HAZUS-MHฎ Hurricane Wind Model Validation Study – Florida Hurricanes Charley and Ivan HAZUS-MHฎ MR-2 April 2007 ABBREVIATIONS ARC – American Red Cross CECBL – Concrete, Engineered Commercial Building, Low-Rise CECBM – Concrete, Engineered Commercial Building, Mid-Rise FDEM – Florida Division of Emergency Management FDOIR – Florida Department of Insurance Regulation FDOT – Florida Department of Transportation FEMA – Federal Emergency Management Agency FHA – Florida Hospital Association GIS – Geographic Information System HAZUS-MH – Hazards United States-Multi-Hazard HMTAP – Hazard Mitigation Technical Assistance Program JFO – Joint Field Office MECBL – Masonry, Engineered Commercial Building, Low-Rise MECBM – Masonry, Engineered Commercial Building, Mid-Rise MLR1 – Masonry, Low-Rise, Industrial/Warehouse/Factory Buildings NCDC – National Climatic Data Center NEMIS – National Emergency Management Information System NOAA – National Oceanic and Atmospheric Administration PA – Public Assistance PDA – Preliminary Damage Assessment PW – Project Worksheet SBA – Small Business Administration SECBL – Steel, Engineered Commercial Building, Low-Rise SECBM – Steel, Engineered Commercial Building, Mid-Rise SECBH – Steel, Engineered Commercial Building, High-Rise TO – Task Order FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY CONTENTS ABBREVIATIONS ................................................................................ ........................................ iii EXECUTIVE SUMMARY ................................................................................ ............................ vii INTRODUCTION ................................................................................ ..........................................1 Validation Study Objectives .............................................................................. 2 Overview of HAZUS-MH ................................................................................ ... 2 1.0 METHODOLOGY..................................................................... ...............................................4 2.0 DATA COLLECTED....................................................................... .......................................17 3.0 ANALYSES RESULTS AND OBSERVATIONS ...................................................................20 3.1 HAZUS-MH Macro Level Results for Aggregate Inventory at the County Level ................................................................................ ............................... 21 3.2 HAZUS-MH Micro Level Results for Site Specific Critical Facilities .......... 37 3.3 HAZUS-MH Loss of Functionality for Hospitals......................................... 46 4.0 CONCLUSION AND RECOMMENDATIONS .......................................................................48 4.1 Conclusions ................................................................................ .............. 48 4.2 Post-disaster Data Collection Recommendations ..................................... 49 4.3 Model Improvement Recommendations.................................................... 50 4.4 Software Functionality Enhancement Recommendations.........................51 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY TABLES 1a. HAZUS-MH and Observed Data Agreement by Hurricane, County, and Category .............ix 1b. Prior Validation Study - Summary of Initial ISO Insured Loss Estimates................................x 2. List of Hurricanes Considered in TO 440............................................................................. .....7 3a. Charlotte County Building Exposure ................................................................................ ....14 3b. Escambia and Santa Rosa Counties Building Exposure .....................................................14 4. Summary of Data Received and Used for Validation Study ...................................................18 5. Summary of Data Received for Validation Study....................................................................19 6a. Hurricane Charley Residential Damage - HAZUS-MH Wind Results ...................................22 6b. Hurricane Ivan Residential Damage - HAZUS-MH Wind Results.........................................24 7a. Hurricane Charley Shelter Demand - HAZUS-MH Wind Results..........................................25 7b. Hurricane Ivan Shelter Demand - HAZUS-MH Wind Results ...............................................27 8a. Hurricane Charley Residential Economic Loss (FDOIR) – HAZUS-MH Wind Results.........28 8b. Hurricane Ivan Residential Economic Loss (FDOIR) – HAZUS-MH Wind Results...............30 8c. Hurricane Charley Residential Economic Loss (ISO) – HAZUS-MH Wind Results ..............30 8d. Hurricane Ivan Residential Economic Loss (ISO) – HAZUS-MH Wind Results....................32 9a. Hurricane Charley Commercial and Industrial Economic Loss (ISO) - HAZUS-MH Wind Results......................................................................... ..............................................................32 9b. Hurricane Ivan Commercial and Industrial Economic Loss (ISO) - HAZUS-MH Wind Results ................................................................................ ......................................................34 10a. Hurricane Charley Damage - HAZUS-MH Site Specific Hurricane Wind Results...............37 10b. Hurricane Ivan Damage - HAZUS-MH Site Specific Hurricane Wind Results ....................37 11a. Hurricane Charley Damage - Comparison of Observed Site Specific Critical Facility Damage and HAZUS-MH Wind Damage Curve.........................................................................40 11b. Hurricane Ivan Damage - Comparison of Observed Site Specific Critical Facility Damage and HAZUS-MH Wind Damage Curve........................................................................... .............41 12. Hurricane Ivan Economic Loss - Comparison of Observed Site Specific Critical Facility Economic Loss and Loss Ratios and HAZUS-MH Wind Loss Curve…….……………………….44 13a. Hurricane Charley Hospital Loss of Functionality - HAZUS-MH Site Specific Hurricane Wind Results......................................................................... .....................................................46 13b. Hurricane Ivan Hospital Loss of Functionality - HAZUS-MH Site Specific Hurricane Wind Results......................................................................... ..............................................................47 April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY FIGURES 1. Hurricane Charley Study Region ................................................................................ ..............5 2. Hurricane Ivan Study Region ................................................................................ ...................6 3a. Hurricane Charley HAZUS-MR2 Peak Gust Wind Speeds ....................................................7 3b. Hurricane Ivan HAZUS-MR2 Peak Gust Wind Speeds ..........................................................8 4. Hurricane Charley Windfield ................................................................................ ....................9 5. Hurricane Ivan Windfield ................................................................................ ........................10 6. HAZUS Validation Results Range ................................................................................ ..........13 7. Default Mapping Schemes in HAZUS- MH.............................................................................. 15 8. Hurricane Charley Residential Damage (Total Damaged Building Count) - HAZUS- MH Wind Results......................................................................... ..............................................................23 9. Hurricane Charley Shelter Demand - HAZUS-MH Wind Results............................................26 10. Hurricane Charley Residential Economic Loss (FLDOIR) - HAZUS-MH Wind Results........29 11. Hurricane Charley Residential Economic Loss (ISO) - HAZUS-MH Wind Results...............31 12. Hurricane Charley Commercial and Industrial Economic Loss (ISO) - HAZUS-MH Wind Results......................................................................... ..............................................................33 13. Hurricane Charley Damage - HAZUS-MH Site Specific Wind Results .................................38 14. Hurricane Ivan Damage - HAZUS-MH Site Specific Wind Results.......................................39 15. Hurricane Charley Damage - Comparison of Observed Site Specific Critical Facility Damage and HAZUS-MH Wind Damage Curve........................................................................... .............42 16. Hurricane Ivan Damage - Comparison of Observed Site Specific Critical Facility Damage and HAZUS-MH Wind Damage Curve........................................................................... .............43 17. Hurricane Ivan Economic Loss - Comparison of Observed Site Specific Critical Facility Loss and HAZUS-MH Wind Loss Curve........................................................................... ...................45 APPENDICES A. Templates for Collecting Damaged Data............................................................................ .. A-1 B. Detailed Raw Data Obtained for Observed Damage............................................................ B-1 C. ARA Prior HAZUS Validation Report.......................................................................... .......... C-1 D. Acknowledgements: Contact Log............................................................................. ............D-1 E. H*Wind Speeds and ARA Modeled Wind Speeds................................................................ E-1 F. Glossary........................................................................ ....................................................... F-1 April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY EXECUTIVE SUMMARY This study was performed for FEMA by URS Corporation and PBS&J, as Task Order 440, under HMTAP contract EMW-2000-CO-0247. This report presents the findings of a FEMA validation study of the HAZUS-MH MR-2 (Build 45) Hurricane Wind Model. The validation study involved the comparison of HAZUS-MH modeled results with observed hurricane wind hazards and impact data. To make comparisons to the HAZUS-MH modeled results from observed data, data collection activities were conducted from November 2005 (more than a year after the storms) through August 2006. Data were collected from local, state, and federal agencies and the private sector after Hurricane Charley in Charlotte, DeSoto, Hardee, Lee, Orange, Osceola, and Polk Counties; and Hurricane Ivan in Escambia County in the State of Florida. In addition, data were also used from prior HAZUS validation studies conducted in 2004 for Hurricanes Charley (HMTAP TO – 332) and Ivan (HMTAP TO – 348). The purpose of this project was to benchmark the best modeled runs of HAZUS-MH (MR2 version) for wind and compare those runs to the observed and recorded damage and loss in various counties and jurisdictions in Florida. A primary goal was to test run HAZUS-MH’s functionality and utility against “real world” historical field data to support disaster operations. A secondary goal was to develop standardized data collection process and analysis for HAZUSMH for long-term recovery operations. This report describes data collection, modeling improvement, and software functionality enhancement recommendations for future HAZUS-MH applications, including validation studies. This validation study was intended to help provide a systematic assessment of how well the model performed in several categories compared with readily available observed data from these specific events. However, some comparisons could not be made due to limitations with existing data or a lack of data. For example, some readily available observed data sets did not distinguish damage caused by wind versus flood; and commercial, industrial and most critical facility qualitative damage count data are not collected. HAZUS-MH modeled results for residential, commercial, and industrial occupancy classes compared well with the observed data for the Hurricane Charley study region. Overall, HAZUSMH modeled results were in good agreement with observed data for residential qualitative damage and economic loss. This good agreement can be attributed to the updated and improved residential building stock in the MR2 version HAZUS-MH default inventory. Even with the uncertainties and limitations with the national default inventory for commercial and industrial occupancy classes, HAZUS-MH economic loss estimates compared well with the observed data from both insurance data sources -- FDOIR and ISO -- for the Hurricane Charley study region. Also, HAZUS-MH estimates for shelter demand were in good agreement for southwest Florida. Prior validation studies conducted for FEMA by Applied Research Associates (ARA) indicated that HAZUS-MH estimates compared well with preliminary ISO losses for Hurricanes Charley, and Jeanne. HAZUS-MH consistently and significantly underestimated economic loss for public and critical facilities. However, this may be attributed to the age and source of the critical facilities default inventory in HAZUS-MH, which is not as current as the general building stock, specifically the residential building stock. The general building stock square footage for schools (EDU1-2) and government essential facilities (GOV1-2) are not well-represented in the commercial real estate datasets purchased for HAZUS. The residential building stock valuations are current as of 2005, while the square footage information is based on the 2000 census, and the critical facilities data is current as of 2001. The critical facilities default data were collected from national and state data providers and most likely do not reflect the most current data (i.e., number and location of facilities) that are available from local governments. The level of agreement between HAZUS-MH and observed data for site specific critical facilities varied. HAZUS-MH estimates for critical facility damage states were in good agreement with more than half of the observed data for site specific facilities. The HAZUS-MH wind damage curve more accurately predicted the type of damage, the closer the facility was located to the hurricane track in nearly all cases for Hurricane Charley. However, the accuracy of the damage curve varied for facilities that were located farther away from the hurricane path, as seen for Hurricane Ivan. The wind damage curve estimates were in good agreement for more than half of the sites for Hurricane Charley, but underestimated the damage for Hurricane Ivan for all but one site. The HAZUS-MH wind loss curve underestimated economic loss and loss ratio for all four of the site specific critical facilities in Escambia County for Hurricane Ivan. The summary of the comparisons of HAZUS-MH modeled results and observed data and observations about these comparisons are listed in Table 1. Overall, there was better agreement in the Hurricane Charley study region than there was in the Hurricane Ivan study region. This is to be expected, as HAZUS-MH estimates are most accurate when used at a multi- county regional level such as Hurricane Charley. The accuracy decreases when the HAZUS-MH modeled results are compared with observed data at a smaller geographic region, such as the one county study region of Escambia County for Hurricane Ivan. Additionally, HAZUS-MH estimates compared the best with observed data for the Hurricane Charley study region, as Charley caused predominantly wind damage and loss. Therefore hurricane wind modeled results generated by HAZUS-MH could be more accurately compared with observed data for predominantly a wind event. This correlation was not as strong for Hurricane Ivan since it was both a wind and flood event. Based on a comparison of the economic loss data received for Ivan, it appears that 35 percent was attributed to flood damage. Table 1a. HAZUS-MH and Observed Data Agreement by Hurricane, County, and Category Hurricane County Agreement of HAZUS-MHEstimate With Observed Data Charley (Residential Damage) Hurricane Charley Study Region Agreement at low end Ivan (Residential Damage) Escambia HAZUS appears to underestimate damage. However, this is not the case since Ivan caused flood damage that is not accounted for in the Hurricane Model. Therefore, direct comparisons were not able to be made as the observed data included both wind and flood damage. Charley (Shelter Demand) Hurricane Charley Study Region Good agreement Ivan (Shelter Demand) Escambia Underestimates Charley (Residential Economic Loss) – FDOIR data Hurricane Charley Study Region Good agreement Ivan (Residential Economic Loss) – FDOIR data Escambia Underestimates wind loss Charley (Residential Economic Loss) – ISO data Hurricane Charley Study Region Good agreement Ivan (Residential Economic Loss) – ISO data Escambia Agreement at the high end Charley (Commercial and Industrial Economic Loss) – ISO data Hurricane Charley Study Region Good agreement Ivan (Commercial and Industrial Economic Loss) – ISO data Escambia Underestimates wind loss Charley (Public Facilities Economic Loss) Hurricane Charley Study Region HAZUS underestimates wind loss. This is likely due to the generalizations made about the building inventory, size (sf), and replacement value for these types of facilities. Ivan (Public Facilities Economic Loss) Escambia Charley (Hospital Economic Loss) Charlotte, DeSoto, and Orange Ivan (Hospital Economic Loss) Escambia Ivan (School Economic Loss) Escambia Charley (Site Specific Critical Facility Damage) Hardee, Osceola, and Polk 80% of Sites in Good Agreement or Agreement at Low End Ivan (Site Specific Critical Facility Damage) Escambia 50% of Sites in Good Agreement Charley (Wind Damage Curve) Hardee, Osceola, and Polk 60% of Sites in Agreement at Low End Ivan (Wind Damage Curve) Escambia 88% of Sites were Underestimated Ivan (Wind Loss Curve) Escambia Underestimates Previous research has shown that the HAZUS-MH Wind Model results have compared well with observed data. A prior validation study of the HAZUS-MH Wind Model was conducted in November 2004 by ARA for FEMA. Initial estimates of industry-wide insured losses released by ISO for Hurricanes Charley, Frances, Ivan, and Jeanne were used in comparison with HAZUSMH modeled economic loss estimates. ARA performed hurricane wind field modeling and hurricane wind loss estimates before, during, and after each of the four hurricanes. The modeled results compared well with insurance economic loss data for Hurricanes Charley and Jeanne, but there were significant differences in the estimates for Hurricanes Frances and Ivan. More recent runs for Hurricane Frances have produced results between $3.0B and $5.8B, through updates in wind model parameters brought about model updates that accounted for FCMP tower wind speeds for terrain effects. The comparison of the modeled results and the observed data is shown in Table 1b and Appendix C. Table 1b. Prior Validation Study - Summary of Initial ISO Insured Loss Estimates Hurricane Landfall Date ISO Press Release Date Initial ISO Insured Loss Estimate ($B) HAZUS-MH Estimate Based on Final ARA Tracks from Tables 3-6 ($B) States Included Charley 8/13/04 8/25/04 6.7 7.1 FL Frances* 9/5/04 9/23/04 4.1 1.8 FL Ivan 9/16/04 10/14/04 5.3 1.6 FL, AL, GA Jeanne 9/26/04 10/26/04 2.8 2.8 FL 2004 Total 18.9 13.3 Source: ARA * More recent runs for Hurricane Frances have produced results between $3.0B and $5.8B, through updates in wind model parameters brought about through correction of the FCMP tower wind speeds for terrain effects The specific conclusions from this report include: • The observed data for the Hurricane Charley study region compared well with HAZUS-MH (1) residential qualitative damage; (2) residential, commercial, and industrial economic loss; and (3) short-term shelter demand estimates. There was better agreement at the regional level, as seen in the Hurricane Charley study region versus the results for one county, Escambia County, in the Hurricane Ivan study region. Hurricane Charley was predominantly a wind event. Therefore, it is more appropriate to compare HAZUS-MH wind results with observed wind data for Charley, than it is for Ivan that was both a wind and flood event. • HAZUS-MH public and critical facilities qualitative damage (i.e., for schools) and economic loss estimates did not compare well with observed data. HAZUS-MH consistently and significantly underestimated economic loss for public and critical facilities. This is most likely because the HAZUS-MH default inventory for public and critical facilities was collected at the national level in 2001, and does not include the most current number and location of facilities. • HAZUS-MH site specific qualitative damage estimates were in good agreement for 80 percent of the sites for Hurricane Charley and 50 percent of the sites for Hurricane Ivan. Considering that HAZUS-MH was designed to be used at a larger scale (i.e., region, county), it appears that the analysis showed that HAZUS-MH estimates compared reasonably well with the observed damage at the site specific level. However, site specific economic loss estimates were underestimated by HAZUS-MH. • HAZUS-MH wind damage curve estimates were in good agreement at the low end for 60 percent of the sites for Hurricane Charley, but underestimated 88 percent of the sites for Hurricane Ivan. • HAZUS-MH wind loss curve estimates underestimated for all sites for Hurricane Ivan. • HAZUS-MH hospital loss of functionality estimates did not compare well with the observed data. The model significantly overestimated the loss of functionality (i.e., number of days). However, it is important to consider that HAZUS-MH estimates loss of functionality based on building damage. It is possible that a facility can be operational, if key parts of the building are not damaged. The lessons learned and next steps for FEMA include understanding the challenges, and recommendations and benefits for HAZUS-MH. This information is organized by: (1) data collection, (2) modeling capabilities, and (3) software functionality. Some of the lessons learned and next steps are listed below. All lessons learned and next steps are included in Section 4. HAZUS-MH Data Collection Challenges: • Modeled estimates for a variety of categories such as: displaced households; debris; damage and loss at the jurisdictional level; damage to site specific facilities; and vulnerability reduction measures through mitigation options were not compared with observed data because the observed data was incomplete, unavailable, or incomparable. • Data for comparison with HAZUS-MH estimates are not always separated by hazard; and damage data are not collected for commercial, industrial, public, and critical facilities. • Currently there is no coordination with counties prior to hurricane season to determine what data are already being collected that are useful for future HAZUSMH validation studies and to determine which additional data are required for the local level. There is limited staff available to collect detailed HAZUS-MH data after an event. Recommendations and Benefits: • It is recommended that FEMA convene a focus group to determine the appropriate data to collect to enhance the default inventory and to document observed disaster impacts. The focus group should develop a pre-disaster local data gap analysis and data gathering coordination groups for enhanced planning and mitigation efforts. • It is recommended that FEMA encourage state and local jurisdictions to use the data collection template (refer to Appendix A) developed for this project to help consistently collect data. This data collection template should be examined by the focus group to ensure that all necessary data are being requested. • It is recommended that future validation studies should be conducted at a regional level for counties that only experienced wind damage, and should include data for only wind impacts. Focused data collection efforts enhance the value, and usefulness of HAZUS-MH validation. • It is recommended that FEMA prepare prewritten rapid response HMTAP task orders for the types of HAZUS-MH data needed to be collected so deployment to collect the data is executed shortly after the event. • It is recommended that coordination occur with counties to determine which data sets are available for comparison with HAZUS-MH, and which data sets need to be collected. It might be necessary to provide assistance to counties to collect data that is not typically collected (e.g., commercial, industrial and public facilities damage). HAZUS-MH Modeling Capabilities Challenges: • When the study regions were exported, all of the results were not automatically exported with the study region. As such, sharing the results between HAZUS-MH users involved exporting individual results tables. • The damaged building count is generated from the general building stock for general occupancy classes, but not specific occupancy classes. Damage counts are not provided for critical facilities in HAZUS-MH. • The default general building stock for the grade school occupancy class (i.e., EDU1) does not include all of the schools that are in the critical facilities default inventory. Recommendations and Benefits: • Develop the capability to retain the results that have been run in HAZUS-MH when it is exported to allow for user sharing. • Develop the capability to provide damage counts for critical facilities. • Develop a process to permit local governments to submit updated building stock inventory for use with HAZUS-MH. Software Functionality Challenges: • Damage and loss estimates are not provided at the zip code, jurisdictional, or site specific levels by HAZUS-MH; this would improve the usefulness of the model for comparative analysis. • HAZUS-MH analysis parameters revert to the default parameters after an analysis is run. Therefore, it can not determine which parameters were selected for analysis after the analysis has been run. Also, when all parameters are not selected for analysis, HAZUS-MH does not always generate all the results specified. • Damaged building counts are not provided for specific occupancy classes or critical facilities. Recommendations and Benefits: • Develop the capability in HAZUS-MH to produce results at the zip code, jurisdictional, and site specific levels. Develop capability to select zip code or jurisdiction level data attributes from the map view. This functionality will allow for more detailed comparisons and validations of results. • Develop the capability in HAZUS-MH to identify which analysis parameters have been run, instead of the analysis screen reverting back to default analysis parameter settings. Also, develop the capability within HAZUS-MH to run individual analysis parameters. Although it appears that this can be done, sometimes the results are not provided consistently. • Develop the functionality to provide the damaged building count for specific occupancy classes and critical facilities in the wind model. This functionality will allow for more detailed comparisons and validations of results. INTRODUCTION This validation study examines the impacts caused by Hurricane Charley in Charlotte, DeSoto, Hardee, Lee, Orange, Osceola, and Polk Counties, and Hurricane Ivan in Escambia County using HAZUS-MH. These two devastating hurricanes provided an unprecedented opportunity to validate the HAZUS-MH Hurricane Wind Model. According to the National Climatic Data Center (NCDC), the general impacts caused by Hurricanes Charley and Ivan are as follows: • Hurricane Charley made landfall in Southwest Florida on August 13, 2004, as a Category 4 storm, causing over $15 billion in losses and at least 34 deaths in Florida, South Carolina, and North Carolina. • Hurricane Ivan made landfall near Gulf Shores, Alabama, on September 16, 2004, as a Category 3 storm, causing over $14 billion