- Email: [email protected]
Ecological indicators of flood risk along the Gulf of Mexico
Ecological Indicators 18 (2012) 493–500
Contents lists available at SciVerse ScienceDirect
Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind
Ecological indicators of flood risk along the Gulf of Mexico Samuel D. Brody a,∗ , Walter Gillis Peacock b , Joshua Gunn b a Texas A&M University at Galveston, Department of Marine Sciences, Ocean and Coastal Studies Building, Building 3029, Room # 366, 200 Seawolf Parkway, Galveston, TX 77553, USA b Texas A&M University at College Station, Department of Landscape Architecture and Urban Planning, TAMU 3731, Texas A&M University, College Station, TX 77843-3137, USA
a r t i c l e
i n f o
Article history: Received 3 July 2011 Received in revised form 2 January 2012 Accepted 4 January 2012 Keywords: Ecological indicators Gulf of Mexico Flood damage Social–ecological systems
a b s t r a c t Despite mounting economic losses from both acute and chronic flood events in coastal areas of the U.S., little empirical research has been conducted on the importance of existing landscape-level ecological components in mitigating the economic impacts to vulnerable coastal communities over the long term. In recognition of this lack of knowledge base, we examine several ecological indicators across 144 counties bordering the Gulf of Mexico. Specifically, we identify and measure the following four indicators: floodplain area, soil porosity, naturally occurring wetlands, and pervious surfaces. We then statistically test the degree to which these indicators reduce insured flood losses observed across the study area over a five-year period from 2001 to 2005. Results based on multiple regression models controlling for various environmental and socioeconomic characteristics support the notion that certain features of the natural environment help mitigate the negative economic consequences that arise from floods. The findings provide guidance to local and regional policy makers on where to guide future development. Reducing the amount of flood losses helps building more flood-resilient human communities along the Gulf coast not only in terms of economic savings but also to reduce human loss. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction The Gulf of Mexico coastal margin has long been one of the regions in the U.S. most vulnerable to the adverse impacts of storms and associated flooding. Rapid population growth and development in low-lying areas have only exacerbated flood-related losses. For example, from 2000 to 2005 Gulf coast counties incurred over $52 billion in overall flood damage with $19 billion reported through the National Flood Insurance Program (NFIP). With mounting losses from both acute and chronic flood events, decision makers are beginning to focus on how to construct coastal communities that are more resilient to economic loss over the long term. Despite the increasing impacts of floods on localities, little research has been conducted on defining, measuring, and testing the effectiveness of ecological indicators related to flood risk. We address this issue by examining ecological conditions indicative of flood risk across 144 counties bordering the Gulf of Mexico. Specifically, we identify and measure four ecological indicators critical to protecting local communities against the adverse impacts of floods, then statistically test the degree to which they reduce flood losses observed across the study area over a five-year
∗ Corresponding author. Tel.: +1 409 740 4939; fax: +1 409 740 4787. E-mail addresses: [email protected] (S.D. Brody), [email protected] (W.G. Peacock), [email protected] (J. Gunn). 1470-160X/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.ecolind.2012.01.004
period. Study findings identify certain features of the natural environment that work to moderate the effects of floods and provide a foundation for building less flood-prone communities over the long term. Local and regional policy makers can use this information to develop more flood-resilient communities and reduce the amount of flood-related economic and human losses accruing along the Gulf coast. 2. Deriving ecological indicators of flood risk Earlier work in the field of disaster reduction and hazard mitigation recognizes community vulnerability as a place-based concept (Longhurst, 1995). In this sense, susceptibility to the adverse effects of natural hazards is based on the interaction of biophysical risk and socioeconomic conditions within a specific geographic domain (Cutter, 1996). When conditions associated with ecological risk coincide with human settlements, social and economic functions are disrupted (Smith and Petley, 2009). Recognizing that the social component critically interacts with the ecological component in social–ecological systems provides the basis for building more hazard-resilient communities (Berkes et al., 2003; Folke et al., 2005). The concept of an integrated socio-ecological system provides insights into how to facilitate community development while minimizing risk (Lebel et al., 2006; Paton and Johnston, 2006). It is important for planners and decision makers to
494
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
reduce pre-existing vulnerabilities by reducing exposure to hazards, such as flooding and storm surge (Mileti, 1999; Godschalk, 2003). Ecological components essentially moderate the impacts of natural hazards on human populations and the built environment (Beatley, 2009). Precisely where human communities persist within ecological and geophysical landscapes thus become a central aspect in determining the degree of risk and vulnerability. Jurisdictions that locate or build in risk-prone areas will be more likely to incur adverse impacts associated with natural hazards unless they can adopt appropriate coping strategies. While scholars recognize the importance of location-based factors when seeking to develop communities in vulnerable areas, they often stop short when it comes to identifying and measuring specific features of the natural environment that provide a foundation for minimizing flood hazard impacts. In response to this lack of study, we propose four baseline ecological indicators that serve as essential moderators between flood events and their degree of economic impact on local communities. Decision makers along the Gulf coast can use these measures to better understand their level vulnerability to flooding and help guide future development that is less damage-prone over the long term.
