” and “voluntary evacuation.” These items were used to assess students’ knowledge of hurricane risk. The respondent’s evacuation likelihood was assessed by asking survey respondents under which warning situations (hurricane watch, hurricane warning, voluntary evacuation and mandatory evacuation) they would evacuate. Survey respondents were asked to identify the determinants of evacuation decision from nine variables (warning message, personal experience, advice of family or friends, difficulty of the last evacuation, cost, knowing where to go, wind speed, actions of neighbors, and floods near home). Respondents were asked to report their evacuation experience during Hurricane Rita in 2005. It has been reported that half of Houston-area residents (about 2.5 million) fled Hurricane Rita during its approach (Mack, 2005). What were international students’ responses to Hurricane Rita? Did they evacuate? If so, where did they go? How did they identify the location of a public shelter? What were the major evacuation impediments they encountered during their evacuation? All of these questions will help to reveal the vulnerability of international students in hurricane evacuations.

88

Chapter Eight

Figure 8-2. Innternational studdents and emergency events

International Students Vulnerability to Emergency Events

89

In addition, questions were asked about English proficiency, residential status (roommates and room facilities), automobile ownership, and demographic factors (citizenship, age, sex, education, marital status, and number of children). Most questions were close-ended. Several questions were open-ended to provide opportunities for respondents to comment in their own words.

Survey Participants The survey was distributed to international students in the Houston area through international-student associations and university administrations with an incentive offered for reply. Universities and colleges involved in the online survey included the American College of Acupuncture & Oriental Medicine, Baylor College of Medicine, Houston Baptist University, Rice University, Texas Southern University, University of St. Thomas, and Galveston College. A total of 111 completed responses were received. The survey respondents were predominantly female (63 percent). Chinese (including Taiwanese and Hong Kong) students accounted for 30 percent. Students from India and South Korea accounted for 12 percent and 8 percent respectively. This distribution matches the top three countries of origin of international students in the U.S. The average age of respondents was 26 years old. Thirty-nine percent of them were undergraduate students. Most international students reported being single (78 percent) and over 80 percent said they live in the U.S. without family members. Half of the students said they share rooms with others. Forty-five students have lived in the Houston area for less than one year. The remaining 66 students (around 60 percent) have lived in Houston for more than 2 years, which means that they experienced Hurricane Rita.

Results Environmental Familiarity The relationship between international students’ environmental familiarity and their length of residence is presented in Table 8-1. The result indicates that there is not a clear pattern between students’ environmental familiarity and their length of residence. This unexpected finding might not reflect the truth. All analyzed data were self-reported, and responses were not verified objectively. Therefore, the respondents who are aware of their floodplain or evacuation zone may in fact be wrong

90

Chapter Eight

and unfamiliar with their environments. Table 8-1. Length of residence vs. environmental familiarity (Respondents who are aware or know their surrounding environments) Residence 0-1 year 2-3 years 4 or more

Live in Floodplain N % 17 37.8 19 50.0 10 35.7

Live in Evac. Zone N % 19 42.2 14 36.8 10 35.7

Know Evac. Route N % 7 15.6 8 21.0 6 21.4

Knowledge of Hurricanes International students’ knowledge of hurricanes changes with their tenure of residence in the Houston area (Table 8-2). Pearson correlation tests indicate that the length of residence is positively related to knowledge of hurricanes. Table 8-2. Length of residence vs. knowledge of hurricanes

Residence 0-1 year 2-3 years 4 or more Correlation Coefficient

Hurricane Warning More Severe N % 26 57.8 29 76.3 22 78.6 R = 0.203, p0, if for k = 0, 1, 2, ... the probability mass function of X is given by Pr(X = k) = Ȝke-Ȝ/k!

[5]

where e is the base of the natural logarithm (e = 2.71828...), k! the factorial of k, and Ȝ the positive real number equal to the expected value of X. For this study, the selected data covered a 100-year period. If the cumulative frequency is denoted as CF, the expected annual average number of the storm events over this period is Ȝ = CF/100. If we expect a specific category of storms to occur only once per year, k = 1 and k! = 1! = 1. Then the expected probability (EP) Equation 5 can be operationalized as follows EP = (F/100) (e-(F/100)) [6] where CF is the cumulative frequency. Again, the output matrix of calculated probabilities was saved in the aforementioned file and mapped out in ArcGIS 10.

Chapter Twelve

144

Results Historical Hurricanes Totally 15 hurricanes made landfall in South Carolina between 1906 and 2005 and six more within the 50-nautical mile-buffer in Georgia and North Carolina. As shown in Table 12-1, of these historical hurricanes that made landfall in South Carolina, one third formed in the western part (WA) of the North Atlantic Basin. Although the state was hit once in 1894 by a hurricane that originated in the Gulf of Mexico (GM), none was found from this region over the study period. Most hurricanes from the Gulf region have been degraded to tropical storms or depressions after passing through Florida or Georgia. Meanwhile, 80% of the hurricanes made landfall in the area directly from the open Atlantic Ocean (AO) and the remaining 20% made landfall first in the Florida Peninsula (FP) or Georgia coast before traveling partially over land to South Carolina. Among the 15 hurricanes are three major hurricanes, Hugo (H4, 1989), Gracie (H4, 1959), and Hazel (H3, 1954). Table 12-1. Hurricanes making landfall in South Carolina, 1906-2005 Lande d

Breeding Region

Landed Month

Path

Aug Sep. Oct. AO FP . 5 0 0 0 0 0 0 0 0 0 0 4 2 1 1 0 0 0 2 0 2 0 3 1 0 1 0 0 0 0 1 1 0 2 3 1 2 0 1 2 0 0 3 0 1 9 3 6 0 2 4 2 1 6 3 Total 15 5 10 0 3 6 4 2 12 3 Notes: EA = East Atlantic; WA = West Atlantic; GM = Gulf of Mexico; AO = Atlantic Ocean; and FP = Florida Peninsula. No hurricanes occurred in June or November during this period. Cat.

N

EA

WA

GM

Jul.

Officially, the hurricane season in the North Atlantic basin is between June and November. However, the records of the study area show only one hurricane happened during the month of June (in 1867); this hurricane predates the study period so there is no column for "June" in the tables. Usually, August is the warmest month of the year in the North Atlantic Ocean but not the peak month in hurricane activity. This is not the case in South Carolina because hurricane strikes over the study period peaked in the warmest month of the year. However, hurricanes developing in August

Spatial Prediction of Hurricane Strikes in South Carolina

145

or earlier all fall into the relatively weak categories (H1 and H2), whereas all three major hurricanes (H3 and H4) occurred in the latter half of the season. Figure 12-1 depicts the tracks of these hurricanes.

Figure 12-1. Tracks of hurricanes that affected South Carolina, 1906-2005

Cumulative Frequencies and Expected Probabilities Figure 12-2 depicts the temporal spacing of hurricanes over the onehundred year period. Hurricanes that made landfall in or had an effect on the state tended to strike cyclically with three active periods lasting 19-25 years (1906-1928, 1940-1959 and 1979-2004) and dormant intervals lasting 8-19 years (1929-1939 and 1960-1978). Although the last two active periods each saw eight hurricanes, no evidence indicates that hurricane frequency was steadily increasing over the entire time span. The intensity, however, demonstrates an increasing trend over the same time span. Spatially, the effect of hurricanes varies in extent, intensity, and frequency as shown in Figure 12-3. Although tropical storm (TS) force winds resulting from the 21 hurricanes have reached the entire state, their frequency decreases with distance from the coastline. Note that within the boundary of South Carolina, places located north of Charleston got hit 12

146

Chapter Twelve

times by hurricane force (H1 or higher) winds over a hundred year period as compared to 15 times aggregated for the state as reported in Table 12-1. In other words, these places received a returning visit by a hurricane every 8.3 years. The coastal area near the northeastern corner, on the other hand, experienced only 5 events of hurricanes at a 20-year returning interval. It should be pointed out that major hurricanes intruded farther inland though both hurricane frequency and spatial extent shrink towards the ocean.

Figure 12-2. Frequency of hurricanes that affected South Carolina, 1906-2005

The expected probabilities of hurricanes follow similar spatial patterns to those of the cumulative frequencies but in different concepts and classification intervals as shown in Figure 12-4. The maximum probability for a hurricane to strike once a year is only 11.25%, while the similar number for a category-4 hurricane to strike is only 5%. Maps in the figure show that along the South Carolina coast, the middle (around Charleston) and southern (around Beaufort) segments are prone to more frequent and intense hurricane strikes, while the northern segment (around Myrtle Beach) suffers the effect of hurricanes making landfall in North Carolina or by-passing offshore, though the effect is relatively weak because it is located to the left side of those hurricanes.

Discussion This study mapped the frequencies and probabilities of hurricanes that affected South Carolina. By applying a modified distance decay function, we were able to delineate the maximum surface wind field of each individual hurricane and thus the affected areas of historical hurricanes by category or intensity. The output makes it possible to understand the spatial distribution of hurricane effects from a zero dimension (data points) to one dimension (linear tracks) and to two dimensions (impacted areas) in space. Unlike those from previous aspatial studies, the resulting maps

Spatial Prediction of Hurricane Strikes in South Carolina

147

allow spatial differentiation of the extent, intensity, frequency, and probability of hurricanes across the state of South Carolina. Furthermore, these maps reflect the total hurricane effect on the border regions more accurately than those generated for South Carolina without consideration of the effect of hurricanes that made landfall in the neighboring states.

Figure 12-3. Cumulative frequencies of hurricane strikes in South Carolina between 1906 and 2005

148

Chapter Twelve

Figure 12-4. Expected probabilities of hurricane strikes once per year in South Carolina.

Findings of the study have multiple implications. Emergency managers can use these maps for mitigation planning and disaster preparedness. The middle and southern segments of the South Carolina coast were identified as the most vulnerable areas prone to frequent and intense hurricane strikes and thus should be targeted for allocating limited resources and prioritizing mitigation efforts, particularly when factoring in the fact that they are among the most populated and fastest growing urbanized regions – Charleston and Beaufort-Hilton Head. If history repeats, the greatest threat is expected to come from major hurricanes making landfall at a large or right angle to the coastline from the open Atlantic Ocean. The

Spatial Prediction of Hurricane Strikes in South Carolina

149

landward boundary of the affected areas for each intensity category can be used as a reference for adjusting or enforcing building code. For insurers, the probability maps are more valuable for determining the insurance rate for wind damages. Limitations of the study can be partially removed if NOAA densifies its best track data points and the intensity decay model is enhanced. The current data record locations of the center of a hurricane once every six hours; a hurricane can travel hundreds of miles over such a time period. Interpolation of the best track data is necessary for studies at the state scale but will inevitably induce error. The intensity decay function used in this study is a simple model that needs to be improved in order to accurately represent the complex hurricane systems yet to be fully understood. Although a simple model is necessary in a data-rare environment when historical hurricanes are involved, a sophisticated model can be developed specially for recently landed hurricanes with relatively rich observation data. For future research, a dual-model approach can be applied to add skill in delineating the hurricane wind fields and to reduce the uncertainty in predicting hurricane frequencies and probabilities.

