FLOOD RISK and INSURANCE STUDY FOR ESCAMBIA COUNTY, FL Report 1: Risk Assessment Howard Kunreuther and Marilyn Montgomery 1 February 28, 2017

FLOOD RISK and INSURANCE STUDY FOR ESCAMBIA COUNTY, FL Report 1: Risk Assessment Howard Kunreuther and Marilyn Montgomery 1 February 28, 2017 Summary This report details an investigation of flood risk assessment and risk-based premiums and compares approaches to FEMA s digital flood insurance rate maps (DFIRMs) and the National Flood Insurance Program (NFIP) rate-setting processes. To assess flood risks for single-family homes in Escambia County Florida with known first floor elevations, we use DFIRMs, current storm surge risk data, future storm surge risk data that include sea level rise, and the April 2014 flood event data that caused severe damage to homes in Pensacola. Estimated NFIP premiums can be high for property outside the Special Flood Hazard Areas (SFHAs) relative to those inside the SFHAs. Surge risk-based rates are generally higher than NFIP premiums but are sensitive to the depth-damage functions used in estimating risk-based rates. The importance of structure-specific flood risk assessment is highlighted in transparency of insurance premium estimates so those residing in flood prone areas can be aware of their risk and thus encouraged to undertake cost-effective mitigation measures to reduce future damage from flooding. Introduction Flood risk to structures and contents is defined as the product of the likelihood of floods with different magnitudes and the resulting water-related damage. The likelihood of the flood hazard refers to the frequency, or return period, of different water levels and water-related damage is defined in terms of replacement cost. To that end, this report provides in-depth analysis of the nature of flood risks in Escambia County, Florida. We begin with an assessment of flood zone maps and premium estimations using the Federal Emergency Management Agency (FEMA) procedures with respect to the National Flood Insurance Program (NFIP) that is under their jurisdiction. Analyses of NFIP hazard maps and the premium pricing processes are complemented by analyses of alternative flood hazard maps, including future conditions, and associated riskbased pricing of premiums. The best way to signal risk is to put a dollar figure on the expected damage; homeowners can understand insurance premiums far better than they can flood hazard maps. The goal of this study is to understand and utilize accurate risk-based flood insurance pricing to examine the tension between risk-based premium prices and the affordability of flood insurance for residents in Escambia County. The flood risk study is comprised of three primary tasks: (1) risk assessment, (2) risk-based premiums, and (3) affordability. Task 1, risk assessment, involves flood hazard mapping that is more granular than the 1% annual chance flood risk zones that FEMA delineates. FEMA s approach to flood hazard mapping is problematic because mandatory flood insurance purchases are based primarily on whether a building is within or outside the 1% annual chance flood hazard zones, also known as Special Flood Hazard Areas (SFHAs) on NFIP maps. FEMA SFHAs are not disaggregated according to flood risks with higher probabilities/frequencies, such as 2% and 4% annual chance events. In other words, all structures within SFHAs are treated as being susceptible to a homogenous 1% annual chance flood risk. The problems with this approach were explained in a recent report by the National Research Council (NRC 2015) that reviewed NFIP methodology for calculating riskbased rates, notably the floodplain analysis and mapping that support insurance rate setting. This NRC report 1 Authors names appear in alphabetical order. 1

concluded that the NFIP methods for setting risk-based rates do not accurately and precisely describe critical hazard and vulnerability conditions that affect flood risk for the following reasons: The determination of the flood hazard, using water surface elevation probability functions referred to by FEMA as probability elevation (PELV) curves, is based on floods having an annual likelihood of occurrence between 1 in 10 and 1 in 100. In reality, a significant portion of potential losses is caused by floods with an annual likelihood of occurrence greater than 10% and less than 1%. The average annual loss component is based on the calculation of 30 PELV curves that represent the flood hazard nationally rather than the flood hazard at the structure s location. As a result, NFIP insurance premiums charged are too high for some policyholders and too low for others. The vulnerability of a structure is determined by a function that relates damage to depth of inundation via a depth-percent damage relationship referred to by FEMA as a damage elevation (DELV) curve. The DELV curve, which expresses damage as a percentage of the structure s replacement value, blends NFIP claims data and U.S. Army Corps of Engineers damage functions weighted according to their credibility. NFIP claims data for a given depth of flooding are highly variable suggesting that other drivers of damage (e.g., duration of inundation, flow velocity, water contamination and debris content) may also be important. In addition, water elevations and lowest floor designations on which the claims were based may be inaccurate. Recognizing the limitations of the NFIP in determining structure-based insurance premiums, the FEMA Technical Mapping Advisory Council proposed a number of recommendations in its 2015 Annual Report including the following: Recommendation 14: FEMA, and its mapping partners including the private sector, should transition to a flood risk assessment focus that is structure specific. Where data are available, FEMA and its partners should contribute information and expertise consistent with their interest, capabilities, and resources towards this new focus. To that end, this research utilizes available structure-specific data to assess flood risk and estimate the resulting risk-based insurance premiums for homes in Escambia County, Florida. One cannot assume that all homes in an SFHA are vulnerable to floods with an annual chance of 1%. For example, suppose two homes that were built in the same year have foundations on pilings with the same elevation and are both located in an SFHA. Within this SFHA, the NFIP flood map shows that the base flood elevation (BFE), which is the water elevation of the 1% annual chance flood, is 9 feet. One of these homes is located within 300 feet of a stream and the other home is 500 feet from the same stream. Since both homes have the same BFE of 9 feet within the SFHA, they are considered to have the same flood risk and their flood insurance premium rates will thus be identical. In contrast, if the SFHA were disaggregated according to 2% and 4% annual chance zones, the home closer to the stream would have a higher flood risk and thus a higher insurance premium. In other words, if SFHAs (1% annual chance zones) were disaggregated into 10%, 4%, and 2% annual chance zones, with the 0.2% (1 in 500) annual chance zone also delineated, it would be possible to calculate a flood insurance premium that more accurately reflects risk. To illustrate, suppose a home with a first-floor elevation of 6 feet implies: an annual chance of 2% of 1 foot of flood water, a 1% annual chance flood of 2 feet of water in the home, and a 0.2% annual chance of 3 feet of water in the home. This home s annual averaged expected losses (AAL) and risk-based premium would be estimated based on equation 14-9 found in the Hazus technical manual (version 2.1) (p. 14-38): 2