2.1. Floodplain area The 100-year floodplain (where there is a 1% change of flooding every year) is a longstanding marker for determining the possibility of an area being inundated by a rainfall event. Communities developing within designated floodplain boundaries are at a greater risk of flood impacts unless mitigation measures are in place. In fact, over the course of a 30-year mortgage, there is a 26% chance that a home in the floodplain will experience flooding (NRC, 2000). Filling-in areas of the floodplain or raising sites above flood elevation may be a protective measure for a specific development, but flooding risks are often increased downstream. The cumulative impacts of seemingly minor alterations can compromise the ability of hydrologic systems to store runoff and literally alter the boundary of a floodplain. For example, Brody et al. (2007a) found that, on average, increasing areas of floodplain in Florida were correlated with larger amounts of property damage from floods. We expect that jurisdictions along the Gulf of Mexico coast with larger percentages of floodplain area will experience significantly greater amounts of flood losses. We statistically test the hypothesis: localities with larger areas outside of the floodplain will experience lower amount of losses from flooding events.
2.2. Soil porosity Soil porosity is another important indicator of vulnerability to flooding because it helps determine the rate of surface water infiltration (Tollan, 2002; Chang and Franczyk, 2008). The amount of water that any given soil will infiltrate and retain depends primarily upon its texture and current moisture condition (Saxton and Shiau, 1990). Porous soils, such as those with high sand content drain much more quickly than low porosity soils, making them a potentially more suitable substrate for development. The potential for higher peak and mean annual flows from basins with low soil permeability is greater than that for basins with higher permeability soils, as higher permeability allows greater infiltration, more storage, and less runoff (Rasmussen and Perry, 2000). We test the hypothesis that: counties or parishes containing soils with higher levels of porosity will incur significantly lower amounts of property damage from floods.
2.3. Naturally occurring wetlands Naturally occurring wetlands are also key ecological indicators when considering the risk to flood damage because of their ability to attenuate floods caused by precipitation and storm surge events (Bullock and Acreman, 2003; Costanza et al., 2008). Both anecdotal and empirical research suggest that wetlands may reduce or slow flooding (Mitch and Gosselink, 2000; Lewis, 2001; Bullock and Acreman, 2003). For example, empirical research in Texas and Florida demonstrates the value of naturally occurring wetlands in reducing the adverse impacts of floods. Brody et al. (2007b) found that the development of wetlands significantly increased the number of exceedances in stream-flow across 85 watersheds in Texas and Florida. When controlling for multiple socioeconomic and geophysical contextual characteristics, Brody et al. (2008) found that the loss of wetlands across 37 coastal counties in Texas from 1997 to 2001 significantly increased observed amount of property damage from floods. Each permit granted by the USACE to alter a naturally occurring wetland under Section 404 of the Clean Water Act translated into an average of $211.88 in added property damage per flood over the five-year period. On average, wetland alteration added over $38,000 in property damage per flood (Brody et al., 2008). A companion analysis for every county in Florida showed similar results (Brody et al., 2007a). In this case, the alteration of wetlands increased the average property damage per flood by over $400,000. Given the evidence, communities that can maintain naturally occurring wetlands should be less vulnerable to flood disasters. Accordingly, we test the hypothesis that: wetland loss will exacerbate property damage from floods even over larger study areas, such as the Gulf of Mexico coast. 2.4. Pervious surfaces Another key ecological indicator moderating flood impacts is the amount of area within a jurisdiction not covered by impervious surfaces, such as roads, rooftops, and parking lots (Arnold and Gibbons, 1996). Pervious surfaces (e.g. green space, protected areas, and coastal prairies) serve important hydrological functions because they absorb, store, and slowly release water (Tourbier and Westmacott, 1981). Conversely, large areas of impervious surface coverage correspond to a decrease in rainfall infiltration and an increase in surface runoff (Paul and Meyer, 2001). Impervious surfaces are especially implicated in increased peak discharge (Brezonik and Stadelman, 2002). Under compromised hydrological conditions, the lag time between the center of precipitation volume and runoff volume is compressed so that floods peak more rapidly (Hirsch et al., 1990). This reduced lag time occurs because runoff reaches water bodies more quickly when rainfall is unable to infiltrate into the soil (Hsu et al., 2000; Hey, 2002). Overall, there is a substantial body of evidence to support the notion that development-based impervious surface increases runoff volume, peak discharges, and associated flood magnitudes. These conditions set the stage for more regular flood events and increasing human losses. For example, Brody et al. (2007b) found that an increase in impervious surfaces (as measured through remote sensing imagery) correlated with a significant increase in stream flow exceedances over a twelve-year period across 85 coastal watersheds in Texas and Florida. A follow-up study in coastal Texas examined the influence of the built environment from 1997 to 2001. The authors observed that over 37 counties, every square meter of additional impervious surface translated into approximately $3602 of added property damage caused by floods per year (Brody et al., 2008). Based on this evidence, undeveloped or pervious areas indicate a certain level of resiliency because they act to moderate flooding events and the degree of impact in terms
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
of property damage and human casualties. Therefore, we specifically test the hypothesis that: increasing areas of pervious surface will result in decreasing amounts of flood loss.