in losses and at least 57 deaths in the eastern United States. Ivan was the most destructive hurricane to impact the area in over 100 years. Two other significantly damaging hurricanes occurred in Florida during 2004 and 2005: • Hurricane Jeanne made landfall in East-Central Florida on September 26, 2004, as a Category 3 storm, causing over $6.9 billion in losses and at least 28 deaths in the Eastern U.S. and Puerto Rico. • Hurricane Dennis made landfall near Pensacola, Florida on July 10, 2005, also as a Category 3 storm, causing $2 billion in losses and at least 12 deaths in the Florida Panhandle and eastern United States. Losses from Hurricane Dennis were substantially less than Ivan, even though both were Category 3 hurricanes. This difference was due to four factors: intensity, size, speed and location. Hurricane Dennis made landfall east of Pensacola, Florida where there was less property than the area that was impacted by Hurricane Ivan. A prior validation study that was conducted for FEMA by ARA in November 2004 showed that the HAZUS-MH Wind Model results compared well with economic loss data for Hurricanes Charley, and Jeanne; as shown in full copy in Appendix C. HAZUS-MH is a risk assessment model developed by FEMA to estimate damage and loss from natural and man-made hazards. Readily available as public domain software on FEMA’s website, HAZUS-MH software is the most frequently downloaded content on the website. The analyses and results are used to help make decisions for disaster preparedness, response, recovery, and mitigation. HAZUS-MH is used to estimate affected populations and infrastructure damage and economic loss, to target response resources, and to evaluate the savings from implementing hazard mitigation measures to reduce impacts from natural hazards. These measures include but are not limited to land use planning, zoning, structural projects, and building code enhancements. In this study, damage is defined as qualitative damage (e.g., minor) and economic loss is defined as the cost to repair or replace structures and contents. The results of this report will be used to understand how well HAZUS-MH estimates impacts for a given event. This understanding will provide a foundation to potentially improve the HAZUS-MH model, and to enhance future data collection efforts for future validation studies. These efforts will support communities who use this loss estimation tool to help identify and prioritize mitigation measures, and perform more accurate analyses of avoided damages resulting from the implementation of mitigation techniques. Validation Study Objectives The objectives of the validation study were to: • Compare HAZUS-MH-modeled estimates of wind damage and loss with actual damage and loss for the general building stock and critical facilities at the county level. • Compare HAZUS-MH-modeled estimates of wind impacts such as displaced populations and debris generated at the county level with observed data. • Compare HAZUS-MH-modeled damage and loss estimates for critical facilities at the site level with observed data. • Compare HAZUS-MH-modeled damage states and resultant loss of functionality (loss of use in days) of hospitals at the site level with actual impacts. • Explore documented vulnerability reduction measures and the potential to mitigate these measures in HAZUS-MH. • Validate the existing HAZUS-MH damage and wind loss curves. • Provide recommendations, as appropriate, to improve the HAZUS-MH Hurricane Wind Model. • Provide recommendations to enhance data collection for future HAZUS-MH validations. Overview of HAZUS-MH HAZUS-MH is a standardized loss estimation software program built upon an integrated Geographic Information System (GIS) platform. HAZUS-MH includes a wide range of inventory data (e.g., demographics, building stock, critical facility, transportation, and utility lifelines) and three models to estimate potential losses from earthquakes, hurricanes, and floods. The system was developed by the Department of Homeland Security’s FEMA to support improved risk assessments to address mitigation, emergency planning, and response. The HAZUS-MH Hurricane Wind Model provides users the ability to estimate potential economic damage and loss to residential, commercial, and industrial buildings. It also allows users to estimate direct economic loss, post-storm shelter needs, and building debris. The model addresses wind pressure, windborne debris, surge and waves, atmospheric pressure change, duration or fatigue, and rain. HAZUS-MH is flexible and can be used in conjunction with third- party models and other hazard and building inventory data to support a range of hazard-related analyses, such as. The results are displayed in a series of tabular reports and maps. It includes the following features: • A building classification system that depends on the characteristics of the building envelope and building frame. • The capability to compute damage based on building classes and the effects of rain and progressive failure. • The capability to compute damage to contents and building interior. • The capability to estimate tree blow-down and structure debris quantities. • Loss estimates that include direct and indirect economic loss, shelter requirements, and casualties. • Modules that facilitate future assessment of mitigation, benefit-cost, and building code issues. Model Releases MR-1 and MR-2 Initially, MR-1 was to be used to run the analyses as it was available at the onset of this project. However, MR-2 was ultimately used, since this version of the model had enhanced functionality as it included the latest modifications to the data and methodology. The following updates have been incorporated in MR-2: • The default inventory data has been updated with 2005 valuation data for all occupancy classes. • Valuation data for single-family residential housing and manufactured housing has been updated and validated based on comparisons with other national databases. • Zeros have been substituted for any negative values calculated for the daytime, nighttime, working commercial, working industrial and commuting populations. 1.0 METHODOLOGY The methodology for this comparative validation study includes conducting two analyses: . Macro-analyses by county of observed data versus HAZUS-MH estimates พ Building damage count for residential structures and critical facilities พ Shelter population requirements พ Economic loss to residential, commercial, industrial, public, and critical facilities พ Wind speed sensitivity analysis . Micro-analysis by jurisdiction of observed data versus HAZUS-MH estimates พ Critical facilities (only damage) พ Hospital loss of functionality พ Comparison of observed data for site specific critical facilities damage and economic loss with HAZUS wind damage and wind loss curves พ Wind speed sensitivity analysis Hurricane wind study regions were created for Hurricane Charley for Charlotte, DeSoto, Hardee, Lee, Orange, Osceola, and Polk Counties, and for Hurricane Ivan hurricane for Escambia County. These two study regions are provided in Figure 1 and Figure 2, respectively. Figure 1. Hurricane Charley Study Region Figure 2. Hurricane Ivan Study Region Hazard Modeling Landfall conditions for Hurricanes Charley and Ivan that were considered in this study are summarized in Table 2. Two types of data were obtained for these hurricanes: (1) Applied Research Associate’s (ARA) best estimate of the hurricane intensity parameters based on data from the following sources: (i) National Hurricane Center (NHC) forecast advisories, (ii) H*Wind surface wind analysis output provided by Hurricane Research Division (HRD), and (iii) surface level wind and pressure observations to determine the values of the profile parameter and radius of maximum winds; and (2) H*Wind swaths which were downloaded directly from the Internet. Comparisons were made for sustained wind because H*Wind data and ARA windfield data both included sustained wind speeds. However, sustained winds were converted to peak gust wind speeds for HAZUS-MH modeling purposes. Table 2. List of Hurricanes Considered in TO 440 Hurricane Landfall Conditions Location Date NHC Saffir- Simpson Category Sustained Wind Speed NHC 1-minute (mph) Charley Charlotte County, FL 8/13/04 4 145 Ivan Baldwin County, AL 9/16/04 3 130 Figures 3a and 3b shows the resulting peak gust wind speeds at the census tract level using the ARA modeled tracks in HAZUS MR-2 for the two hurricanes included in this study. Figure 3a. Hurricane Charley HAZUS-MR2 Peak Gust Wind Speeds Figure 3b. Hurricane Ivan HAZUS-MR2 Peak Gust Wind Speeds Figures 4 and 5 illustrate the comparison of the sustained wind speeds between ARA and H*Wind data for all the jurisdictions within the counties considered in this study, and Appendix E lists this comparison in tabular format. This comparison shows in general small variations in the predicted wind speeds between the two data sources. It seems from the plot of both storms that right along the track the ARA data was very consistent with H wind, but less accurate away from the track. In comparison to ARA windfield data, H*Wind sustained wind speeds were: • 10 percent lower for Charley • 10 percent higher for Ivan Due to the 10-percent difference in the wind speeds, three hazard scenarios were created to account for a 10-percent margin of error for the wind speeds: 1) Default HAZUS-MH using ARA windfield data, which is referred to in the results section as the Optimum HAZUS Results. 2) HAZUS-MH using ARA windfield data decreased by 10 percent, which is referred to in the results section as the Lower End Results. 3) HAZUS-MH using ARA Windfield data increased by 10 percent, which is referred to in the results section as the Higher End Results. Figure 4. Hurricane Charley Windfield Figure 5. Hurricane Ivan Windfield Hazard Modeling Assumptions and Limitations This analysis used the best available wind speed data. As is the case with all models, there are modeling uncertainties that result from incomplete scientific knowledge about hurricanes and their effects on structures. For example, it is difficult to predict all types of flying debris that could damage a structure. HAZUS-MH modeled peak gust wind speeds were converted from sustained wind speeds for the analysis. However, it is always likely that micro-bursts could occur, which would not be represented in the peak gust wind speeds. Or, the location of large buildings in an area could reduce wind speeds in a given area. These wind speed anomalies would not be accounted for and cause uncertainty in modeling the hazard. Study Approach Observed damage and loss data were compared with HAZUS-MH modeled estimates at the county-level, and observed damage data were compared at the site-specific level. Hospital loss- of-functionality-observed data were also compared with HAZUS-MH estimates. HAZUS-MH estimates were generated using three hazard scenarios for the purpose of a sensitivity analysis, which is described in greater detail following this study approach section. These three scenarios included running HAZUS-MH with a hazard scenario using wind speeds provided by ARA, and running two scenarios to account for a +/-10 percent variance in wind speed. HAZUS-MH estimates were compared with the observed data that were collected from a variety of sources that are summarized in Section 2.0, and discussed in detail and provided in tabular format in Appendix B. A county-level (macro-level) analysis was performed to compare HAZUS-MH estimates of the following with observed data: • The number of people requiring short-term shelter. • Qualitative damage (i.e., minor and moderate damage, severe and destroyed) counts for the residential building stock, and public and critical facilities. • Economic loss to the residential building stock, and public and critical facilities. Debris quantities were not received for either county; consequently, debris estimates could not be compared with observed data. Damage and economic loss to structures were modeled in HAZUS-MH based on the vulnerability of the structure. This vulnerability was based on building characteristics such as the type of building materials and roof type. Short-term shelter demand was modeled in HAZUSMH based on a percentage of the population that would require sheltering because their homes were damaged. A detailed (micro-level) analysis was performed to compare HAZUS-MH of the following with observed data: • Qualitative damage (i.e., minor and moderate damage, severe and destroyed) for site specific critical facilities. • Economic loss to the site specific critical facilities (Escambia only). • Wind damage curve for site specific critical facilities. • Wind loss curve for site specific critical facilities. A micro level analysis was intended to be performed at the jurisdictional level to compare HAZUS-MH damage and loss estimates with observed data. However, this could not be completed because observed data were not available for this analysis at the jurisdictional level. Data received were aggregated at the county level. An analysis was also performed using HAZUS-MH to estimate hospital loss of functionality (i.e., loss of use in days). Observed data were compared with HAZUS-MH estimates. Approach Assumptions and Limitations Most of the observed data included damage and economic loss from both wind and flood hazard for Escambia County, and could not be separated. As such, comparing HAZUS-MH results with these data sets did not usually result in good agreement. Comparisons were attempted for the damaged building count for critical facilities such as schools, fires stations, and shelters. The damaged building count is generated from the general building stock for general occupancy classes, but not specific occupancy classes. Damage counts are not provided for critical facilities in HAZUS-MH Sensitivity Analysis Given the nature of uncertainty in HAZUS, it is more practical to compare observed data with a range of HAZUS-MH results instead of simply comparing the observed data with just HAZUS results. This uncertainty may be due to uncertainty in the hazard modeling, the way the inventory is modeled, the valuation in the damage model, or a combination of some or all three of these parameters. The comparison to ARA windfield data with the H*Wind sustained wind speeds were 10 percent lower for Charley and 10 percent higher for Ivan. A range of +/- 10 percent of wind speeds is believed to have the highest impact on results and would cover the valuation in results if the other parameters are varied. The scale shown in Figure 6 provides a consistent and statistically based scale for deriving quantitatively based conclusions as to how well HAZUS-MH estimates compared with observed data. The scale consists of five categories: HAZUS-MH underestimates impacts in comparison with observed data. The observed data are greater than the upper-end results. Observed data agrees with HAZUS low-end results. The range includes numbers that fall within the bounds between the low-end results, and the midpoint of the low-end results and the optimum results. Observed data has good agreement. The range includes numbers that fall within the bounds between the midpoint of the low-end results and the optimum results, and the midpoint of the optimum results and the upper-end results. Observed data agrees with HAZUS upper-end results. This data includes the numbers that fall within the bounds between the midpoint of the optimum results and the high-end results, and the upper-end results. HAZUS-MH overestimates impacts in comparison with observed data. The observed data are less than the lower-end results. These icons, as illustrated, are used as a visual guide to represent the level of agreement in the Results and Observation Section of this report. These icons are correlated with the levels of agreement in Figure 6. Figure 6. HAZUS Validation Results Range Lower-end Results (LR) Optimum (Default) Results (OR) Good Agreement (A) Agreement at the low end (AL) Agreement at the high end (AH) HAZUS Underestimates Losses (U) HAZUS Overestimates Losses (O) Upper-end Results (UR) (LR + OR)/2 (OR + UR)/2 Inventory HAZUS-MH MR-2 Build 45 default inventory obtained from national level sources was used for this study. There are five general building types (e.g., wood, masonry, concrete, steel, and manufactured homes). The general building stock data are in aggregate form, and percentages of building types are assumed for each census tract or block based on average characteristics of the geographic region. The HAZUS-MH default building inventory is based on U.S. Census (2000) and Dun & Bradstreet (D&B) (2002) data. The Census and D&B data provide a range of the year of construction at the census block , beginning with pre-1939 structures and includes each decade up to 1990, as well as structures built during 1998 to 2002 (these are referred to in HAZUS-MH as post-1998 construction). The default general building stock data are further classified into 39 specific building types and 33 occupancy classes, which includes building square footage and building value. General building stock data are grouped by occupancy class (e.g., residential, commercial, industrial, and governmental facilities), critical facilities data (e.g., emergency operation centers, hospitals, police and fire departments, and schools), and population characteristics. Buildings are further categorized by characteristics such as roof shape, roof covering, and opening protection. Default inventory data were used, meaning that no data modifications were made to the general building stock percentage distribution to reflect local building inventory characteristics (e.g., building count, replacement value, building type, roof type). Local building inventory was considered, though none was received for this validation study that was applicable for modifying the general building stock exposure, occupancy class, or detailed building characteristics. Table 3a lists the building exposure values for counties within the Hurricane Charley study region, and Table 3b lists the building exposure for the county within the Hurricane Ivan study region, for HAZUS-MH MR-1 and MR-2. Building valuations for MR-2 are current as of 2005. These tables are provided to illustrate the difference in building valuation between MR-1 and MR-2. Table 3a. Hurricane Charley Study Region - County Building Exposure Counties Residential ($B) Commercial ($B) Industrial ($B) Other Occupancies($B) Total HAZUS-MR2 ($B) Total HAZUS-MH MR1 ($B) Charlotte 8.9 0.8 0.1 0.1 9.8 8.7 Desoto 1.1 0.1 0 0 1.3 1.3 Hardee 0.9 0.1 0 0 1 0.9 Lee 26.9 3.2 0.5 0.3 30.9 26.4 Orange 47 8.9 1.4 1.3 58.6 51.1 Osceola 8.3 1.1 0.1 0.2 9.7 8.3 Polk 22.3 3 0.9 0.6 26.8 22.9 Total for Region 115.4 17.2 3 2.5 138.1 119.6 Table 3b. Hurricane Ivan Study Region - County Building Exposure Counties Affected Total HAZUS-MR1 ($ Billion) Total HAZUS-MR2 ($ Billion) 16.2 Residential ($ Billion) Other Occupancies ($ Billion) Industrial ($ Billion) Commercial ($ Billion) Escambia 13.4 2.1 0.3 0.4 The distribution of the exposure data, building count, and square footage into hurricane-specific building types was done using the default mapping schemes of the hurricane model. Figure 7 illustrates the allocation of these default schemes for the different areas in Florida. Figure 7. Default Mapping Schemes in HAZUS-MH These schemes generally reflect the different building codes requirements where southeast Florida has historically stronger wind provisions while the northern portion of the state reflects more wood frame construction. Southeast Florida is typically exposed to higher wind speeds and buildings are designed to higher wind pressures. Default Inventory Assumptions and Limitations General building stock data are in aggregate form by census block and tract. The percentages of building types are assumed for each census tract or block based on average characteristics of the geographic region. As noted in the HAZUS-MH MR-2 User Manual, the model can be used to estimate losses for a group of similar buildings. However, nominally similar buildings have experienced greatly different damages and losses during a hurricane. This could be due to factors such as the structural quality of construction, enforcement of building codes, and lack of maintenance. Where construction quality is known to be different from the defined norms in the HAZUS-MH model, larger uncertainties in loss projections can occur. Default data were obtained at the national level, and did not include recently constructed building stock and critical facilities after 2000, or contain actual replacement cost values for the building stock. Estimates used are based on a replacement value established with R.S. Means, per the HAZUS-MH Hurricane Wind Technical Manual. This does not reflect actual repair costs that can fluctuate with the economy at the time of the disaster. Local Inventory Data Local inventory data were considered to update the general building stock and critical facilities databases, but minimal data were received. Sample general building stock characteristics and damage data were collected in prior validation studies for Hurricanes Charley (TO – 332) and Ivan (TO – 348) for residential and commercial structures and public and critical facilities. Region I. Southeast Florida Region II. South Florida Region III. Middle Florida Region IV. North Florida Region I. Southeast Florida Region II. South Florida Region III. Middle Florida Region IV. North Florida Local Inventory Data Assumptions and Limitations Sufficient data were not received to update the general building stock or critical facilities inventory. Prior validation study site assessments did not represent statistically significant samples, for which general building stock conditions could be extrapolated at the jurisdictional or county level. There were not enough building characteristic data to fully update the site specific critical facilities for the micro analysis. No critical facilities data were collected for Charlotte County after Hurricane Charley. There is subjectivity in characterizing damage during field observations. Efforts were made to reduce this subjectivity, by providing descriptions of damage states to field inspectors. However, uncertainty is always possible. 2.0 DATA COLLECTED Per the scope of work, this validation study was to include Hurricane Ivan for Escambia County and Hurricane Charley for Charlotte County. To illustrate the regional comparison of HAZUSMH estimates with observed data for Hurricane Charley, results were included for these additional six counties: DeSoto, Hardee, Lee, Polk, Orange, and Osceola Counties. Data were also collected but not presented in this report include six additional counties for Hurricane Frances (i.e., Brevard, Indian River, Martin, Okeechobee, Palm Beach, and St. Lucie Counties), and two counties for Hurricane Jeanne (i.e., Martin and St. Lucie Counties), and one county for Hurricane Ivan (i.e., Santa Rosa). Data were collected for Hurricane Dennis for Escambia County, per the scope of work. Data collected for Hurricane Dennis was not useable as much of the structures were damaged during Ivan and not all had been repaired when Dennis impacted these same counties. Therefore, the default HAZUS-MH building stock inventory would still include the structures as they were initially constructed. Nor, would the default HAZUS-MH building stock inventory have been updated to include structural repairs or reconstruction to current code. Therefore, it was prudent to compare the HAZUS-MH results with observed data in Escambia County for Hurricane Ivan to reduce the margin of error. Data aggregated to the county, jurisdictional, and site levels were requested from various sources for this study for Hurricanes Charley and Ivan. Data were requested from local, state, and federal agencies and organizations during November 2005 through March 2006. Readily available data were received by August 2006. Data collected included hazards data, physical damage to structures and contents, social impacts, and economic loss. Data were also retrieved from prior HAZUS validation studies conducted in 2004 for Hurricanes Charley (TO – 332) and Ivan (TO – 348). The data collection form that was sent to local governments is provided in Appendix A. At the local level, data were provided by DeSoto, Escambia, and Hardee Counties. At the state level, data were provided by the Florida Department of Insurance Regulation (FDOIR) and the Florida Hospital Association (FHA). At the federal level, data were compiled from the American Red Cross (ARC), FEMA HMTAP task orders such as local and regional HAZUS validations, and from the FEMA Public Assistance (PA) program. Private sector level data for the hurricane windfield were provided by ARA. Data used in this validation study includes Charlotte, DeSoto, Hardee, Lee, Polk, Orange, and Osceola Counties for Hurricane Charley, and Escambia County for Hurricane Ivan. Table 4 summarizes which data have been received, its source, identifies the gaps, and summarizes the usefulness of the data for this validation study. Therefore, only the categories with green cells would allow any comparisons. Appendix B includes all data that were collected, and describes the observed data, data source(s), limitations, and modeling usefulness of data that was collected for this study, as summarized in Table 5. Table 4. Summary of Data Received and Used for Validation Study Data Received and Charley Ivan Charlotte DeSotoHardeeLeeOrange OsceolaPolkEscambiaSource Usefulness for Validation Residential Qualitative Damage: American Red Cross (ARC) Preliminary Damage Assessment (PDA) Yes, but includes flood damage for Ivan. County Damage Assessment ------- No, data are very similar to ARC data, which is used instead. Critical Facility Damage: County Damage Assessment ------Yes for schools. Post-disaster Short-term Shelter Population ARC Yes. Residential Economic Loss: Florida Department of Insurance Regulation claims (FDOIR) Yes. Residential, Commercial, and Industrial Economic Loss: Insurance Services Office, Inc. (ISO) Yes. Public Building Economic Loss: FEMA Public Assistance (PA) No,includes flood damage for Ivan. Hospital Economic Loss: Florida Hospital Association (FHA) ----Yes. School Economic Loss: County -------Yes. Debris Generated County Estimate ------No, estimates are not final. FEMA PA --------Not available. Site Specific Damage: HAZUS Validations ----Yes. Site Specific Economic Loss: Escambia County School Board ----Yes. Hospital Loss of Functionality: FHA ---Yes. = Data Received Table 5. Summary of Data Received for Validation Study Data Received and Source Charley Dennis Frances Ivan Jeanne CharlotteDeSotoHardeeLeeOrangeOsceolaPolkEscambia Santa RosaBrevardIndian River MartinOkeechobee Palm Beach St. LucieEscambia Santa RosaMartinSt. Lucie Residential Qualitative Damage ARC PDA -------- County ------------------ Commercial Qualitative Damage ------------------- Industrial Qualitative Damage ------------------- Critical Facility Qualitative Damage Medical Care Facilities ------------------- Schools (County) ------------------ Fire Stations (County) ----------------- Police Stations ------------------ Shelters (County) ----------------- Critical Facility Site Damage (Prior HAZUS Validations) ------------ Critical Facility Loss of Function Medical Care Facilities (FHA) ------------- Schools ------------------- Fire Stations ------------------- Police Stations ------------------- Shelters ------------------- Displaced Households (ARC) Residential Economic Loss Wind (FL DOIR claims) Commercial and Industrial Economic Loss (ISO) Critical Facility Economic Loss Medical Care Facilities (FHA) ------------- Schools (county) ------------------ Fire Stations ------------------- Police Stations ------------------- Shelters (county) ------------------ Public Building Economic Loss (FEMA PA) Debris Generated (County) ---------------- Injuries and Deaths (County) ------------------ Road Damage (FDOT) ----------------- = Data Collected FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY 3.0 ANALYSES RESULTS AND OBSERVATIONS HAZUS-MH results for the macro level are presented by: 1. residential damage 2. critical facilities damage 3. shelter populations 4. residential, commercial, and industrial economic loss 5. public and critical facilities economic loss HAZUS-MH results for the micro level are provided for: 1. Site specific critical facilities damage 2. Site specific critical facilities economic loss 3. Wind damage curve for site specific critical facilities. 