References ASFPM. 1992. Learning from Hurricane Hugo: Implications for Public Policy (An Annotated Bibliography). The Association of Floodplain Managers, Inc. Report to the Federal Emergency Management Agency (FEMA). Contract No. EMW-90-G-33-4 A001. Blake, E.S., C.W. Landsea, and E.J. Gibney. 2011. The Deadliest, Costliest, and Most Intense United States Tropical Cyclones from 1851 to 2010 (And Other Frequently Requested Hurricane Facts). NOAA Technical Memorandum NWS NHC-6. http://www.nhc.noaa.gov/ pdf/nws-nhc-6.pdf. Last accessed 9 June 2012. Crossett, K.M., T.J. Culliton,, P.C. Wiley, and T.R. Goodspeed. 2004. Population Trends Along the Coastal United State: 1980-2008. National Oceanic and Atmospheric Administration. Darling, R.W.R. 1991. Estimating Probabilities of Hurricane Wind Speeds using a Large-Scale Empirical Model. Climate 4: 1035-1046. DeMaria M., J.A. Knaff, R. Knabb, C. Lauer, C.R. Sampson, and R.T. DeMaria. 2010. A New Method for Estimating Tropical Cyclone Wind Speed Probabilities. Weather and Forecasting 24: 1573-1591. Doyle, T.W. 2009. Hurricane Frequency and Landfall Distribution for Coastal Wetlands of the Gulf Coast, USA. Wetlands 29 (1): 35-43.

150

Chapter Twelve

Emanuel, K. 2005. Increasing Destructiveness of Tropical Cyclones over the Past 30 Years. Nature 436:686–88. Gray, W.M., C.W. Landsea, Jr., P.W, Mielke, and K.J. Berryet. 1994. Predicting Atlantic Basin Seasonal Tropical Cyclone Activity by 1 June. Weather and Forecasting 9: 103-115. Holland, G.J. 1980. An Analytic Model of the Wind and Pressure Profiles in Hurricanes. Monthly Weather Review 108: 1212–1218. Holland, G. J., J. I. Belanger, and A. Fritz. 2010. A Revised Model for Radial Profiles of Hurricane Winds. Monthly Weather Review 138: 4393-4401. Kaplan, J., and M. DeMaria. 1995. A Simple Empirical Model for Predicting the Decay of Tropical Cyclone Winds after Landfall. Journal of Applied Meteorology 34: 2499-2512. Kimball, S. K., and M. S. Mulekar. 2004. A 15-year Climatology of North Atlantic Tropical Cyclones. Part I: Size Parameters. Journal of Climate 17: 3555–3575. Klotzbach, P., and W. Gray. 2012. United States Landfall Probability Webpage. A collaborative Tropical Meteorology Research Project at Colorado State University and the GeoGraphics Laboratory at Bridgewater State University. http://www.e-transit.org/hurricane/ welcome.html. Last accessed 8 June 2012. Knaff, J.A., D.P. Brown, J. Courtney, G.M. Gallina, and J.L. Beven. 2010. An Evaluation of Dvorak Technique–Based Tropical Cyclone Intensity Estimates. Weather and Forecasting 25: October 1362-1376. Knaff, J. A., and R. M. Zehr. 2007. Reexamination of Tropical Cyclone Wind–Pressure Relationships. Weather and Forecasting 22: 71–88. Murnane, R.J., C. Barton, E. Collins, J. Donnelly, J. Elsner, K. Emanuel, I. Ginis, S. Howard, C. Landsea, K.B. Liu, D. Malmquist, M. McKay, A. Michaels, N. Nelson, J. O’Brien, D. Scott, and T. Webb III. 2000. Model Estimates Hurricane Wind Speed Probabilities. EOS, Transactions of the American Geophysical Union 81(38): 433-438. Neumann, C.J., B.R. Jarvinen, C.J. McAdie, and J.D. Elms. 1999. Tropical Cyclones of the North Atlantic Ocean, 1871-1998. Historical Climatology 6(2): 206. NHC. 2009. Technical Summary of the National Hurricane Center Track and Intensity Models. National Hurricane Center - National Oceanic and Atmospheric Administration. http://www.nhc.noaa.gov/pdf/ model_summary_20090724.pdf. Last accessed July 2009. Pielke, R.A., Jr., J. Gratz, C.W. Landsea, D. Collins, M.A. Saunders, and R. Muslin. 2008. Normalized Hurricane Damage in the United States: 1900-2005. Natural Hazard Review 9: 29-42.

Spatial Prediction of Hurricane Strikes in South Carolina

151

Rappaport, E.N., J.L. Franklin, L.A. Avila, S.R. Baig, J.L. Beven II, E.S. Blake, C.A. Burr, J-G. Jiing, C.A. Juckins, R.D. Knabb, C.L. Landsea, M. Mainelli, M. Mayfield, C.J. McAdie, R.J. Pasch, C. Sisko, S.R. Stewart, and A.N. Tribble. 2009. Advances and Challenges at the National Hurricane Center. Weather and Forecasting 24: April 395419. Simpson, R.H. 1974. The Hurricane Disaster Potential Ccale. Weatherwise 27: 169-186. Vickery, P. J., P. F. Skerlj, and L. A. Twisdale. 2000. Simulation of Hurricane Risk in the U. S. using Empirical Track Model. Journal of Structural Engineering 126: 1222- 1237. Vickery, V.J., F. J. Masters, M. D. Powell, and D. Wadhera. 2009. Hurricane Hazard Modeling: The Past, Present, and Future. Journal of Wind Engineering and Industrial Aerodynamics 9(7): 392-405. Webster, P.J., G.J. Holland, J.A. Curry, and H.R. Chang. 2005. Changes in Tropical Cyclone Number, Duration, and Intensity in a Warming Environment. Science 309:1844-1846.

PART II TECHNOLOGY

INTRODUCTORY COMMENTS BURRELL E. MONTZ AND GRAHAM A. TOBIN

The technological advances of the last thirty years or so have greatly increased the arsenal for hazard researchers, providing new approaches and techniques never before thought possible. The scene was set, in part, by Ian McHarg’s (1969) ground-breaking work, Design with Nature, which advocated the consideration of layers of data in human endeavors. However, at the time, the challenge to incorporate detailed spatial (and temporal) analyses was overwhelming. Today, of course, the expansion of geographical information systems (GIS) provides new tools for in-depth analysis of a multitude of studies, not least being hazards. Investigations of dynamic forces over time and space can now be visually represented. Along with GIS, the refinement of remote sensing techniques, such as LiDAR, now facilitate development and application of complex models of landforms and human activities. Furthermore, others have utilized webbased resources and sophisticated statistical modeling to address a range of hazard-related topics. With these improvements in mind, we have included in this section a number of papers to illustrate evolving methodological approaches in hazards research. Certainly, GIS and remote sensing technologies have been used extensively and increasingly by hazard researchers to map and analyze hazard-prone areas and disaster impacts. In relatively early work, for example, Walsh (1986) used advanced high resolution satellite data to map drought conditions in Oklahoma, deriving spatial and temporal relationships through multiple regression models of the Crop Moisture Index, and Monmonier (1993) provided a general narrative on the importance of mapping and graphic narratives for emergencies. The concepts introduced by these two papers set the scene for later studies. Several papers in this section have been included to demonstrate the mapping potential of remote sensing and GIS techniques in hazard research. For example, such techniques were employed by Pearson (1996) to map flood-prone agricultural land in Mississippi; by Showalter and Lu (2010) to record flash flood crossings and swift water rescues in Texas; by Watson et al. (2003) to monitor weather hazards, including flash flooding

Introductory Comments

155

on rural roads in New Mexico; and by Firouzabadi et al. (2012) to assess flood damage in Iran. In addition, different hazards have been mapped by Wachal (2000) who looked at landslide susceptibility in Travis County, Texas, and Brinkmann et al. (2004) who mapped sinkholes in Pinellas County, Florida. An innovative approach to modeling hazards was adopted by Hu (2004) who used a combination of GIS, remote sensing, and virtual reality techniques to assess the flood hazard in Piasa Creek watershed in southwestern Illinois. The inclusion of virtual reality approaches to understanding hazards has further potential. Similarly, Montz (2000) applied various techniques to map hazardousness of place, in this case Hillsborough County, Florida which is subject to many different hazards. Finally, we included a couple of papers that demonstrate the power of remote sensing and GIS based-work to provide meaningful solutions to disaster managers. Stimers (2006), for example, used a GIS-based approach to analyze the appropriateness of tornado warning siren networks in two Kansas counties, and Prasad (2012) examined evacuation needs and shelter locations in Broward County, Florida. Together, these papers provide insight into the rapidly accumulating literature on hazards and technology, and we urge the reader to consider the broader perspectives as we strive to fully comprehend hazards and disasters. Hazard problems are complex and present many challenges; it is the technology that has greatly facilitated our research. Hazard problems are not easily defined because they interact with all aspects of society, suggesting a complexity that makes understanding, modeling, and mitigation difficult. All of these are complicated by scale and the spatial organization of society…Hazards, therefore, must be examined through filters that account for these spatial differences in physical, political, and economic contexts, as well as broader social needs. The advances in technology, especially GIS, remote sensing, and computerized databases, therefore, have expanded our research endeavors. —Tobin and Montz (2004, p. 566).





156

Part II

References 

McHarg, I. 1969. Design with Nature. New York: Natural History Press. Tobin, G.A. and B.E. Montz. 2004. Natural hazards and technology: Vulnerability, risk and community response in hazardous environments. In Geography and Technology, S.D. Brunn, S.L. Cutter, and J.W. Harrington, eds. Dordrecht, Netherlands: Kluwer Academic Publishers. Chapter 23, pp. 547-570.

CHAPTER THIRTEEN DROUGHT ASSESSMENT THROUGH ADVANCED VERY HIGH RESOLUTION RADIOMETER SATELLITE DATA STEPHEN J. WALSH

Introduction Drought has a variety of meanings. Both agricultural and meteorological droughts exist. A meteorological drought may be defined as deviations of precipitation below some mean value, whereas an agricultural drought is primarily determined by the moisture content of the soil and its availability for plant utilization throughout the growing season. Agricultural drought is a function of the interplay between rainfall, temperature, topography, evapotranspiration, and the ability of the soil to store moisture. Formulation and recognition of drought indicators are a complex problem due to the nature of agricultural drought and often the lack of detailed meteorological observations. The Advanced Very High Resolution Radiometer (AVHRR) on board the National Oceanic and Atmospheric Administration's (NOAA) satellites has been used successfully to monitor and analyze vegetation parameters over extensive regions. A vital characteristic of data secured from such satellites is its high temporal resolution. The daily radiometric coverage of the earth afforded by the NOAA satellites provides a capability to assess the dynamic characteristics of vegetation. The relatively low spatial resolution of the satellites (1.1 km) precludes detailed spatial analyses of vegetation, but is well suited to inventories requiring less detail, especially over extensive regions. During periods of drought conditions, satellite sensors are capable of discerning changes in physiognomic characteristics of vegetation seen on

Papers and Proceedings of the Applied Geography Conferences, (1986) 9: 45-55

158

Chapter Thirteen

the landscape, many of which can be evaluated through satellite-derived vegetation indices. When a rainfall event does occur during a drought period, the "effectiveness" of the precipitation relative to vegetation change is related to the soil, terrain, condition of the vegetation, and antecedent soil moisture conditions of the landscape. In addition the amount, duration, and rate of precipitation enter into the measure of rainfall "effectiveness".