3. Research methods 3.1. Study area We selected as our study area 144 coastal counties and parishes along the Gulf of Mexico as identified by the National Oceanic and Atmospheric Administration (NOAA) (Crossett et al., 2004). This region extends from the Florida Keys westward to the Southern tip of Texas, and includes counties from the following six states: Florida, Georgia, Alabama, Mississippi, Louisiana, and Texas (Fig. 1). The NOAA special projects office defines a county (or parish if in Louisiana) as coastal if one of the following two criteria is met: (1) at a minimum, 15% of the county’s total land area is located within a coastal watershed or, (2) a portion of, or an entire county accounts for at least 15% of a coastal cataloging unit. Coastal counties associated with the Gulf of Mexico provide an ideal area in which to study ecological risk and flooding for several reasons. First, this low-lying coastal margin is extremely vulnerable to the adverse effects of rainfall and surge-based flooding events, resulting in a long history of losses to examine. For example, data on insured losses from 1996 to 2007 show that jurisdictions bordering the Gulf coast experienced the largest amount of property damage in the U.S. (Brody et al., 2011). Louisiana, which suffered extensively from Hurricane Katrina in 2005, reported the highest amount of property damage with over $13.8 billion over a twelve-year period, followed by Mississippi (over $2 billion), Florida (over $2 billion), Texas (over $1.8 billion), and Alabama (over $793 million). Second, this study area contains a wide range of ecological conditions that may act as moderators between storm events and impacts to human communities. For example, some of the largest areas of naturally occurring wetlands and coastal floodplain are located within the study area, providing an ideal test bed for examining their influence on flood losses. Third, these ecological features have been heavily impacted by human settlement. Localities fringing the Gulf of Mexico have a legacy of rapid population growth and a diversity of development patterns in flood-prone areas. For example, a NOAA survey of land use change along U.S. coasts from 1996 to 2001 revealed that 53% of new development occurred in the southeast between Texas and North Carolina. Finally, the study area and unit of analysis is a principle target for NOAA policy and planning for coastal resiliency and hazard reduction. The results of our study can therefore be directly used by decision makers as they work to reduce the adverse impacts of floods and other natural hazards in the Gulf coast.
3.2. Concept measurement Flood loss, the dependant variable for the study, was measured using National Flood Insurance Program (NFIP) property damage claims aggregated from 2001-2005. The NFIP was established in 1968 and is administered through FEMA to provide flood insurance to residents and businesses. In participating communities, residential structures within the 100-year floodplain with a mortgage are required to purchase flood insurance. For properties outside the floodplain, insurance coverage is optional. However, throughout the study area, we found that approximately 30% of claims are from properties outside the 100-year floodplain. The dollar value of individual insured claims was aggregated to the county or parish level and log-transformed to better approximate a normal distribution (see Table 1).