4. Wind loss curve for site specific critical facilities. HAZUS-MH results are also provided for hospital loss of functionality. Each of these categories of results presents observations about the level of agreement between the HAZUS-MH results and the observed data. Observed data were compared with three HAZUS-MH estimates. 1. The ARA wind speeds were decreased by 10 percent to model the lower-end HAZUS estimate. 2. The ARA wind speeds were used to model the optimum HAZUS estimate. 3. The ARA wind speeds were increased by 10 percent to model the upper-end HAZUS estimate. The icons below are used as a visual guide to represent the level of agreement in this section. HAZUS-MH underestimates impacts in comparison with observed data. The observed data are greater than the upper-end results. Observed data agrees with HAZUS low-end results. The range includes numbers that fall within the bounds between the low-end results, and the midpoint of the low-end results and the optimum results. Observed data has good agreement. The range includes numbers that fall within the bounds between the midpoint of the low-end results and the optimum results, and the midpoint of the optimum results and the upper-end results. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Observed data agrees with HAZUS upper-end results. This data includes the numbers that fall within the bounds between the midpoint of the optimum results and the high-end results, and the upper-end results. HAZUS-MH overestimates impacts in comparison with observed data. The observed data are less than the lower-end results. 3.1 HAZUS-MH Macro Level Results for Aggregate Inventory at the County Level Hurricane Charley and Ivan Results and Observations Residential Damage HAZUS-MH was used to obtain wind-damaged building-count estimates for residential structures (i.e., all residential general building stock occupancy classes in HAZUS-MH). HAZUS-MH results were compared with ARC (American Red Cross) Preliminary Damage Assessment (PDA) data, as this data set included qualitative damage estimates. HAZUS-MH estimated damage counts for each qualitative damage state (e.g., minor, moderate, severe, and destroyed) and were compared with ARC data. ARC data were the best available data that was received for the study region. ARC PDAs are collected by windshield survey performed by Red Cross volunteers for the purpose of estimating shelter needs and are not detailed site inspections. PDA data might under- or over-estimate the damage counts for Hurricanes Charley and Ivan. Ivan damage was due to hurricane wind and coastal flooding, which is not differentiated in the PDA data. The following number of structures was inaccessible during the PDA data collection process: Charlotte – 200, Escambia – 684, Hardee – 461, and Polk – 5,000. Residential damage comparisons are provided in Table 6a for the Hurricane Charley Counties and Table 6b for Escambia County. In addition, not all County ARC estimates included estimates for buildings that had minor damage; therefore, the total damaged structures for those Counties will not be readily comparable for the minor damage state or total damage. In HAZUS-MH, lower wind speeds resulted in more minor damage and fewer moderate, severe, and destroyed structures. Higher wind speeds resulted in fewer structures estimated to have minor damage and more moderate, severe, and destroyed structures. ARC PDA qualitative damage definitions are very comparable to those in HAZUS-MH. Detailed ARC PDA qualitative damage definitions are included in Appendix B. For reporting purposes, the following correlation between damage states has been established for comparison of HAZUS-MH estimates with observed ARC PDA data: ARC PDA HAZUS-MH Damage State Affected = Minor Damage Minor Damage = Moderate Damage Major Damage = Severe Damage Destroyed = Destruction April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 6a. Hurricane Charley Residential Damage - HAZUS-MH Wind Results HAZUS Damage HAZUS-MH Validation County DamageState Lower- End Optimum HAZUS Estimate Upper- End ARC PDA (wind and flood) Minor 20,499* 19,027* 12,963* 8,373 x Moderate 10,128 17,478 18,240 12,457 x Severe 2,031 7,644 15,004 12,006 x Destroyed 722 3,109 8,274 5,013 x Charlotte Total 33,380 47,258 54,481 37,849 x Minor 2,922 2,792 2,024 0** x Moderate 1,578 2,826 2,999 2,020 x Severe 262 1,108 2,213 2671 x Destroyed 130 593 1,489 3644 x DeSoto Total 4,892 7,319 8,725 8,335 x Minor 1,627 2,424 2,473 1,578 x Moderate 493 1,305 2,093 2,488 x Severe 63 331 877 1,052 x Destroyed 40 225 682 367 x Hardee Total 2,223 4,285 6,125 5,485 x Minor 8,174 14,067 21,187 9,648 x Moderate 2,898 5,634 9,111 6,817 x Severe 691 2,090 4,017 654 x Destroyed 230 908 2,377 331 x Lee Total 11,993 22,699 36,692 17,450 x Minor 9,593 26,495 48,698 0** x Moderate 1,185 5,096 14,366 2,036 x Severe 23 197 1,268 177 x Destroyed 4 80 501 2 x Orange Total 10,805 31,868 64,833 2,215 x Minor 2,840 7,435 12,923 0** x Moderate 361 1,549 4,158 149 x Severe 9 85 491 487 x Destroyed 3 46 250 137 x Osceola Total 3,213 9,115 17822 773 x Minor 7,209 15,136 22,223 0** x Moderate 1,393 4,825 10,454 1,700 x Severe 102 681 2,401 1,782 x Destroyed 146 751 2,380 2,012 x Polk Total 8,850 21,393 37,458 5,494 x Minor 52,864 87,376 122,491 19,599 x Moderate 18,036 38,713 61,421 27,667 x Severe 3,181 12,136 26,271 18,829 x Destroyed 1,275 5,712 15,953 11,506 xTotal for Region Total 75,356 143,937 226,136 77,601 x *Lower wind speeds resulted in more minor damage and fewer moderate, severe, and destroyed structures. Higher wind speeds resulted in fewer minor damage and more moderate, severe, and destroyed structures. **DeSoto, Orange, Osceola, and Polk County ARC estimates did not include estimates for buildings that had minor damage. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 8 illustrates the geographic trend of the comparison of the modeled and observed results. HAZUS-MH predicted the damage type that actually occurred, more accurately in counties that were located closer to the point of the landfalling hurricane. Figure 8. Hurricane Charley Residential Damage (Total Damaged Building Count) HAZUS-MH Wind Results April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 6b. Hurricane Ivan Residential Damage - HAZUS-MH Wind Results HAZUS Wind Damage HAZUS-MH Validation County Damage State Lower- End Optimum HAZUS Estimate Upper- End ARC PDA (wind and flood damage) Minor 8,702 21,131 33,765 29,898 x Moderate 1,073 4,385 11,539 22,926 x Severe 22 196 1,204 9,385 x Destroyed 2 50 336 5,224 x Escambia Total 9,799 25,762 46,844 67,433 x Observations: • Minor damage data was not collected for DeSoto, Orange, Osceola, and Polk Counties. The HAZUS-MH estimates were in good agreement for Charlotte, Hardee, and Lee Counties. • The observed damage data, which are based on windshield surveys performed by Red Cross volunteers, might under- or overestimate the damage states through study region. • There were 5,000 inaccessible structures in Polk County, which is nearly equal to the number of structures that were reported to be damaged. It is likely that this data limitation tainted the comparison for Polk County, but it did not seem to affect the level of agreement for the study region. The study region level of agreement was computed without Polk County and there was still a 60 percent level of good agreement. • It appears that HAZUS-MH significantly underestimated damage in Escambia County. This is due to the fact that Hurricane Ivan caused coastal flood damage, and this model only estimates wind damage. Based on a comparison of ARC PDA data and National Flood Insurance Program (NFIP) data, it appears that about 10 percent of the damage was attributed to flood damage. However, NFIP data only includes claims for insured properties. • It looks like HAZUS also did a better job of predicting losses when the damage is severe or destroyed and was not as good at damage predicting when the loss was minor. However, it is important to consider that not all County ARC estimates included estimates for buildings that had minor damage. Therefore, it is not appropriate to directly compare the HAZUS-MH estimated number of structures with minor damage with the ARC data for minor damage. • It appears that the model did a better job closer to the landfall spot then it did inland as the windfield decays. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Critical Facility Damage HAZUS-MH was used to obtain hurricane wind damage building count estimates for Escambia County schools. Comparisons were attempted for damage for Escambia County schools, fires stations, and shelters. However, HAZUS-MH does not provide damaged building counts by qualitative damage state for critical facilities. The damaged building count is generated from the general building stock for general occupancy classes, but not specific occupancy classes. For that reason, results could not be generated for fire stations and shelters. Results were provided for the damaged building count for schools, using the estimates provided for the general occupancy for schools (EDU). It was determined that the schools included in the EDU general occupancy class only contained grade schools (EDU1). Therefore the observed data, which was of the same occupancy class, was suitable for comparison. Observations: • The HAZUS-MH model consistently and significantly underestimated the damage to the schools. This underestimate is due to the underestimated educational (EDU) building stock. The HAZUS-MH default inventory contains 25 schools in Escambia County. There were 72 schools for which damage was reported for by the county. The total number of schools in the county was not reported. Shelter Population HAZUS-MH was used to obtain hurricane wind estimates for short term post- landfall shelter population and compared with ARC data collected at the hurricane shelters. Pre- landfall shelter population is included to show the perception of the evacuees needs to evacuate. Post-landfall shelter population comparisons are provided in Table 7a for the Hurricane Charley Counties and Table 7b for Escambia County. Table 7a. Hurricane Charley Shelter Demand - HAZUS-MH Wind Results County Shelter Needs HAZUS Short Term ARC Pre- landfall Shelter Population ARC Post- landfall Shelter Population Shelter Capacity HAZUS-MH Validation Lower- End Optimum HAZUS Estimate Upper- End Charlotte 652 2,475 5,617 425 425 1,500 x DeSoto 102 453 1,081 1,374 1,400 4,245 x Hardee 26 140 415 537 537 1,267 x Lee 142 474 1,044 8,129 1,191 17,768 x Orange 104 406 1,112 727 727 6,320 x Osceola 29 109 321 1,714 102 10,284 x Polk 64 327 1,072 3,390 14 11,172 x Total for Region 1,055 4,057 9,590 12,906 4,382 41,384 x April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 9 illustrates the geographic trend of the comparison of the modeled and observed results. HAZUS-MH predicted the shelter requirements more accurately for the inland counties. Figure 9. Hurricane Charley Shelter Demand - HAZUS-MH Wind Results April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 7b. Hurricane Ivan Shelter Demand - HAZUS-MH Wind Results County Shelter Needs HAZUS Short Term ARC Pre- landfall Shelter Population ARC Post- landfall Shelter Population Shelter Capacity HAZUS-MH Validation Lower- End Optimum HAZUS Estimate Upper- End Escambia 72 252 698 6,707 978 16,827 x Observations: • HAZUS-MH short-term shelter estimates were in very good agreement at the regional level and for Orange and Osceola Counties in the Hurricane Charley Study region. Shelter needs were underestimated for DeSoto, Hardee, and Lee Counties, and overestimated for Charlotte and Polk Counties. It is possible that people evacuated away from Charlotte County to DeSoto, Hardee, and Lee because they were located farther away from where Charley was predicted to make impact. • The pre-landfall and the post-landfall population for Charlotte County are the same, and the shelter was operating at less than 30 percent of its capacity. It is important to note that the observed data for shelter use was low in comparison to shelter estimates developed in the Hurricane Evacuation Studies (HES). As stated in the FEMA 2004 Hurricane Assessments Executive Summary, low shelter use was due to low evacuation participation. It is also likely that people needing shelter in one county may have sought shelter in adjacent counties, or stayed with family or friends, or at a hotel. • It appears that HAZUS-MH underestimated the short term shelter needs in Escambia County. Observed shelter data could include transient or tourist populations, which are not reflected in the HAZUS-MH default inventory, although they can be added where the data are available. People may have stayed in shelters after landfall because they could not return to their homes due to electric, water, and sewage outages. This could explain why HAZUS-MH underestimated the shelter need for Escambia County. This could also be related to the fact that Ivan was also a coastal flood event. • HAZUS-MH estimates short term shelter needs based on damaged building stock. It is also possible that the estimates did not agree well with the observed data due to uncertainties in the default building characteristic data. • The behavior of whether people who chose to evacuate could have been caused by their perception of the need to evacuate based on experience with prior hurricanes. For example, people in Escambia County had possibly experienced hurricane impacts during the 1990’s (i.e., Erin and Opal), and were more likely to evacuate. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Residential Economic Loss HAZUS-MH was used to obtain wind economic loss estimates for residential structures found in the general building stock occupancy classes given in HAZUS-MH. HAZUS-MH estimated economic loss was compared with FDOIR as well as ISO wind insurance claim data. These insured-loss estimates provide a useful benchmark for the HAZUS-MH wind loss estimate comparisons. However, FDOIR and ISO data cannot be directly compared with HAZUS-MH estimates. FDOIR and ISO insurance claims data includes losses for automobiles, boats, and appurtenant structures, and additional living expenses, yet do not include deductibles or uninsured properties. The FDOIR and ISO data includes fewer inventories (i.e., fewer structures) than HAZUS, as it only includes insured properties. However, FDOIR and ISO data include more property (e.g., automobiles and boats) than is included in HAZUS. In addition, the ISO claims data do not include manufactured housing. As such, the total economic loss for the residential building stock less the manufactured housing loss was compared with the ISO data. Raw FDOIR data were compared with HAZUS-MH estimates, as the penetration rate (i.e., the percent of structures that are insured) was not provided. However, since the penetration rate was provided for ISO data, the ISO raw data were converted for comparison. The penetration rate for the ISO data was 45 percent. Therefore the ISO losses were converted to 100 percent by multiplying the ISO raw losses by 2.22. Residential loss comparisons using FDOIR data are provided in Table 8a for the Hurricane Charley Counties and Table 8b for Escambia County. Residential loss comparisons using ISO data are provided in Table 8c for the Hurricane Charley Counties and Table 8d for Escambia County. Table 8a. Hurricane Charley Residential Economic Loss (FDOIR) - HAZUS-MH Wind Results County HAZUS Residential Loss ($M) FDOIR Data ($M) HAZUS-MH Validation Lower-End Optimum HAZUS Estimate Upper-End Charlotte 884 2,322 4,317 2,561 x DeSoto 109 300 555 283 x Hardee 42 130 285 138 x Lee 362 874 1657 1,014 x Orange 311 716 1597 991 x Osceola 77 183 425 560 x Polk 150 430 1045 554 x Total for Region 1,935 4,955 9,881 6,101 x Note: HAZUS-MH inventory includes more buildings. FDOIR claims include fewer buildings, but also include losses for automobiles, boats and appurtenant structures, and additional living expenses. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 10 illustrates the geographic trend of the comparison of the modeled and observed results. HAZUS-MH modeled residential economic loss results were in good agreement with FLDOIR observed data for six of the seven counties. Figure 10. Hurricane Charley Residential Economic Loss (FLDOIR) - HAZUS-MH Wind Results April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 8b. Hurricane Ivan Residential Economic Loss (FDOIR) – HAZUS-MH Wind Results County HAZUS Residential Loss ($M) FDOIR Data ($M) HAZUS-MH Validation Lower-End Optimum HAZUS Estimate Upper-End Escambia 205 448 968 1,698 x Note: HAZUS-MH inventory includes more buildings than insurance data. FDOIR claims include fewer buildings, but also include losses for automobiles, boats and appurtenant structures, and additional living expenses. Table 8c. Hurricane Charley Residential Economic Loss (ISO) – HAZUS-MH Wind Results County HAZUS Residential Loss ($M) ISO Data ($M) HAZUS-MH Validation Lower-End Optimum HAZUS Estimate Upper-End Charlotte 809 2,139 4,020 1,763 x DeSoto 88 239 448 87 x Hardee 37 113 244 53 x Lee 340 823 1,562 384 x Orange 308 709 1,579 591 x Osceola 74 174 401 226 x Polk 121 341 835 226 x Total for Region 1,777 4,538 9,089 3,330 x Note: HAZUS-MH inventory includes more buildings. ISO claims include fewer buildings, but also include losses for automobiles, boats and appurtenant structures, and additional living expenses. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 11 illustrates the geographic trend of the comparison of the modeled and observed results. HAZUS-MH modeled residential economic loss results were in good agreement with ISO observed data for six of the seven counties. Figure 11. Hurricane Charley Residential Economic Loss (ISO) - HAZUS-MH Wind Results April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 8d. Hurricane Ivan Residential Economic Loss (ISO) – HAZUS-MH Wind Results County HAZUS Residential Loss ($M) ISO Data ($M) HAZUS-MH Validation Lower-End Optimum HAZUS Estimate Upper-End Escambia 200 435 933 895 x Note: HAZUS-MH inventory includes more buildings. ISO claims include fewer buildings, but also include losses for automobiles, boats and appurtenant structures, and additional living expenses. Observations: • It appears that HAZUS-MH estimates were in very good agreement with the observed residential economic loss data for Charlotte County. • It appears that HAZUS-MH significantly underestimated residential economic loss for Escambia County. It is likely that the underestimation occurred, because Escambia County experienced economic loss from both hurricane wind and coastal flood. It is also possible that the estimates did not agree well with the observed data due to uncertainties in the default inventory and building characteristic data. Commercial and Industrial Economic Loss HAZUS-MH was used to obtain wind economic loss estimates for commercial and industrial structures found in the general building stock occupancy classes in HAZUS-MH. HAZUS-MH estimated economic loss was compared with ISO wind insurance claim data. These insured-loss estimates provide a useful benchmark for the HAZUS-MH wind loss estimates. The commercial and industrial losses are combined for comparison with ISO data, as provided in Table 9a for the Hurricane Charley Counties and Table 9b for Escambia County. Table 9a. Hurricane Charley Commercial and Industrial Economic Loss (ISO) - HAZUSMH Wind Results County HAZUS Commercial and Industrial Loss ($M) ISO Data ($M) HAZUS-MH Validation Lower-End Optimum HAZUS Estimate Upper-End Charlotte 118 340 609 202 x DeSoto 20 60 107 24 x Hardee 6 20 42 18 x Lee 25 71 145 64 x Orange 26 115 367 131 x Osceola 6 23 78 47 x Polk 18 75 208 67 x Total for Region 219 704 1,556 553 x April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 12 illustrates the geographic trend of the comparison of the modeled and observed results. HAZUS-MH modeled commercial and industrial economic loss results were in good agreement with ISO observed data for all seven counties. Figure 12. Hurricane Charley Commercial and Industrial Economic Loss (ISO) - HAZUSMH Wind Results April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 9b. Hurricane Ivan Commercial and Industrial Economic Loss (ISO) - HAZUS- MH Wind Results County HAZUS Commercial and Industrial Loss ($M) ISO Data ($M) HAZUS-MH Validation Lower-End Optimum HAZUS Estimate Upper-End Escambia 13 53 163 311 x Observations: • HAZUS-MH is in good agreement for every county in the Hurricane Charley study region. However, HAZUS-MH underestimated loss for Escambia County. • Perhaps the reason the model appears to underestimate the results for Escambia County is that the building inventory is not as accurate as it is in southwest Florida. Public and Critical Facilities Economic Loss Public Facilities Economic Loss HAZUS-MH was used to obtain economic loss estimates for public and critical facilities. HAZUS-MH results for government buildings, hospitals, and schools (i.e., GOV1, GOV2, COM6, and EDU1 in the general building stock) were compared with a summary of PA funds that were aggregated at the county level. PA funding was generated from NEMIS for Category “E” Public Facilities for structure and content loss. Category E covers uninsurable losses to repair or restore publicly owned and maintained structures, equipment (e.g., electrical, mechanical, telecommunications, etc.), and contents (e.g., furniture, books, computers, etc.). PA loss amounts could be overestimated or underestimated, as insurance claims were still being settled by local governments at the time that this report was prepared. PA loss amounts might also include associated debris removal and mold remediation costs, or costs for code and standard upgrades. Observations: • It appears that HAZUS-MH underestimated the economic loss of public and critical facilities in comparison to the PA data for most of the Hurricane Charley study region and for Escambia County. • This underestimation appears to be due to the fact that Hurricane Ivan caused both hurricane wind and coastal flood economic loss. In comparing the optimum HAZUSMH results with the observed PA data, the Hurricane Charley study region results were underestimated by 62 percent, whereas, Escambia County results were underestimated by 800 percent. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY • This underestimation appears to be caused by the fact that the default general building stock is incomplete. For example, the grade school occupancy class (i.e., EDU1) does not include all of the schools that are in the critical facilities default inventory. It is assumed that the default inventory is also inaccurate for the governmental and hospital structures. • It is also plausible that the underestimation occurred because of an underestimation of content value. For example, hospitals are considered commercial structures, for which HAZUS-MH estimates the content value to be between 100 to 150 percent of the replacement value of the structure. As such, if the default inventory is not accurate for replacement value for commercial structures, HAZUS-MH would not be able to accurately calculate the content replacement cost. Hospital Economic Loss HAZUS-MH was used to obtain hurricane wind economic loss estimates for hospitals. HAZUSMH results for hospitals (i.e., COM6 in the general building stock) were compared with losses that were reported by the FHA, which included loss to structures, equipment, contents, and “other damages” as listed in the following observations. Observations: • It appears that HAZUS-MH underestimated economic loss for hospitals for both Charlotte and Escambia Counties. • It is possible that the estimates did not agree well with the observed data, as it included losses for items not modeled in HAZUS-MH. The FHA-observed data included costs for non-structural damage, referred to as “other damages” (e.g., debris removal, signage, landscaping, fencing, screens, canopies and awnings, and compressors). FHA reported that of $67.4 million in losses for all 2004 hurricanes, $10.9 million or 16 percent of economic loss was attributed to other damages. • It is possible that the estimates did not agree well with the observed data due to uncertainties associated with missing facilities or the assumptions made for the default building characteristic data in the general building stock for commercial structures. • It is also plausible that the underestimation occurred because of an underestimation of content value. Hospitals are considered commercial structures, for which HAZUSMH estimates the content value to be between 100 to 150 percent. As such, if the default inventory is not accurate for commercial structures, HAZUS-MH would not be able to accurately calculate the content replacement cost. School Economic Loss HAZUS-MH was used to obtain hurricane wind economic loss estimates for schools. HAZUSMH results for schools (i.e., EDU1 in the general building stock) were compared with losses that April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY were reported by Escambia County. No data were received for southwestern counties that were impacted by Hurricane Charley. Observations: • It appears that HAZUS-MH significantly underestimated economic loss for schools for Escambia County. • It is possible that the estimates did not agree well with the observed data due to uncertainties associated with missing facilities or the assumptions made for the default building characteristic data in the general building stock. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY 3.2 HAZUS-MH Micro Level Results for Site Specific Critical Facilities Site Specific Damage HAZUS-MH was used to obtain hurricane wind damage estimates for site specific critical facilities in Escambia County. Data were not received for critical facility damage in Charlotte County. Site specific comparisons are provided in Table 10a for Hardee, Osceola and Polk Counties, and Table 10b for Escambia County. Table 10a. Hurricane Charley Damage - HAZUS-MH Site Specific Hurricane Wind Results Damage State HAZUS-MH Validation County Type of Facility Facility Name Lower- End Optimum HAZUS Estimate Upper- End Observed Wind Damage Hardee School Hardee High School Minor Severe Severe Minor x Hardee School Wauchula Elementary Minor Severe Severe Minor x Osceola School Poinciana High School Minor Minor Minor Moderate x Osceola School Thacker Elementary Minor Minor Minor Minor x Polk Fire Station Haines City Fire Dept. Minor Minor Moderate Minor x Table 10b. Hurricane Ivan Damage - HAZUS-MH Site Specific Hurricane Wind Results Damage State HAZUS-MH Validation County Type ofFacility Facility Name Lower- End Optimum HAZUS Estimate Upper- End Observed Wind Damage Escambia Hospital Baptist Hospital Minor Minor Minor Minor x Escambia Hospital Naval Hospital Pensacola Minor Minor Minor Minor x Escambia Hospital Sacred Heart Health System Minor Minor Minor Minor x Escambia Hospital West Florida Hospital Minor Minor Minor Severe x Escambia School Bellview Elementary School Minor Minor Minor Minor x Escambia School Longleaf Elementary Minor Minor Minor Severe x Escambia School Pine Forest High School Minor Minor Minor Moderate x Escambia School West FL High School/Beggs Educational Center Minor Minor Minor Moderate x April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figures 13 and 14 illustrates the geographic trend of the comparison of the modeled and observed results for the Hurricane Charley and Hurricane Ivan study regions, respectively. The maps show the levels of agreement of HAZUS-MH damage estimates with the observed damage estimates for the site specific critical facilities. Figure 13. Hurricane Charley Damage - HAZUS-MH Site Specific Wind Results April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 14. Hurricane Ivan Damage - HAZUS-MH Site Specific Wind Results Observations: • HAZUS-MH estimates were in good agreement with 80 percent of the observed damage states for the counties in the Hurricane Charley study region, and were in good agreement with 50 percent of the observed damage to the critical facilities in Escambia County even though, HAZUS-MH was not intended to be used for site- specific assessments. • HAZUS-MH modeled damage results were generally in good agreement with the observed damage for critical facility sites that were located along the path of the hurricane track. However, damage to facilities that were located farther away from the hurricane track were also compared for Hurricane Ivan to determine if there is any fluctuation in the modeling accuracy, depending on the proximity of the facilities to April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY the hurricane track. There was a lot more variance in the agreement of the HAZUSMH estimates in comparison with the observed data for the Hurricane Ivan study region, than there was for the Hurricane Charley study region. • All three hazard scenarios produced the same level of damage for all sites in Escambia County. It appears that varying the wind speed did not have an effect on the damage states at the site-level, as it did at the aggregate level. As such, it does not appear that HAZUS-MH is valid for site-specific analysis. Wind Damage Curve Comparison for Site Specific Critical Facilities The type of observed critical facility damage (e.g., minor, moderate, etc.) provided by the Escambia County School Board was compared with the probability of that same of type of damage predicted by the HAZUS-MH wind damage curve. The damage curve shows the likelihood of the type of qualitative damage (e.g., minor) associated with a particular wind speed for each model building type, expressed as a percentage. These model building types are spelled out in the Abbreviations List at the beginning of this report. Tables 11a and 11b show the observed damage compared with the probability of that predicted type of damage for 13 critical facilities in Hardee, Osceola, Polk, and Escambia Counties. For example, the first data record in Table 12a shows that HAZUS-MH estimated there to be a 60 percent chance that Hardee High School would experience minor damage. Table 11a. Hurricane Charley Damage - Comparison of Observed Site Specific Critical Facility Damage and HAZUS-MH Wind Damage Curve County Type of Facility Facility Name HAZUSMH Wind Class Peak Gust Observed Wind Damage HAZUS-MH Wind Damage Curve Estimated Percent of Type of Building Damage Level of Agreement Hardee School Hardee High School MECBL 133 Minor 60% - minor Agreement at Low End Hardee School Wauchula Elementary SECBL 133 Minor 60% - minor Agreement at Low End Osceola School Poinciana High School MLR1 107 Moderate 55% - moderate Agreement at Low End Osceola School Thacker Elementary MECBL 104 Minor 19% - minor Underestimates Polk Fire Station Haines City Fire Dept. MECBL 107 Minor 20% - minor Underestimates April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 11b. Hurricane Ivan Damage - Comparison of Observed Site Specific Critical Facility Damage and HAZUS-MH Wind Damage Curve HAZUS-MH Wind Damage Curve Level of Agreement Agreement at Low End County Escambia Type ofFacility Hospital Facility Name Baptist Hospital HAZUSMH Wind Class MECBM Peak Gust 109 Observed Wind Damage Minor Estimated Percent of Type of Building Damage 50% - minor Escambia Hospital Naval Hospital Pensacola CECBM 96 Minor 20% - minor Underestimates Escambia Hospital Sacred Heart Health System SECBM 106 Minor 40% - minor Underestimates Escambia Hospital West Florida Hospital SECBH 106 Severe 0% - severe Underestimates Escambia School Bellview Elementary School MECBL 109 Minor 25% - minor Underestimates Escambia School Longleaf Elementary CECBL 108 Severe 0% - severe Underestimates Escambia School Pine Forest High School CECBL 113 Moderate 12% - moderate Underestimates Escambia School West FL High School/Beggs Educational Center CECBL 113 Moderate 12% - moderate Underestimates April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figures 15 and 16 illustrate the geographic trend of the comparison of the modeled and observed results for the Hurricane Charley and Hurricane Ivan study regions, respectively. The maps show the percent chance of the HAZUS-MH damage curve predicting the type of observed damage for each of the site specific critical facilities. Figure 15. Hurricane Charley Damage - Comparison of Observed Site Specific Critical Facility Damage and HAZUS-MH Wind Damage Curve April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 16. Hurricane Ivan Damage - Comparison of Observed Site Specific Critical Facility Damage and HAZUS-MH Wind Damage Curve Observations: • The HAZUS-MH wind damage curve more accurately predicted the type of damage, the closer the facility was located to the hurricane track in nearly all cases for Hurricane Charley. However, this was not the case for Hurricane Ivan. In Escambia County, the critical facilities were located 30 to 50 miles east of the hurricane track. There did not appear to be a correlation between accuracy of damage type estimates for facilities located closer to the path of hurricane. Critical facility damage that was farther away from the path was predicted with less accuracy than damage to facilities that were closer to the track. It is recommended that this be studied for future analysis, collecting more data for damage to critical facilities throughout the county. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY •HAZUS-MH wind damage curve estimates were in good agreement at the low end for 60 percent of the sites for Hurricane Charley, but underestimated 88 percent of the sites for Hurricane Ivan. Wind Loss Curve Comparison for Site Specific Critical Facilities The economic loss and economic loss ratios provided by the Escambia County School Board were also compared with the HAZUS-MH wind loss curve. The wind loss curve shows the percentage of economic loss associated with a particular wind speed for each modeled building type, expressed as a percentage. The economic loss ratio percentages generated by HAZUS-MH were multiplied by the replacement costs (provided by the School Board) to calculate the economic loss values. Table 12 shows the observed economic loss and loss ratios compared with HAZUS-MH estimates. Table 12. Hurricane Ivan Economic Loss - Comparison of Observed Site Specific Critical Facility Economic Loss and Loss Ratios and HAZUS-MH Wind Loss Curve Observed Economic Loss Reported by School Board ($) HAZUS-MH Wind Loss Curve County Typeof Facility Facility Name HAZUS- MH Wind Class Peak Gust Total Building Value (Structure and Contents) ($) Loss ($) Loss Ratio Estimated Building Economic Loss ($) Estimated Building Economic Loss Ratio Bellview Escambia School Elementary School MECBL 109 4,700,096 500,000 0.1064 58,751 0.0125 Longleaf Escambia Elementary School CECBL 108 15,454,013 10,000,000 0.6471 6,439 0.0004 Pine Forest High Escambia School School CECBL 113 18,755,925 1,000,000 0.0533 15,630 0.0008 West FL High School/Beggs Escambia School Educational Center CECBL 113 34,710,840 6,000,000 0.1729 28,926 0.0008 Observations: • HAZUS-MH wind loss curve estimates underestimated for all sites for Hurricane Ivan. April 2007 44 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 17. Hurricane Ivan Economic Loss - Comparison of Observed Site Specific Critical Facility Loss and HAZUS-MH Wind Loss Curve Vulnerability Reduction Measures for Mitigation Options Of the 13 site-specific critical facilities for which data were collected, two of them reportedly had shutters, and two had hurricane strapping. These were the only two vulnerability reduction measures that were provided in the available data. These facilities are commercial facilities. HAZUS-MH currently allows users to select two mitigation options for commercial facilities which include shutters, and the use of superior metal deck attachment for roofs. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY 3.3 HAZUS-MH Loss of Functionality for Hospitals HAZUS-MH was run to obtain hurricane wind induced loss of functionality estimates for hospitals for Hurricane Charley in Charlotte, DeSoto, Hardee, and Orange Counties and for Hurricane Ivan in Escambia County. An attempt was made to compare HAZUS-MH estimates with FHA data. The FHA provided loss of functionality by county, not by individual hospital. These are not direct comparisons since the observed data were aggregated by county and HAZUS-MH estimates were provided by hospital (e.g., FL000091). Comparisons are provided for Charlotte, DeSoto, Hardee, and Orange Counties in Table 13a and for Escambia County in Table 13b. Table 13a. Hurricane Charley Hospital Loss of Functionality - HAZUS-MH Site Specific Hurricane Wind Results County HAZUSLower-End EUpper-End Closed FHA - Days Closed HAZUS-MH Validation Optimum HAZUS Estimate stimated Days Charlotte 10* x FL000091 161 71 257 FL000195 96 31 175 FL000196 96 31 175 DeSoto 2* x FL000104 43 118 200 FL000108 52 138 223 Hardee 0 0 0 0 x Orange 0 x FL000071 1 5 20 FL000072 1 5 15 FL000074 1 7 27 FL000131 1 5 19 FL000138 1 2 7 FL000219 0 0 0 FL000229 0 0 0 * One hospital closed for each period of time. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 13b. Hurricane Ivan Hospital Loss of Functionality - HAZUS-MH Site Specific Hurricane Wind Results County HAZUSLower-End EUpper-End Closed FHA - Days Closed HAZUS-MH Validation Optimum HAZUS Estimate stimated Days Escambia 0 x FL000082 2 8 24 FL000183 1 5 13 FL000184 1 7 20 Observations: • Even at the aggregate level, HAZUS-MH appears to have overestimated the loss of functionality for hospitals. • Although there were several hospitals that did close, overall the hospitals remained functional. This is due in part to good building performance, operational capability from auxiliary power, and site design of a hospital to protect key “functional” areas from wind damage. • HAZUS-MH estimates loss of functionality based on building damage, such as damage to the roof and building openings. This is not a clear indication of complete loss of function, as part of the building could be damaged and the hospital could still be operational. The FHA reported that there was $46.3 million in economic loss to hospitals in the Hurricane Charley study region and Escambia County. Despite the significant economic losses that occurred, only two hospitals closed for brief periods of time. One hospital in Charlotte County closed for 10 days and one hospital in DeSoto County closed for two days. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY 4.0 CONCLUSION AND RECOMMENDATIONS The conclusions and recommendations are based on the evaluation of how the HAZUS-MH default building inventory was used to provide results and the observations about the comparison of HAZUS-MH estimates with observed historical data from past hurricane events. This section presents general and detailed conclusions and recommendations for post-disaster data collection, model improvements, and software functionality enhancements. To provide potential future HAZUS-MH improvements, the recommendations are relatively ranked and listed in order of importance under the data collection, modeling, and software functionality headers below. 4.1 Conclusions To achieve better agreement, the HAZUS-MH default inventory can be improved with more current and accurate data, especially for public and critical facilities, such as schools and government facilities. Also, the some of the observed data received for this study did not allow for direct comparisons with HAZUS-MH results, as data that were collected for past disasters did not differentiate between wind versus flood damage, or the data attributes needed for comparison with HAZUS-MH results were not always presented in a manner for direct comparisons. For example, ARC PDA data included damage from both wind and flood for Hurricane Ivan, and commercial qualitative damage data did not appear to have been collected. HAZUS- MH does not generate results that can be compared with the attributes of some of the damage and loss data that it is currently collected. For example, the damaged building count by qualitative damage states are not generated for all critical facilities, only those which are represented in the general occupancy class such as educational facilities (EDU). The observed data for the Hurricane Charley study region compared well with HAZUS-MH (1) residential qualitative damage; (2) residential, commercial, and industrial economic loss; and (3) short-term shelter demand estimates. There was better agreement at the regional level, as seen in the Hurricane Charley study region versus the results for one county, Escambia County, in the Hurricane Ivan study region. This appears to be due to several factors. It appears that the HAZUS-MH results consistently underestimated for Escambia County since Hurricane Ivan caused both flood and wind damage and loss. Another possible explanation, could be that the there is better default inventory data for the southwest Florida region versus the northwest Florida region. HAZUS-MH public and critical facilities qualitative damage (i.e., for schools) and economic loss estimates did not compare well with observed data. HAZUS-MH consistently and fairly significantly underestimated economic loss for public and critical facilities. This appears to be due to several factors. The HAZUS-MH underestimates may be attributed to the age and source of the critical facilities default inventory in HAZUS-MH, which is not as current as the general building stock, specifically the residential building stock. The residential building stock valuations are current as of 2005. However, the critical facilities data are current as of 2001. The critical facilities default data were collected from national and state data providers and likely does not reflect more accurate observed data that is available from local governments. Also, the HAZUS-MH general national building stock inventory is not current or accurate for those specific occupancy classes that were used for the critical facility comparisons (i.e., EDU for schools, GOV for public facilities, and COM6 for hospitals). Observed data such as FEMA PA data can include costs to repair the structure beyond its pre-disaster condition. For example, April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY additional funding can be provided to upgrade public facilities using 406 hazard mitigation, or upgrade to current building codes and design standards. Therefore, the public assistance economic losses will be greater than the HAZUS-MH estimates. The level of agreement between HAZUS-MH and observed data for site specific critical facilities varied. HAZUS-MH site specific qualitative damage estimates were in good agreement for 80 percent of the sites for Hurricane Charley and 50 percent of the sites for Hurricane Ivan. Considering that HAZUS-MH was designed to be used at a larger scale (i.e., region, county), it appears that the analysis showed that HAZUS-MH estimates compared reasonably well with the observed damage at the site specific level. However, site specific economic loss estimates were underestimated by HAZUS-MH. HAZUS-MH wind damage curve estimates were in good agreement at the low end for 60 percent of the sites for Hurricane Charley, but underestimated 88 percent of the sites for Hurricane Ivan. HAZUS-MH wind loss curve estimates underestimated for all sites for Hurricane Ivan. HAZUS-MH hospital loss of functionality estimates did not compare well with the observed data. The model significantly overestimated the loss of functionality (i.e., number of days). 4.2 Post-disaster Data Collection Recommendations Below are the challenges, recommendations, and benefits for future validation studies. The specific data collection challenges included: • Modeled estimates for a variety of categories such as: displaced households; debris; damage and loss at the jurisdictional level; damage to site specific facilities; and vulnerability reduction measures through mitigation options were not compared with observed data because the observed data was incomplete, unavailable, or incomparable. • Data for comparison with HAZUS-MH estimates are not always separated by hazard; and damage data are not collected for commercial, industrial, public, and critical facilities. • Vulnerability reduction measures were reviewed and data were insufficient to explore further vulnerability measures for the mitigation options in HAZUS-MH. Further research could be conducted to determine what measures are being used to reduce vulnerability. For example, post-disaster damage assessment could be conducted for structures that have been mitigation through the Hazard Mitigation Grant Program. • Currently there is no coordination with counties prior to hurricane season to determine what data are already being collected that are useful for future HAZUSMH validation studies and to determine which additional data are required for the local level. There is limited staff available to collect detailed HAZUS-MH data after an event. • After a disaster event, limited staff is available to collect detailed HAZUS- MH data, because of numerous response and recovery activities. As such, it was a strain on LTRO staff resources to compile the data needed for this validation study. If a county is severely impacted, it is suggested to send support staff to assist with data collection April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY efforts, as several of the counties contacted during this study had a limited number of staff that was very involved with response and recovery activities. Specific data collection and analysis recommendations include: • It is recommended that FEMA convene a focus group to determine the appropriate data to collect to enhance the default inventory and to document observed disaster impacts. The focus group should develop a pre-disaster local data gap analysis and data gathering coordination groups for enhanced planning and mitigation efforts. • It is recommended that FEMA encourage state and local jurisdictions to use the data collection template (refer to Appendix A) developed for this project to help consistently collect data. This data collection template should be examined by the focus group to ensure that all necessary data are being requested. • The value and usefulness of these validation studies requires more than readily data and could be focused for further study. Many data sets were incomplete, not available or too specific to be used to compare with an aggregate model. It is recommended that future validation studies should be conducted at a regional level, for counties that only experienced wind damage, and should include data for only wind impacts. • It is recommended that FEMA prepare prewritten rapid response HMTAP task orders for the types of HAZUS-MH data needed to be collected so deployment to collect the data is executed shortly after the event. • The benefit for future studies is that better data collection enhances the value and usefulness of HAZUS-MH to support all areas of disaster operations. 4.3 Model Improvement Recommendations The following were the model challenges for this validation: • When the study regions were exported, all of the results were not automatically exported with the study region. As such, sharing the results between HAZUS-MH users involved exporting individual results tables. • The damaged building count is generated from the general building stock for general occupancy classes, but not specific occupancy classes. Damage counts are not provided for critical facilities in HAZUS-MH. • The default general building stock for the grade school occupancy class (i.e., EDU1) does not include all of the schools that are in the critical facilities default inventory. • For the site-specific critical facilities analysis, all three hazard scenarios produced the same level of damage for all sites. It appears that varying the wind speed did not have an effect on the damage states at the site-level, as it did at the aggregate level. As such, further analysis should be conducted to determine whether HAZUS-MH is valid for site-specific analysis. Below are the recommendations for model improvements for future validation studies: April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY • Develop the capability to retain the results that have been run in HAZUS-MH when it is exported. • Develop the capability to provide damage counts for critical facilities, and economic loss for critical facilities at the site specific level. • Develop a process to permit local governments to submit updated building stock inventory for use with HAZUS-MH. • When possible, develop the functionality in HAZUS-MH models to display and export results the same way for all models. 4.4 Software Functionality Enhancement Recommendations The following were the software challenges for this validation: • Damage and loss estimates are not provided at the zip code, jurisdictional, or site specific levels by HAZUS-MH; this would improve the usefulness of the model for comparative analysis. • HAZUS-MH analysis parameters revert to the default parameters after an analysis is run. Therefore, it can not determine which parameters were selected for analysis after the analysis has been run. Also, when all parameters are not selected for analysis, HAZUS-MH does not always generate all the results specified. • Damaged building counts are not provided for specific occupancy classes or critical facilities. Below are the recommendations for future validation studies: • Develop the capability in HAZUS-MH to produce results at the zip code, jurisdictional, and site specific levels. Develop capability to select zip code or jurisdiction level data attributes from the map view. This functionality will allow for more detailed comparisons and validations of results. • Develop the capability in HAZUS-MH to identify which analysis parameters have been run, instead of the analysis screen reverting back to default analysis parameter settings. Also, develop the capability within HAZUS-MH to run individual analysis parameters. Although it appears that this can be done, sometimes the results are not provided consistently. • Develop the functionality to provide the damaged building count for specific occupancy classes and critical facilities in the wind model. This functionality will allow for more detailed comparisons and validations of results. • Develop the capability to increase or decrease wind speeds by a percentage, instead of having to enter this data manually, and have the option to modify the maximum radius winds and central pressure that corresponds with the wind speed modification. Automate this process instead of having to enter this data by census block or tract or through sequel server. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY • In the wind speed map legend that is generated by HAZUS-MH, develop breaks in the range that do not overlap. For example, instead of having the range of wind speeds overlap (<50, 50 – 65, 65 – 80, etc.), do not have them overlap (<50, 51 – 65, 66 – 80, etc.). Currently when the wind speed maps are created with the wind speed labels listed over the census tracts on the map, there are cases where different colored tracts show up with the same wind speed label. • Allow more standard ArcGIS features in HAZUS-MH such as ability to sum columns in attribute tables, without first having to export as a data layer. April 2007 FEMA HMTAP 440 – HURRICANE WIND MODEL VALIDATION STUDY 5.0 SOURCES Federal Emergency Management Agency (FEMA), 2006. HAZUS-MH Models. April 6, 2006. . Downloaded 4/19/06. National Climatic Data Center (NCDC), 2006. Billion Dollar U.S. Weather Disasters. March 15, 2006. . Downloaded 3/17/06. NCDC, 2004. Climate of 2004 Atlantic Hurricane Season. December 13, 2004. . Downloaded 3/20/2006. NCDC, 2004. Climate of 2005 Atlantic Hurricane Season. January 13, 2006. . Downloaded 3/20/2006. April 2007 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY APPENDIX A Templates for Collecting Damaged Data Appendix A provides data templates that were used to collect data. Data were collected for damages and losses to residential, commercial, and industrial structures and critical facilities. A-1. County-Level Data Collection Form Data were requested from Escambia and Santa Rosa Counties for Hurricanes Dennis and Ivan; and Charlotte, Hardee, DeSoto, and Orange for Hurricanes Charley, Frances, and Jeanne, using the following data collection form: Local Government Damage Survey Data Collection __________ County Hurricane ___________: Flood An analysis is being conducted to compare HAZUS-MH estimates with known damages and losses caused during the 2004 and 2005 hurricane seasons. Results will be used to improve the model’s accuracy. Please provide as much data that is available in the tables below, or in your spreadsheet format (if already available). Please list separately by residential, commercial, and industrial, as data are available. Also, please send Residential Substantial Damage Estimates and local National Flood Insurance Claims Payments data for flood damages for county and municipal damages. Please contact at for more information. Thank you. General Occupancy Damage and Loss The following clarifications pertain to residential, commercial, and industrial tables: (1) “Total Replacement Value” is the assessed value of structures in the county. (2) “Minor” includes minor damage to roofs, siding, decking, few broken windows, etc. (3) “Major” requires substantial repairs to house before it is safe for use. Repairs will take a few weeks. (4) “Destroyed” means total loss/must be demolished. April 2007 A-1 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY (5) “Business interruption loss” refers to disaster impacts such as revenue loss. 1.0 RESIDENTIAL STRUCTURES Counties Affected Total # of Residential Structures Total Replacement Value of Residential Structures (1) Total # Residential Structures Damaged Total # Residential Structures with Minor Damage (2) Total # Residential Structures with MajorDamage (3) Total # Residential Structures Destroyed (4) Economic Loss to Residential Buildings (repair cost) Unincorporated Add Cities, Towns Countywide Total 2.0 COMMERCIAL STRUCTURES Counties Affected Unincorporated Total # of Commercial Structures Total Replacement Value (1) Total # Commercial Structures Damaged Total # Commercial Structures with Minor Damage Total # Commercial Structures with Major Damage Total # Commercial Structures Destroyed (4) Economic Loss to Commercial Buildings Building & Content Damage Business Interruption Loss (5) Add Cities, Towns Countywide Total 3.0 INDUSTRIAL Counties Affected Unincorporated Total # of Industrial Structures Total Replacement Value (1) Total # Industrial Structures Damaged Total # Industrial Structures with Minor Damage Total # Industrial Structures with Major Damage Total # Industrial Structures Destroyed (4) Economic Loss to Industrial Buildings Building & Content Damage Business Interruption Loss (5) Add Cities, Towns Countywide Total April 2007 A-2 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY 4.0 INJURIES AND DEATHS Counties Affected Residential Injuries Commercial Industrial Total Residential Deaths Commercial Industrial Total Unincorporated Add Cities, Towns Countywide Total 5.0 ALL OCCUPANCIES (COMBINED) ONLY fill this out if data are not available by occupancy classes in the previous tables. Counties Affected Total # of Structures Total Replacement Value (1) Total # Residential Structures Damaged Total # Residential Structures with Minor Damage Total # Residential Structures with Major Damage Total # Residential Structures