Objectives This paper investigates the relationship between the Crop Moisture Index (CMI), a meteorologically based drought index, and vegetation characteristics assessed through remotely-sensed satellite data sampled during a growing season and situated in the state of Oklahoma, an environment which normally experiences conditions of moisture stress. Vegetation indices are applied to the AVHRR data for characterization of the landscape for the four 1980 time periods studied in this analysis. Statistical models are derived through multiple regression analysis to assess the spatial and temporal relationship between drought conditions evaluated through the Crop Moisture Index and remote sensing techniques. AVHRR data will be utilized in this study because of its applicability to global vegetation analysis. Moreover, the temporal and spatial resolution essential for evaluation of dynamic vegetation conditions over broad areas and on a systematic and repetitive basis can be evaluated with AVHRR data.

Study Area The state of Oklahoma covers over 180,000 square kilometers in area, is astride the forest/grassland ecotone, and is prone to substantial climatic inter-seasonal and intra-seasonal variations. Precipitation varies from a mean of 130 cm annually in the southeast to less than 40 cm in the northwest portion of the state. A period of maximum precipitation occurs in the spring with a second, lesser maximum occurring in the early fall. Precipitation in Oklahoma is strongly tied to the advection of moisture from the Gulf of Mexico. Temperatures vary less across the state than does precipitation. Mean annual temperatures range from 17.5°C at Idabel, in the extreme southeast corner of the state, to 12.0°C at Boise City, in the western part of the Oklahoma Panhandle. Average July temperatures range from 25°C in the Panhandle to 27.5°C in the southeastern quadrant of the state. January

Drought Assessment through Advanced Radiometer Satellite Data

159

average temperatures range from 0oC in the Panhandle to 6.5°C in the southeast. Maximum temperatures of 37.8°C or higher may be expected in Oklahoma from June to September.

Literature Review The visible and near-infrared radiation bands have been extensively used in vegetation studies (Richardson and Wigand, 1977). Visible and near-infrared reflectance from plant canopies can provide information about the condition of the vegetation. Healthy plant leaves exhibit low reflectance of near-infrared energy. Wavelengths of red light are strongly absorbed by chlorophylls and, therefore, the reflectance of red light is related to the amount of green vegetation present. As the amount of vegetation increases, more near-infrared energy is reflected away from the plant canopy, but at an exponentially decreasing rate as the leaf-area index increases (Brakke et al., 1981). Kanemasu (1974) found the longer the wavelength, the higher the reflectance observed, with the highest reflectance occurring in the near-infrared. Tucker (1979) found the reflectance of red light exhibited a non-linear inverse relationship between integrated spectral reflectance and green biomass, while the near-infrared wavelengths indicated a non-linear direct relationship. Frequently, information from the visible and near-infrared spectral regions is combined and manipulated in order to construct vegetation indices for determining estimates of vegetation conditions, such as density, vigor, biomass, greenness, and moisture stress. Curran (1980) indicated that indices are sensitive to the rate of plant growth, as well as the amount of growth. Such indices would also be sensitive to the changes in vegetation affected by moisture stress. Landsat MSS data secured in the visible and near-infrared wavelengths have been used by Kanemasu (1978) to estimate the Leaf Area Index (LAI) of wheat. The Perpendicular Vegetation Index (PVI), Green Vegetation Index (GVI), and the Transformed Vegetation Index (TVI) were favorably compared at an OS significance level by Kanemasu to sets of ground based measurements of LAI. Kauth and Thomas (1976) developed a multi-dimensional model of seasonal crop development using the four MSS channels as axes. They identified four vectors related to four environmental plant characteristics from which they defined four MSS band transformations. One of these transformations has subsequently been developed into an index of the amount of chlorophyll present. Thompson and Wehmanen (1979) used the Green Index Number (GIN) developed by Kauth and

160

Chapter Thirteen

Thomas to detect moisture stress in the Great Plains. Research by Norwine and Greegor (1983) and Brown and Bernier (1982) indicate that the results of vegetation analyses achieved through the use of the first two channels of AVHRR data are very similar to the results secured from the Landsat Multispectral Scanner (MSS) channels 2 and 4 (Bands 5 and 7). The spectral response of chlorophyll pigment in the visible and near infrared portions of the electromagnetic spectrum provides a good measure of density and vigor of green vegetation. Gray and McCrary (1981) have shown, however, that NOAA-6 AVHRR and Landsat MSS vegetation indices are in close agreement when compared over a primary crop growing region along the Brazil-Argentina border. Since early 1982, the Vegetation Index (VI) and the Normalized Vegetation Index (NVI) have been routinely calculated by the U.S. National Climatic Data Center. The NVI is recommended for global vegetation monitoring because it partially compensates for changing illumination conditions, surface slope, and viewing angle (US National Climate Data Center, 1983). Norwine and Greegor (1983) used the NVI to sample vegetation types along an east-west transect in Texas. They equated this index with surface greenness and indicated that AVHRR data could detect changes in the greenness of vegetation associated with changes in vegetation type. Barnett and Thompson (1983) related both MSS and AVHRR to wheat yields over the Great Plains. Their results showed a relationship between yield and both types of satellite data. They recommended that AVHRR data be used for monitoring large areas of agricultural crops. Townshend and Tucker (1984) have shown that in three contrasting regions, AVHRR data represented approximately 70 percent of the variation in Landsat MSS bands 5 and 7 and over 50 percent of the variation in the normalized difference vegetation index, which is the AVHRR NVI versus the MSS NVI. The results were obtained despite tremendous differences in the spatial resolution of MSS and AVHRR data. AVHRR data are useful in the assessment of vegetation conditions over extensive regions because of the satellite sensor's temporal, spatial, spectral, and radiometric resolution.

Crop Moisture Index The Crop Moisture Index (CMI) utilized in this study considers surface, as well as subsurface conditions when estimating a moisture availability index for agricultural crops. The index is designed to evaluate large geographic regions and not individual field conditions. The CMI is essentially a measure of evapotranspiration anomalies. The Crop Moisture

Drought Assessment through Advanced Radiometer Satellite Data

161

Index assesses the severity of agricultural drought by computing conditions at the end of any week. The index considers conditions at the start of the week and the evapotranspiration deficit or soil moisture recharge during the week (Palmer, 1968). Negative values indicate deficient evapotranspiration levels. The Crop Moisture Index calculates crop moisture conditions by examining the interrelationship between the deviations of precipitation levels from normal, soil moisture supplies, and evapotranspiration demands.

Satellite Regional Land Cover Assessment Four AVHRR Local Area Coverage data sets were acquired over Oklahoma. The nadir of all data sets is located over central Oklahoma. The dates of the four data sets are: 26 June 1980, 14 July 1980, 23 July 1980, and 23 August 1980. All data sets were geographically referenced to the Universal Transverse Mercator (UTM) coordinate system and aggregated to a cell size of 1609 m x 1609 m. All data sets are composed of the four spectral regions of the NOAA AVHRR satellite: channel 1 (0.58-0.68 micrometers), channel 2 (0.725-1.10 micrometers), channel 3 (3.55-3.93 micrometers), and channel 4 (10.50-11.50 micrometers). In addition to employing the four channels of spectral response values for use in statistical analysis, four indices were calculated. The indices employed are the Gray-McCrary Index (GMI = Ch 2-Ch 1), Transformed Vegetation Index (TVI = (Ch 2-Ch 1/ Ch 2+Ch 1)+0.5)), Normalized Difference Index (NDI = Ch 2-Ch 1/ Ch 2+Ch 1), and the Difference Vegetation Index (DVI = 0.26+2.73 (Ch 2-Ch 1)).

Methods Each of the 181 Oklahoma cooperative weather stations distributed throughout the state (Figure 13-1) were located in the Universal Transverse Mercator (UTM) coordinate system. Crop Moisture Index values for weeks inclusive of the four AVHRR data sets were acquired from the Oklahoma Climatological Survey and computed for each of the Oklahoma cooperative weather stations. Each of the four AVHRR data sets (26 June 1980, 14 July 1980, 23 July 1980, and 23 August 1980) also were geographically referenced to the UTM coordinate system. A computer program was written to locate all Oklahoma cooperative weather stations in the appropriate AVHRR cell (kilometer x kilometer), which were processed for the entire state of Oklahoma. AVHRR spectral responses in channel 1, channel 2, channel 3, and channel 4 were extracted

162

Chapter Thirteen T

from the fouur satellite daata sets and co orresponding CMI values also a were extracted foor further anaalysis. In addiition to the fo four spectral responses r sensed by thhe AVHRR, four vegetatio on indices, N Normalized Difference, Transformedd Vegetation, Gray-McCrarry, and Vegettation Differen nce, were calculated fo for each satelllite data set and a paired witth correspond ding Crop Moisture Inndex values based upon the location of each co ooperative weather stattion. Statisticaal modeling off CMI and rem motely sensed variables is limited too only four Oklahoma O clim matic divisionns: north centrral, south central, eastt central, and west central. Figure 13-2 shows the lo ocation of the climaticc divisions in Oklahoma in ncluding the ffour to be analyzed in this paper.

Figure 13-1. L Location of Okklahoma cooperaative weather sttations

Analy ysis The purp rpose of this research is to t identify reelationships between a meteorologiical-based droought measurre, the Crop Moisture In ndex, and NOAA sateellite spectral responses of landscape el ements senseed in four spectral regiions by the onn-board AVHRR sensor. Thhe overall objjective of the analysis is to identifyy meteorologiical-based meeasures sensitiive to the spatial and temporal flucctuations in drought d condiitions and to compare satellite speectral responses of vegettation conditiions to such h drought indices. Thhe Crop Moiisture Index, Drought Seeverity Index x, and a Hydrologic Accounting Model M have been used as m meteorologicaally-based measures off drought connditions. This paper only reeports on relaationships between the Crop Moisturre Index and satellite s data.

Drouught Assessmennt through Advaanced Radiometter Satellite Daata

163

Oklahoma clim matic divisions Figure 13-2. O

The Croop Moisture Index is caapable of quuantifying spaatial and temporal flluctuations inn moisture conditions c ovver extensivee regions (Palmer, 1965). Figures 13-3 through 13-6 show thhe areal distriibution of CMI valuess across the state of Oklahoma for tthe four stud dy weeks correspondinng to the proocessed AVHRR satellite ddata. The CM MI values indicate thatt crop moisturre conditions can range froom extreme wetness w to extreme droought. The fouur study week ks of this anallysis provide a sample of moisture conditions exxtending throu ugh the summ mer of 1980, a time of drought initiiation and spaatial diffusion. Figures 13-3 through 13-6 show th hat the CMI foor the first stu udy week, June 22-28, 1980, was generally po ositive in naature indicatin ng above normal cropp moisture coonditions. A month m later, July 13-19, 1980, 1 the drought of 1980 had beecome quite harsh with C CMI values indicating i severe to exxtreme droughht in the west and southwesst portions of the state. A week later, July 20-26, 1980, the CM MI values werre generally neegative in nature indiccating moderaate to extrem me drought coonditions thro oughout a significant pportion of Okllahoma. By August A 17-23, 1980 (the fou urth study week), a riddge of moderaate to extreme drought had developed thrrough the central portiion of the statte oriented in n a southwest to northeast direction; d the severity of drought coonditions alon ng this ridge ddecreased in a northeast direction. Figures 13-3 through t 13-6, collectively,, indicate thatt drought conditions w were becominng more severre with time aand that harsh her, more severe drougght conditionss were becomiing more spatiially widespreead.