The AAL premium for this house would be calculated as follows: AAL = (0.02 0.01) * [(losses attributable to 1 foot of flood water) + (losses attributable to 2 feet of flood water in home)]/2 + (0.01 0.002) * ((losses attributable to 2 feet of flood water) + (losses attributable to 3 feet of flood water in home))/2 + 0.002 * (losses attributable to 3 feet of flood water) In calculating the AAL, losses attributable to floods with given frequencies and probabilities are expressed in percentages of the home s building and contents replacement values, and are derived from various depthdamage curves that come with the Hazus software. Consequently, if flood hazards are modeled more accurately than only the 1% annual chance zones, the AAL approach to determine risk-based premiums is more accurate than the NFIP approach since AAL employs depth-damage functions that express flood losses as percentages of total insured values. Specifying a risk-based premium for structures based on the AAL is thus transparent in contrast to the NFIP that specifies different rates for the first $60,000 and $25,000 of coverage for buildings and contents respectively on a national basis. This report presents the analyses associated with risk assessment of homes in Escambia County (Task 1). Task 2 of this study will further investigate risk-based insurance premiums with premiums that are based on current NFIP methodology. Task 3 examines affordability concerns that arise from charging households riskbased insurance premiums and proposes strategies for incentivizing financially burdened households to adopt risk reduction measures that would lower flood risk-rated insurance premiums while at the same time addressing affordability issues. Thus the reports for Tasks 2 and 3 will expand on the efforts detailed in this report. Our analysis begins with estimates of NFIP premiums for single-family homes (SFHs) within Escambia County and Pensacola city limits according to the 2015 parcel data from Escambia County and the October 2016 version of the NFIP manual (https://www.fema.gov/media-library/assets/documents/127010) and the effective DFIRMs for Escambia County. Effective DFIRMs are the regulatory NFIP maps currently in place for a community that guide mandatory NFIP policy purchases for homes with federally-backed mortgages. Given the available data for our study area, we demonstrate that we can estimate NFIP premiums for a very large sample of owner- and renter-occupied single-family homes in the County and City. 3

We then assess the proposed changes in SFHAs for homes in our study area and discuss the potential implications of the newly mapped procedure 2 on NFIP premiums for these homes. Escambia County is currently having their effective DFIRM updated by FEMA, and a new preliminary DFIRM was issued by FEMA on January 27, 2017. Although the preliminary DFIRM does not become effective until the community formally adopts it, homes in areas changing from X zones to SFHAs according to the preliminary DFIRM are identified so that they can be investigated further for affordability concerns in our subsequent tasks. A home that is newly mapped into a SFHA will be required to purchase NFIP insurance if the home has a federally-backed mortgage, and this premium can increase annually until the full-risk rate is realized. We also explore the differences between the NFIP SFHAs and flood risk data that is more granular than the 1% annual chance SFHA zones, by assessing flood risks and estimating risk-based premiums for single-family homes based on storm surge data (called U-Surge) provided by Marine Weather and Climate (http://www.u- Surge.net/pensacola.html). Furthermore, we examine the Long Hollow tract in terms of flooding from the April 2014 event, and estimate insurance premiums based on April 2014 flood hazard data. The next section provides a broad overview of the data and methodology for the analyses followed by the results and discussion section. We then provide conclusions and recommendations, and directions for future analyses to implement Tasks 2 and 3. Appendix A is a very detailed explanation of the methodology used to estimate NFIP premiums from the parcel and building data. It also describes the U-Surge surge risk data and the April 2014 flood data. Appendix B explains the NFIP methods to calculate premiums based on the most recent version of the NFIP manual. Appendix C presents several illustrative examples of NFIP premiums estimations. Appendix D explains the method of estimating surge risk-based premiums and Appendix E includes several tables of statistical results. Our investigation of the flood risk and risk-rated insurance premiums on houses and their contents in Escambia County is an in-depth case study that is designed to inform the national dialog concerning reauthorization of the NFIP in September 2017. Data and Methodology To assess flood risk for Escambia County single-family homes, data was obtained from the Escambia County property appraiser s office, Escambia County GIS and Building Inspections Department, the City of Pensacola, FEMA, the Northwest Florida Water Management District (NWFWMD), surge risk data from Marine Weather & Climate, and April 2014 flood data from Atkins engineering. Estimating NFIP premiums Our analyses begin with estimating NFIP premiums for a large sample of Escambia County SFHAs based on the most recent version of the NFIP manual (October 2016) and available data. Appendix A provides a detailed explanation of our data sources and methodology including geospatial analyses Based on available data, we estimated NFIP premiums for all single-family homes in the County and City for which we were able to compute premiums. The single-family homes for which we computed premiums include: all X zone properties as first floor elevation is not used in X zone rates; pre-firm A zones properties as first floor elevation is not used in pre-firm rates; 2 The newly mapped procedure factsheet can be viewed at this link: https://www.fema.gov/media-librarydata/1428947341380-23a056704409206c86cc89ac72f9f070/fema-hfiaa_newlymappedfs_041015.pdf. The newly mapped procedure chapter of the April 2016 version of the NFIP manual can be viewed at this link: https://www.fema.gov/media-library-data/1458758632214- e139ecc72b1d9564f10f27bc30b8cbb9/10_newly_mapped_508_apr2016.pdf 4