Destroyed (4) Economic Loss to Residential Buildings (estimated or actual repaircost) Unincorporated Add Cities, Towns Countywide Total Critical Facilities Damage and Loss The following clarifications pertain to critical facilities tables: (1) Total Replacement Value = Assessed value of structures in the county. (2) Minor – repairs can be made in 1-2 week(s); school back in use within 3-4 weeks. (3) Major – repairs take 90 days or more. (4) Destroyed – needs to be rebuilt. (5) Estimated Cost of Repairs – include repair or replacement costs, as applicable. April 2007 A-3 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY 6.0 CRITICAL FACILITIES – SCHOOLS Counties Affected Total # of Public Schools Total ReplaceSchoBuilding ment Value of ols (1) Content Total # Schools Damaged Total # Schools with Minor Damage (2) Total # Schools with Major Damage (3) Total # Schools Destroyed (4) EstimateRepaEconomic Loss: d Cost of irs (5) Building Content Unincorporated Add Cities, Towns Countywide Total 7.0 CRITICAL FACILITIES – MEDICAL CARE FACILITIES (hospitals and nursing homes, etc.) Counties Affected Total # of Hospitals Total Replaof HospBuilding cement Cost itals (1) Content Total # Hospitals Damaged Total # Hospitals with Minor Damage (2) Total # Hospitals with MajorDamage (3) Total # Hospitals Destroyed (4) EconomEstimateRepaic Loss: d Cost of irs (5) Building Content Unincorporated Add Cities, Towns Countywide Total 8.0 CRITICAL FACILITIES – FIRE STATIONS (FS) Total Replacement Cost of FS (1) Economic Loss: Estimated Cost of Repairs (5) Counties Affected Total # of FS Building Content Total # FS Damaged Total # FS with Minor Damage (2) Total # FS with Major Damage (3) Total # FS Destroyed (4) Building Content Unincorporated Add Cities, Towns Countywide Total April 2007 A-4 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY 9.0 CRITICAL FACILITIES – POLICE STATIONS (PS) Counties Affected Total # of PS Total ReplacePSBuilding ment Cost of (1) Content Total # PS Damaged from Wind Total # PS with Minor Damage (2) Total # PS with MajorDamage (3) Total # PS Destroyed (4) Estimated Co(5) Economic Loss: st of Repairs Building Content Unincorporated Add Cities, Towns Countywide Total 10.0 CRITICAL FACILITIES – SHELTERS (primarily schools, churches, civic centers, senior centers, etc.) Counties Affected Total # of Shelters ToCost of SBuilding tal Rephelter (1) Content lacement Total # Shelters Damaged from Wind Total # Shelters with Minor Damage(2) Total # Shelters with Major Damage(3) Total # Shelters Destroyed (4) Estimated Co(5) Economic Loss: st of Repairs Building Content Unincorporated Add Cities, Towns Countywide Total 11.0 DEBRIS ESTIMATES Please provide estimates or known quantities (volume – cubic yard, weight – tons) for occupancy-class generated debris and or total debris (whichever is available): Counties Affected ResidVegetative ential C&D CommVegetative Debrercial C&D is GeInduVegetative nerated strial C&D ToVegetative tal C&D Unincorporated Add Cities, Towns Countywide Total April 2007 A-5 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY APPENDIX B Detailed Raw Data Obtained for Observed Damage The data included in Appendix B was collected from local, state, and federal government agencies and national programs and organizations. Data were provided for damage and economic loss for Hurricanes Ivan and Dennis in Escambia and Santa Rosa Counties; Hurricane Charley in Charlotte, DeSoto, Hardee, Lee, Polk, Orange, and Osceola Counties; Hurricane Frances in Brevard, Indian River, Martin, Okeechobee, Palm Beach, and St. Lucie Counties; and Hurricane Jeanne in Martin and St. Lucie Counties. B-1. Qualitative Damage Estimates B-1.1 Residential Damage Residential damage data were requested from the American Red Cross (ARC), the Florida Department of Insurance Regulation (FLDOIR), and from the counties that are being assessed for this study. Data: PDA Source: ARC Usefulness: Provided qualitative damage estimates of minor and major damage and destruction to residential buildings (e.g., single-family dwelling, apartment/multi-family unit, and manufactured homes). Limitations: These data were collected through windshield survey, not a physical inspection of the damaged structures. Qualitative damage types are not the same as HAZUS damage types; however, an analysis can be performed of the damage descriptions to match the ARC’s damage types with HAZUS damage types. Damages are not separated for wind versus flood damage. Hurricanes/County(ies): Hurricanes Ivan in Escambia and Santa Rosa Counties; Hurricane Charley in Charlotte, DeSoto, Hardee, Lee, Polk, Orange, and Osceola Counties; and Hurricane Frances in Brevard, Indian River, Martin, Okeechobee, and Palm Beach Counties. Note: PDA data were not received for Dennis or Jeanne. Whenever damage states are “0”, it is assumed that this data were not available. Table 1. Preliminary Damage Estimates - Hurricane Charley – Charlotte County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 5,013 12,006 12,457 8,373 200 38,049 Apartment/Multi-Family Unit 0 0 0 0 0 0 Mobile Home 0 0 0 0 0 0 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 5,013 12,006 12,457 8,373 200 38,049 April 2007 B-1 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 2. Preliminary Damage Estimates - Hurricane Charley – DeSoto County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 149 968 981 0 0 2,098 Apartment/Multi-Family Unit 21 96 70 0 0 187 Mobile Home 3,474 1,607 969 0 0 6,050 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 3,644 2,671 2,020 0 0 8,335 Table 3. Preliminary Damage Estimates - Hurricane Charley – Hardee County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 122 775 1,791 1,054 153 3,895 Apartment/Multi-Family Unit 0 0 84 0 0 84 Mobile Home 245 277 613 524 308 1,967 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 367 1,052 2,488 1,578 461 5,946 Table 4. Preliminary Damage Estimates - Hurricane Charley – Lee County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 326 630 6,321 8,665 0 15,942 Apartment/Multi-Family Unit 0 3 158 780 0 941 Mobile Home 5 21 338 203 0 567 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 331 654 6,817 9,648 0 17,450 Table 5. Preliminary Damage Estimates - Hurricane Charley – Orange County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 2 177 2,036 o o 2,215 Apartment/Multi-Family Unit 0 0 0 0 0 0 Mobile Home 0 0 0 o o 0 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 2 177 2,036 0 0 2,215 Table 6. Preliminary Damage Estimates - Hurricane Charley – Osceola County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 137 487 149 0 5,000 5,773 Apartment/Multi-Family Unit 0 0 0 0 0 0 Mobile Home 0 0 0 0 0 0 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 137 487 149 0 5,000 5,773 April 2007 B-2 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 7. Preliminary Damage Estimates - Hurricane Charley – Polk County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 2,012 1,782 1,700 0 0 5,494 Apartment/Multi-Family Unit 0 0 0 0 0 0 Mobile Home 0 0 0 0 0 0 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 2,012 1,782 1,700 0 0 5,494 Table 8. Preliminary Damage Estimates - Hurricane Ivan – Escambia County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 2,699 6,084 17,280 23,973 684 50,720 Apartment/Multi-Family Unit 2,217 2,373 3,212 3,259 0 11,061 Mobile Home 308 928 2,434 2,666 0 6,336 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 5,224 9,385 22,926 29,898 684 68,117 Table 9. Preliminary Damage Estimates - Hurricane Ivan – Santa Rosa County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 796 2,469 5,641 10,054 245 19,205 Apartment/Multi-Family Unit 33 261 414 445 0 1,153 Mobile Home 109 296 959 1,106 44 2,514 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 938 3,026 7,014 11,605 289 22,872 Table 10. Preliminary Damage Estimates - Hurricane Jeanne – Martin County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 32 762 1,639 1,662 38 4,133 Apartment/Multi-Family Unit 8 192 531 130 1 862 Mobile Home 141 389 540 462 3 1,535 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 181 1,343 2,710 2,254 42 6,530 Table 11. Preliminary Damage Estimates - Hurricane Jeanne – St. Lucie County Dwelling Type Destroyed Major Minor Affected Inaccessible Total Single Family Dwelling 263 781 2,019 12,031 0 15,094 Apartment/Multi-Family Unit 37 56 0 0 0 93 Mobile Home 428 1,031 1,206 631 0 3,296 Unknown Dwelling Type 0 0 0 0 0 0 Sub-Total 728 1,868 3,225 12,662 0 18,483 April 2007 B-3 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Background information regarding PDA qualitative damage states versus HAZUS damage states: Table 12. PDA Damage State Definitions Destroyed – Structure is a total loss or damaged to such an extent that repairs are not economically feasible. Any one of the following may constitute a status of destroyed: . Repair of structure is not economically feasible. . Structure is permanently uninhabitable. . There is a complete failure of major structural components (collapse of walls or roof). . Unaffected structure will be required to be removed or demolished due to ordinance (e.g., beachfront homes removed due to severe beach erosion). Major – Structure has sustained structural or significant damage, is uninhabitable, and requires extensive repairs. Any of the following may constitute major damage: . Substantial failures to structural elements of the residence (e.g., walls, floors, foundations). . Damage to the structure exceeds the Disaster Housing Program, Home Repair Grant maximum ($10,000). . General exterior property damage exceeds the Disaster Housing Program Home Repair Grant maximum (e.g., roads and bridges, wells, earth movement) and has more than 50 percent damage to the structure. . Damage will take more than 30 days to repair. Minor – Structure is damaged and uninhabitable, but may be made habitable in a short period of time with home repairs. Any of the following may constitute minor damage: . Structure can be repaired within 30 days. . Structure has more than $100 of eligible habitability items through the Disaster Housing Program, Home Repair Grant; or has less than $10,000 of eligible habitability items through the Disaster Repair Program, Home Repair Grant. . Damage repair costs are less than 50 percent of total value of house. Affected – Sustained some damage to structure and contents, but is habitable without repairs, and damage to habitability items is less than Disaster Housing Program, Home Repair Grant minimum. The PDA definitions are very comparable to the ones in HAZUS. For reporting purposes, the following alignment provides the appropriate mapping between the two: Human Services PDA HAZUS-MH Damage States Affected = Minor Damage Minor Damage = Moderate Damage Major = Severe Destroyed = Destruction Data: Damaged Residential Structures Source: DeSoto County Usefulness: Provided qualitative damage estimates of minor and major damage and destruction to residential buildings. Limitations: These data were indicative of what has been reported as of January 2006. DeSoto County provided qualitative damage estimates for residential structures that were slightly lower than the PDA data. Hurricanes/County(ies): Hurricane Charley in DeSoto County. April 2007 B-4 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 13. Residential Damage - Hurricane Charley – DeSoto County County Total # of Buildings Total # Damaged Buildings Minor Damage Major Damage Destroyed DeSoto 10,700 9,672 *3,587 2,589 3,496 * Minor 2,095 plus affected 1,492 Data: Wind Loss Insurance Claims Data Source: FLDOIR Usefulness: Provided insurance data for total claims and total losses (i.e., destroyed) for residential properties. Limitations: These data only reflect insured property losses, but includes damage to appurtenant structures and automobiles, which are not modeled in HAZUS-MH. Hurricanes/County(ies): All hurricanes and all counties involved in this study. Table 14. Residential Damage - Hurricane Charley, Charlotte County and Hurricane Ivan, Escambia County (FLDOIR) County Disaster # of Claims Reported # of Claims Total Loss Escambia Ivan 101,715 3,859 Santa Rosa Ivan 43,785 2,025 Escambia Dennis 14,330 57,416,279 Santa Rosa Dennis 21,648 127,077,866 Charlotte Charley 83,085 8,601 DeSoto Charley 13,209 1,714 Hardee Charley 8,101 638 Lee Charley 78,317 1,945 Orange Charley 99,164 1,017 Osceola Charley 42,204 843 Polk Charley 49,595 1,857 Brevard Frances 61,321 2,838 Indian River Frances 31,627 1,730 Martin Frances 24,064 1,288 Okeechobee Frances 6,783 587 Palm Beach Frances 107,926 2,534 St. Lucie Frances 56,722 4,053 Martin Jeanne 20,966 626 St. Lucie Jeanne 31,866 960 * Claims payments for Dennis only reflect those made as of 10/7/2005. Modeling Usefulness: PDA data for qualitative residential damage estimates (e.g., number of structures with minor or major damage, or destroyed) were compared with the damage estimates generated by the HAZUS-MH Hurricane Wind Model. Although it is widely known that PDA data are based on windshield survey and estimates can be somewhat subjective, this appears to be the best available data set for comparison purposes for this validation study. Insurance claims were much higher than the HAZUS-MH estimates for total damage and the PDA estimates, as April 2007 B-5 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY the FDOIR data included claims paid for items not modeled in HAZUS-MH such as boats and cars. Therefore the number of claims is not suitable for comparison with HAZUS- MH estimates. General Limitations: Residential damage data likely contains damage caused by both flood and wind, which are not able to be separated. This does not appear to be a concern for counties that were impacted by Hurricane Charley, as this was predominantly a wind event. However, this is a concern for counties that were impacted by Frances, Ivan, Jeanne, and Dennis. In addition, it will be difficult to determine which damages occurred from Ivan versus Dennis or Jeanne versus Frances, since these storms hit the same areas. As such, all prior damages had not yet been repaired from the first hurricane when the second hurricane caused damage. B-1.2 Commercial and Industrial Damage Commercial and industrial damage data were requested from the ARC, the Insurance Services Office, Inc. (ISO), and from the counties that are being assessed for this study. No data were received. ISO data were received for residential, commercial, and industrial damage. These data are not reproduced per licensing agreement requirement. B-1.3 Critical Facilities Damage Aggregate Critical Facilities Damage Critical facilities damage data were requested for medical care facilities, schools, fire stations, police stations, and shelters. Data were requested from the Florida Hospital Association and from the counties that are being assessed for this study. Data: Damaged Critical Facilities Sources: Escambia County School Board Risk Manager, Escambia County, and DeSoto County. Usefulness: Provided qualitative damage estimates of minor, moderate, and major damage to critical facilities from wind. Limitations: Qualitative damage types are not the same as HAZUS damage types; however, an analysis can be performed of the detailed damage descriptions to match the county’s damage types with HAZUS damage types. The shelter data are not a comprehensive list of all shelter damage as it only includes schools that served as shelters. Hurricanes/County(ies): Qualitative damage were provided for Escambia County schools, fire stations, and shelters for Hurricane Ivan, and for DeSoto County fire stations and shelters for Hurricane Charley. April 2007 B-6 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 15. Critical Facilities Damage – Hurricanes Ivan and Charley County Disaster Critical Facility Type Total # of Facilities Total # of Facilities Damaged Minor Moderate Major Destroyed Escambia Ivan Schools n/a n/a 25 17 30 Escambia Ivan Fire Stations 23 23 21 0 2 Escambia Ivan Police Stations n/a n/a n/a n/a n/a Escambia Ivan Shelters 10 n/a 0 3 2 DeSoto Charley Fire Stations 2 2 2 0 0 DeSoto Cities Total Charley Fire Stations 2 2 1 1 DeSoto Charley Shelters 1 1 n/a = not available Schools Damage Table 16. Detailed Qualitative Damage Estimates – Hurricane Ivan – Escambia County Schools School Name Major Moderate Minor Jim Allen Portables-Major damage; Minor roof leaks over kitchen. Bellview Elementary Roof missing; major water damage; siding & facial; portables-minor damage. Beulah Damaged bus canopy; roof damage; major water damage. Bibbs, Spencer Siding 100-200 rooms; front entrance awnings; awnings. Blue Angels, (Shelter) Major damage east side structural; minor damage throughout; portables damaged. Bratt Roof blown off; major water damage; siding/awnings; bldg. 3-severe water damage. Brentwood Elementary Roof missing main hallways; extensive water damage all spaces. Byrneville Portables - tree damage Hellen Caro Roof leaks-major in some areas; broken windows 615; portables A & D-windows. Century/Carver 2 trees on new building; 1 tree on #2 portable. April 2007 B-7 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY School Name Major Moderate Minor Cook, N. B. Main hall skylight gone; major water damage; roof damage; second floor- extensive damage; trap door (over stage) open; kitchen exhaust could fall. Cordova Park Main bldg.-ceiling; A/C damaged by walkway; rms. 111,114-water damage. Edgewater Major roof damage - bldg. #2; windows; rms. 138,135,130 extensive. Ensley Roof is good; awnings; trees; *power line Ferry Pass Elementary No assessment Hallmark Minor leaks; many trees down. Holm Minor roof damage; kitchen, rm. 325; awnings. Lincoln Park Roof leaks throughout; awnings; walkways. Lipscomb (Shelter) Lift station over flowing; classroom roof leaks all wings; windows & siding on portables. Longleaf Major roof damage; water damage all spaces except new wing. McArthur Roof damage-819; siding743; bldg. 99; roof leaks- various minor; awnings. Molino Park (Shelter) Awnings; roof leaks; major water damage (admin. spaces). Montclair Trees; portables; roof vents. Myrtle Grove Extensive roof damage; shingles; leaks in most spaces; trees on portables. Navy Point Major walls rms. 30-33; cafeteria-pine tree; walkways; 2nd floor Windows in 2 Classrooms. Oakcrest Roof damage all spaces; awnings; portables-major damage. Pine Meadow Major trees down; roof damage minor-104, 101, 204, 105, 642, 256; water damage; office-major damage. Pleasant Grove Building 3-roof damage; portables electrical lines. April 2007 B-8 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY School Name Major Moderate Minor Scenic Heights 300 bldg.-tree on roof; awnings; vent fan off. Semmes Siding-portable; bldg 1 & bldg. 9-awning, siding; media center-major damage; backflow preventer box- structural damage. Sherwood Bldg. 3-major roof damage; bldg. 9-no roof; major water damage. Suter, A.K. Roof-118, 112, 119, 100D, 111, 108, 124, 100F, 107, 107C, 205, 203, 125A, 125B; media addition; #99 portable-roof sagging; 8 trees down. Warrington Elementary Roof gone; ceilings; floors; trees; awnings; covered walks. Weis, C. A. Extensive water damage all spaces; awnings; roof damage all spaces. West Pensacola Extensive debris clean-up. Yniestra, Allie Gutters; portables settled. Molino (Vacant) Walkways bldgs. 2 & 3 Bailey (Shelter) Roof system severe; windows out; portable damage. Bellview Middle Cafeteria/kitchen damage; portable damage; water damage; ceiling tiles & water all over; ese/admin. windows, flooding. Brentwood Middle Siding; roof peeling on trailers; minor water damage. Brown Barge Roof damaged over mechanical room; major roof damage-108,109,106,104, 063,cafeteria 208, library, 314,317; awnings, portables. Brownsville Gas leak-secure 14:30; minor water damage; portables-siding & windows; large tree on main building. Ferry Pass Middle Main building-shingles; skylights; awnings; siding; office flooded; major water damage-618, 619, 500, 501, 516, kitchen, cafeteria, 201, 202, bookkeepers' office. April 2007 B-9 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY School Name Major Moderate Minor Ransom Portables siding damaged- all; windows portables-all; roof-admin.; library-roof, windows; trees total school; water main Ernest Ward Minor roof damage; walkways & awnings. Warrington Middle Major roof damage bldgs. 1,2,4,5; windows on portables; extensive water damage. Wedgewood Major flooding; library; major roofing damage. Workman Main building-roof; library- water damage; all awnings; girls' p.e.-water damage; cafeteria-water damage; cafeteria-windows; 200 wing- roof. Escambia Gym major roof damage; major leaks; awnings; broken windows. Northview Roof leaks all spaces; minor water damage. Pensacola High Bleachers destroyed; gym destroyed; awnings/ramps; water damage; surveillance cameras stolen. Pine Forest Roof damage; cooling tower destroyed. Tate Lift station down; major water damage-main bldg.; windows-main bldg.; all walkways; cafeteria, gym. Washington Music/choral/band roof damage and water damage; minor roof damage over atrium; water damage-north classrooms. Woodham Cafeteria; media center destroyed; major roof damage; water damage. West Florida (Shelter) Major debris clean-up; major roof damage; bldg. 7-lost roof; awnings/gutters. Hall Center McDaniel Administration 2nd floor windows w/ extensive water damage; supt. office-water damage. Data Center Perimeter offices water damage; computer room dry. End User Support Roof gone; major water damage. April 2007 B-10 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY School Name Major Moderate Minor Walnut Hill Bus Garage Shingles; tree damage. Transportation Main Garage Major roof damage; 2 roll up doors; water leaks all over wash rack and electronic equipment. Andrews, Judy Broken windows; no roof or water damage. Dixon 3 broken windows. Clubbs, A. V. 3 trees w/ roof penetration; major roof leaks; water damage all spaces. Escambia Westgate Roof missing - 505, 506; hallway leaks @ new bldg. seam; cafeteria-broken windows, water damage. ESEAL Quansit huts gone; tool shed gone. Sid Nelson Roof damage all buildings, cafeteria kitchen-300, 500, 400, 200; all awnings. McMillan/Title I Portables - heavy damage; minor leaks Petree Pre-K Wiring down; shingles blew off roof; tree on covered walkway; tree on fence. Pickens Windows-board covers missing. Environmental Center Many Trees on Roof. Warehouse Freezer up; 1 unit down; roof over 70 degree room. George Stone SEE WFHSAT Awnings/gutters; bldg. 3flooded; paint booth destroyed; portable 99destroyed; portable 17-blown off foundation; administration wing-roof destroyed. Gibson No assessment Old Carver No assessment Totals 30 17 25 April 2007 B-11 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Shelter Damage Table 17. Shelter Detailed Qualitative Damage Estimates – Hurricane Ivan – Escambia County School Name Major Moderate Minor Blue Angels (Shelter) Major damage east side structural; minor damage throughout; portables damaged. Lipscomb (Shelter) Lift station over flowing; classroom roof leaks all wings; windows & siding on portables. Molino Park (Shelter) Awnings; roof leaks; major water damage (admin. spaces). Bailey (Shelter) Roof system severe; windows out; portable damage. West Florida (Shelter) Major debris clean-up; major roof damage; bldg. 7-lost roof; awnings/gutters. Totals 2 3 0 Site Specific Critical Facilities Damage Critical facility qualitative damage data were provided from prior HAZUS-MH validation studies conducted in 2004 for Hurricanes Charley , Ivan, and Dennis. Data: Damaged Critical Facilities Sources: Hurricanes Charley (TO – 332), Ivan (TO – 348), and Dennis (TO – 406). Usefulness: Provided qualitative damage estimates of minor, moderate, and major damage to critical facilities. Limitations: Observed damage states collected in the field can be subjective. Hurricanes/County(ies): Qualitative damage were provided for Osceola and Hardee schools, and a Polk County fire station; and for Escambia County schools, fire stations, and shelters for Hurricane Ivan. Table 18. Prior HAZUS Validation – Hurricane Charley County Type of Facility Facility Name Latitude Longitude HAZUS Wind Class Observed Damage Osceola EDU1 Poinciana High School 28.239121 -81.486760 SECBL Moderate Hardee EDU1 Wauchula Elementary 27.542352 -81.817834 MLR1 Minor Hardee EDU1 Hardee High School 27.529024 -81.834071 MECBL Minor Osceola EDU1 Thacker Elementary 28.289289 -81.424193 MECBL Minor Polk Fire Station Haines City Fire Dept. 28.108560 -81.622275 MECBL Minor April 2007 B-12 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 19. Prior HAZUS Validation – Hurricane Ivan HAZUS Observed Observed County Type of Facility Facility Name Latitude Longitude Wind Class Wind Damage Flood Damage - Escambia EDU1 Longleaf Elementary 30.486689 87.291219 CECBL Severe None - Escambia EDU1 Pine Forest High School 30.293898 87.452102 CECBL Moderate None Escambia EDU1 West FL High School/Beggs Educational Center 30.293879 - 87.451672 CECBL Moderate None Santa Rosa Fire Station Navarre Beach VFD #18 30.379285 - 86.879664 SECBL Minor Minor Santa Rosa EDU1 Gulf Breeze Middle School 30.370639 - 87.176129 MECBL Minor None Santa Rosa Hospital The Friary of Baptist Health Care Center 30.3976 -87.03941 WMUH1 Minor None Santa Rosa Hospital Gulf Breeze Hospital 30.36104 -87.15677 MECBL Moderate None Escambia Hospital Sacred Heart Health System 30.4753 -87.21374 SECBM Minor None Escambia Hospital Baptist Hospital 30.43083 -87.23097 MECBM Minor None Escambia Hospital West Florida Hospital 30.51439 -87.21784 SECBH Major None Table 20. Prior HAZUS Validation – Hurricane Dennis HAZUS Observed Observed County Type ofFacility Facility Name Longitude Latitude Wind Class Wind Damage Flood Damage Escambia Fire Station Pensacola Beach FD 30.33813 87.11544 CECBL Moderate Moderate Santa Rosa EDU2 Pensacola Jr College - Milton Campus 30.60474 -87.07495 MECBM very minor none Santa Rosa EDU1 Hobbs Middle School 30.62813 -87.06482 MECBL moderate none Santa Rosa COM6 Santa Rosa Medical Center 30.63452 -87.06714 MECBM very minor none Santa Rosa EDU1 Pearidge Elementary School 30.60832 -87.11166 MECBL moderate none Santa Rosa Fire Station Holley-Navarre Vol. Fire Dept. 30.43208 86.87494 S5 None none Santa Rosa EDU1 West Navarre Elem School 30.40642 86.93256 MLR1 Very Minor none Santa Rosa EDU1 T.R. Jackson Pre-K 30.61542 87.04339 MLR1 None none Santa Rosa COM1 School of Readiness 30.62382 87.0368 MECBL None none Santa Rosa Fire Station Fire Station 18 30.37941 86.87975 MECBL Severe Moderate Santa Rosa COM6 Gulf Breeze Hospital 30.47532 87.21374 SECBM Minor Minor B-1.4 County Structures Damage Data: Damaged County Structures April 2007 B-13 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Source: Escambia County Usefulness: Provided number of county structures damaged. Limitations: Data are not separated by occupancy class, or by hazard (e.g., wind versus flood). Hurricane/County: Hurricane Ivan for Escambia County Modeling Usefulness: There was not enough information known about which building occupancy classes were included in the observed data. These data were not suitable for comparison with damage estimates generated by the HAZUS-MH Hurricane Wind Model. Table 21. County Structure Damage – Hurricane Ivan – Escambia County Community Total # of Buildings (county structures) Total # of Buildings Damaged in Hurricane Ivan Total # of Buildings Sampled in Survey Total # of Buildings Surveyed that Sustained at Least Minor Damage Total # of Buildings Surveyed that Sustained at Least Moderate Damage Escambia County 240 192* n/a n/a n/a n/a = not available County structures damaged (this does not include parks and recreation) Data: Damaged Structures (countywide) Source: Santa Rosa County Property Appraiser Usefulness: Providec qualitative damage estimates of minor and major damage for all buildings (occupancies). Limitations: County data were not useable as they were not separated by occupancy class, or by hazard (i.e., wind versus flood). As such, ARC PDA data were used for analysis purposes, so that all qualitative damage