164

Chapter Thirteen T

Crop moisture index i – June 22 2-28, 1980 Figure 13-3. C

Figure 13-4. C Crop moisture index i – July 13-19, 1980

Satellite spectral respponses for eacch of the fourr AVHRR waavelength channels and the four derrived vegetatiion indices haave been calcu ulated for all 181 metteorological stations in Ok klahoma for eeach of the fo our study weeks. The Crop Moisturre Index (dependent variabble) has been regressed against the satellite-derivved variables (independent variables) for each of the nine clim matic divisionns in Oklahom ma. The resullts of only fo our of the climatic divvisions or reggions are rep ported in thiss paper due to space limitations. The selected regions do prrovide an apprropriate spatiaal sample for assessinng the value of remote sensing for ddrought appraaisal. The

Drouught Assessmennt through Advaanced Radiometter Satellite Daata

165

dependent aand independeent variables calculated c for each study week w have been combined into onne variable pool p for regrression analy ysis. The combinationn of variabless into one regression pooll, regardless of week, facilitates aappraisal of remote r sensin ng spectral rresponses and d derived indices for aassessing drouught througho out the entire analysis perio od. Table 13-1 presentts 12 step moddels of the Crop Moisture iindex regresseed against remote senssing spectral responses and a derived-in indices for th he North Central, Souuth Central, East E Central, and a West Cenntral climatic divisions of Oklahom ma, respectivelyy.

Figure 13-5. C Crop moisture index i – July 20 0-26, 1980

Remote sensing varriables calcullated for thee regression analysis include: chaannel 1, 2, 3, 4; 4 TVI, GVI, GMI, NDI; 22nd, 3rd, 4th powers p of the eight caalculated sateellite values; and cross-prooducts of alll variable combinationns. All the vaariables were entered into a regression pool for calculation oof the maximuum R2 regresssion model innvolving 12 stteps or 12 iterations. T The remote seensing regresssion model ffor the North h Central, South Centrral, East Cenntral, and Weest Central reegions explain ned 89.9, 82.6, 87.8, aand 88.0 perceent of the variation observe d in the Crop Moisture index, respeectively. Each of the four models m utilizess various com mbinations of the remotte sensing speectral responsses and the deerived indicess. Time is consistently an importantt variable in the t regressionns. While the selection of 12 steps or iterations for f the calculaation of the reegression anaalysis was somewhat aarbitrary (R2 continued c to increase i and Sum of Squaares Error continued too decrease), substantially s fewer f steps, annd hence few wer model variables aree required to explain e a "satisfactory" am mount of CMI variation. v All regressioons presented in Table 13-1 1 are significan ant at the 0.05 level.

Chapter Thirteen T

166

Crop moisture index i – Augustt 17-23, 1980 Figure 13-6. C Twelve step muultiple regression model of C CMI and remotely-sensed Table 13-1. T variables for O Oklahoma regioons North Centtral Beta IInd. Value V Var.

Soutth Central Beta Ind. Valuee Var.

East Centraal Beta In nd. Value Vaar.

West Central Beta Ind. Value Var.

+597.50 -75.176 +2.627 -0.001 +0.050 -0.001 +0.003 -4.375 +0.577

Inter. WK WK2 WK4 C23 C24 C43 G GM12 G GM13

+596.668 -74.8445 +8.180 -1.724 +3.2005 -0.0446 +0.453 -0.0002 +0.0002

Inter. WK C1 C4 WK2 WK3 C12 C24 DVI3

+832.95 -99.743 +3.390 -0.001 -0.002 +1.154 -0.357 +0.019 +1.039

Innter. WK W WK2 W WK4 C24 M12 GM WK KC1 WK KC4 C C2C2

+465.22 -64.492 -3.178 +2.269 -0.001 -0.002 +0.170 +0.028 -0.499

-0.333 -0.050 -0.482 +4.410 R2 .899

G GM14 W WKC4 C CIDVI TV VIDVI

-0.327 +0.118 -0.179 -2.912 R2 .826

WKC1 WKC4 C1C4 GMIND

-0.360 -13.463 -0.304 +0.145 R2 .878

C C2C4 C22ND C33ND C4D DVI

+4.215 +3.102 -0.069 +2.669 R2 .880

Inter. WK GM1 WK2 WK4 C14 C22 C42 WKGM I WKND C2DVI C3C4 C3TVI

Note: R2 signnificant at 0.05 level; l Crop moiisture index vs. satellite spectrral response for aall four weeks combined. c

Conclusions The Croop Moisture Inndex is a meaasure of availlable soil moisture for plant utilizaation. The CM MI is used to t provide a spatial and temporal assessment oof moisture coonditions overr broad areas. The calculatiion of the

Drought Assessment through Advanced Radiometer Satellite Data

167

index, however, is dependent upon the availability of a significant amount of meteorological information. In geographic areas of the world where meteorological data are not sufficiently available to calculate the Crop Moisture index, either from a spatial or temporal perspective, NOAA AVHRR satellite data can be used as a substitute for the CMI. This research has shown that from nearly 83 to 90 percent of the variation in the CMI can be explained by satellite spectral responses and indices manipulation of the spectral channels. While additional research needs to be completed regarding the analysis of remotely-sensed data for estimation of landscape conditions measured by the Crop Moisture Index, these reported preliminary results are promising. They suggest that drought conditions can be assessed over broad areas and on a repetitive temporal basis without access to meteorological variables.

References Barnett, T. L. and D.R. Thompson. 1983. Larger-Area Relation of Landsat MSS and NOAA-6 AVHRR Spectral Data to Wheat Yields. Remote Sensing of Environment 13: 277-290. Brakke, T. W., E.E. Kanemasu, and J.L. Steiner. 1981. Microwave Radar Response to Canopy Moisture, Leaf-Area Index, and Dry Weight of Wheat, Corn and Sorghum. Remote Sensing of Environment 11:207220. Brown, R. J. and M. Bernier. 1982. NOAA AVHRR Crop Condition Monitoring. Canadian Journal of Remote Sensing S: 107-109. Curran, P. 1980. Multispectral Remote Sensing of Vegetation Amount. Progress in Physical Geography 4: 315-341. Gray, T. I. and D.C. McCrary. 1981. Meteorological Satellite Data: A Tool to Describe the Health of the World's Agriculture, AgRISTARS Report, WE-NI-04042, Johnson Space Center, Houston, Texas, P. 1-7. Kanemasu, E. T. 1974. Seasonal Canopy Reflectance Patterns of Wheat, Sorghum, and Soybean. Remote Sensing of Environment 3:43-47. —. 1978. Estimating Winter Wheat Yield from Crop Growth Predicted by Landsat. NASA Final Report, Kansas State University, Evapotranspiration Laboratory, NASA-14899-3S. Kauth, R. J. and G.S. Thomas. 1976. A Geographic Description of the Spectral-temporal Development of Agricultural Crops as seen by Landsat. Machine Processing of Remotely Sensed Data, Laboratory for Applied Remote Sensing, West Lafayette, Indiana, Catalog 76, Ch 1103-1-MPRSD.

168

Chapter Thirteen

Norwine, J. and D.H. Greegor. 1983. Vegetation Classification Based on Advanced Very High Resolution Radiometer (AVHRR) Satellite Imagery. Remote Sensing of Environment 13: 69-87. Palmer, W. C. 1965. Meteorologic Drought. U.S. Weather Bureau, Research Paper No. 45. —. 1968. Keeping Track of Crop Moisture Conditions Nationwide: The New Crop Moisture Index. Weatherwise August: 156-161. Richardson, A. J. and C.L. Weigand. 1977. Distinguishing Vegetation from Soil Background Information. Photogrammetric Engineering and Remote Sensing 43: 1541-1552. Thompson, D. R. and O.A. Wehmanen. 1979. Using Landsat Digital Data to Detect Moisture Stress. Photogrammetric Engineering and Remote Sensing 45: 201-207. Townshend, J. R. and C.J. Tucker. 1984. Objective Assessment of Advanced Very High Resolution Radiometer Data for Land Cover Mapping. International Journal of Remote Sensing 5(2): 497-504. Tucker, C. J. 1979. Red and Photographic Intra-red Linear Combinations for Monitoring Vegetation. Remote Sensing of Environment 8: 127150. U.S. National Climate Data Center. 1983. Global Vegetation Index Users' Guide, National Environmental Satellite Data and Information Service, Washington, D.C. pp. 1-6.

CHAPTER FOURTEEN GRAPHIC NARRATIVES FOR EMERGENCY MAPPING MARK MONMONIER

Introduction Emergencies accompanying natural or technological disasters are among the most stressful and error-prone types of map use. The size and complexity of the problem, the dire consequences of mistakes, and pressures to act promptly obviate the more thoughtful, cautious and vigilant use of cartographic materials possible in facilities management and transportation planning. Designing maps for disaster response can be especially difficult because map users can vary widely not only in their experience in managing and carrying out evacuations and relief efforts but also in their knowledge of the area and the data as well as their skills in map reading and interpretation (Dymon and Winter, 1991). Without specific efforts to tailor presentations to both the stressful context of a crisis and the variable abilities of users, increased use of geographic information systems and electronic graphics is unlikely to meet the cartographic needs of emergency management agencies. Graphic narratives offer a promising strategy for improving cartographic communication in a crisis. A multimedia solution that recognizes an inherent need for more than one map, graph, diagram, or text screen, the graphic narrative promotes interpretation and understanding through an ordered, coherent sequence of displays. This paper discusses the graphic script and other concepts of narrative graphics, reports on the Atlas Touring Project and the development and evaluation of two prototype graphic scripts, and explores the relevance of the project's findings to emergency management.

Proceedings of the Applied Geography Conferences (1993)16: 64-70

170

Chapter Fourteen

Principles of Graphic Narration Graphic narratives are an extension of the concept of a script developed by artificial intelligence experts Roger Schank and Robert Ableson and (1977) who observed standard sequences of questions, answers, comments, actions, and reactions in such task-oriented conversations as ordering a meal in a restaurant. Schank (1991) defines the script as "a kind of play that we can engage in where our lines are prepared by a kind of general societal agreement, where we anticipate the lines of our partner in their likely place in the play and react accordingly." It is necessary, of course, that scripts vary with circumstances as well as with the performers' backgrounds. Programming a computer to converse intelligently with humans thus depends upon a comprehensive library of scripts and a mechanism for identifying and adapting the right one. Although extending the notion of scripts to cartographic communication requires replacing much of the natural-language code of human speech with graphic symbols, a graphic narrative also depends for its effectiveness on the system's ability to anticipate the user's information needs by presenting relevant displays in an appropriately intelligible sequence.