post-firm A and VE zones properties with slab on-grade foundations because first floor elevation of slab on-grade foundations is a negligible amount above ground elevation, and ground elevations are available from Lidar-derived data; and houses with geocoded elevation certificates in any flood zone. For these homes we used post-firm rating tables since post-firm premiums require house elevation data. See Appendix B for the NFIP premium calculation methodology, and Appendix C for several examples of NFIP premiums calculations based on different flood zones and pre- or post-firm houses. Assessing houses using preliminary DFIRM for Escambia County The preliminary DFIRM for Escambia County became available on January 27, 2017, and it was intersected with the 2006 effective DFIRM to assess changes in SFHAs. Then we overlaid the areas of changing SFHAs with the building footprints used in our study to assess changes in flood zones for these homes. Homes that are newly mapped into SFHAs are likely to be subject to the NFIP newly mapped procedure which would impose mandatory NFIP policy purchases for these homes if they have a federally-backed mortgage. Relatively few BFEs were given in the preliminary DFIRM SFHAs so we could not estimate new premiums for these homes. It is important to note that the preliminary DFIRM does not become effective until after the affected communities have adopted it. This adoption process can take one year or more. Estimating storm surge flood risk and premiums based on U-Surge data The surge data, called U-Surge, were provided to the Wharton Risk Center by Marine Weather and Climate (https://www.u-surge.net/pensacola.html). Based on observations from the National Oceanic and Atmospheric Administration (NOAA), tide gauges and other data sources, storm surge data from 1900 to 2016 for Pensacola were analyzed and used to develop the U-Surge dataset for our study area. In this analysis, time (return period) is the independent variable and water level (storm tide height) is the dependent variable. Therefore, one can choose a time period (for example, 100 years) and the analysis will provide the most extreme water level that should be reached. These estimations do not guarantee that a 100-year water level will be reached within the next 100 years, but rather, over a period of centuries, this water level should be reached or exceeded every 100 years, on average. For further information on the methodology involved in deriving the U-Surge data, see these three sources: Needham and Keim (2012), Needham et al. (2013), and Needham et al. (2015). These three sources established the scientific credibility of a storm surge database called SURGEDAT. A more recent database, U-Surge improves upon SURGEDAT by localizing datasets and providing datum conversions. Whereas SURGEDAT listed the datums for high water marks, U-Surge takes this work one step farther by converting different datums into one common datum referenced by all high water marks in a data set. Since U-Surge has converted all high water marks to one common datum, or vertical reference system, statistical analysis is possible. In contrast, SURGEDAT involved more than five different datums and these data could not be easily compared since they were not converted to a common datum. U-Surge data for present-day surge risks (water elevations and return periods/probabilities) was utilized for analysis, as well as storm surge risks with sea level rise for 25, 50, 75, and 100 years into the future. The analysis to derive the future U-Surge data utilized three sea level rise scenarios that represent fast, medium and slow rates for the Gulf of Mexico region, following Tissot (2016). Tissot s (2016) work provides sea level rise estimates from 2011 to 2100 using A1FI, RCP 4.5 and RCP 6.0 projections. This source provides a sea-level rise estimate of 60 cm between 2011 and 2100 for the region, an estimate that is relevant for Pensacola because this area observes little subsidence. The U-Surge data included GIS data representing the surge water elevations for annual events with 10%, 4%, 2%, 1%, and 0.2% probabilities. These GIS data were intersected with the building footprints to estimate flood risks from current and future surge and compute surge risk-based premiums. This U-Surge data are 5