estimates for residential units will come from one source to be consistent. Hurricanes/County(ies): Hurricane Dennis in Santa Rosa County Table 22. Countywide Damaged Structures - Hurricane Dennis – Santa Rosa County Community Total # of Buildings (all occupancies) Total # of Buildings Damaged in Hurricane Dennis Total # of Buildings Sampled in Survey Total # of Buildings Surveyed that Sustained at Least Minor Damage Total # of Buildings Surveyed that Sustained at Least Moderate Damage Navarre 11,024 200 41 10 2 Bagdad 2,400 29 9 9 7 Milton 3,246 28 164 69 27 Navarre Beach 1,729 0 156 126 36 Pensacola Beach 0 0 5 0 0 Gulf Breeze 2,850 0 12 0 0 Totals 21,249 257 387 214 72 April 2007 B-14 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Loss of Function B-2.1 Critical Facilities 2.3.1.6 Critical Facilities Loss of Function Critical facilities loss of function data (e.g., loss of use in days) was requested from the FHA and the counties included in this study. Data: Hospital closures Source: Florida Hospital Association Usefulness: Providec economic loss estimates by county. Limitations: Does not indicate whether closures were due to physical damage or loss of power/water. Does not list damage for each hospital, only by county. Hurricanes/County(ies): Hurricane Ivan for Escambia County, and for Charlotte, Hardee, DeSoto, and Orange Counties for Hurricane Charley. Modeling Usefulness: All relevant available data for loss of functionality for critical facilities (e.g., loss of use in days) were compared with the loss of functionality estimates generated by the HAZUS-MH Hurricane Wind Model. B-2.1.1 Medical Care Facilities Table 23. Hospital Loss of Function – Hurricane Ivan and Charley County Disaster Number of Hospitals in County (Acute/Other) Number Responding to FHA survey Closures Days Closed Escambia Ivan 3/3 2 0 0 Santa Rosa Ivan 3/1 DNR DNR DNR Charlotte Charley 3 2 1 10 DeSoto Charley 1 1 1 2 Hardee Charley 1 1 0 0 Orange Charley 10/3 6 0 0 DNR = Did not report B-2. Shelter Demand B-3.1 Shelter Population Pre- and post-disaster shelter population data were requested from the ARC. Data: Pre- and post-disaster shelter counts – Hurricanes Ivan and Dennis Source: Ivan: ARC; Dennis: FL Division of Emergency Management (FDEM) Usefulness: Provided shelter population, from which to derive the number of displaced households by county. April 2007 B-15 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Limitations: Does not include population that evacuated outside the county. This data might include population that were displaced from their homes short-term due to electric and water/sewage damage, not structural damage. Data does not indicate which populations sought shelter to avoid coastal flooding versus wind impacts. Hurricanes/County(ies): All hurricanes and all counties involved in this study. Modeling Usefulness: Post-disaster shelter population counts were compared with the shelter population estimates generated by the HAZUS-MH Hurricane Wind Model. Table 24. Shelter Populations - Hurricanes Charley, Dennis, Frances, Ivan, and Jeanne – All Counties County Disaster Pre-Landfall Shelter Population Post-Landfall Shelter Population* Escambia** Ivan 7,692 978 Santa Rosa** Ivan 636 349 Escambia*** Dennis 2,472 100 Santa Rosa*** Dennis 878 35 Charlotte Charley 425 425 DeSoto Charley 1,374 1,400 Hardee Charley 537 537 Lee Charley 8,129 1,191 Orange Charley 727 70 Osceola Charley 1,714 102 Polk Charley 3,390 14 Brevard Frances 9,701 10,654 Indian River Frances 4,055 1,205 Martin Frances 5,558 275 Okeechobee Frances 1,595 250 Palm Beach Frances 4,348 17,585 St. Lucie Frances 4,896 2,745 Martin Jeanne 1,873 118 St. Lucie Jeanne 2,118 536 * Data includes post-land fall shelter population the night after the hurricane made landfall. ** Total provided by FDEM. *** Total provided by the ARC, estimated on post-landfall population. B-4 Economic Loss B-4.1 Residential Economic Loss Residential economic loss data were requested from the FLDOIR, FEMA’s Individual Assistance (IA) Program, ISO, and the counties included in this study. FDOIR and ISO data were received for residential damage. This data are not reproduced per licensing agreement requirement. Data: Wind Loss Insurance Claims Data; value of repair or replacement cost Source: FLDOIR April 2007 B-16 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Usefulness: Provided insurance claims for residential properties. Limitations: These data only reflect insured property losses. Hurricanes/County(ies): All hurricanes and all counties involved in this study. Modeling Usefulness: Insurance losses were compared with loss estimates generated by the HAZUS-MH Hurricane Wind Model. It did not appear that IA data were be useful for this validation study, as the program provides loans to homeowners for $5,200 for property loss, which typically represents only a portion of the loss. Table 25. Department of Insurance Regulation Wind Loss Claims Payments – Hurricanes Charley, Dennis, Frances, Ivan, and Jeanne – All Counties County Disaster Claims Payments* Escambia Ivan 1,698,123,360 Santa Rosa Ivan 668,074,159 Escambia Dennis 38,336,681 Santa Rosa Dennis 56,755,861 Charlotte Charley 2,561,459,188 DeSoto Charley 283,494,623 Hardee Charley 138,479,116 Lee Charley 1,014,468,730 Orange Charley 990,660,827 Osceola Charley 560,179,016 Polk Charley 554,140,262 Brevard Frances 988,571,302 Indian River Frances 766,420,342 Martin Frances 423,664,555 Okeechobee Frances 94,125,899 Palm Beach Frances 1,165,038,320 St. Lucie Frances 1,127,106,791 Martin Jeanne 201,448,046 St. Lucie Jeanne 246,490,019 * Claims payments for Dennis only reflect those made as of 10/7/2005. B-4.2 Commercial and Industrial Economic Loss No observed data were used for this study Commercial economic loss data were requested from the Small Business Administration (SBA), ISO, and the counties included in this study. It was decided that SBA data would not reflect accurate economic loss, as SBA provides business loans to those who apply. Commercial and industrial damage data were requested from the ARC, the Insurance Services Office, Inc. (ISO), and from the counties that are being assessed for this study. No data were received. ISO data were received for residential, commercial, and industrial damage. This data are not reproduced per licensing agreement requirement. April 2007 B-17 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY B-4.3 Critical Facilities Economic Loss B-4.3.1 Medical Care Facilities Critical facilities economic loss data requested from the FHA and the counties included in this study. Data: Hospital Economic Loss Source: Florida Hospital Association Usefulness: Provided economic loss estimates for hospitals by county. Limitations: The data does not list economic loss for each hospital; only by county. Hurricanes/County(ies): Economic loss data were provided for Hurricane Ivan for Escambia County, and for Charlotte, Hardee, DeSoto, and Orange Counties, for Hurricane Charley. Table 26. Hospital Economic Loss - Hurricanes Ivan and Charley Actual Cost Number of Estimates County Hurricane Hospitals in County (Acute/Other) Number Responding to FHA survey Total Cost for Repairs Related to Patient Care Staffing Costs Escambia Ivan 3/3 2 $20,335,000 $4,600,000 $1,040,800 Santa Rosa Ivan 3/1 DNR DNR DNR DNR Charlotte Charley 3 2 $23,000,000 DNR DNR DeSoto Charley 1 1 $2,501,000 DNR DNR Hardee Charley 1 1 DNR DNR DNR Orange Charley 10/3 6 $284,978 $571,500 $874,666 DNR = Did not report B-4.3.2 Schools Data: School Economic Loss Source: Escambia County School Board Risk Manager Usefulness: Provided economic losses to school buildings and educational facilities. Limitations: Economic losses are not separated by building and contents. Hurricane/County: Economic loss data were provided for Hurricane Ivan for Escambia County. Table 27. School Economic Loss – Hurricane Ivan – Escambia County Location Name Building Value Contents Value Est. Damages Jim Allen Elementary 4,841,890 718,190 250,000 Judy Andrews Pre-K 995,805 149,371 20,000 Bailey Middle 12,818,250 1,922,738 6,000,000 Bellview Elementary 4,087,040 613,056 500,000 Bellview Middle 8,999,640 1,349,946 450,000 Beulah Elementary 5,042,480 756,372 750,000 Spencer Bibbs Elementary 3,848,720 577,308 150,000 Blue Angels Elementary 8,858,407 1,763,000 400,000 April 2007 B-18 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Location Name Building Value Contents Value Est. Damages Brentwood Elementary 3,756,400 563,460 700,000 Brentwood Middle 6,766,740 1,015,011 250,000 Brown Barge Middle 4,067,990 610,199 1,000,000 Brownsville Middle 8,117,820 1,217,673 150,000 Byrneville Elementary 673,840 101,076 75,000 Hellen Caro Elementary w Cafeteria Expansion 6,491,506 966,580 400,000 Carver Century K-8 5,253,120 911,482 100,000 Carver (Middle Old) 6,109,809 916,471 no data Cook Elementary - AV Clubbs Ctr. 2,149,440 322,416 300,000 N.B. Cook Elementary 11,000,000 1,650,000 1,000,000 Cordova Park Elementary 3,800,400 570,060 100,000 Dixon Elementary 3,093,680 464,052 50,000 Edgewater Elementary 3,513,120 526,968 200,000 Ensley Elementary 3,945,286 636,560 200,000 Escambia High w Elevator 19,025,048 2,825,690 1,000,000 Escambia Westgate Center & Snoezelen Bldg. 6,101,580 1,093,965 1,500,000 Ferry Pass Elementary 4,012,400 601,860 200,000 Ferry Pass Middle 7,110,540 1,066,581 400,000 McMillan Learning Ctr. Goulding (Pre-K) 2,256,072 338,411 100,000 Hallmark Elementary 2,517,200 377,580 75,000 Holm Elementary 10,447,800 2,134,818 150,000 Lincoln Park Elementary 3,503,600 525,540 150,000 RC Lipscomb Elementary 6,726,320 1,008,948 150,000 Lakeview Center no data no data no data Longleaf Elementary 14,587,950 866,063 10,000,000 McArthur Elementary 4,305,760 645,864 150,000 Molino Elementary 1,758,080 263,712 50,000 Molino Park Elementary 10,861,556 1,200,000 200,000 Montclair Elementary 3,445,840 516,876 75,000 Myrtle Grove Elementary 3,869,760 580,464 300,000 Navy Point Elementary 4,287,440 643,116 500,000 Northview High 11,017,400 1,652,610 450,000 Oakcrest Elementary & Media Center 4,420,534 695,348 500,000 Pensacola High School 21,210,800 3,181,620 1,500,000 Petree Pre-K 940,969 141,145 75,000 Pine Forest High 16,309,500 2,446,425 1,000,000 Pine Meadow Elementary 4,731,280 709,692 250,000 Pleasant Grove Elementary 3,420,640 513,096 150,000 Ransom Middle 11,023,200 1,653,480 50,000 Scenic Heights Elementary 4,902,720 735,408 100,000 Semmes Elementary & Media Center 4,065,964 665,762 350,000 Sherwood Elementary 4,671,517 701,606 1,500,000 Sidney W. Nelson Pre-K 3,658,052 548,708 300,000 George Stone Vocational (AKA) West FL High/Beggs Educational Center 29,918,122 4,792,718 6,000,000 Suter Elementary 2,800,320 405,019 400,000 Tate High 25,367,100 3,805,065 8,000,000 April 2007 B-19 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Location Name Building Value Contents Value Est. Damages Ernest Ward Middle & Pump House 6,337,530 950,630 100,000 Ernest Ward Bus Facility no data no data no data Warrington Elementary 3,821,680 573,252 5,000,000 Warrington Middle 7,102,620 1,065,393 1,000,000 Washington High 19,319,100 2,897,865 450,000 Wedgewood Middle 6,552,270 982,841 4,000,000 Weis Elementary 6,325,760 948,864 2,000,000 West Pensacola Elementary 4,524,400 678,660 350,000 Woodham High School 17,401,200 2,610,180 4,000,000 Workman Middle & ESE Suite 9,146,601 1,363,132 4,000,000 Yniestra Elementary 2,692,800 403,920 100,000 Administrative Office 2,359,474 353,921 150,000 Gibson School Federal Project 470,015 70,502 no data Maintenance and Transportation 1,727,933 259,190 no data Kirskey Warehouse 344,677 51,702 n/a Administrative Bldg. (End-User) 208,440 31,266 400,000 Data Center 1,252,493 187,874 100,000 Central Warehouse 3,408,912 511,337 75,000 Transportation Facility 2,626,683 394,002 550,000 Roy L. Hyatt Environmental Studios 175,805 26,371 200,000 J.E. Hall Center-Support Svc. Facilities Planning 9,131,215 1,369,682 500,000 Storage - Pickens Book Dept. 1,645,050 246,758 50,000 ESEAL 1,263,731 189,560 100,000 Bratt Elementary 3,162,640 474,396 800,000 Occupied by Plumbers & Pipefitters 137,000 no data no data Occupied by USO 1,800,000 no data 200,000 Occupied by Property Appraiser 372,000 no data no data Same as above Building 2 135,000 no data no data Data Center-EDP Only no data no data no data Portables to be added effective 7/1/04 no data no data no data Freezer Food-All Schools no data no data 1,500,000 Vehicles-Various: no data no data no data Trucks no data no data 100,000 Cars no data no data no data Buses no data no data 250,000 Total $484,951,476 $72,300,474 $74,645,000 B-4.3.3 Fire Stations Data: Fire Station Replacement Cost Source: Escambia County Usefulness: Provided replacement cost of fire stations, but does not break the replacement costs down by fire station, which would have been useful to update the HAZUS default inventory. So, this data are not useful by itself. Limitations: Economic loss was not provided, which is necessary to determine the loss ratio. April 2007 B-20 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Hurricane/County: Replacement cost data were provided for Hurricane Ivan for Escambia County. Table 28. Fire Station Replacement Cost – Hurricane Ivan – Escambia County Disaster County Total Replacement Cost of Fire Stations Ivan Escambia $10,414,593 B-4.3.4 Police Stations No observed data were received. B-4.3.5 Shelters Data: Shelter (schools) economic loss Source: Escambia County School Board Risk Manager Usefulness: Provided economic loss to shelter buildings. Limitations: This was not a comprehensive list of all shelter damage as it only included schools that served as shelters. Hurricane/County: Shelter economic loss was provided for Hurricane Ivan for Escambia County. Table 29. Shelter Economic Loss – Hurricane Ivan – Escambia County School Name Cost of Reconstruction or Remediation Blue Angels, (Shelter) $400,000 Lipscomb (Shelter) $150,000 Molino Park (Shelter) $200,000 Bailey (Shelter) $6,000,000 West Florida (Shelter) $6,000,000 Totals $12,750,000 B-4.3.6 County Structures Data: County structures total replacement value Source: Escambia County Usefulness: Provides replacement cost of all county structures (except for parks and recreation), but does not break the replacement costs down by structure, which would have been useful to update the HAZUS default inventory. So, this data are not useful by itself. Limitations: Economic loss was not provided, which is necessary to determine the loss ratio. Hurricane/County: Replacement cost data were provided for Hurricane Ivan for Escambia County. April 2007 B-21 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 30. County Structures Replacement Cost – Hurricane Ivan – Escambia County Disaster County Total Replacement Cost of County Structures* Ivan Escambia $175,648,373 * This does not include Parks and Recreation. Data: Public Assistance Category E (Buildings) Summary Data Source: FEMA Public Assistance Program Usefulness: This data provided the total PA costs for buildings, prepared as Category E project worksheets (PWs). Costs Covered By PA: Category E PWs cover losses not covered by insurance for the repair or restoration or publicly owned and maintained structures, equipment (e.g., electrical, mechanical, telecommunications), and contents (e.g., furniture, books, computers). Sometimes vegetative or construction and demolition (C&D) debris removal costs, or mold remediation costs are included in Category E PWs, and sometimes these costs are covered under Categories A or B PWs. As such, it is indiscernible as to whether the PA summary data includes these costs and to what degree (e.g., none, some, or all). This information could be identified by perusing each Category E PW, which is a very time consuming process. As stated on FEMA.gov, FEMA may pay for upgrades that are required by certain codes and standards, such as roof bracing installed following a hurricane, and upgrades to meet standards regarding use by the disabled. For repairs, upgrades are limited to damaged elements only. If a structure must be replaced, the new facility must comply with all applicable codes and standards regardless of the level of FEMA funding. If a damaged building must be replaced, FEMA has the authority to pay for a building with the same capacity as the original structure. However, if the standard for space per occupant has changed since the original structure was built, FEMA may pay for an increase in size to comply with that standard while maintaining the same occupant capacity. A federal or state agency or statute must mandate the increase in space; it cannot be based only on design practices for an industry or profession. Limitations: This data does not reflect costs that were covered by insurance or insurance deductibles, which can grossly underestimate building repair/replacement costs. This data does reflect economic loss to critical facilities, but includes costs for most public buildings. Sometimes building economic loss can be covered under other PA categories. For example temporary repairs/mold remediation could be covered under Category B – Emergency Protective Measures, or repairs could be covered under Category G, Parks, Recreation, and Other. This data includes both wind and flood loss, which does not allow for direct comparison, but is a useful benchmark. Hurricanes/County(ies): All hurricanes and all counties involved in this study. April 2007 B-22 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 31. Public Buildings Economic Loss – FEMA Public Assistance County Hurricane Total Obligated Charlotte Charley $25,156,724 DeSoto Charley $3,557,181 Hardee Charley $1,989,438 Lee Charley $4,546,783 Orange Charley $10,398,422 Osceola Charley $4,931,922 Polk Charley $2,946,422 Brevard Frances $5,324,603 Indian River Frances $3,192,071 Martin Frances $10,889,585 Okeechobee Frances $1,044,711 Palm Beach Frances $12,765,742 St. Lucie Frances $20,342,900 Escambia Ivan $38,761,211 Santa Rosa Ivan $4,797,993 Martin Jeanne $4,182,316 St. Lucie Jeanne $5,523,031 Escambia Dennis $2,741,666 Santa Rosa Dennis $2,181,305 Modeling Usefulness: FHA, Escambia County schools, and PA economic loss data could not be directly compared with the loss estimates for critical facilities, generated by the HAZUS-MH. Hurricane Wind Model. Shelter loss cannot be compared with HAZUS-MH estimates because there are no building characteristic data (e.g., building type) to run a user defined scenario. B-4 Debris Generated Debris quantities for both vegetative and construction and demolition debris types were requested from the FEMA Public Assistance Program and the counties included in this study. Some data were received from the jurisdictions. Data: Debris quantities data for Santa Rosa County for Hurricane Dennis; and DeSoto and Hardee Counties, and Hardee cities for Hurricane Charley. Source(s): Santa Rosa, DeSoto, and Hardee Counties Usefulness: Provides vegetative debris estimates as of January 2006. This is an incremental quantification of debris, and this quantity will very likely increase in the future. This data set does not appear to be accurate for comparison with HAZUS-MH debris quantity estimates. Limitations: Debris quantities are not separated by wind versus flood hazard. Does not include concrete and demolition debris estimates. Debris operations are still underway in some of these April 2007 B-23 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY counties; therefore, these estimates will not reflect the total quantity once all debris has been removed from eligible areas. It was difficult to determine the amount of debris generated for Orange County, as not all of the debris had been collected for Hurricane Charley when Hurricane Frances made landfall. Hurricane(s)/County(ies): Debris quantity estimates were provided by Santa Rosa County for Hurricane Dennis, and DeSoto and Hardee Counties for Hurricane Charley. Modeling Usefulness: Debris quantities provided by these counties were not compared with the debris quantity estimates generated by the HAZUS-MH Hurricane Wind Model, as the observed quantities appear to be incremental. Comparing this observed data with HAZUS-MH would not result in a fair comparison. Table 32. Debris Estimates – Hurricanes Ivan, Dennis and Charley – Escambia, Santa Rosa, DeSoto, and Hardee Counties County Disaster Total C&D Veg Escambia Ivan NDR NDR NDR Santa Rosa Ivan NDR NDR NDR Escambia Dennis NDR NDR NDR Santa Rosa Dennis 17M TN NDR NDR DeSoto Charley 95K CY NDR 95K CY Hardee Charley 11K CY NDR 11K CY Hardee Cities Charley 165K CY NDR 165K CY NDR = No data received. Note: It was difficult to determine the amount of debris generated for Orange County, as not all of the debris had been collected for Charley when Frances made landfall. B-5 Injuries and Deaths Table 33. County Structures Replacement Cost – Hurricane Charley – DeSoto County Disaster County Injuries (Residential) Deaths (Residential) Charley DeSoto 12 1 April 2007 B-24 APPENDIX C ARA Prior HAZUS Validation Report DRAFT Final Report – November 22, 2004 HAZUS-MH Support for Hurricanes Charley, Frances, Ivan and Jeanne Prepared for: Federal Emergency Management Agency Mitigation Division 500 C Street, SW Washington, DC 20472 Prepared by: Applied Research Associates, Inc. 8540 Colonnade Center Drive, Suite 307 Raleigh, NC 27615 November 2004 ARA 16567 April 2007 C-1 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY 1. Introduction In August and September of 2004, four hurricanes made landfall in Florida and Alabama (see Table 1). In response to a request for support from the Federal Emergency Management Agency (FEMA) Response Division, FEMA’s Mitigation Division contracted with Applied Research Associates, Inc. (ARA) to provide near real-time loss estimation support using HAZUS-MH. ARA performed hurricane wind field modeling and hurricane wind loss estimates before, during, and after each of the four hurricanes. In addition, ABS Consulting performed coastal and inland flooding loss estimates for the first three hurricanes under a subcontract to ARA. Table 1. Summary of 2004 Hurricanes Assessed Using HAZUS-MH Hurricane Landfall Conditions Location Date NHC Saffir- Simpson Category NHC 1-minute Sustained Wind Speed (mph) Charley Charlotte County, FL 8/13/04 4 145 Frances Martin County, FL 9/5/04 2 105 Ivan Baldwin County, AL 9/16/04 3 130 Jeanne Martin County, FL 9/26/04 3 115 This report summarizes the analyses performed by ARA and ABS and lessons learned. The hurricane wind loss estimates performed by ARA are discussed in Section 2. The coastal and inland flooding loss estimates performed by ABS Consulting are discussed in Section 3. April 2007 C-2 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY 2. HAZUS-MH Hurricane Wind Loss Estimates This section summarizes the approach and lessons learned by ARA in developing hurricane storm tracks, wind fields, and wind loss estimates for Hurricanes Charley, Frances, Ivan, and Jeanne. At the time of the analyses, two versions of the HAZUS-MH hurricane model were available: 1. HAZUS-MH version 1.0 (Build 31) – the first and only official FEMA release of HAZUS-MH to date. 2. HAZUS-MH Build 36A – a developmental version of HAZUS-MH Maintenance Release 1 (MR1). The completed version of HAZUS-MH MR1 is expected to be delivered to FEMA in December 2004. With respect to near real-time estimates of hurricane wind losses, there are three key differences between Build 31 and Build 36A: 1. Build 36A includes an updated database of general building stock valuations. The baseline building valuations in the southeastern United States are approximately 18% lower than the valuations given in Build 31.1 In this report, the baseline building valuation data in Builds 31 and 36A are referred to as the “old” and “new” valuations, respectively. 2. Build 36A includes a new capability to automatically download and import National Hurricane Center (NHC) Forecast/Advisories (F/A) from the HurrEvac ftp site. In Build 31 this information had to be manually input into the HAZUS hurricane scenario wizard. 3. Build 36A includes a new capability to accept radii to 50 knot or 34 knot winds at points on the storm track where the maximum wind speeds (either observed or forecast) are less than hurricane force (i.e., the radius to 64 knot winds is undefined). This improvement allows better modeling of damage and loss as the system intensifies from a tropical storm to a hurricane and as it weakens again after landfall. Initially during Hurricane Charley, loss estimates were computed using both Build 31 and Build 36A and using both the old and the new building valuation data. Since Builds 31 and 36A produced essentially the same loss estimates given the same hurricane scenario definition and the same building inventory data, we decided to stop using Build 31 after Forecast/Advisory 17 (FA17) on August 13, 2004.2 Furthermore, because the “new” valuation data provide a better represent the true replacement value of the building stock in Florida and its neighboring states, we also decided to stop using the old valuation data after FA19 on August 13, 2004. In general, three categories of hurricane track models were evaluated for each storm: 1. FA. The Forecast/Advisory tracks directly model the information provided in the NHC Forecast/Advisory. These tracks were generated by mapping the FA parameters (i.e., central pressure, maximum sustained wind speed, and maximum radius to 64, 50, or 34 knot winds) to the fundamental HAZUS hurricane storm track parameters (i.e., central 1 The changes in general building stock valuation between Build 31 and Build36A are the result of a more refined implementation of the RSMeans regional cost multipliers. 2 See, for example, Hurricane Charley runs 3 and 5 in Table 3. April 2007 C-3 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY pressure, Holland profile parameter (B), and radius to maximum winds). This mapping was accomplished either by: (a) directly using the HurrEvac download and validation option in Build 36A, or (b) running the mapping algorithms off-line and manually inputting the resulting parameters into the HAZUS hurricane scenario wizard. 2. ARA. The ARA tracks use the NHC FA represent our best estimate of the hurricane intensity parameters based on data from the following sources: (1) NHC Forecast/Advisories, (2) H*Wind surface wind analysis output provided by Hurricane Research Division (HRD), and (3) surface level wind and pressure observations to determine the values of the profile parameter and radius to maximum winds. 3. Hwind. H*Wind tracks (actually peak 3-second wind gust estimates by census tract centroid) were provided to ARA as research products by the HRD. Two swaths were made available for each of the four storms: an initial estimate made near the time of landfall and a final estimate generally made within 2-4 days after landfall. Each of the three track models used the actual or forecast position and time information from the latest official NHC Forecast/Advisory. In a few cases, minor adjustments to the overland track positions were made in the ARA tracks to provide a smoother track because the storm coordinates in the Forecast/Advisories are rounded to tenths of degrees. The evolutions of HAZUS wind loss estimates for Hurricanes Charley, Frances, Ivan, and Jeanne are summarized in Tables 3-6.3 The following paragraphs highlight several observations and recommendations resulting from our analysis: Loss Estimates Based on NHC Tracks As with Hurricane Isabel in 2003, the general trend seen in Tables 3-6 is that the tracks based directly on NHC Forecast/Advisories tend to produce wind speeds and loss estimates that are significantly higher than the ARA or H*Wind tracks. Recommendation: Although the Forecast/Advisory import capability provides a simple and efficient method for estimating losses for an approaching hurricane, emergency managers should not rely solely on the direct use of NHC Forecast/Advisories for response planning and mobilization decisions. Loss Estimates Based on H*Wind Swaths With the exception of Hurricane Charley, the final loss estimates based on H*Wind swaths provided by HRD are very comparable to the loss estimates from the ARA tracks. To a large degree, this is to be expected because the ARA tracks rely heavily on the same off-shore and coastal observations used by H*Wind. However, the models do tend to diverge inland from the coast, primarily because of differences in the two filling models. This difference was most pronounced in Hurricane Charley. Recommendation: FEMA should continue to encourage the NHC to operationalize the H*Wind surface wind analysis capability as a needed decision support tool for response planning. 