Graphic Scripts and Graphic Phrases The Atlas Touring Project is attempting to integrate Schank's (1991) concept of a script with the notion of a twenty-first century electronic atlas able to offer customized tours of a geographic database. A guided tour can be customized in a variety of ways, for example, to provide an overview of the data, to address concerns about data quality, to explore a particular explanatory relationship, or to identify and describe geographic patterns the viewer might find interesting. In principle, a tour can respond not only to the user's inherent need for multiple graphic views but also to preferences for specific measurements and designs as well as concerns about particular regions and patterns. A key concept is the graphic script, defined as a purposefully ordered sequence of maps, text screens, diagrams, statistical graphs, and other images selected and integrated to address a specific problem or query (Monmonier, 1990). A shorter, more focused and fundamental sequence is called a graphic phrase. As an example, a graphic phrase useful in the early stages of a flooding emergency might generate a series of maps identifying roads likely to remain passable as the river rises one foot, two feet, and so on above flood stage. Roads that remain passable longer

Graphic Narratives for Emergency Mapping

171

provide relatively reliable evacuation routes, and areas severed early from the overall road network would require early evacuation. This graphic phrase might be an appropriate component in a more comprehensive graphic script useful in identifying and describing neighborhoods and isolated residents at risk and in developing an evacuation plan. Work to date has focused on the design of arguably logical sequences of automatically generated dynamic maps and statistical graphs. Two prototype scripts developed in 1991 proved useful in identifying a number of principles for composing graphic scripts (Monmonier, 1992). The first prototype, called the "Correlation Script," examines the relationship between a dependent and an independent variable in a three-act play that progresses from a detailed description of each variable individually in the first act, to an exploration of their covariation in the second act, to the introduction of a regression model and an examination of residuals in the third act. Fill patterns for bar graphs, choropleth maps, and other displays reinforce the identity of each variable with a distinctive, highly contrasting signature hue: red for the dependent variable, blue for the independent variable, and magenta for the residuals from linear regression. Each act is subdivided into scenes, and each scene presents a variety of closely-related text, maps, and graphs. To minimize the distraction of needlessly complex transitions, the script adds one new element at a time, and its various text windows, maps, and statistical diagrams have standard screen locations. As demonstrated by research on reader expectations, linking new material to old information promotes coherence and clarity (Gopen and Swan, 1990). To encourage understanding, a text screen defines each variable before its introductory scene in the first act, and a small text window introduces each dynamic sequence by announcing its point, or purpose. At various places in the script, graphic phrases orchestrate an analysis of geographic patterns by guiding the viewer through a systematic examination of a linked map and graph. Called "canvass" sequences, these phrases follow a narrative thread based on places grouped by data category or region and highlighted in turn with blinking symbols or a temporarily different color. In the Correlation Script, a typical canvass-by-category sequence progresses from the highest to the lowest data category, whereas a canvass-by-region sequence moves roughly from east to west. The second prototype script, the Historical Script, includes canvass sequences with a third kind of narrative thread: time. Designed to explore spatial-temporal data using complementary absolute and relative measurements, this second Atlas Touring prototype demonstrates the potential of dynamic graphics for exploring the stability of rates of change by varying the length, starting point, and direction of movement of time

172

Chapter Fourteen

periods. All three narrative threads -- magnitude, region, and time -- are relevant to emergency management. Natural processes, evacuation plans, and remediation processes have their own narrative threads, mostly based on time.

Focus Group Evaluation Focus-group interviews conducted in 1992 provided information useful in refining principles for composing graphic scripts as well as in making the prototypes more effective. An experienced focus-group facilitator moderated the sessions -- however thick-skinned, no designer should expect fully frank comments from viewers coaxed or dared to criticize his work. The facilitator collaborated with the principal investigator in developing a focus-group protocol that integrated discussion issues and "probe" questions with two software demonstrations. The first part of the group interview consisted of an approximately 11-minute demonstration of the Correlation Script, narrated by the facilitator, followed by a moderated discussion of the script's “informativeness” and coherence as well as its good points and bad points. In the second portion, designed to solicit more general opinions about customization and user interaction, the facilitator demonstrated two dynamic spatial-temporal maps from the Historical Script and an interactive version of selected graphic phrases from the Correlation Script's first act. To assure thoroughness and uniformity, the facilitator and the principal investigator rehearsed both the protocol and the facilitator's oral presentation of the demonstrations. Each session lasted about one hour. To make the results meaningful, the principal investigator recruited 26 participants with experience in map design, cartographic analysis, or human-computer interaction. The facilitator conducted group interviews at four sites: Syracuse University's School of Information Studies (IST), the U.S. Geological Survey's National Mapping Division (in Reston, Virginia), the National Geographic Society's Cartographic Division (in Washington, D.C.), and the GIS applications unit at IBM Corporation's research and development center in Kingston, New York. The four groups ranged in size from five to ten. The IST group consisted of faculty and doctoral students; participants in the other three groups were full-time employees, most with more than five years of professional experience. For each session, the facilitator and the principal investigator studied the audio recordings independently, took notes, and collaborated in preparing two lists: one that identified consensus opinions and illustrated participants' views with revealing quotes, and a second that collected suggestions

Graphic Narratives for Emergency Mapping

173

potentially useful in the next design cycle. Unlike the primary list, this second list included the principal investigator's and the facilitator's own ideas stimulated by the recorded discussion. Like other qualitative data-gathering methods, the group interview can be highly subjective as well as non-rigorous. Its usefulness in developing products or concepts reflects the synergetic setting of a small group of appropriately selected participants encouraged to think about specific issues and to react to and amplify each other's insights and opinions (Krueger, 1987). Although investigators can minimize bias in the wording of discussion questions and by employing a comparatively neutral facilitator, interpretation of the results is innately imprecise. Systematic content analysis used in marketing research to promote consistency is unwarranted with a small number of participants surveyed in a few groups by a single facilitator. Confirmatory statistics are irrelevant as well because group interviews are intended to reveal a broad range of attitudes and opinions, not to estimate a population's characteristics from measurements collected for a small, desirably random sample. Nonetheless, consistency in the opinions registered across the four groups provided a rational basis for confidence in conclusions drawn from the taped discussions. The focus groups were revealing. Most participants found the graphic scripts engaging, correctly detected salient patterns in the data, and appreciated the underlying graphic logic. They understood the signature hues but objected to the visually harsh blinking symbols used to highlight portions of some maps and statistical graphs. Participants also recognized the need for a voice-over or oral narrative similar to the facilitator's running commentary. Although screens with text might appropriately precede a dynamic sequence of graphics, viewers cannot easily watch a moving display and simultaneously read titles, labels, and other accompanying text. Synthesized or recorded speech will become an important component of visualization support systems. Of greater significance, participants unanimously wanted to control the pace of the presentation--once started, the prototype scripts run without interruption to their final scene, but viewers desperately wanted to slow down, stop, back up, and repeat selected portions. Although most agreed that a non-interactive, "closed" graphic script could be useful as a presentation medium or as a training tool to introduce the individual user to a new dataset, region, or analytical technique, participants felt strongly that efficient learning and understanding depend upon direct interaction with the data. In short, once introduced to a tool or data set by a training script, the experienced user demands full control. Participants also

174

Chapter Fourteen

recognized that an intelligent visualization support system able to anticipate an analyst's or a manager's needs by providing or suggesting the next most appropriate display could be valuable to both neophytes and experienced users. Even though the context of the Atlas Touring prototypes and the group interviews was hypothesis generation and data analysis, these findings are relevant to emergency management.

Openness, Interaction, and Narrative Menus Graphic narratives need not be as closed as the prototype Correlation Script. A rudimentary yet effective way to make Atlas Touring scripts more "open" would be to add pause points after all graphic phrases, at appropriate intermediate places within some graphic phrases, and at all text screens. At these points in a graphic narrative, the system would pause and await the viewer's command either to repeat the preceding view or sequence, or to continue immediately to the next. A second, more open modification is to provide interaction points, at which the viewer may manipulate the current graphic phrase. Of course, the system might also provide a "player-piano mode" to let the user run a graphic script as a closed presentation. A more flexibly open strategy would provide access to the script's text blocks and graphics phrases through one or more narratively organized menus. At the lowest level of menu control, all graphic phrases and text screens might be linked to a set of pull-down menus. For the Correlation Script and other narratives organized like a play, a menu bar at the top of the screen might present the three acts as a group of adjacent pull-down menus, arranged in sequence from left to right. Scenes or graphic phrases ordered from top to bottom within each menu would help the viewer recreate the script in its intended order as well as jump back freely to a particular segment. In a different menu configuration, a command window could list in narrative order a series of short descriptors representing the script's text blocks and graphics phrases. In its fully automated, player piano mode, the system would progress down this list, highlighting individual descriptors in the command window while presenting the corresponding text block or graphic sequence in the view window. A. two-dimensional version is the toolbox, a grid of cells containing icons or acronyms organized sequentially by row or column. If the narrative has branch points at which the viewer must choose among two or more optional sequences, the system might present the menu as a flow diagram useful as well for highlighting progress. To offset the awkwardness of making frequent

Graphic Narratives for Emergency Mapping

175

manual choices from pull-down menus, command windows, graphic toolboxes or flow menus, a "next" key such as the keyboard's downward arrow might allow the user to conveniently advance to the narrative's next logical step. In this case, the menu only need appear when the user reaches a branch point or decides to consider another path. Whatever its degree of openness, a graphic narrative is a statement of a preferred or canonical sequence of views or techniques. The statement might be a simple sequence or a complex web, and its basis might be logic, common practice, administrative mandate, or a knowledgeable author's pragmatic instincts. Because narrative graphics is a new concept, cartographers and information experts have little if any experience in designing or detecting canonical graphic sequences -- indeed, the classic cartographic problem is the single optimum map, not the pragmatical1y effective set of complementary maps and graphs. Yet if Schank and Ableson (1977) are right, attempts to identify and formalize graphic scripts might reveal surprisingly little variety among demonstrably successful experts, credentialed or otherwise. In emergency management, of course, these canonical sequences can vary considerably by type of hazard or disaster. For example, maps and other graphics useful in dealing with a severe winter storm should differ significantly, if not dramatically, from emergency graphics for coping with nuclear accident -- and determining exactly how different might be useful in identifying an amalgamated script appropriate, however unlikely, for a nuclear emergency during a heavy snowstorm.

User Profiles, Disaster Profiles, and Area Profiles The success of graphic narratives in emergency management will depend upon user profiles, that is, files describing each user's interests, experience, training, and preferences for specific types of presentations, details, and symbols. This information would allow a visualization support system to customize graphic narratives for individual users---delivering background information important to those unfamiliar with the area, the data, and the problem, but not needlessly antagonizing those users more fully aware of both the situation and an appropriate response. Especially relevant is the user's experience in disaster management and cartographic competence. Developed during a query session when the viewer first accesses the system, the user profile could be updated at the end of the session, saved for later use, and updated as needed thereafter. User profiles could also be evaluated and modified during periodic training sessions. In addition, residential information in the user profile could help the system

176

Chapter Fourteen

promote morale during an emergency by assuring workers of the safety of their own families--and advise officials about workers likely to perceive added stress as a result of family members still at risk. Other metadata useful in customizing a graphic narrative include the disaster profile, the data profile, and the area profile. A catalog of disaster profiles could not only guide a visualization support system in adapting a master script to a wide variety of emergencies but also help program managers evaluate the adequacy of their databases, their equipment, and their personnel resources. Similarly, a data profile would help the system brief the user on the assets and limitations of the data, and thereby discourage unwarranted inferences. And an area profile would inform the user about significant landmarks and reference features, local slang, nonofficial place names, and other unique occurrences with which state and federal officials might not be aware. In addition to supporting training sessions and actual emergencies, these profiles are valuable planning tools, useful for informing and guiding periodic evaluations of geographic information, personnel, and equipment.