495
Ecological indicators of flood risk were measured using a variety of data sources. Non-floodplain area was derived by first calculating the percentage of a county or parish occupied by FEMA-defined 100-year floodplains. Information was acquired from FEMA Q3 data. On average, 35% of jurisdictions in the study area are within the 100-year floodplain. We then subtracted this percentage from 1 to calculate the percent area for a locality outside of the floodplain, indicating areas least at risk. Indicators of porosity and potential run-off were measured using the State Soil Geographic Database (STATSGO). Average permeability, based inches per hour of water infiltration, was generated for each jurisdiction. Higher averages of porosity indicate a greater degree of soil infiltration for rainwater. The amount of naturally occurring wetlands across the study area was measured as a percent change score over a five-year time period. We chose this approach because it is easier to detect a statistical signal representing the effect of wetlands on flood damages as these areas change over time (loss of ecological services) rather than examining only one time period. Wetland alteration was measured by calculating the percent change in wetland landcover classified by NOAA C-CAP using Landsat Thematic Mapper data at 30 m resolution from 2001 to 2005. This change measure represents loss (or in some cases gains) of wetlands within a county/parish over the five-year study period. While our measurement approach captures the amount of wetland change relative to the total area of wetland cover, other measures, such as timeaverages could better capture the impact of change. We also do not examine the specific type of wetland cover for each jurisdiction. Future work should address these measurement issues. Pervious surfaces were measured by subtracting the percent of impervious surface within a jurisdiction from 1. Impervious surfaces were calculated using NOAA C-CAP landcover classifications derived from Landsat Thematic Mapper data at 30 m resolution. Percentages of land cover were calculated in a GIS by using a shapefile of the 144 counties in the study area to compute zonal statistics. The area of all impervious surface types (low, medium, and high intensity) were aggregated and divided by the area of each jurisdiction to derive a percentage. We then measured the area within a county/parish not classified as impervious. Several contextual control variables were also measured and included in the explanatory model. Precipitation is an essential control variable as it is usually considered the most important factor contributing to local flooding and associated property damage. Generally, the more rainfall, the greater the likelihood streams and rivers will overflow due to excessive runoff. We calculated precipitation based on annual rainfall amounts from the PRISM dataset. Annual precipitation data for the five-year period was mapped at a scale of 30 arc second normals and then averaged annually in millimeters for each jurisdiction. To capture precipitation events most likely to result in flooding, we totaled the number of times per year in a jurisdiction rainfall amounts exceeded the 75th percentile for the study period. This variable was thus measured as a count variable, where higher values represent a greater number of extreme rainfall events. Surge-based flooding events also affected the study area from 2001 to 2005. During this time period, there were several waveinduced surge events, most notable the one caused by Hurricane Katrina in 2005. As with precipitation, surge inundates the landscape, but more from overland flow, which can cause sudden and catastrophic damage. We measured the control variable storm surge by summing the number of storm surge events over the study period as reported by the Spatial Hazard Events and Losses Database for the United States (SHELDUS version 8.0). This count variable ranges from 0 to 7 surge events per jurisdiction during the study period. Two additional socioeconomic controls were included in the final model. First, we measured the number of housing units in each
496
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
Fig. 1. Gulf of Mexico Coastal Study Area.
county/parish based on 2000 U.S. Census data. We expect that jurisdictions with more housing units are more likely to report higher amounts of flood-related property loss. Second, we controlled for relative wealth of the population using per capita income levels for each county or parish in the sample based on 2000 U.S. Census data. We expect that higher incomes are associated with the presence of more expensive structures and resulting flood damage. Finally, we measured a flood mitigation variable based on the FEMA Community Rating System (CRS) activity for floodplain management and planning (activity 510) for the year 2000. This variable was constructed based on the number of points a community receives by participating in the CRS and receiving credit for the specific activity. Jurisdictions received a 0 if they do not participate or receive credit for the activity. The highest scoring jurisdiction in the sample received 24 points, although the majority was coded as 0.
3.3. Data analysis Data on ecological indicators of flood risk are analyzed in two stages. First, we map and describe the spatial pattern for each indicator across the study area. Visual and descriptive analysis provides insights on the variation in landscape conditions and where the most resilient communities are situated. Second, we use robust regression analysis to explain the variation in insured flood losses across the study area. While we pursue a cross-sectional research design with which to analyze the data, we temporally position independent variables to reduce validity threats. For example, independent variables, such as landcover, and socioeconomic characteristics were measured at the beginning of the flood damage record (the year 2000 or 2001) to reinforce a statistical interpretation that these factors predict flood impacts. As with any study of this type, it is not possible to eliminate
Table 1 Concept measurement dependent and independent variables. Variable
Measurement
Source
Flood loss
Logged NFIP paid claims for flood losses from 2001 to 2005. Proportion area outside of the FEMA-defined 100-year floodplain. Average soil permeability in inches per hour.