3 Complete sets of summary reports (quick, global, economic, shelter, and debris) and peak gust wind swath maps have been archived for each run and can be furnished to FEMA in electronic format upon request. April 2007 C-4 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Loss Estimates Based on ARA Tracks Over the course of the four hurricanes, ARA developed a repeatable process for extracting the storm track information from multiple sources, mapping this information into the format required by the HAZUS hurricane model, generating peak gust estimates, and iterating once or twice, if necessary. This process was generally carried out during the first 30 minutes or so following the release of each Forecast/Advisory and the results were immediately transmitted to FEMA via e- mail or ftp. Once the track was finalized and sent to FEMA, loss calculations were performed by ARA using the HAZUS-MH tool. In general, copies of five summary reports (quick, global, economic, shelter, and debris) and a map of the peak gust wind swath were archived for each run. The remaining time between advisories was used to perform comparisons of modeled wind speed traces, wind direction traces, and atmospheric pressure traces to observations at as many locations a possible. This process was repeated through landfall. A final “best estimate” ARA track was generally available within 2-3 days of landfall. Recommendation: The process for estimating the track parameters should be automated as much as practical in order to ensure timely, accurate, and cost-effective loss estimates for future hurricanes. However, we should not attempt to fully automate this process. Expert knowledge and judgment will always be necessary to identify and address errant observational data and other unusual situations. Storm Track Time Step Hurricane Charley was an intense, small, and fast-moving storm. The current time interval used in HAZUS-MH for evaluating peak gust wind speeds is once every 15 minutes. This time step was too large for Hurricane Charley and resulted in an underestimate of peak gust wind speeds in some census tracts. This was confirmed in Run #31, which was a re-analysis of Run #29 using a 3-minute time step instead of a 15-minute time step. The loss estimate for Run #31 increased by over 75%. Recommendation: An algorithm should be developed and implemented in HAZUS-MH to allow a variable time step based on the forward translation speed, size, and intensity of the storm. Set-up, Analysis, and Post-Processing Times The time required to create the run the analysis and process the results after the release of a new Forecast/Advisory and the development of the new storm track parameters was generally on the order of 20-40 minutes (depending on the size of the study region) on a 3 GHz Pentium 4 with 1 GB of RAM. When the forecast track shifted outside the bounds of the previous study region, an additional time of 10-20 minutes was usually required to generate a new study region. Although these times were generally viewed to be acceptable, the following opportunities for reducing the turnaround time have been identified. Recommendation: A tool should be added to the hurricane model to automatically generate a pre-selected set of reports and maps at the end of a scenario analysis. This task has been funded under the current HAZUS contract and will be implemented in MR2. Recommendation: A tool should be added to the hurricane model to automatically generate the list of counties needed to encompasses a user-specified storm track. There should be an option to review this information and pass it directly to the HAZUS shell to create a new study region. Loss Estimates Initial estimates of industry-wide insured losses have been released by ISO for each of the four hurricanes. The ISO estimates cannot be directly compared to the estimates produced by HAZUS-MH because the ISO estimates include losses for automobiles and boats, appurtenant April 2007 C-5 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY structure losses, and additional living expenses, yet do not include deductibles or uninsured properties. In spite of these differences, insured loss estimates do provide a useful benchmark for the HAZUS-MH wind loss estimates. Table 2. Summary of Initial ISO Insured Loss Estimates Hurricane Landfall Date ISO Press Release Date Initial ISO Insured Loss Estimate ($B) HAZUS-MH Estimate Based on Final ARA Tracks from Tables 3-6 ($B) States Included Charley 8/13/04 8/25/04 6.7 7.1 FL Frances 9/5/04 9/23/04 4.1 1.8 FL Ivan 9/16/04 10/14/04 5.3 1.6 FL, AL, GA Jeanne 9/26/04 10/26/04 2.8 2.8 FL 2004 Total 18.9 13.3 Also shown in Table 2 are the HAZUS-MH loss estimates based on the final ARA storm tracks. It can be seen that the ISO and HAZUS-MH estimates for Hurricanes Charley and Jeanne are very similar, but there are significant differences in the estimates for Hurricanes Frances and Ivan. Further investigation is required to better understand these differences.4 Recommendation: A follow-on task should be planned to perform a more detailed analysis of the HAZUS-MH loss estimates relative to the final ISO loss estimates for all four hurricanes. 4 Upon further review of the Hurricane Frances results, it now appears that the final ARA track model developed for Run #31 may have given too much weight to the surface level wind speeds measured at the Florida Coastal Monitoring Program towers. Therefore, the previous ARA best track estimate developed for Run #29 may be a better representation of the actual storm track than the track that was subsequently developed for Run #31. The resulting loss estimate from Run #29 was $3.2B. More recent runs for Hurricane Frances have produced results between $3.0B and $5.8B, through updates in wind model parameters brought about through correction of the FCMP tower wind speeds for terrain effects. Major updates to the Ivan estimates due to hurricane modeling issues are not expected April 2007 C-6 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 3. Wind Loss Estimates for Hurricane Charley Peak Displaced Gust Households Run Date Build Valuation Study Region Scenario Advisory Track (mph) Loss ($B) (1,000's) 1 13-Aug 31 Old Charley_FL31Cnty_B31_OldVal Charley_B31_16_90pct_PJV 16 ARA 119 5.0 9.9 2 13-Aug 36A New Charley_FL31Cnty_B36A_NewVal Charley_2004_160 16 FA 126 12.0 33.8 3 13-Aug 31 Old Charley_FL31Cnty_B31_OldVal Charley_FLCnty31_B31_OldVal_CliffFA16 16 FA 122 11.4 25.5 4 13-Aug 36A Old Florida26 Charley_Cliff_FA16_Try2 16 FA 122 11.3 26.6 5 13-Aug 36A Old Florida31 Charley_Cliff_FA16_Try2 16 FA 122 11.4 26.6 6 14-Aug 31 Old Charley_16 Charley_16 16 ARA100 127 13.9 33.3 7 13-Aug 36A Old Florida26 Charley_2004_160 16 FA 126 15.7 37.4 8 13-Aug 36A Old Florida31 Charley_2004_160 16 FA 126 15.7 37.4 9 13-Aug 36A New Charley_FL31Cnty_B36A_NewVal C_FL31Cnty_B36A_New_AF17_90_PJV 17 ARA 118 4.3 8.5 10 13-Aug 36A Old Florida31 Charley_2004_170 17 FA 134 32.1 89.4 11 13-Aug 36A New Charley_FL31Cnty_B36A_NewVal C_PJV_Adv17 17 ARA 131 24.6 73.5 12 13-Aug 31 Old Charley_FL31Cnty_B31_OldVal C_FL31Cnty_B31_old_AF17Cliff 17 FA 125 27.5 64.4 13 13-Aug 36A New Charley_FL28Cnty_B36A_NewVal Charley_2004_180 18 FA 160 17.1 68.1 14 13-Aug 36A New Charley_FL28Cnty_B36A_NewVal Adv18_Reduced_B 18 ARA 142 9.7 28.6 15 13-Aug 36A New Charley_FL28Cnty_B36A_NewVal Charley_2004_190 19 FA 163 10.9 44.1 16 13-Aug 36A New Charley_FL28Cnty_B36A_NewVal Charley_AF19_Reduced_Rmax 19 ARA 147 7.9 24.9 17 13-Aug 36A New Charley_FL31Cnty_B31_NewVal Charley_2004_180 18? FA 134 24.4 79.2 18 13-Aug 36A Old Florida31 Charley_FA17_Cliff 17 FA 125 27.5 66.2 19 13-Aug 36A Old Florida28 Charley_2004_180 18 FA 160 21.1 72.0 20 13-Aug 36A New Charley_FL26Cnty Charley_2004_160 16 FA 126 12.0 33.8 21 13-Aug 36A Old Florida26 Charley_2004_160 16 FA 126 15.7 37.4 22 13-Aug 36A Old Florida26 HWIND_FA17 17 Hwind 136 8.4 21.2 26 13-Aug 36A Old Florida28 Charley FA19 Cliff 19 FA 167 16.6 58.5 27 14-Aug 36A New Charley_FL28_B36A_NewVal Charley_2004_211 21 FA 161 5.2 20.0 28 14-Aug 36A New Charley_FL28_B36A_NewVal Charley_2004_211Corrected 21 FA 161 5.2 20.0 29 14-Aug 36A New Charley_FL28_B36A_NewVal FA22_ARA4 22 ARA 146 6.4 18.0 30 14-Aug 36A New Charley_FL28_B36A_NewVal hwind_5pmSat_charley 22 Hwind 159 3.2 12.3 31 23-Aug 36A New Charley_FL28_B36A_NewVal FA22_ARA4 22 ARA 146 11.4 34.6 32 27-Aug 36A New Charley_FL28_B36A_NewVal hwind_charley_27aug04 22 Hwind 147 2.7 7.7 33 30-Aug 36A New Charley_B36A_NewVal_FL28Cnty Charley_ARA_30Aug2004 22 ARA 150 7.1 22.3 Landfall at ~5pm EDT (2100Z) on 8/13 (same time as FA19) Advisory FA ARA Hwind FA18 17.1 9.7 Hurricane Charley Loss Estimates FA19 10.9 7.9 FA21 5.2 FA22 6.4 3.2 FA22-3min 11.4 18 FA18 FA19 FA21 FA22 FA22-3min Final Advisory Final 7.1 2.7 FA 16 ARA Hwind 14 12 10 8 6 4 2 0 Landfall at FA19 Direct Economic Loss ($B) April 2007 C-7 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 4. Wind Loss Estimates for Hurricane Frances Displaced Peak Gust Households Run Date Build Valuation Study Region Scenario Advisory Track (mph) Loss ($B) (1,000's) 1 1-Sep 36A New Frances_B36A_NewData_15WF_FL Frances_PJV_20040901_1718_32_Alt 32 ARA 147 31.7 114.9 2 1-Sep 36A New Frances_B36A_NewData_15WF_FL Frances_PJV_20040901_1718_32_FA 32 FA 164 61.9 273.1 3 2-Sep 36A New Frances_B36A_NewData_15WF_FL Frances_PJV_2004_340 4 2-Sep 36A New Frances_B36A_NewData_15WF_FL Frances_PJV_34_Alt 5 2-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_Adv35Alt 6 2-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_Adv35 7 2-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_2004_350 8 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_2004_380 9 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_38 10 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_38_ALT 11 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_39_ALT 12 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_Hurrevac_39 13 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_39 14 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrancesHwind39 15 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_2004_400 16 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_40 17 3-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_40_ALT 18 4-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_42_ALT 19 4-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_42 20 4-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_2004_421 21 4-Sep 36A New Frances_B36A_NewVal_20040904_57Cnty Frances_PJV_43_Alt 22 4-Sep 36A New Frances_B36A_NewVal_20040904_57Cnty Frances_2004_430 26 4-Sep 36A New Frances_B36A_NewVal_20040904_57Cnty Frances_PJV_43 27 5-Sep 36A New Frances_B36A_NewVal_20040904_57Cnty Frances_PJV_47_Alt 28 6-Sep 36A New Frances_B36A_NewVal_20040902_1004_49CnFrances_PJV_50_Alt2 29 6-Sep 36A New Frances_B36A_NewVal_FL_Final Frances_PJV_51_Alt 30 10-Sep 36A New Frances_B36A_NewVal_FL_Final Frances_HWIND_091004 31 28-Sep 36A New Frances_B36A_NewVal_20040904_57Cnty Frances_28Sep04_ARA Landfall at ~1am EDT (0500Z) on 9/5 (between FA45 at 0300Z and FA46 at 0900Z) Direct Economic Loss ($B) 120 100 80 60 40 20 0 Hurricane Frances Loss Estimates FA ARA Hwind 34 FA 169 116.1 501.0 34 ARA 151 41.1 160.6 35 ARA 148 24.3 89.4 35 FA 167 78.1 340.7 35 FA 170 79.7 339.9 38 FA 159 65.7 251.7 38 FA 162 69.7 287.4 38 ARA 148 21.3 79.4 39 ARA 120 4.3 8.6 39 FA 136 18.1 52.2 39 FA 137 16.8 49.8 39 Hwind 114 2.6 4.4 40 FA 142 22.7 80.7 40 FA 145 29.5 102.9 40 ARA 129 10.2 25.0 42 ARA 118 6.7 13.2 42 FA 131 17.2 45.7 42 FA 125 18.8 46.0 43 ARA 117 6.4 12.3 43 FA 125 9.2 21.4 43 FA 131 17.8 48.7 47 ARA 115 3.1 5.9 50 ARA 115 3.4 6.1 51 ARA 115 3.2 6.0 Final Hwind 111 2.7 5.6 Final ARA 106 1.8 2.3 Advisory FA ARA Hwind FA32 61.9 31.7 FA34 116.1 41.1 FA35 79.7 24.3 FA38 65.7 21.3 FA39 18.1 4.3 2.6 FA40 22.7 10.2 FA42 18.8 6.7 FA43 9.2 6.4 FA47 3.1 FA50 3.4 FA51 3.2 Final 1.8 2.7 FA32 FA34 FA35 FA38 FA39 FA40 FA42 FA43 FA47 FA50 FA51 Final Advisory Landfall between FA45 and FA46 April 2007 C-8 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 5. Wind Loss Estimates for Hurricane Ivan Displaced Peak Gust Households Run Date Build Valuation Study Region Scenario Advisory Track (mph) Loss ($B) (1,000's) 1 14-Sep 36A New Ivan_FL_GA_AL_MS_LA_Expanded Ivan_2004_480 48 FA 165 48.6 227.8 2 14-Sep 36A New Ivan_FL_GA_AL_MS_LA_Expanded Ivan_PJV_Adv48 48 FA 164 38.1 174.6 3 14-Sep 36A New Ivan_FL_GA_AL_MS_LA Ivan_PJV_Adv48_Alt 48 ARA 146 11.6 40.2 4 14-Sep 36A New Ivan_FL_GA_AL_MS_LA_Expanded Ivan_PJV_Adv49 49 FA 159 37.8 184.3 5 14-Sep 36A New Ivan_FL_GA_AL_MS_LA Ivan_PJV_Adv49_Alt 49 ARA 149 21.2 97.8 6 15-Sep 36A New Ivan_FL_GA_AL_MS_LA_Expanded Ivan_2004_520 52 FA 147 13.6 52.6 7 15-Sep 36A New Ivan_FL_GA_AL_MS_LA_Expanded Ivan_PJV_Adv52 52 FA 156 22.1 98.3 8 15-Sep 36A New Ivan_FL_GA_AL_MS_LA Ivan_PJV_Adv52_Adj 52 ARA 137 7.8 25.1 9 15-Sep 36A New Ivan_FL_GA_AL_MS_LA_Expanded Ivan_PJV_Adv53 53 FA 152 30.2 131.7 10 15-Sep 36A New Ivan_FL_GA_AL_MS_LA Ivan_PJV_Adv53_Alt 53 ARA 142 11.4 42.9 11 16-Sep 36A New Ivan_FL_GA_AL_MS_LA_Expanded Ivan_2004_560 56 FA 149 19.6 72.5 12 16-Sep 36A New Ivan_FL_GA_AL_MS_LA Ivan_PJV_Adv56_Adj 56 ARA 118 1.4 2.0 13 16-Sep 36A New Ivan_FL_GA_AL_MS_LA_Expanded Ivan_HWND_091604_1107am 56 Hwind 121 1.7 2.9 14 16-Sep 36A New Ivan_FL_GA_AL_MS_LA Ivan_PJV_Adv57_Adj 57 ARA 118 1.6 2.6 15 17-Sep 36A New Ivan_FL_GA_AL_MS_LA Ivan_PJV_Adv60_Adj 59 ARA 118 1.6 2.6 16 17-Sep 36A New Ivan_FL_GA_AL_MS_LA_Expanded Ivan_HWND_091704_1352 59 Hwind 121 1.7 2.9 Landfall at ~2am CDT (0700Z) on 9/16 (between FA55 at 0300Z and FA56 at 0900Z) Advisory FA ARA Hwind FA48 38.1 11.6 Hurricane Ivan Loss Estimates FA49 37.8 21.2 FA52 22.1 7.8 FA53 30.2 11.4 FA56 19.6 1.4 1.7 FA57 1.6 Final 1.6 1.7 FA48 FA49 FA52 FA53 FA56 FA57 Final Advisory Landfall just before FA56 0 5 10 15 20 25 30 35 40 Direct Economic Loss ($B) FA ARA Hwind April 2007 C-9 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Table 6. Wind Loss Estimates for Hurricane Jeanne Displaced Peak Gust Households Run Date Build Valuation Study Region Scenario Advisory Track (mph) Loss ($B) (1,000's) 1 24-Sep 36A New Jeanne_FL_NoPanhandle Jeanne_PJV_adv44_Adj 44 ARA 85 0.3 0.1 2 24-Sep 36A New Jeanne_FL_NoPanhandle Jeanne_PJV_adv48 44 ARA 115 3.8 7.0 3 25-Sep 36A New Jeanne48 Jeanne_48_ARA 48 ARA* 130 12.6 33.5 4 25-Sep 36A New Jeanne48 Jeanne_48_ARA_NoIntensification 48 ARA 117 5.1 9.6 5 25-Sep 36A New Jeanne48 Jeanne_49_HRD_Based 49 ARA* 140 25.1 77.7 6 25-Sep 36A New Jeanne48 Jeanne_49_ARA 49 ARA 134 17.1 46.3 7 25-Sep 36A New Jeanne48 Jeanne_49_Hwind 49 Hwind 123 2.6 5.1 8 26-Sep 36A New Jeanne_51 9 27-Sep 36A New Jeanne_51 10 27-Sep 36A New Jeanne_51 11 28-Sep 36A New Jeanne_51 12 28-Sep 36A New Jeanne_51 Jeanne_51_ARA2 Jeanne_52_Hwind Jeanne_56a_ARA Jeanne_56a_ARA2 Jeanne_28Sep04_Hwind Landfall just after 11pm EDT on 9/25 (i.e., just after FA50 at 0300Z on 9/26) Direct Economic Loss ($B) 18 16 14 12 10 8 6 4 2 0 Hurricane Jeanne Loss Estimates FA44 FA48 FA49 FA51 FA56 Final Advisory FA ARA Hwind Landfall at FA50 51 ARA 110 2.5 4.1 52 Hwind 105 0.8 0.9 56 ARA 105 1.5 1.9 56 ARA 113 2.9 5.1 56 Hwind 114 2.7 4.2 Advisory FA ARA Hwind FA44 3.8 0.3 FA48 5.1 FA49 17.1 2.6 FA51 2.5 FA56 1.5 Final 2.9 2.7 April 2007 C-10 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY 3. HAZUS-MH Flood Loss Estimates 3.1 Hurricane Charley This section documents the work performed by ABS Consulting using the HAZUS-MH floodmodel for Hurricane Charley. The work began on August 13, the day Charley made landfall on the western side of Florida. HAZUS flood model study regions were constructed for the coastal counties predicted to be in Charley’s path. Inputs Using the results of SLOSH runs posted on the National Hurricane Center (NHC) website, the coastal flood hazard was computed based on the SLOSH storm surge elevations. By the time the hurricane made landfall, the actual storm track was south of the predicted track. As such, additional flood model study regions were built, and the coastal hazard was recomputed based on the most recent SLOSH results. The DEM for each county was downloaded from the National Elevation Dataset (NED) on the USGS website. Following are pertinent details regarding the flood model runs: Charley hurricane advisory 18 HAZUS build number 31 HAZUS DVD data version Build 36 Counties analyzed Florida: Charlotte, Collier, Lee, Monroe, Sarasota Approach/Methodology We ran each county with the storm surge elevation that we received from the NHC. The projected surge elevations showed little variation across each county, so a uniform surge elevation was applied to each county. In HAZUS, only General Building Stock (Damage and Loss) and Essential Facilities analyses were run in the interest of time. Results The projected surge elevations and flood discharges for each county is shown in the table below. The resulting loss estimates are also shown below. County Surge Elevation (ft) Building Loss ($) Processing Time Hazard Analysis Charlotte 4.5 471,520 3 min 9 hours Collier 8 6,318,000 3 min 12 hours Lee 13.5 48,472,000 6 min 18 hours Monroe 5.5 2,537,000 2 hr 10 min 9 hours Sarasota 3 704,000 10 min 3 hours Lessons Learned The exercise of applying the HAZUS flood model for coastal flooding has brought forth some lessons that can be applied to future hurricanes: . Study Region Size – The number of census blocks impacted over a county due to a large coastal flood can be quite high. As such, it’s recommended to build single county study regions for coastal analysis. . Storm Track – In the final hurricane advisories prior to landfall, the storm track changed rather significantly. Due to this, the affected counties and associated storm April 2007 C-11 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY surges can also change in a short period of time. Hence, it is recommended to delay running of the coastal model until the true storm track is known. . Validation – For future hurricanes, it would be useful to acquire measured storm surge elevations and apply them to the coastal model, in order to help validate the accuracy of the model results. At this time it is not clear if the surge elevations computed in the final SLOSH runs mimicked actual surge elevations. . Computer Requirements – At this time, coastal analysis in the HAZUS flood model is much more computer-intensive than riverine analysis. It is important to use the most powerful machine available in terms of processor speed and RAM. . Relative Impact – For Hurricanes Charley, Frances, and Ivan, hurricane winds and riverine flooding had a greater impact on damage than coastal storm surge. Figure 1. NHC Storm Surge Map 3.2 Hurricane Frances This section documents the work performed by ABS Consulting using the HAZUS-MH flood model for Hurricane Frances. The work began with a flood analysis of a swath of counties in Central Florida. As the storm moved west, another group of counties in and around the Florida panhandle were run through the flood model. Later the storm moved north and stalled in the region along the North Carolina – Tennessee border. A third group of counties were run for this area. This was the first effort to use the HAZUS Flood Model over a large area in a real-time April 2007 C-12 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY setting. In an effort to reduce processing times, these first three groups of study regions were analyzed for dollar exposure, not for damage and loss. Later in the process, discharge values for selected streams were obtained by FEMA for two counties in central Florida, and a few areas in western North Carolina. Study regions for these areas were also constructed. For these study regions, dollar losses were computed. Inputs A map of the area to be analyzed was posted on the National Hurricane Center (NHC) website and is attached at the end of this report. John Ingargiola, of the Department of Homeland Security, provided a document with the peak discharges from gages of the affected area. The DEM for each county was downloaded from the National Elevation Dataset (NED) on the USGS website. Following are pertinent details regarding the flood model runs: Frances hurricane advisory 50 HAZUS build number 37+ (Build 37 with updated flood software) HAZUS DVD data version Build 36 Counties analyzed Florida: Hillsborough, Pasco SR1-Brevard, Indian River, St. Lucie (all in FL) SR2-Seminole, Orange, Osceola, Okeechobee (all in FL) SR3-Polk, Hardee, DeSoto, Highlands (all in FL) SR4-Pinellas, Hillsborough, Manatee, Sarasota (all in FL) SR5-Hernando, Pasco, Sumter, Lake (all in FL) SR6-Calhoun, Gulf, Liberty, Franklin (all in FL) SR7-Gadsen, Leon, Wakulla (all in FL) SR8-Walton, Holmes, Washington, Bay (all in FL) SR9-Geneva, AL, Houston, AL, Jackson, FL SR10-Early, Miller, Seminole, Decatur (all in GA) SR11-Graham, Swain, Jackson, Haywood (all in NC) SR12-Blount, Sevier, Jefferson, Cocke (all in TN) SR13-Hambler, Greene, Washington, Unicoi (all in TN) SR14-Madison, Buncombe, Yancey, McDowell (all in NC) SR15-Mitchell, Avery, Burke, Caldwell (all in NC) Florida: Manatee, Pasco, Hillsborough North Carolina: Buncombe, Madison, Transylvania, Gaston, Stanly Approach/Methodology Given the relatively low storm surges predicted for Hurricane Frances, the focus was on riverine flooding. The 15 study regions in Florida were analyzed for dollar exposure, not for damage and loss. The selection of specific counties to be analyzed was based on the flood outlook from NOAA’s Southeast River Forecast Center, obtained through a link on the FEMA website. Areas identified on the map as “Significant River Flooding Likely” were included in flood study regions. The maps are shown as Figures 2 and 3. Due to the large amount of inventory data and cell size of the depth grids, the flood model is ideally run with one county per region. To cut down analysis time, river networks were defined based on a 25-square mile drainage area, resulting in fairly large streams. Peak rainfall estimates were received but those data were difficult to apply to HAZUS because the flood model does not perform rainfall-runoff modeling. The flood hazard was based on current/peak discharges at gages provided by FEMA. The gages had to be located in each study region, in which the information was provided or was April 2007 C-13 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY found on NOAA’s website. In HAZUS, the hydrologic analysis was skipped because the discharges were already provided. The discharges were distributed downstream of each respective gage. For study regions in which damage analyses were performed, the results include the General Building Stock (Damage and Loss) and Essential Facilities. The results also contain exposure estimates for each of the analysis areas, in terms of full replacement costs and depreciated costs for residential and government occupancies, along with population totals. The totals include any census blocks that intersect the floodplain boundary. Losses were only computed for the Florida and North Carolina single county regions. Results The projected flood discharges for each county are shown in the following table. The resulting loss estimates are not available due to problems in the analysis portion of the program. April 2007 C-14 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY County Flood Discharge River Name Processing Time Hazard Analysis SR1 100-yr Entire SR NA SR2 100-yr Entire SR 9 hrs NA SR3 100-yr Entire SR NA SR4 100-yr Entire SR 6.5 hrs NA SR5 100-yr Entire SR 9 hrs NA SR6 100-yr Entire SR 6.5 hrs NA SR7 100-yr Entire SR 4 hrs NA SR8 100-yr Entire SR NA SR9 100-yr Entire SR NA SR10 100-yr Entire SR 6 hrs NA SR11 100-yr Entire SR 6.5 hrs NA SR12 100-yr Entire SR 10 hrs NA SR13 100-yr Entire SR NA SR14 100-yr Entire SR NA SR15 100-yr Entire SR 7.5 hrs NA Hillsborough, FL 5,000, 7,000, 11,500 cfs Hillsborough, Alafia, & Little Manatee 4 hrs 1.5 hrs Pasco, FL 3,000 cfs Anclote 4 hrs 30 min Buncombe, NC 50,000 cfs French Broad 1.5 hrs 20min Madison, NC 43,000 and 64,500 cfs French Broad 1.5 hrs 20min Transylvania, NC 22,000 cfs French Broad 1 hr 25 min Gaston, NC 10,200 cfs South Fork Catawba 1 hr 20 min Stanly, NC 52,500 cfs Rocky 12 min 20 min Lessons Learned Through this analysis, several lessons were learned that can be applied to the use of the HAZUS flood model during future hurricanes: . Computers – There is a 4-county limit on the size of flood model study regions due to the system limits on the size of study region databases. As such, in order to analyze a large area, it is necessary to build multiple study regions. With time crunches typical with emergency response, one of the most important factors in the analysis for these study regions was the availability of multiple computers, which allowed for simultaneous analysis. . Study Region Creation – Based on the river to be analyzed, care should be exercised during study region creation. If a river to be analyzed forms a border between counties, the study region should include the counties on both sides of the river. Otherwise, the synthetic stream network may not be continuous. . Processing Time – The computers used for this exercise had 3.2 GHz processors, with 2 GB of RAM, and tens of gigabytes of hard drive space. For HAZUS flood analysis, the processor speed and RAM are very important. The total processing times are listed in the tables above. April 2007 C-15 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 2. Florida Forecasted Affected Areas April 2007 C-16 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 4. North Carolina Forecasted Affected Areas Figure 5. Central Florida Counties Analyzed April 2007 C-17 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 6. Florida Panhandle Counties Analyzed Figure 7. North Carolina – Tennessee Border Counties Analyzed April 2007 C-18 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 8. North Carolina Study Regions 3.3 Hurricane Ivan This section documents the work performed by ABS Consulting using the HAZUS-MH flood model for Hurricane Ivan. Inputs Using the results of SLOSH runs posted on the National Hurricane Center (NHC) website, the coastal flood hazard was computed based on the SLOSH storm surge elevations, as seen in Figure 9. For riverine analyses, John Ingargiola, of the Department of Homeland Security, provided documents with the peak discharges from gages of the affected area. The DEM for each county was downloaded from the National Elevation Dataset (NED) on the USGS website. April 2007 C-19 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Following are pertinent details regarding the flood model runs: Ivan hurricane advisory 56 HAZUS build # 37+ (Build 37 with updated flood software) HAZUS DVD data version Build 36 Counties analyzed Coastal: Baldwin, AL, Escambia, FL, Santa Rosa, FL Riverine: SR1-Buncombe, NC, Madison, NC SR2-Transylvania, NC, Henderson, NC SR3-Graham, NC, Swain, NC SR4-Allegheny, PA, Beaver, PA SR5-Allegheny, PA, Westmoreland, PA, Armstrong, PA SR6-Jefferson, OH, Brooke, WV, Hancock, WV SR7-Ohio, WV, Marshall, WV, Belmont, OH, Monroe, OH Approach/Methodology In order to run the coastal analysis, the storm surge elevation for each county’s coast was estimated from the SLOSH results. The projected surge elevations showed little variation across each county, so a uniform surge elevation was applied to each county. In HAZUS, the General Building Stock (Damage & Loss) and Essential Facilities analyses were computed. For the riverine analysis, aggregating flood study regions in the flood model only allows for a maximum of 4 counties per region, so five study regions