more granular than the NFIP SFHAs. Although surge risk does not include flood risk from precipitation, the methods employed to derive the U-Surge data are transparent and not subject to the NFIP map adoption processes. To compute current and future surge risk-based premiums, we used two different depth-damage functions that come with the Hazus software: one from the Federal Insurance & Mitigation Administration (FIA), and the other from the US Army Corps of Engineers Institute for Water Resources (USACE IWR). We compare surge risk-based premiums per $100 of coverage for present and future conditions to NFIP premiums per $100 of coverage. We normalize all risk-based AAL and NFIP premiums per $100 of coverage in this analysis to address the fact that NFIP premiums have limits on coverage and different rates for basic and additional coverages. AAL premiums do not have any coverage limits imposed on them and apply the same rates no matter how much is covered. April 2014 flood Pensacola experienced a major storm in April 2014 that the National Weather Service (NWS) estimated to be between a 1 in 100 to 1 in 200 year event (https://www.weather.gov/mob/20140429_flashflood). Atkins engineering firm provided flood elevation GIS data designed to replicate the April 2014 event. These data were intersected with building footprints to assess which homes may have been flooded in the April 2014 event. The methodology involved modeling the April 2014 event is detailed in Atkins report Long Hollow Drainage Basin Analysis. Twenty homes in our Long Hollow tract sample were flooded in the 2014 storm (http://studeri.org/wpcontent/uploads/2015/02/long_hollow_drainage_report_compressed_201501231141166055.pdf). A sensitivity analysis was implemented with the FIA and USACE IWR depth-damage functions to estimate flood insurance premiums if one were to insure against a flood similar to the April 2014 storm. We use annual flood probabilities of 0.2%, 0.5%, 0.67%, and 1% to estimate premiums in our analyses. We then compare premiums per $100 of coverage based on the April 2014 data to estimated NFIP premiums per $100 of coverage for these 20 homes, all of which are located in X zones (at risk of 0.2% annual chance flood or less according to the effective DFIRM). Results and Discussion NFIP premiums estimations: single-family homes in Escambia County and the City of Pensacola In this section mean NFIP premiums for single-family homes in Escambia County and the Pensacola are depicted using bar charts. Tables with the relevant statistics can be found in Appendix E. Figure 1 shows mean building values for homes by flood zones and pre- and post-firm designation (total of 77,834 homes) without geocoded elevation certificates. These 77,834 homes include all X zone and pre-firm homes because there is no consideration of elevation for these homes when developing insurance premiums and post- FIRM homes in SFHAs with slab on-grade foundations where the elevation of these structures can be estimated by using ground elevation data. For Escambia County and the City of Pensacola, homes built before 1978 are considered pre-firm and those built during or after 1978 are post-firm. For Pensacola Beach, homes built before 1974 are treated as pre-firm and those built during or after 1974 are post-firm. For the City of Century, homes built prior to 1988 are pre-firm and those built during or after 1988 are post-firm. Homes in VE zones often have very desirable and expensive water view amenities, which explains why they have the highest mean property values. Homes in A zones are generally more expensive than those in X zones, and pre-firm homes are less valuable than post-firm homes in all zones. Newer construction is usually more valuable than older construction. In our sample of single-family homes shown in Figure 1, we have 75,255 homes in X zones, 2,565 homes in A zones, and 14 homes in VE zones. Note that A zones include all unnumbered A zones, AE zones, AH zones, and AO zones. 6

Figure 1. Mean building values of homes for which NFIP premiums were estimated by flood zone and pre- or post-firm (n = 77,834 single-family homes). There are many more homes in X zones in Escambia County than in SHFAs. This is expected since SFHAs comprise a much smaller area than X zones in Escambia County, since NFIP SFHAs largely follow natural floodplains that are adjacent to streams, bayous and open water where there are relatively few homes. VE zones, the 1% annual chance zones with 3 feet or more of wave action hazards, are very small and contain very few homes relative to other zones. We were unable to calculate several VE zone post-firm premiums because these homes are slab foundations with FFEs that are over 10 feet below BFE for their zones. Further, for VE properties with slab foundations over 1 foot below BFE, one must refer to the Specific Rating Guide (page 1-15) which is a supplement to the NFIP manual and the table in the Specific Rating Guide only goes to -10 feet difference from BFE. It is unknown whether this implies that post-firm homes with slab foundations in VE zones with over 10 less than BFE are simply uninsurable with the NFIP. Figure 2 depicts mean estimated NFIP building premiums for owner-occupied homes without geocoded elevation certificates, and Figure 3 shows contents premiums for all homes without geocoded elevation certificates. The sample reflected in Figures 2 and 3 includes all X zone and pre-firm homes, and post- FIRM homes in SFHAs with slab on-grade foundations. As expected, X zone building premiums are significantly lower than A and VE zones premiums; premiums for contents for pre- and post-firm homes are quite similar for each flood zone. X zone premiums for contents and buildings are the same for pre- and post- FIRM property. For A zone homes without geocoded elevation certificates we used the very costly premiums for A zone homes without elevation certificates (Table 3C on page RATE 9 of the NFIP manual. See Appendix C Examples 8 through 10 for A zone property examples). Homeowners in A zones interested in purchasing NFIP insurance for property without BFEs are likely to reduce their premiums significantly by obtaining an elevation certificate. 7

Figure 2. Mean estimated NFIP building premiums for owner-occupied homes. Figure 3. Mean estimated contents premiums for renter- and owner-occupied homes. 8