Concluding Remarks However unattainable and unrealistically futuristic graphic scripts might seem at present, graphic narratives are fundamental elements in the inevitable (if not imminent) "new cartography" of dynamic maps and visualization support systems. Implementation of graphic narratives, user profiles, and related concepts will be especially valuable in emergency management to assist users ignorant of the impacted area, the data, or the hazard, or with limited skills in map reading and interpretation. In addition, by promoting a more vigilant and conscientious use of maps, graphic narratives can offset the effects of stress during a crisis. Moreover, graphic narratives should prove useful in evaluating information resources, in developing evacuation and response plans, and in training local officials to deal with disasters. There is little formal knowledge of canonical sequences of maps and map-reading tasks upon which graphic narratives should be based. Yet even though information systems are unlikely to implement graphic scripts for another decade, it is not too early for emergency management agencies to address issues of cartographic complementarity and informative sequencing. Assessment of information needs, particularly the front-line user's need for geographic information, can identify gaps in existing data that diminish coherence and understanding. Because they represent serious defects in traditional, map-based information systems as well as

Graphic Narratives for Emergency Mapping

177

impediments to the adoption of graphic narratives, such gaps warrant immediate attention.

References Dymon, U.J., and N.L. Winter.1991. Emergency Mapping in Grassroots America: A Derailment Evacuation Case Study. Geoforum 22: 377389. Gopen, G.D. and J.D. Swan. 1990. The Science of Scientific Writing. American Scientist 78: 550-558. Krueger, R.A. 1987. Focus Groups: A Practical Guide for Applied Research. Beverly Hills, CA.: Sage. Monmonier, M. 1990. Atlas Touring: Concepts and Development Strategies for a Geographic Visualization Support System. CASE Center Technical Report No. 9011. Syracuse, N.Y.: New York State Center for Advanced Technology in Computer Applications and Software Engineering, Syracuse University. —. 1992. Authoring Graphic Script: Experiences and Principles. Cartography and Geographic Information Systems 19: 247-260. Schank, R. C. 1991. The Connoisseur's Guide to the Mind. New York: Summit Books. Schank, R. and R. Ableson. 1977. Scripts. Plans. Goals, and Understanding. Hillsdale, N.J.: Erlbaum.

CHAPTER FIFTEEN THE APPLICATION OF REMOTE SENSING AND GIS FOR MAPPING FLOOD-PRONE AGRICULTURAL LAND RANDALL S. PEARSON AND STEPHEN A. KAY

Introduction The catastrophic nature of flooding was emphasized in the spring of 1993 as the floodwaters of the Mississippi, the Missouri, and the Illinois rivers breached numerous levees flooding millions of acres of highly productive farmland, scores of small towns, and literally thousands of homes and farmsteads. Although floods of this magnitude are undeniably devastating to all involved, floods of a lesser magnitude can have a positive impact on the environment. Seasonal floodwaters deposit new sediment on farm fields, recharge groundwater supplies, and often generate temporary reservoirs for migrating water fowl. For example, Tiner (1984) reported that approximately 80 percent of the three million mallards and essentially 100 percent of the four million wood ducks of the Mississippi Flyway use the (seasonally flooded) bottomland hardwoods and agricultural land of the Lower Mississippi River Valley as overwintering areas (Frayer et al., 1983). In 1985, Congress recognized frequently flooded agricultural land as an important component to wetland ecosystems with the wetland conservation provisions of the Food Security Act (termed "Swampbuster"). Although mapping of these frequently flooded agricultural lands has (for the most part) been completed, recent activity within agriculture has rekindled efforts to understand and monitor the extent of these seasonally flooded systems. The following discussion details methods developed for

Papers and Proceedings of the Applied Geography Conferences (1996) 19: 265273

Remote Sensing and GIS for Mapping Flood Prone Agricultural Land

179

mapping the extent of these frequent flooding events. Additionally, this study raises questions on the importance the flooding limits could have on agriculturally related ecological risk assessments. The flooding events being mapped in this study conform to the Farmed Wetland category (agriculturally flooded land) defined in the 1985 Food Security Act and carried out by the USDA Soil Conservation Service (SCS) in each state. Specifically, Farmed Wetlands are defined as wetland areas that were used to produce agricultural commodities before December 23, 1985, but have not been altered by the farmer so that frequent flooding of a moderate duration (50 percent frequency of 15 consecutive days during the growing season) still occurs. Although the definition is quite specific, by using flood recurrence intervals calculated from stream gauge data, actual dates can be identified that corresponded to seasonal flooding events. Once a list of dates is developed, satellite imagery (acquired from archives) can be procured and digitally processed to map agriculturally active land that meets the Farmed Wetland criteria (Huffstatler, 1989; Jones, 1989; US Department of Agriculture, 1989).

Study Area The study area selection was based primarily on the availability of historical stream gauge data and historical satellite coverage. Major consideration was also given to those areas that, according to SCS officials, had a high probability of containing flooded agricultural lands. An area satisfying the above criteria is the southwestern portion of the Yalobusha watershed located in west-central Mississippi (Figure 15-1). The study area covers 83 square kilometers (51.93 square miles) and encompasses a portion of the Money and Greenwood 7.5 minute topographic quadrangles (US Geological Survey 1982a,b). Four major basins make up the Lower Mississippi Valley: the St. Francis Basin, the Yazoo Basin, the Tensas Basin, and the Atchafalaya Basin. The Yalobusha watershed is located in the east-central portion of the Yazoo Basin and abuts the Quaternary aged loess bluff along the eastern portion of the Mississippi Embayment. The Yalobusha River flows westward from the highly dissected uplands of the coastal plains into the Yazoo Basin where it turns abruptly to the south and meanders for about twenty miles. At Greenwood, Mississippi the Tallahatchie River and the Yalobusha River join to form the Yazoo River. The Yazoo River flows southward from Greenwood along the eastern edge of the Yazoo Basin until it joins with the Mississippi River near Vicksburg, Mississippi (Hunt, 1973).

180

Chapter Fifteen F

The topoography of thee study area is flat to gentlyy sloping. Th he overall slope of the study area is from north to south with a gradient of ab bout 11.4 cm per killometer. Thee greatest lo ocal relief inn the study area is predominanttly the resultt of relatively y steep slope s associated with the natural levees that com mmonly occur along strreams, sloug ghs, and abandoned stream channnels (Kolb ett al., 1967). Natural levees grade quickly to a nearly flat terrain with su ubtle ridges aand swales. Th his ridge and swale toopography (thhat occurs as parallel, p archeed ridges com mposed of predominanttly coarse teextured sedim ment) develooped from point bar deposits forrmed during the meandering process oof the local drainage system. Thee micro relieef within thee ridge and swale topography is generally lesss than 1.5 meeters (Saucier,, 1974).

Figure 15-1. S Study area

Remote Sensing and GIS for Mapping Flood Prone Agricultural Land

181

Flooding in the study area has been a problem since human settlement of the delta. According to the USDA Soil Conservation Service, between 1900 and 1950, there was a major flood in the Yazoo basin on the average of every two years (US Department of Agriculture, 1957). Though many flood control programs have been introduced (i.e., dredging, channelization, channel cutoffs, and flood control reservoirs), the study area continues to suffer from frequent flooding. Most flooding is the result of runoff from the steep-sloped loess bluff. Flooding also results from reservoir releases from the three major lakes located on the bluff. Sands Lake and Enid Lake flow into the Tallahatchie River, and Grenada Lake flows into the Yalobusha River. Additionally, Arkabulta Lake in the northern part of Mississippi feeds the Coldwater River which in turn flows south into the Tallahatchie River. Approximately one-sixth of the total area of Mississippi is drained by the Yalobusha and the Tallahatchie Rivers. The fact that the rivers merge just south of the study area enhances the flooding potential (Harrison, 1961; Jones, 1989).

Methods The primary objective of this study was to use archived satellite imagery for mapping the aerial extent of seasonal flooding events along the Yalobusha River. Although the satellite imagery was the primary mapping tool in this study, it was the detailed hydrologic analysis of stream gauge data that helped define which satellite images to process. For this study, the flood frequency was set at the two year flood. Coupled with this frequency, a flooding duration of 15 consecutive days was also required. Simply put, this study was attempting to use satellite imagery to map the agricultural areas that flooded on the order of every other year with a flooding duration of at least fifteen consecutive days.

Hydrological Analysis In order to identify precise hydrologic events meeting the Farmed Wetland criteria, nineteen years of stream gauge data were obtained for two different gauging stations. Only nineteen years of data were used as a result of changing hydrology during the early seventies due to stream channelization projects sponsored by the U.S. Army Corps of Engineers. The gauging stations were located on the Yalobusha River at Whaley, Mississippi, and on the Yazoo River at Greenwood, Mississippi. Each station was processed independently. For each year, the highest readings for fifteen consecutive days were located in the stream gauge data. From

Chapter Fifteen

182

this highest fifteen consecutive days, the minimum stage was recorded. This minimum value was used because it is the level the stream was at or above for fifteen consecutive days. This process was repeated for all nineteen years. These minimum stage values were then ranked highest to lowest, one to nineteen, respectively (CERL, 1989). Once ranked, the flood recurrence interval (Table 15-1) was calculated using equation [1] from Dunne and Leopold (1978): RI = (N+1) M

[1]

where, RI = Recurrence Interval N = Number of years of data M = Rank from highest to lowest Table 15-1. Flood recurrence intervals

1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980 1979 1978 1977 1976 1975 1974 1973

Stage (ft) 25.6 24.2 23.3 16.4 22.2 12.1 21.2 24.1 24.5 14.7 13.3 24.8 22.6 19.5 15.7 19.1 23.6 24.5 25.6

Yalobusha River Rank RI (years) 1.5 13.33 6.0 3.33 9.0 2.22 15.0 1.33 11.0 1.80 19.0 1.05 12.0 1.66 7.0 2.85 4.5 4.44 17.0 1.17 18.0 1.11 3.0 6.66 10.0 2.00 13.0 1.50 16.0 1.25 14.0 1.42 8.0 2.50 4.5 4.44 1.5 13.33

Stage (ft) 34.7 32.5 32.1 24.7 28.6 17.7 27.6 31.6 34.6 22.9 21.l 34.7 31.4 28.8 24.2 27.2 32.7 34.3 36.9

Yazoo River Rank 2.5 7.0 8.0 15.0 12.0 19.0 13.0 9.0 4.0 17.0 18.0 2.5 10.0 11.0 16.0 14.0 6.0 5.0 1.0

RI (years) 8.00 2.85 2.50 1.33 1.66 1.05 1.50 2.22 5.00 1.17 1.11 8.00 2.00 1.80 1.25 1.42 3.33 4.00 20.00