were constructed for the exposure analysis. The riverine analysis consisted of multi-county study regions, seen below in Figures 10 through 12. Specific study region maps are included in the zip file. Due to the large amount of inventory data and cell size of the depth grids, the flood model is ideally run with one county per region. To cut down analysis time, river networks were defined based on a 25-square mile drainage area, resulting in fairly large streams. River discharges were computed using hydrologic regression analysis. As such, floodplains were determined assuming a 100-year return period flood. The flood hazard was based on current/peak discharges at gages provided by FEMA. The gages had to be located in each study region, in which the information was provided or was found on NOAA’s website. In HAZUS, the hydrological analysis was skipped because the discharges were already provided. The discharges were distributed downstream of each respective gage. The analyses ran in HAZUS for each study region includes the General Building Stock (Damage and Loss) and Essential Facilities. Exposure estimates for each of the analysis areas were computed, in terms of full replacement costs and depreciated costs for residential and government occupancies, along with population totals. The totals include any census blocks that intersect the floodplain boundary. The global summary reports were produced for each of the study regions. The exceptions are study regions 6 and 7 along the Ohio River. There were technical difficulties with the flood model that didn’t allow for loss computations at the time of this report. Results The projected surge elevations and flood discharges for each county is shown in the table below. The resulting loss estimates are not available due to problems in the analysis portion of the program. April 2007 C-20 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Coastal Analysis County Surge Elevation (ft) Coast Name Processing Time Hazard Analysis Baldwin, AL 7 Baldwin 5 min 7 hr 41min Escambia, FL 7 Escambia 7 min 3 hr 8 min Santa Rosa, FL 7.5 and 3.6 Santa Rosa 2 min 6 hr 2 min Riverine Analysis County Flood Discharge (cfs) River Name Processing Time Hazard Analysis Buncombe, NC 30,300 French Broad 1 hr 32 min 16 min Madison, NC 48,000 and 74,600 French Broad 1 hr 32 min 16 min Transylvania, NC 19,400 French Broad 45 min 14 min Graham, NC 30,750 Tuskasegee 58 min 22 min Swain, NC 30,750 Tuskasegee 58 min 22 min Henderson, NC 19,400 French Broad 45 min 14 min Allegheny, PA 197,000 and 193,000 and 178,000 and 325,000 Allegheny, Ohio 1 hr 23 min, 1 hr 25 min 1 hr 31min, 47 hr 46min Westmoreland, PA 194,000 Allegheny 1 hr 23min 1 hr 31min Armstrong, PA 174,000 Allegheny 1hr 23min 1 hr 31min Beaver, PA 361,000 Ohio 1 hr 25min 47 hr 46min Jefferson, OH 356,000 and 351,000 and 346,000 Ohio 36 min 4 min Belmont, OH 353,000 Ohio 1 hr 32 min 8 min Monroe, OH 346,000 Ohio 1 hr 32 min 8 min Brooke, WV 368,000 Ohio 36 min 4 min Hancock, WV 368,000 Ohio 36 min 4 min Ohio, WV 353,000 Ohio 1 hr 32min 8 min Marshall, WV 357,000 Ohio 1 hr 32min 8 min Lessons Learned Through this analysis, several lessons were learned that can be applied to the use of the HAZUS flood model during future hurricanes: . Computers – There is a 4-county limit on the size of flood model study regions due to the system limits on the size of study region databases. As such, in order to analyze a large area, it is necessary to build multiple study regions. With time crunches typical with emergency response, one of the most important factors in the analysis for these study regions was the availability of multiple computers, which allowed for simultaneous analysis. . Processing Time – The computers used for this exercise had 3.2 GHz processors, with 2 GB of RAM, and tens of gigabytes of hard drive space. For HAZUS flood analysis, the processor speed and RAM are very important. The total processing times are listed in the tables above. April 2007 C-21 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY . Riverine Study Region Creation – Based on the river to be analyzed, care should be exercised during study region creation. If a river to be analyzed forms a border between counties, the study region should include the counties on both sides of the river. Otherwise, the synthetic stream network may not be continuous. . Coastal Study Region Creation – At this time it is recommended to build a single county study region for coastal analysis. The number of census blocks impacted over a county due to a large coastal flood can be quite high, resulting in long general building stock processing times. . Storm Track – In the final hurricane advisories prior to landfall, the storm track changed rather significantly. Due to this, the affected counties and associated storm surges can also change in a short period of time. Hence, it is recommended to delay running of the coastal model until the final storm track is known. . Validation – For future hurricanes, it would be useful to acquire measured storm surge elevations and apply them to the coastal model, in order to help validate the accuracy of the model results. At this time it is not clear if the surge elevations computed in the final SLOSH runs mimicked actual surge elevations. Figure 9. SLOSH Output Map April 2007 C-22 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 10. Alabama-Florida Study Regions Figure 11. North Carolina Study Regions April 2007 C-23 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Figure 12. Ohio-Pennsylvania-West Virginia Study Regions April 2007 C-24 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY APPENDIX D ACKNOWLEDGEMENTS: CONTACT LOG Caller Contact Name Agency Or Organization County Data Requested For Requested Date Of Phone Or Email Contact Date Data Received Comments Stacy Robinson Rick Burgess Charlotte County GIS Charlotte Follow-up on Tom's request for county damage data 12/2/2005; 12/6/05 No response Stacy Robinson Ron Peacock Charlotte Co Schools Charlotte Follow-up on Tom's request for school damage data 12/2/2005 No response Stacy Robinson Danny Killcollins FDEM ALL Statewide shelter info for Charley, Ivan, and Dennis 12/2/2005 Referred to Barbara Bratcher Stacy Robinson Barbara Bratcher FDEM ALL Statewide shelter info for Charley, Ivan, Dennis, Francis, and Jeanne 12/2/2005 Referred to Michael Whitehead Stacy Robinson Michael Whitehead FLDBPR ALL Statewide shelter info for Charley, Ivan, and Dennis 12/5/2005; 1/11/2006 12/5/2005 and ongoing Rec'd pre- and post-landfall shelter estimates for all counties for Charley, Ivan, Dennis, Jeanne, and Frances. Stacy Robinson AC Castello Florida Hospital Assoc ALL Statewide hospital damage info for Charley, Ivan, and Dennis 12/2/2005 Sent a report and was referred to Debbie Hegerty Stacy Robinson Debbie Heggerty Florida Hospital Assoc ALL Statewide hospital damage info for Charley, Ivan, and Dennis 12/2/2005 Rec'd what limited data they had Stacy Robinson Karen Thornhill Santa Rosa Co Santa Rosa Follow-up on Tom's request for county damage data 12/2/2005 Rec'd limited data (apparently what Tom asked for) April 2007 D-1 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Caller Contact Name Agency OrOrganization County DataRequested For Requested Date Of Phone Or Email Contact Date Data Received Comments Lisa Flax Karen Thornhill Santa Rosa Co Santa Rosa Sent e-mail to request data per revised collection forms 1/17/2006 Lisa Flax Karen Thornhill Santa Rosa Co Santa Rosa 1/18/2006 Left follow-up message Lisa Flax Karen Thornhill Santa Rosa Co Santa Rosa 1/19/2006 Left follow-up message Lisa Flax Gregg Cotton Santa Rosa Co PW Santa Rosa Debris estimate 1/18/06 1/18/06 17.M TN for Dennis Stacy Robinson NW FL Red Cross NW FL Red Cross Escambia & Santa Rosa Shelter counts, damaged housing counts and displaced HH 12/2/2005; 1/12/06 No response Dent follow-up email 1/12 Stacy Robinson Charlotte Co Red Cross (mike Lee) Charlotte Co Red Cross Charlotte Shelter counts, damaged housing counts, and displaced HH 12/2/2005; 1/12/06 No response Sent follow-up email 1/12 Stacy Robinson Becky Sebren Central FL Red Cross Orange Shelter counts, damaged housing counts, and displaced HH 12/2/2005; 1/11/2006 12/5; awaiting another response Referred to Orange County EM for damaged housing counts Stacy Robinson Luke Wood Manatee Red Cross Hardee Shelter counts, damaged housing counts, and displaced HH 12/2/2005; 1/11/2006 12/5/2006 Requested Frances and Jeanne 1/11/2006 Stacy Robinson SW FL Red Cross SW FL Red Cross DeSoto Shelter counts, damaged housing counts, and displaced HH 12/2/2005 No response Sent follow-up email 1/12 to Bill Sullivan (see contact below) Stacy Robinson Pam Miner PBS&J Chipley Help ID FDOT3 contacts 12/1/2005 12/5/2006 Stacy Robinson Denny Wood FDOT3 Escambia & Santa Rosa 12/5/2005 12/5/2006 Referred to Tina Hegan Stacy Robinson John Locke FDOT3 Escambia & Santa Rosa 12/5/2005 12/5/2005 Stacy Robinson Tina Hegan FDOT3 Escambia & Santa Rosa 12/5/2005 12/7/2005 Rec'd spreadsheets of all road repair and DOT debris removal costs April 2007 D-2 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Caller Contact Name Agency OrOrganization County DataRequested For Requested Date Of Phone Or Email Contact Date Data Received Comments Stacy Robinson Mike Healy FDOT3 Orange Debris estimates 12/5/2005 12/5/2005 Stacy Robinson Bill Sullivan DeSoto County Red Cross DeSoto Displaced HH and damaged housing counts 1/12/2006 1/17/2006 Lisa Flax Rick Burgess City of Punta Gorda Charlotte County damage and loss data 1/17/2006 Left message about data collection. Lisa Flax Rick Burgess City of Punta Gorda Charlotte 1/18/2006 Asked about status of data collection; he will check and call tomorrow Lisa Flax Angela Whitehead Charlotte County / Solid Waste Charlotte Debris estimates 1/18/2006 She will e-mail available data Lisa Flax Greg Colton Santa Rosa Co / Public Works Santa Rosa Debris estimates 1/18/2006 He will e-mail available data Lisa Flax Sherry Babcock Escambia County Escambia County damage and loss data 1/17/2006 Lisa Flax Sherry Babcock Escambia County Escambia 1/18/2006 Left message about data collection Lisa Flax Sherry Babcock Escambia County Escambia 1/19/2006 Spoke with Sherry and her Director. They are very busy with hurricane recovery effort and it wlll take a lot of effort and time to provide requested data. Will see if they can work on it. Lisa Flax Susan Holt Escambia County/Solid Waste Escambia 1/18/2006 Left message about debris data request April 2007 D-3 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Caller Contact Name Agency OrOrganization County DataRequested For Requested Date Of Phone Or Email Contact Date Data Received Comments Stacy Robinson Steve Detwiler and John Witcher Orange County EM and Orange Co. Property Appraiser Orange Emailed data survey 1/17/2006 Referred to them by Becky Sebren. Received damage totals by damage state for Charley, Frances, and Jeanne on 1/23. OCEM is working on other info requested. Stacy Robinson and Nathan Slaughter Don Daniels St. Lucie EM St. Lucie Stacy emailed data survey, Nathan followed up with phone call 1/19/2006 Spoke to Mr. Daniels - he indicated that he was very busy but would try to make time for this. Did not give an indication of when he would have data ready. Stacy Robinson Keith Holman Martin EM Martin Emailed data survey 1/17/2006 1/18-he is gathering the data Stacy Robinson and Darrin Punchard Catherine Furr, Dawn Ballard, Larry Hilton, Adrian Cline DeSoto County--EM, Chamber of Commerce, Bldg&Zoning, School superintendent DeSoto Stacy emailed data survey; Darrin followed up with phone calls 1/17/2006 1/24/2006 (schools) Received schools data from Florence Gobble. Catherine Furr currently collecting other data. Stacy Robinson and Darrin Punchard Rich Shepard, Janet Gilliard, Kathy Crawford, Park Winter, Betty Croy, Joann McCray, and County Chamber of Commerce Hardee County--EM, Community Development, Property Appraiser, Economic Development, School Superintenden t's office, and chamber of commerce Hardee Stacy emailed data survey; Darrin followed up with phone calls 1/17/2006 1/19/06 (partial) Received updated critical facility listing from Rich S. Damage survey data are still being gathered by others. April 2007 D-4 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Appendix E H*Wind Wind Speeds and ARA Modeled Wind Speeds This appendix shows the variation between the Hwind and ARA-modeled tracks in sustained wind speed prediction at the jurisdiction level for each of the Hurricanes considered in this report. (1) Hurricane Charley Hwind ARA City/Place County MIN MAX MEAN MIN MAX MEAN Charlotte Harbor Charlotte 119 126 122 105 108 106 Charlotte Park Charlotte 93 122 108 104 111 108 Cleveland Charlotte 85 117 102 65 116 99 Englewood Charlotte 42 55 47 54 75 65 Grove City Charlotte 44 44 44 59 106 80 Harbour Heights Charlotte 105 122 113 108 108 108 Manasota Key Charlotte 44 56 52 59 59 59 Port Charlotte Charlotte 94 126 112 87 108 104 Punta Gorda Charlotte 76 137 112 104 111 108 Rotonda Charlotte 58 79 67 75 106 91 Solana Charlotte 117 117 117 104 116 110 Arcadia Desoto 75 102 85 72 106 97 Southeast Arcadia Desoto 73 91 80 72 105 94 Bowling Green Hardee 86 105 96 89 100 95 Wauchula Hardee 104 113 110 87 105 93 Zolfo Springs Hardee 103 110 107 87 105 96 Alva Lee 43 48 45 0 56 22 Bokeelia Lee 73 137 94 118 118 118 Bonita Springs Lee 40 63 52 0 56 32 Buckingham Lee 46 49 48 0 57 33 Burnt Store Marina Lee 65 89 77 91 108 101 Cape Coral Lee 48 89 57 69 108 78 Charleston Park Lee 43 43 43 0 0 0 Cypress Lake Lee 48 48 48 62 67 65 East Dunbar Lee 48 48 48 60 62 61 Estero Lee 43 62 48 0 56 41 Fort Myers Lee 48 49 48 0 69 59 Fort Myers Shores Lee 48 49 49 53 56 55 Gateway Lee 46 48 47 0 51 17 Harlem Heights Lee 48 48 48 66 66 66 Iona Lee 48 49 48 65 74 69 Lehigh Acres Lee 40 49 45 0 54 7 Lochmoor Waterway Estates Lee 48 48 48 69 73 71 April 2007 E-1 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Hwind ARA City/Place County MIN MAX MEAN MIN MAX MEAN Matlacha Lee 65 65 65 118 118 118 McGregor Lee 48 48 48 64 70 66 North Fort Myers Lee 48 52 49 53 104 71 Olga Lee 47 48 48 0 56 22 Page Park Lee 48 48 48 62 62 62 Palmona Park Lee 48 48 48 68 72 70 Pine Island Center Lee 71 101 81 100 118 109 Pine Manor Lee 48 48 48 62 64 63 Pineland Lee 92 101 96 118 118 118 San Carlos Park Lee 46 48 47 54 59 56 Sanibel Lee 51 99 73 81 109 95 St. James City Lee 59 97 76 100 118 109 Suncoast Estates Lee 48 48 48 66 72 68 Three Oaks Lee 46 47 47 0 56 41 Tice Lee 48 48 48 59 62 60 Villas Lee 48 48 48 56 65 61 Whiskey Creek Lee 48 48 48 62 66 64 Apopka Orange 26 40 30 0 55 28 Azalea Park Orange 63 88 77 77 80 79 Bay Hill Orange 38 50 44 63 73 68 Bay Lake Orange 33 52 39 62 72 67 Belle Isle Orange 83 91 87 79 81 80 Bithlo Orange 37 43 40 59 71 64 Christmas Orange 33 34 33 59 59 59 Citrus Ridge Orange 30 35 32 0 72 45 Conway Orange 82 88 84 79 81 80 Doctor Phillips Orange 43 50 47 67 73 69 Eatonville Orange 57 71 64 64 71 68 Edgewood Orange 72 88 80 79 81 81 Fairview Shores Orange 57 76 66 64 74 70 Goldenrod Orange 74 87 80 78 79 79 Gotha Orange 33 37 35 59 65 62 Holden Heights Orange 66 88 75 78 81 80 Hunters Creek Orange 81 93 87 78 84 82 Lake Buena Vista Orange 38 58 47 62 73 67 Lake Butter Orange 31 43 35 50 67 61 Lake Hart Orange 48 50 49 58 73 66 Lockhart Orange 44 57 51 55 67 62 Maitland Orange 57 71 65 65 76 71 Meadow Woods Orange 63 93 81 70 83 78 Oak Ridge Orange 58 79 69 76 81 78 Oakland Orange 27 28 28 0 50 17 Ocoee Orange 29 34 31 0 62 48 Orlando Orange 37 91 64 58 81 75 Orlovista Orange 37 52 46 63 68 66 Paradise Heights Orange 30 30 30 0 51 17 April 2007 E-2 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Hwind ARA City/Place County MIN MAX MEAN MIN MAX MEAN Pine Castle Orange 72 91 82 81 82 81 Pine Hills Orange 32 48 40 59 69 64 Sky Lake Orange 72 79 75 79 82 81 South Apopka Orange 30 32 31 0 55 39 Southchase Orange 88 89 88 78 82 81 Taft Orange 91 91 91 73 82 79 Tangelo Park Orange 65 65 65 77 77 77 Tangerine Orange 24 24 24 0 0 0 Tildenville Orange 28 31 29 50 50 50 Union Park Orange 53 68 58 71 78 76 Wedgefield Orange 34 42 36 59 71 65 Williamsburg Orange 65 72 68 77 82 79 Windermere Orange 33 43 37 50 63 57 Winter Garden Orange 27 33 30 0 55 30 Winter Park Orange 71 88 80 71 80 76 Zellwood Orange 24 29 27 0 0 0 Campbell Osceola 85 95 90 80 84 82 Celebration Osceola 39 70 53 72 72 72 Citrus Ridge Osceola 31 55 40 52 74 65 Kissimmee Osceola 57 95 85 70 84 80 Poinciana Osceola 50 98 79 72 85 80 St. Cloud Osceola 44 55 50 62 72 66 Yeehaw Junction Osceola 63 88 77 70 81 77 Auburndale Osceola 32 42 36 54 70 63 Babson Park Osceola 50 59 56 73 84 80 Bartow Osceola 36 61 47 55 90 73 Citrus Ridge Osceola 31 41 37 0 74 52 Combee Settlement Osceola 28 31 29 0 52 10 Crooked Lake Park Osceola 63 63 63 84 84 84 Crystal Lake Osceola 28 31 30 0 52 21 Cypress Gardens Osceola 67 83 75 78 89 83 Davenport Osceola 59 85 72 74 85 80 Dundee Osceola 90 100 95 89 91 90 Eagle Lake Osceola 51 60 55 76 81 79 Fort Meade Osceola 78 106 90 90 100 94 Frostproof Osceola 52 54 53 57 73 65 Fussels Corner Osceola 29 34 32 0 61 46 Gibsonia Osceola 27 27 27 0 0 0 Haines City Osceola 54 95 74 67 89 81 Highland City Osceola 35 35 35 55 59 57 Highland Park Osceola 59 69 64 82 87 85 Hillcrest Heights Osceola 57 57 57 73 84 79 Inwood Osceola 42 51 46 68 75 71 Jan Phyl Village Osceola 38 51 43 59 76 69 Kathleen Osceola 26 27 26 0 0 0 Lake Alfred Osceola 40 58 46 65 78 71 April 2007 E-3 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Hwind ARA City/Place County MIN MAX MEAN MIN MAX MEAN Lake Hamilton Osceola 90 90 90 89 91 90 Lake Wales Osceola 62 103 89 78 94 87 Lakeland Osceola 27 32 28 0 55 4 Lakeland Highlands Osceola 29 32 30 0 58 27 Loughman Osceola 55 79 69 72 85 79 Medulla Osceola 28 29 29 0 0 0 Mulberry Osceola 31 34 33 0 58 38 Polk City Osceola 28 31 29 0 52 17 Wahneta Osceola 60 77 68 81 84 83 Waverly Osceola 83 103 96 87 94 90 Willow Oak Osceola 28 29 29 0 55 11 Winston Osceola 27 28 28 0 0 0 Winter Haven Osceola 40 100 63 65 91 78 (2) Hurricane Ivan Hwind ARA City/Place County MIN MAX MEAN MIN MAX MEAN Bellview Escambia 89 91 90 84 87 85 Brent Escambia 87 90 88 83 85 84 Century Escambia 89 90 89 73 77 75 Ensley Escambia 87 91 89 83 84 84 Ferry Pass Escambia 86 90 88 82 84 83 Gonzalez Escambia 90 92 91 82 84 83 Goulding Escambia 86 87 87 84 85 85 Molino Escambia 92 93 92 80 81 81 Myrtle Grove Escambia 88 91 89 85 87 86 Pensacola Escambia 84 103 92 82 87 84 Warrington Escambia 88 90 89 86 89 87 West Pensacola Escambia 88 90 89 85 87 86 Bagdad Santa Rosa 80 83 82 69 78 75 Gulf Breeze Santa Rosa 83 103 97 79 82 81 Jay Santa Rosa 85 87 86 77 77 77 Milton Santa Rosa 82 84 83 69 76 74 Pace Santa Rosa 85 88 87 78 82 80 April 2007 E-4 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY APPENDIX F Glossary At Risk – Exposure values that include the entire building inventory or population value in a census block or tract that lie within, or bordering the inundation areas or any area potentially exposed to a hazard based on location. Building – A structure that is walled and roofed, principally above ground and permanently fixed to a site. The term includes a manufactured home on a permanent foundation on which the wheels and axles carry no weight. Census Block– A subdivision of a census tract (or, prior to 2000, a block numbering area), a block is the smallest geographic unit for which the U.S. Census Bureau tabulates 100-percent data. Many blocks correspond to individual city blocks bounded by streets, but blocks – especially in rural areas – may include many square miles and may have some boundaries that are not streets. Census Tract – A small, relatively permanent statistical subdivision of a county delineated by a local committee of census data users for the purpose of presenting data. Census tract boundaries normally follow visible features, but may follow governmental unit boundaries and other non- visible features in some instances; they always nest within counties. Designed to be relatively homogeneous units with respect to population characteristics, economic status, and living conditions at the time of establishment, census tracts average about 4,000 inhabitants. They may be split by any sub-county geographic entity. Critical Facility – Facilities that are critical to the health and welfare of the population and that are especially important following a hazard. Critical facilities include essential facilities, transportation systems, lifeline utility systems, high-potential loss facilities, and hazardous materials sites. As defined for the Portland risk assessment, this category includes: schools, hospitals, fire stations, police stations, and hazardous materials sites. Content Value – The value of a building’s content include all the items in a building, excluding the structure itself. The values are estimated to be 50 percent of the residential structural value and 100 percent of the commercial building replacement value. Duration – The length of time a hazard occurs. Erosion – Wearing away of the land surface by detachment and movement of soil and rock fragments, during a flood or storm or over a period of years, through the action of wind, water, or other geologic processes. Exposure – The number and dollar value of assets that are considered to be at risk during the occurrence of a specific hazard. April 2007 F-1 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Extent – The size of an area affected by a hazard or the occurrence of a hazard. Federal Emergency Management Agency (FEMA) – Independent agency (now part of the Department of Homeland Security) created in 1978 to provide a single point of accountability for all federal activities related to disaster mitigation and emergency preparedness, response, and recovery. Geographic Information Systems (GIS) – A computer software application that relates data regarding physical and other features on the earth to a database to be used for mapping and analysis. Hazard – A source of potential danger or an adverse condition that can cause harm to people or cause property damage. For this risk assessment, priority hazards were identified and selected for the pilot project effort. A natural hazard is a hazard that occurs naturally (such as flood, wind, and earthquake). A man-made hazard is one that is caused by humans (for example, a terrorist act or a hazardous material spill). Hazards are of concern if they have the potential to harm people or property. Hazard Mitigation – Sustained actions taken to reduce or eliminate the long-term risk and effects that can result from the occurrence of a specific hazard. For example, building a retaining wall can mitigate potential hazards. HAZUS (Hazards U.S.) – A GIS-based nationally standardized earthquake loss estimation tool developed by FEMA. HAZUS was replaced by HAZUS-MH (see below) in 2003. HAZUS-MH (Hazards U.S. - Multi-Hazard) – A GIS-based nationally standardized earthquake, flood, and wind loss estimation tool developed by FEMA. The purpose of this pilot project is to demonstrate and implement the use of HAZUS-MH to support risk assessments. HAZUS-MH Provided Data – The databases included in the HAZUS-MH software that allow users to run a preliminary analysis without collecting or using local data. HAZUS-MH Risk Assessment Methodology – This analysis uses the HAZUS-MH modules (earthquake, wind [hurricane] and flood) to analyze potential damages and losses. For this pilot project risk assessment the earthquake and flood hazards were evaluated using this methodology. HAZUS-MH Supported Risk Assessment Methodology – This analysis involves using inventory data in HAZUS-MH combined with knowledge such as (1) information about potentially exposed areas, (2) expected impacts, and (3) data regarding likelihood of occurrence for hazards. Infrastructure – The public services of a community that have a direct impact on the quality of life. Infrastructure includes communication technology such as phone lines or Internet access, vital services such as public water supplies and sewer treatment facilities, and transportation systems (such as airports, heliports, highways, bridges, tunnels, roadbeds, overpasses, railways, April 2007 F-2 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY bridges, rail yards, depots; and waterways, canals, locks, seaports, ferries, harbors, dry docks, piers, and regional dams). Intensity – A measure of the effects of a hazard occurring at a particular place. Inventory – The assets identified in a study region. The inventory assessment addresses what can be lost when a disaster occurs, that is, what community resources are at risk. Assets include people, buildings, transportation, and other valued community resources. Lifelines – Critical facilities that include utility systems (potable water, wastewater, oil, natural gas, electric power facilities, and communication systems) and transportation systems (airways, bridges, roads, tunnels, and waterways). Loss Estimation – The process of assigning hazard-related damage and loss estimates to inventory, infrastructure, lifelines, and population data. HAZUS-MH can estimate the economic and social loss for specific hazard occurrences. Loss estimation is essential to decision making at all levels of government and provides a basis for developing mitigation plans and policies. It also supports planning for emergency preparedness, response, and recovery. Occupancy Classes – Categories of buildings used by HAZUS-MH (for example, commercial, residential, industrial, government, and “other”). Replacement Value – The cost of rebuilding or repairing a structure. This cost is usually expressed in terms of cost per square foot and reflects the present-day cost of labor and materials to construct a building of a particular size, type, and quality. Risk – The estimated impact that a hazard would have on people, services, facilities, and structures in a community; the likelihood of a hazard occurring and resulting in an adverse condition that causes injury or damage. Risk is often expressed in relative terms such as a high, moderate, or low likelihood of sustaining damage above a particular threshold due to occurrence of a specific type of hazard. Risk also can be expressed in terms of potential monetary losses associated with the intensity of the hazard. Risk Assessment – A methodology used to assess potential exposure and estimated losses associated with priority hazards. The risk assessment process includes four steps: (1) identifying hazards, (2) profiling hazards, (3) conducting an inventory of assets, and (4) estimating losses. This pilot project report documents this process for selected hazards addressed as part of the pilot project. Scale – A proportion used in determining a dimensional relationship; the ratio of the distance between two points on a map, and the actual distance between the two points on the earth’s surface. Structure – Something constructed (for example, a residential or commercial building). April 2007 F-3 TASK ORDER 440 – HURRICANE WIND MODEL VALIDATION STUDY Substantial Damage – Damage of any origin sustained by a structure in a SFHA, for which the cost of restoring the structure to its pre-hazard event condition would equal or exceed 50 percent of its pre-hazard event market value. Vulnerability – Description of how exposed or susceptible an asset is to damage. This term depends on an asset’s construction, contents, and the economic value of its functions. Like indirect damages, the vulnerability of one element of the community is often related to the vulnerability of another. For example, many businesses depend on uninterrupted electrical power. If an electric substation is flooded, it will affect not only the substation itself, but a number of businesses as well. Often, indirect effects can be much more widespread and damaging than direct ones. Vulnerability Assessment – Evaluation of the extent of injury and damage that may result from a hazard event of a given intensity in a given area. The vulnerability assessment should address impacts of hazard occurrences on the existing and future built environment. April 2007 F-4