A notable observation in Figures 2 and 3 is that post-firm X zone premiums are very close to pre-firm A zone premiums. Since pre- and post-firm X zones rates are identical, the higher mean estimated post-firm X zone premiums are due to the higher building and contents values of post-firm property relative to pre- FIRM. Because pre-firm A zone rates are subsidized, pre-firm A and X zone premiums are virtually the same. All X zone premiums and pre-firm A zones premiums do not consider home elevation or the number of floors in the home. The NFIP should consider a more aggressive plan for phasing out A zone pre-firm subsidies and develop X zone rates based on house elevation and number of floors. Rates for X zones based on home elevation and numbers of floors would be more risk-based and equitable for policy holders. Three NFIP communities in our study area participate in the Community Rating System (CRS) of the NFIP, which is a program to encourage NFIP communities to exceed minimum NFIP standards for floodplain management. Based on a complex system of points awarded to communities for different floodplain management activities, NFIP policies are discounted according to the communities overall CRS score. Property on Pensacola Beach is within the Pensacola Beach/Santa Rosa Island Authority community, and because Pensacola Beach is a Class 5 community its NFIP policies within SFHAs are discounted by 25%, and non-sfha policies are discounted by 10%. Unincorporated Escambia County is a Class 6 and thus its SFHA NFIP policies are discounted by 20% and non-sfha policies discounted by 10%. The City of Pensacola is a Class 7 so its SFHA policies are discounted by 15% and non-sfha policies are discounted 10%. In Figures 4 and 5 we show mean building values and estimated NFIP premiums with CRS discounts for the homes with geocoded elevation certificates. For estimating premiums, we assumed that all homes are owner-occupied and used post-firm rating since they have elevation certificates. While building values for homes in VE zones are higher than those in AE zones, the similarity in mean premiums for AE and VE zone homes is striking. The homes with geocoded elevation certificates in VE zones are on Pensacola Beach and all were built since 2001. These Pensacola Beach homes are all built on elevated foundations with at least 3 feet over BFE for their zones, so their NFIP premiums are quite low for VE zones. This is a significant finding in light of the fact that the estimated NFIP premiums for the homes with geocoded elevation certificates should be more accurate than those estimated for the slab on-grade foundation homes. Additionally, Pensacola Beach is a higher CRS class than City of Pensacola and Escambia County so the VE premiums that are all on Pensacola Beach are discounted at a greater rate than the AE policy premiums located in the City and County. 9

Figure 4. Mean building values for homes with geocoded elevation certificates. (n=96 homes) Figure 5. Mean estimated total NFIP premiums including CRS discounts for homes with geocoded elevation certificates (n=96). All VE policy premiums are for property located on Pensacola Beach, while AE policy premiums are for property located in the City of Pensacola and Escambia County. 10

Homes subject to preliminary DFIRM changes in their flood zone Based on our analyses of the preliminary DFIRM for Escambia County, we identified 927 homes that are currently in X zones but will now be in SFHAs, and 24 homes on Pensacola Beach changing from VE to AE zones based on the preliminary DFIRM. The mean and median values of homes in areas changing from X zones to SFHAs are $90,497 and $80,639 respectively, so these are moderately priced homes that could become more costly due to new NFIP premiums for SFHAs that they will be required to purchase if these homes have federally-backed mortgages and the preliminary DRFIRMS are adopted. As stated above, we did not estimate new premiums for the homes in X zones changing to SFHAs because the BFEs were not given for most of the preliminary zones within which they fall. The home values for those on Pensacola Beach in zones changing from VE to AE are high: mean and median values are $499,325 and $461,074 respectively. Further research is needed to assess whether homeowners of the 927 homes newly mapped into SFHAs according to the preliminary DFIRM may be financially burdened by mandatory NFIP policy purchase. If the first floor elevations of homes newly mapped into SFHAs are lower than the BFE for the new SFHAs, the NFIP premiums will be costly. We will attempt to obtain information on the new BFEs in areas that are changed to SFHAs if these data become available, so we can estimate new NFIP premiums for these homes in our subsequent tasks for this project. Storm surge flood risk and premiums based on U-Surge data Turning to an analysis of surge risk-based premiums based on the U-Surge data, Figure 6 shows means of estimated pre- and post-firm NFIP total premiums with and without CRS discounts, 2017 surge risk-based AAL premiums based on the FIA and USACE IWR depth-damage curves, and future surge risk-based AAL premiums for 2042, 2067, 2092, and 2117 based on the FIA and USACE IWR depth-damage curves for 3,455 slab on-grade foundation homes that coincide with surge risk zones. All premiums are expressed as a rate per $100 coverage. They are normalized in this way to enable one to compare NFIP premiums where coverage is capped at $250,000 and $150,000 for building and contents coverage respectively with surge risk-based AAL premiums where there are no caps on coverage. The surge risk-based AAL rates are based on full replacement values and two different depth-damage functions from Hazus. As one can see from Figure 6, pre-firm NFIP rates with and without CRS discounts are both higher than post-firm rates with and without CRS discounts. This is because post-firm homes in our sample of slab ongrade foundation homes have first floor elevations that make the rates more favorable than pre-firm rates which do not account for home elevation. Realistically, homeowners with pre-firm homes should get elevation certificates to see whether post-firm ratings based on home elevation are more favorable than pre- FIRM rates. These findings demonstrate the importance of structure-specific elevation data used for rating flood risk. Figure 7 depicts the means of estimated post-firm NFIP premiums per $100 coverage with and without CRS discounts, surge risk-based AAL premiums per $100 coverage for 2017 based on the FIA and USACE IWR depth-damage curves, and future surge risk-based AAL premiums per $100 coverage for 2042, 2067, 2092, and 2117 based on the FIA and USACE IWR depth-damage curves for the 37 homes with geocoded elevation certificates that coincide with surge risk zones, all of which are in VE zones on Pensacola Beach.. There are no pre-firm homes in Figure 7 because the premiums are based on houses having elevation certificates. 11