In order to calculate the flood exceedence probability and its upper and lower confidence bounds, the data were converted from stage (feet above local datum) to discharge (cubic feet per second). The conversion

Remote Sensing and GIS for Mapping Flood Prone Agricultural Land

183

graphs (rating curves) were obtained from the United States Army Corps of Engineers (COE) located at Vicksburg, Mississippi. The conversion from stage to discharge is given in Table 15-2. Using Gumble Type I graph paper, the annual fifteen consecutive day discharge (cubic feet per second) was plotted against recurrence interval (years) for both gauging stations. The best fit line was then drawn through the data points. Using the Confidence Band Coefficient Table adopted from Dunne and Leopold (1978), 90 percent confidence bands were calculated for each gauging station. Confidence bands are important because of the possibility the true discharge for a given recurrence interval may lie slightly above or below the best fit line. The graph of the fifteen consecutive day flood recurrence intervals (and 90 percent confidence bands) for the Yalobusha River at Whaley and the Yazoo River at Greenwood are given in Figures 15-2 and 15-3, respectively. Table 15-2. Stage to discharge conversion Year 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980 1979 1978 1977 1976 1975 1974 1973

Yalobusha River Stage (ft) CFS 25.6 12800 24.2 10000 23.3 8200 16.4 2200 22.2 6800 12.1 500 21.2 5500 24.1 9700 24.5 10500 14.7 1500 13.3 900 24.8 11000 22.6 7100 19.5 4200 15.7 1900 19.1 3900 23.6 8700 24.5 10300 25.6 12800

Yazoo River Stage (ft) CFS 34.7 29000 32.5 26500 32.1 25500 24.7 15800 28.6 20800 17.7 8300 27.6 19500 31.6 25000 34.6 28900 22.9 13800 21.1 11900 34.7 29000 31.4 24900 28.8 21000 24.2 15500 27.2 19000 32.7 26300 34.3 28800 36.9 33500

By using the fifteen day flood recurrence interval graphs (with the 90 percent confidence bands), one can locate the high and low discharge values that correlate with the two year recurrence interval (50 percent occurrence). The maximum discharge observed for the Yalobusha River

184

Chapter Fifteen F

at Whaley w was approxim mately 8,600 cubic feet per second (stagee of 23.9 feet) while tthe minimum discharge vaalue was approoximately 5,4 400 cubic feet per secoond (stage of 21.0 feet). Th he maximum ddischarge obseerved for the Yazoo R River at Greennwood was 26,750 cubic feeet per second (stage of 33.0 feet) w while the minim mum discharg ge value was approximately 19,000 cubic feet pper second (sstage of 27.0 feet). In ordder to accurattely map Farmed Wettlands, an imaage must havee been acquirred when the observed o discharge ffor both gauuging stationss was withinn the maxim mum and minimum diischarge outliined above. Table T 15-3 ideentifies one sh hort time period durinng 1989 in which w both riv vers were at thhe optimum stage for mapping thee two year floood with a fllooding duratiion of 15 con nsecutive days. Fortunnately, a Lanndsat Thematiic Mapper im mage was acq quired on March 10, 1989 during thhis optimum window w (Tablee 15-4).

d recurrence innterval for the Yalobusha Y Figure 15-2. Fifteen conseccutive day flood River at Whaaley, Mississippi

Remotte Sensing and GIS G for Mappin ng Flood Prone Agricultural Land L

185

Table 15-3. R River stages from m March 8, 198 89 through Marrch 10, 1989

Date 3/08/89 3/09/89 3110/89

Yaalobusha Riverr Sttage LB90% 233.6' 21.0' 233.9' 21.0' 244.0' 21.0'

UB 23.9' 23.9' 23.9'

Date 3/08/89 3/09/89 3/10/89

Y Yazoo River Staage LB90% % 32..6' 27.0' 32..5' 27.0' 32..5' 27.0'

UB 33.0' 33.0' 33.0'

Table 15-4. Im mages selected for this study Image # 1 2 3

Date 05/08/87 09/08/89 03/10/89

Path/Row w 23/37 23/37 23/37

Scen ne ID Y51163166010XO Y51163166010XO Y51835166065XO

% Cloud 00% 00% 10%

Fifteen consecuutive day flood recurrence inteerval for the Yaazoo River Figure 15-3. F at Greenwoodd, Mississippi

Althoughh an image acquired durin ng a flooding event was paaramount for this studdy, it was equaally importantt to have imaggery that was acquired

186

Chapter Fifteen

during a different part of the season to identify the land cover types inundated by the flood waters. Table 15-4 also lists other dates of the Landsat Thematic Mapper data used in this study.

Processing of Satellite Imagery Earth-Resources Laboratory Application Software (ELAS), version 8.1, was used in this study for processing the digital satellite imagery into land cover types. The ELAS modules NVAR and TMTR were used in an unsupervised approach for statistics generation. NVAR and TMTR use a semi-intelligent 3 by 3 matrix to search through the digital data attempting to identify homogenous training sites (NASA, 1989). The ELAS statistics generation modules identified 79, 75, and 59 clusters for the May 1987, the September 1989, and the March 1989 images, respectively. Once the homogenous clusters were identified and labeled, the digital data were processed through a maximum likelihood classifier to generate the final classifications for each image date. Following classification, all layers were spatially rectified (nearest neighbor) to the Universal Transverse Mercator (UTM) projection, zone 15 (Jensen, 1986; NASA, 1989; Swain, 1976; Swain and Davis, 1978). For final layer manipulation and flood limit boundary mapping, each classified data set was converted to GRASS, version 4.0 (CERL, 1989). Simple boolean logic was applied to determine the location and aerial extent of permanent water, flooded agriculture, frequently flooded agriculture, non-flooded agriculture, non-flooded forested lands, and flooded forested lands. Ground verification was difficult, especially for the flooded agricultural areas. However, acquisition of color infrared slides in April of 1989 as well as weeks of ground truthing with the Mississippi SCS determined the mapping accuracy of the flooded agricultural areas to be 88 percent with a lower confidence bound of 81 percent (Fitzpatrick-Lins, 1978; Jensen, 1986; US Department of Agriculture, 1989).

Conclusions The methods described in this paper set the stage for establishing a baseline, as well as monitoring through time the losses and gains of flooded agricultural lands (Farmed Wetlands). This is especially important with the lack of current financial incentives for farmers to continue allowing these lands to flood. Many of the Farmed Wetlands are easily altered through leveling, installation of tile drainage, and

Remote Sensing and GIS for Mapping Flood Prone Agricultural Land

187

modification of drainage networks around the fields. By using stream gauge data to drive the dates of satellite images used in processing, large area monitoring can be more easily performed. As well, the method of image date determination using stream gauge data enables the analyst to look at different flooding scenarios (i.e., the two year flood, the five year flood, etc.). Finally, the methods of mapping seasonally flooded agricultural lands described above may have implications for new analytical methods being considered by agricultural companies as well as the US Environmental Protection Agency. Remote sensing and Geographic Information Systems are currently being studied as potential means of providing exposure analyses (in terms of a risk assessment) for certain crop protection products. Typical analyses would include proximity of a target crop to aquatic systems. These proximity analyses may yield important information with regard to potential contamination due to runoff and/or drift. If one considers aquatic systems to be spatially static (especially those in an alluvial system), there is a good chance of underestimating exposure. This can be emphasized by the fact that along the Yalobusha River, seasonal flood waters introduce a dynamic land/water interface that, in some areas, varies up to one quarter of a mile in a given season.

References CERL, 1989. GRASS Users Guide, Version 3.1. U.S. Army Corps. of Engineers, Construction Engineering Research Laboratory, Champaign, IL. Dunne, T., and L. Leopold. 1978. Water in Environmental Planning. New York: W.H. Freeman and Company. Fitzpatrick-Lins, K. 1978. Accuracy and Consistency Comparisons of Land Use and Land Cover Maps Made from High Altitude Photographs and Landsat Multispectral Imagery. Journal of Research U.S. Geological Survey 6: 23-40. Frayer, W.E., T.J. Monahan, D.C. Bowden, and F.A. Graybill. 1983. Status and Trends of Wetlands and Deep Water Habitats in the Conterminous United States, 1950's to 1970's. Department of Forest and Wood Sciences, Colorado State University. Harrison, R. 1961. Alluvial Empire. Little Rock, AK: Pioneer Press. Huffstatler, H. 1989. Personal Communications. State Biologist, Mississippi. Jackson, MS: U.S. Department of Agriculture, Soil Conservation Service. Hunt, C.B. 1973. Natural Regions of the United States and Canada. San

188

Chapter Fifteen

Francisco: W.H. Freeman and Company. Jensen, J.R. 1986. Introductory Digital Image Processing. Englewood Cliffs: Prentice Hall. Jones, P. 1989. Personal Communications. State Engineer, Mississippi. Jackson, MS: U.S. Department of Agriculture, Soil Conservation Service. Kolb, C., P. Mabrey, W. Steinriede, E. Krinitsky, R. Saucier, and F. Smith. 1967. Geologic Investigations of the Yazoo Basin, Lower Mississippi Valley. U.S. Army Engineers Waterway Experiment Station; Vicksburg, Mississippi. NASA 1989. National Aeronautical Space Administration. ELAS Users Reference, Volume II. Science and Technology Laboratory, John C. Stennis Space Center. Report No. 183. Saucier, R. 1974. Quaternary Geology of the Lower Mississippi Valley. U.S. Army Engineer Waterways Experiment Station, Vicksburg, Mississippi. Swain, P.H. 1976. Layered Classification Adapted to Multi-temporal Data Sets. LARS Final Report. NASA contract NASA-14016. LARSPurdue University. West Lafayette, IN. Swain, P.H. and S.M. Davis. 1978. Remote Sensing: The Quantitative Approach. New York: McGraw-Hill Book Company. Tiner, R.W. Jr. 1984. Wetlands of the United States: Current States and Recent Trends. Fish and Wildlife Service, U.S. Department of the Interior, Washington, D.C. US Department of Agriculture-Soil Conservation Service. 1957. Soil Survey of Leflore County, Mississippi. Soil Conservation Service, Washington, D.C. US Department of Agriculture. 1989. Draft Wetlands Mapping Conventions for 1985 Food Security Act (FSA) SNTL. Soil Conservation Service, Fort Worth, TX. US Geological Survey. 1982a. Greenwood Quadrangle, Mississippi. Reston, VA: U.S. Geological Survey. US Geological Survey. 1982b. Money Quadrangle, Mississippi. Reston, VA: U.S. Geological Survey.