1.00 Mean estimated premiums per $100 coverage: slab on-grade foundation homes (n=3,455) 0.90 premiums per $100 coverage 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 Figure 6. Means of estimated total premiums (buildings and contents) for pre- and post-firm NFIP rates with and without CRS discounts, and surge risk-based AAL premiums based on the FIA and USACE IWR depth-damage functions for 2017, 2042, 2067, 2092, and 2117 for homes that coincide with surge risk zones and slab on-grade foundations normalized by $100 of total coverage (building and contents coverage). (n=3,455) The differences in mean premiums per $100 of coverage between Figures 6 and 7 are significant, due to the location of the homes: the 37 homes represented in Figure 7 are all vulnerable to high surge hazards and future sea level rise. In contrast, those represented in Figure 6 are all on mainland Escambia County and City of Pensacola and are not as vulnerable to surge and sea level rise. We expect that future surge risk-based premiums would be higher than NFIP rates since the future data accounts for sea level rise and NFIP maps and premiums do not. However, the fact that current surge risk-based premiums per $100 coverage are higher than post-firm premiums per $100 coverage in Figures 6 and 7 based on the USACE IWR depth-damage function is interesting because surge risk does not include flood risk due to precipitation. Composite flood risk that includes both surge and precipitation should be higher than that with only surge risk. Additionally, the data shown in Figure 7 are more realistic than that shown in Figure 6 because the homes represented in Figure 7 have elevation certificates while the elevation of homes in Figure 6 are based on the ground elevation within the building footprint according to the Lidar data of homes with a slab on-grade foundation according to Escambia County Property Appraiser s data. The data from the elevation certificates should be more accurate than the property appraiser s foundation-type data because elevation certificates are produced from a professional survey that is specifically intended for NFIP insurance rating. 12

premiums per $100 coverage 5.00 4.50 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50 0.00 Mean estimated premiums per $100 coverage: geocoded elevation certificates (n=37) Figure 7. Means of estimated total premiums (buildings and contents) for pre- and post-firm NFIP rates with and without CRS discounts, and surge risk-based AAL premiums based on the FIA and USACE IWR depth-damage functions for 2017, 2042, 2067, 2092, and 2117 for homes that coincide with surge risk zones and with geocoded elevation certificates normalized by $100 of total coverage (building and contents coverage). (n=37) In Figure 7, a linear increase with years into the future in premiums per $100 coverage is exhibited in premiums estimated from both the FIA and USACE IWR functions. This is expected since surge hazards with sea level rise that we used also increases with each year into the future, but the peak for present day surge risk-based premiums based on the USACE IWR function in Figure 6 is unexpected. In Figure 6, the highest values for surge risk-based AAL per $100 coverage is first based on the FIA function for 2117 followed by the USACE IWR function for 2017. These peak values for surge risk-based premiums per $100 coverage observed in Figure 6 are due to the location of the homes relative to present and future storm surge hazards, their first floor elevations, and the number of floors in the homes. Furthermore, the high value of surge riskbased premiums per $100 coverage based on the FIA function for 2117 in Figure 6 is partly due to the large number of homes in VE zones, which have a higher percent damage from floods than A zone homes. There are a number of differences between the FIA and USACE IWR depth-damage functions that we used to estimate surge risk-based AAL premiums that explains their significant differences in the premiums per $100 coverage. The USACE IWR functions vary according to single-floor homes and those with more than one floor, while those from FIA have functions for homes with one floor, two floors, and three or more floors. The FIA functions differ for A and V zones, while those for USACE IWR do not. In our calculations, we treated homes in X zones as A zone homes when applying functions from FIA since FIA functions do not include X zones. The damage estimates for the same flood elevations and number of floors are higher in the USACE IWR functions than those from FIA, which explains why the surge risk-based premium rates based 13

on the USACE IWR functions are usually higher than those based on FIA functions. For example, the damage attributable to two feet of flood water over FFE in a one-floor home is 32% according to the USACE IWR function but only 14% based on the FIA function. Coupled with the high surge risks of Pensacola Beach homes reflected in Figure 7, the greater percent damages due to flood water elevations of the USACE IWR functions relative to FIA explains the large differences between FIA and USACE IWR premiums per $100 coverage in Figure 7. Another significant difference between the depth-damage functions from FIA and USACE IWR that often makes the premiums based on the USACE IWR functions higher is the threshold at which no damage occurs: the FIA functions have very small damages when there is no difference between FFE and flood elevation (e.g., 4% or 5% of building values are attributable to flood elevations same as FFE). However, the functions from USACE IWR have 10% and 13% of building value lost to damages when flood elevation is the same as FFE (depending on whether the home is one or more than one floor), and 3% of building value loss incurred when flood elevations are one foot less than FFE. An insurance company would undoubtedly use USACE IWR functions over those from FIA to compensate for uncertainty in estimating flood losses. These findings align with de Moel and Aerts (2010), who assert that is difficult to specify flood damages very precisely due to the uncertainties in characterizing depth-damage curves and estimating the value of exposed assets. April 2014 flood Our last analysis involves the modeled April 2014 flood data from Atkins. The April 2014 flood hazard data coincided with 20 homes in the Long Hollow tract. Figure 8 is a location map showing the Long Hollow and Sanders Beach tracts, the modeled April flood elevations from Atkins, actual residential damage locations recorded by Escambia County (red triangles), and flooded homes according to the model (cyan polygons). No homes in the Sanders Beach tract coincided with the April 2014 flood data. Flood insurance premiums were estimated based on flooded homes and the modeled April 2014 data. 14

Figure 8. Long Hollow and Sanders Beach tracts with modeled April 2014 flood elevations, locations of residential damage reports (red triangles) from Escambia County building inspections department, and homes flooded by the April 2014 event according to the model data (in cyan). Long Hollow tract is north center and Sanders Beach tract is southwest of Long Hollow. 15

Figure 9 shows mean estimated NFIP premiums per $100 of coverage with and without CRS discounts for these 20 potentially flooded homes in the Long Hollow tract, and premiums based on April 2014 flood data if the event had an annual likelihood of 0.2%, 0.5%, 0.67%, or 1%. We tested several different annual event probabilities in a sensitivity analysis of the FIA and USACE IWR depth-damage functions that we used to estimate losses because the return period/probability of the April 2014 event is uncertain. As pointed out above, the NWS estimates that the precipitation of the April 2014 event was between a 1 in 100 and 1 in 200 year event. Even though the entire Long Hollow tract is in the X zone, the mean NFIP premiums per $100 coverage with and without CRS discounts are much higher than the premiums based on April 2014 flood data. Since we have only one flood water surface elevation for the April 2014 case rather five flood elevations used for the surge risk-based AAL analysis (corresponding to the 10%, 4%, 2%, 1%, and 0.2% annual chance of surge hazards), the premiums based on the April 2014 event are not AAL rates. Nevertheless, we show how much it might cost to insure these homes against an event similar to the April 2014 flood if it had a probability of a 0.2%, 0.5%, 0.67%, or 1% annual chance event. 1.20 Mean estimated premiums per $100 coverage: April 2014 flood slab on-grade foundation homes (n=20) premiums per $100 coverage 1.00 0.80 0.60 0.40 0.20 premiums based on the April 2014 flood data 0.00 NFIP premium NFIP premium CRS discount Apr. 2014 FIA (0.2%) Apr. 2014 IWR (0.2%) Apr. 2014 FIA (0.5%) Apr. 2014 IWR (0.5%) Apr. 2014 FIA (0.67%) Apr. 2014 IWR (0.67%) Apr. 2014 FIA (1%) Apr. 2014 IWR (1%) Figure 9. Mean premiums per $100 of coverage for 20 homes in the Long Hollow tract that coincided with the modeled April 2014 flood hazard data. Conclusions and Recommendations Our analysis of flood risk and insurance in Escambia County demonstrates the importance of estimating structure-specific flood risks and risk-based insurance premiums based on the AAL method. Not only must the NFIP SFHAs be mapped with more granular representations of flood hazards, but the ratings processes of the NFIP should be revised. We have estimated surge risk-based AAL premiums in a number of ways as 16

discussed above, and a very salient finding is that all AAL premium estimations use depth-damage functions that express flood losses as percentages of total insured values attributable to floods with given elevations. The following recommendations follow from our analyses designed to improve risk assessment and risk-rated premium rates for the NFIP: 1. NFIP rates should be based on a percentage of total insured value, not flat rates for basic and additional coverage. Basic building and contents coverage rates (for the first $60,000 and $25,000 of building and contents coverage respectively) are designed to maximize premium income for the NFIP, but they are inequitable to those insuring moderately-priced property. Rates based on the percentage of total property value that is insured or having some coinsurance clause would be more equitable. 2. Pre-FIRM rates should be risk-based. Pre-FIRM rates are problematic because there is no consideration of FFE, the single most important element for structure-specific flood risk assessment. While homeowners can obtain an elevation certificate for a pre- FIRM home and use post-firm rates based on the elevation certificate (if more favorable to the insured), the fact that pre-firm rates are referred to as subsidized is misleading because they could be actually higher than post-firm rates. 3. Homeowners should obtain elevation certificates. All homeowners should be encouraged to obtain an elevation certificate for their home and get an estimate of post-firm rates. This would not only result in more accurate structure-specific risk assessments and associated risk-based premiums, but it would enable homeowners to understand the importance of the elevation of their property in determining a risk-based insurance premium. 4. The NFIP should determine risk-based premium rates for houses outside SFHAs that consider home elevation (FFE) because floods often occur outside the boundaries of SFHAs. Again, home FFE is the most important element of structure-specific risk-based premiums no matter what flood zone the home is located. 5. Differentiate homes with more than 1 floor from those with one floor in all flood zones and all premium rates. Post-FIRM ratings for both contents and buildings in SFHAs are lower for homes with more than one floor than those with one floor, because flood water that exceeds the FFE damages a smaller percentage of the home. 6. CRS discounts should be the same for policies inside and outside SFHAs. Although CRS activities are targeted within SFHAs, their benefits are not limited to areas inside SFHAs. Furthermore, if FEMA wants to promote insurance penetration outside SFHAs, the same CRS discounts should be extended to non-sfha properties. The above recommendations would make the NFIP more risk-based and equitable. The link between home elevation and flood risks must be made more transparent in specifying premiums and through communicating the importance of elevation certificates. Future Research For Tasks 2 and 3 of this research based in Escambia County, we will be implementing the following analyses. 17