CHAPTER SIXTEEN MAPPING LANDSLIDE SUSCEPTIBILITY: A TRAVIS COUNTY, TEXAS CASE STUDY DAVID J. WACHAL

Introduction Whether occurring naturally or triggered by human activities, landslides result in enormous financial damages. In 1996, it was estimated that landslides in the United States caused $1.5 billion in property damage. Due to the widespread occurrence of landslides nationwide, the first step toward mitigation is mapping the areal extent of landslide susceptibility (Howell et al., 1999). Landslides damage transportation networks, buildings and structures, public works projects, and personal property. Moreover, erosion caused by landslides can lead to sedimentation problems in nearby streams and reservoirs. Landslide susceptibility maps can provide valuable information to planners and engineers who devise or implement land use strategies. Using Travis County, Texas as a case study, this research presents a method for indirectly mapping zones potentially susceptible to landslides. Indirect landslide mapping involves inferring the presence, absence, or potential of landslides through observations of geology, geomorphology, topography, and land cover (Atkinson and Massari, 1998). A grid based Geographic Information System (GIS) was used to analyze factors known to play a role in landslides including slope, geology, vegetation, and proximity to faults. Specifically, this research will: 1) provide an overview of landslide types, factors associated with landslides, and the use of GIS in landslide research, 2) apply a raster GIS methodology to delineate zones potentially prone to landslides in Travis County, and 3) discuss

Papers and Proceedings of the Applied Geography Conferences (2000) 23:378387

190

Chapter Sixteen

applications of landslide information to land use planning.

Background Landslide is a general term used for a number of different mass movement processes. A mass movement process can be defined as the downward movement of earth or surface material due to gravity. Most mass movements can be classified in a matrix of materials and movement types. Material types are most commonly rock, debris, or soil. Primary movement types are creep, falls, slides, or flows (Varnes, 1984). Mass movements can occur in a wide variety of geologic materials, on gradual or steep slopes, and under diverse climatic conditions. Locating and mapping these areas where mass movements may occur can provide useful information to planners, developers, and local communities. A GIS provides a tool for analyzing several factors simultaneously, delineating areas potentially prone to landslides. Depending on the type of movement and the kind of material involved, mass movements can vary in their shape, rate of movement, and how they affect the land surface. Four of the most common types of mass movements are creep, falls, slides, and flows. Soil or bedrock creep involves extremely slow downward or gradual deformation of soil and/or bedrock. Expansion and contraction of the surface layer resulting from natural processes, such as changes in soil temperature or moisture, is one of the main causes of creep. Falls are most often characterized as an abrupt free fall of bedrock or soil. Usually this type of movement occurs along steep slopes and road cuts. Jointing, fracturing, and weathering of bedrock are important precursors to falls. Slides are characterized by a fairly cohesive unit of rock or soil that slips downward along a clearly defined surface or plane. Slides are classified as either transitional or rotational. Transitional slides occur on nearly planar surfaces, whereas rotational movements produce a single or series of curved slip surfaces. Rotational type movements are usually referred to as slumps. Flows often involve rapid but viscous movement of soil, bedrock, or debris. A high water content in surface materials can initiate this type of movement. The magnitude, frequency, location, and extent of damage from these four major forms of mass movements can vary greatly. It is possible to avoid, minimize the effects of, and even control mass movements if there is a thorough understanding concerning slope instability. Factors influencing mass movements vary from one location to another. Evaluating these factors can lead to a more comprehensive understanding of a particular area. Two of the most important factors

Mapping Landslide Susceptibility: A Travis County Case Study

191

affecting mass movements are slope and geology. Mass movements occur whenever the downward pull of gravity overcomes resisting forces. Downward pull is related to material mass and slope angle. When this pull, or shearing stress, exceeds frictional resistance, movement occurs. Generally, the steeper the slope the greater the chance of slope failure (Montgomery, 1997). It is important to note that, although the chance of a landslide increases with an increase in slope, slopes need not be steep for landslides to occur. Changes in slope can be caused either by natural or human processes. Along with slope, geology plays a significant role in slope stability (Sarkar et al., 1995). Because most slope failures are shallow involving deposits near the surface, particular attention must be given to the geology of an area. Geologic units can be grouped and ranked according to their observed or inferred slope stability (Varnes, 1984). Unconsolidated materials generally have less cohesion and friction, and have higher infiltration rates, resulting in less shear strength than consolidated rocks. The stability of unconsolidated materials varies, as coarse fragments can maintain a steeper slope angle than finer ones. For sedimentary rocks, important characteristics determining strength and stability are grain size, shape, area, and the amount and type of cement. In fined grained sedimentary rock like clay, an increase in water content can result in a loss of shear strength. An understanding of slope and geology is paramount to assessing mass movement potential in a particular study area. In addition to slope and geology, land use and proximity to faults are important factors to consider in landslide research. Land use variables can affect slope stability. Vegetation coverage tends to increase the stability of slopes. Interlocking root networks absorb soil moisture, provide shear strength, and reduce erosion. Regions with dense vegetation were found to be less susceptible to slope failure compared to areas with sparse vegetation (Gokceoglu and Aksoy, 1996; Larsen and Torres-Sanchez, 1998; Pachauri et al., 1998). Agriculture and urbanization can have a negative effect on slope stability. Irrigation of cropland can cause changes in the soil water regime of a slope, which may cause instability (Blight, 1997). Steep grading, undercutting, and loading slopes are the most common human activities increasing the chance for landslides to occur (Griggs and Gilchrist, 1983). Fault zones contribute to instability by creating steep slopes and sheared, weakened rocks. Landslide frequency has been correlated with distance to major faults in a number of studies (Gokceoglu and Aksoy, 1996; Larsen and Torres-Sanchez, 1998; Pachauri et al., 1998) Thus, it is important to include factors such as vegetation and proximity to faults when attempting to delineate areas potentially susceptible to landslides.

192

Chapter Sixteen

Recently, GIS has emerged as an important tool for landslide susceptibility mapping (Atkinson and Massari, 1998; Howell et al., 1999). Geologists, engineers, and planners can use a GIS to store and manipulate geospatial data sets. The overlay operation commonly applied within a GIS can simultaneously analyze important factors related to slope stability such as slope, geology, vegetation and proximity to faults. Using basic principles to delineate landslide susceptibility, hazard, or risk, a GIS can provide useful information that can be applied to land use planning, preparing and evaluating environmental impact reports, and designing public and private facilities.

Study Area – Travis County Travis County is located in south central Texas (Figure 16-1). The region consists of two major geologic provinces divided by a fault zone. To the west is a plateau of Cretaceous limestone and exposed Precambrian Llano uplift. Cretaceous limestone units of the western region strike generally northeast with a gentle southeastward dip. Along the fault zone, the gently dipping Cretaceous units are broken and displaced by normal faults. To the east, soft shales, sandstones, and marls of the Upper Cretaceous Taylor and Navarro Formations, and Quaternary alluvial deposits of the Colorado River, form the Black Prairies. Dividing these two distinct geologic regions is the Balcones Fault zone. Somewhat centered in the county, the fault zone runs on a slight angle from northeast to southwest. Straddling the zone is the city of Austin, the state capital of Texas. Austin is a rapidly growing city with a metropolitan area population of 800,000 (625,000 in the city limits) (Villani, 2000). On a nationwide scale, Travis County is located in a region considered to have a moderate landslide potential if precipitation or construction disturbs slopes (Montgomery, 1997). Elevations above mean sea level in Travis County range from 110 meters in the east to 400 meters in the west. Travis County's topography varies from relatively flat in the east, to steeply sloping along the fault zone and to the west. The climate is humid subtropical with long hot summers and mild winters (National Cooperative Soils Survey, 1974). In the spring, rains are short in duration but high in intensity, causing occasional flooding and severe erosion. Land cover is primarily woodland/forest and shrubland in the west, urban and shrubland areas in the central portion, and cropland and woodland/forest in the east. Previously noted types of mass movement in the urban areas of Travis County include slumps and creep in clay and limestone falls (Trippet and Garner, 1976). Primarily, slope failures in the urban areas are due to

Mappping Landslide Susceptibility y: A Travis Couunty Case Study y

193

constructionn and developpment practicees that either remove mateerial from the base of tthe slope, or overload o on the head of the slope.

Travis County, Texas Figure 16-1. T

Methods An emppirical approaach was used d to evaluatte and map landslide susceptibilitty in Travis County. C A num merical rating system was applied a to assess the fa factors that coontribute to slope instabilityy. Factors numerically assessed inccluded slope, geology, g vegetation, and prroximity to fau ults. Each of the four factors consissted of four variables v (Figuure 16-2). Eaach of the variables waas assigned a value of 1 through 4 (R R), one being the least susceptible to landslides and 4 being the most sus ceptible. The use of a numerical raating system has h been noteed by Varnes (1984) as an effective approach foor evaluating landslide l hazaard and risk. B Based on theiir relative importance to slope insttability in thee study area, the four facttors were weighted annd assigned a landslide susceptibility weight (W)). In this research, sloope is considdered to be th he most impoortant factor related r to landslides; iit was assigneed the largestt weight of 335 percent. Geology is considered tthe second moost important factor and waas assigned a weight w of 30 percent. V Vegetation annd proximity to o faults were assigned weig ghts of 20 percent andd 15 percent, respectively. A raster baased GIS wass used to overlay the four 30X30 meter m resolutiion grids andd calculate a Landslide L Susceptibilitty Index (L LSI) for each individual cell (Figurre 16-2). Additional iinformation pertaining p to the t factors annd variables applied a in the analysis is presented below. b

194

Chapter Sixteen

Figure 16-2. Landslide susceptibility index (*R=Rating based on 1 being least susceptible and 4 being most susceptible)

Slope Slope gradient is best modeled using a grid-based Digital Elevation Model (DEM) in a raster environment (Sarkar et al., 1995). A Digital Elevation Model is a digital representation of the earth's surface derived from a sampled array of elevations for a number of ground positions at regularly spaced intervals. For this research, a DEM with a spatial resolution of 30 meters was acquired from the Texas Natural Resource Information System. Slope gradient for Travis County was derived from the DEM. Slope gradient is a numerical value defined as the largest angle between a slope surface and the horizon. Slope identifies the maximum rate of change in value (elevation) from each cell to its neighbors. Slope gradient in Travis County ranges from less than 1° to 48° (Figure 16-3).

Mappping Landslide Susceptibility y: A Travis Couunty Case Study y

195

Slope gradieents were grouuped into fourr categories annd rated (Table 16-1).

Travis County slope s Figure 16-3. T

G Geology and d Faulting Geologyy and faultingg, the second d and third fa factors analyzzed, were digitized froom a 1:250,0000 scale geo ologic map innto two sepaarate data layers. Twennty-five differeent geologic units u are depictted at this scalle (Figure 16-4). Travvis County's twenty-five t geologic g unitss can be grou uped and characterizeed as unconsoolidated materrial, shale, m marl and predo ominately limestone. T These four units u were ratted based on relative streength and stability (Taable 16-2). Sixty-six S faultts depicted att the 1:250,000 scale, were buffereed at three inttervals of 500 meters (Figuure 16-5). Gen nerally, as the distancce from fauults increasess, landslide frequency decreases d (Gokceoglu and Aksoy, 1996; Pachau uri et al., 19988; Sarkar et al., a 1995). Buffers weree used for twoo reasons: 1) faults f weaken surrounding rocks r and material; and 2) when refferring to larger scale geoloogic maps, it is i evident that there arre numerous smaller faultss dispersed ouutward from the t major faults illusttrated in the 1:250,000-sccale map at distances ov ver 1,000 meters. Thee buffered reggions were ratted according to their distaance from major faults (Table 16-1)..

196

Chapter Sixteen S

Travis County sllope, fault buffe fers, and land coover Table 16-1. T

Slope (Degreees) 0-11 12-23 24-35 36-48 Buffer Distan nce (m) >1,500 1,000-1,500 500-1,000