In April to June of 2017, the Risk Map products will be made available by FEMA and the Northwest Florida Water Management District. The Risk Map products include GIS data representing flood elevations for 10%, 4%, 2%, 1%, and 0.2% annual chance flood frequencies. Once we obtain these flood risk products, analyses of single-family homes flood risk will be implemented and AAL riskbased premiums will be estimated and compared to results from surge risk and NFIP estimated premiums. Once we have computed AAL risk-based premiums with the Risk Map data, we will assess premiums for affordability concerns (Task 3). This will involve analyses of sociodemographic data from the census. Additional sensitivity analyses will be conducted to assess how many flood elevations with annual probabilities are needed to get an accurate risk-based AAL premium. Using multiple annual flood probabilities with their associated flood water elevations is far better than using only the 1% annual chance of a flood as in current NFIP maps. We will examine homes that are vulnerable to floods with 10% and 4% annual probabilities, since these are homes that would benefit most from mitigation. NFIP data will be assessed for policy penetration, uninsured homes, and how many actual pre- and post-firm policies exist in Escambia County. NFIP insurance penetration data will illuminate which homes are vulnerable to flood risks due to lack of insurance. Further, the NFIP data will add to the discussion of privatizing flood insurance. There might be a number of uninsured homes in X zones that could purchase flood insurance from a private company at rates that are lower than those offered by the NFIP. Long Hollow and Sanders Beach tracts will be further investigated. Field visits will made to estimate home elevations by visual inspection and/or conversations with homeowners. We will design criteria for homeowners who could qualify for a voucher or low-interest loan to assist with risk-based premium costs, and for mitigating their home against flood risks. We could determine how large a voucher and/or low-interest loan homeowners would receive for flood insurance and mitigation. There are neighborhood and individual mitigation projects throughout the County that warrant investigation of how and why certain decisions were made. For example, there is a neighborhood north of Pensacola called Bristol Park where homeowners are being approached for FEMA acquisition. The events leading up to FEMA acquisition should be studied. There are other cases of homeowners elevating their homes to mitigate flood risk with and without public funds. Evaluating why some homeowners mitigate and others do not, and their opinions on the process in hindsight would provide valuable insight as to the feasibility and efficiency of home elevation for mitigating flood risks. The research presented in this report highlights the importance of more accurate flood hazard mapping and structure-specific flood risk determination, as noted by the NRC (2015) report and the TMAC 2105 Annual Report. Additionally, further investigation of how depth-damage functions affect risk-based premium estimations is needed, since depth-damage functions are a source of significant uncertainty. Although the NFIP was designed to be affordable, the national debt attributable to flood losses shows the NFIP is unsustainable in its current state. In our subsequent reports based on Escambia County, we will further investigate flood risk assessment and risk-based premiums, and address the tension between risk-based insurance and affordability, adding value to the national discussion of the NFIP and its reauthorization in September 2017. 18

References Atkins. Long Hollow Drainage Basin Analysis. Report prepared for The City of Pensacola, January 2015. Accessed 24 February 2017 at http://studeri.org/wpcontent/uploads/2015/02/long_hollow_drainage_report_compressed_2015012 31141166055.pdf. De Moel, H., and Aerts, J.C.J.H. 2011. Effect of uncertainty in land use, damage models and inundation depth on flood damage estimates. Natural Hazards 58:407-425 Federal Emergency Management Agency (FEMA) 2015. Technical Mapping Advisory Council (TMAC) TMAC 2015 Annual Report. Last accessed 25 February 2017 at https://www.fema.gov/medialibrarydata/1454954097105a94df962a0cce0eef5f84c0e2c814a1f/tmac_2015_annual_report.pdf. National Research Council (NRC) 2015. Affordability of National Flood Insurance Program Premiums: Report 1. Last accessed 25 February 2017 at https://www.nap.edu/download/21709#. Needham, H.F., and B.D. Keim, 2012: A Storm Surge Database for the U.S. Gulf Coast. International Journal of Climatology 32(14):2108-2123. (DOI: 10.1002/joc.2425). Needham, H.F., B.D. Keim, and D. Sathiaraj, 2015: A Review of Tropical Cyclone-Generated Storm Surges: Global Data Sources, Observations and Impacts. Reviews of Geophysics 53(2): 545-591. Needham, H.F., B.D. Keim, D. Sathiaraj, and M. Shafer, 2013: A Global Database of Tropical Storm Surges. EOS, Transactions American Geophysical Union 94(24): 213-214. Tissot, P., 2016: Relative Sea Level Rise around the Gulf of Mexico and its Impact: from Nuisance Flooding to Large Surges. Presentation for Hydrographic Services Review Panel Meeting, Galveston, Texas, March 15-17, 2016. https://www.nauticalcharts.noaa.gov/ocs/hsrp/galveston2016/tissot_hsrp_march2016_presentation. pdf 19