FEMA
Non-floodplain area Soil permeability
Wetland alteration
Pervious surface
Precipitation
Storm surge Number of housing units Per capita household income Floodplain management & planning
Proportion area change in wetland cover from 2001 to 2005. Based on summing 30 m2 pixels based from Landsat Thematic Mapper remote sensing imagery. Proportion area not covered by impervious surfaces based on summing 30 m2 pixels from Landsat Thematic Mapper remote sensing imagery. Number of times precipitation exceeded the 75th percentile for during the study period. Number of storm surge events per jurisdiction during the study period. U.S. Census 2000 estimates for the number of housing units in each jurisdiction. U.S. Census 2000 estimates for the median household income for each jurisdiction. Number of credit points received through the FEMA CRS.
FEMA
Range 0–22.6 0.03–1
State Soil Geographic Database NOAA, Coastal Change & Analysis Program
0.7–13
NOAA, Coastal Change & Analysis Program
0.40–0.99
−0.038 to 0.002
Mean
Standard deviation
13.66
4.47
0.658
0.245
5.35
4.33
−0.004
0.005
0.943
0.082
PRISM Climate Group
0–5
1.65
1.44
SHELDUS v8.0
0–7
0.68
1.62
U.S. Census
U.S. Census
FEMA
281–1,298,130
7069–31,195
0–24
54,396
126,431
16,194
3,593
1.11
4.21
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
497
Fig. 2. Percent of Gulf Coastal Communities within FEMA’s 100-year floodplain.
history threats, but temporally ordering variables better justifies the stated x–y relationships. Diagnostics for model misspecification and AIC analysis, variance inflation/multicollinearity, and spatial-autocorrelation did not yield any major violations. We did, however, find heteroskedacticity in the data, leading us to estimate the model with robust standard errors. No other violations of OLS regression assumptions were detected. 4. Results Descriptive and visual analysis illustrates extensive variation in the pattern of ecological flood indicators across the study area. The amount of area outside of the 100-year floodplain varies widely across the study area, ranging 3 to over 99% of a jurisdiction. As shown in Fig. 2, the highest proportion of floodplain occurs in the central part of the study area in and adjacent to New Orleans. Areas along the Florida Panhandle (Gulf and Franklin Counties) and upper Texas coast (Jefferson and Brazoria Counties) also contain well-above average (35%) proportions of floodplain area. Generally, communities situated away from the coastline at higher elevations contain a lower percentage of floodplain area, making it easier to develop without the threat of flooding during severe storms. For example, inland counties in Texas and parts of Florida contain the least amount of floodplain area, making them ideal locations for resilient development projects. As indicated in Table 1, the average level of soil permeability for the study area is 5 inches per hour, with a range from 0 to 13 inches per hour. Overall, this indicates poorly drained soils
indicative of coastal plain geography associated with the Gulf of Mexico Coast. Fig. 3 illustrates a well-defined spatial trend for soil permeability across the 144-county study area. Generally, counties in Florida with high sand content are the most permeable while the clay dominated counties in the upper Texas coast are the least permeable. In fact, Galveston and Brazoria counties in Texas have the least permeable soils in the study area while Gulf and Hendry counties in western Florida have the highest levels of porosity. Thus, from an ecological perspective, localities in Florida are preferable for human settlement because surface runoff will, on average, drain much more quickly. The mean calculated proportion of wetlands per jurisdiction in 2001 was approximately 3%, with a wide range from 0.04 to 90%. Monroe and Collier counties contain the most wetland area due to their association with the Everglades Ecosystem in south Florida. Several parishes in Louisiana, including Iberville, St. Martin, and Terrebonne contain over 60% wetlands within their boundaries. In contrast, large percentages of wetlands are scarce among Texas inland counties. As shown in Fig. 4, the loss of wetlands from 2001 to 2005 occurred in rapidly expanding urban and suburban coastal areas. Specifically, wetland areas surrounding New Orleans, LA, Sarasota, FL, and the Houston-Galveston region in TX were converted for development during the study time period. While there exist large population centers, very little of the total study area is developed. The average percent of pervious surface across the entire study area is over 94% (Fig. 5). As expected, areas containing large urban and sprawling suburban development patterns have the least amount of pervious land. Pinellas County, FL
Fig. 3. Soil permeability for Gulf Coastal Communities.
498
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
Fig. 4. Wetland losses and gains in Gulf Coastal Communities.
(40%), Harris County, TX (50%), and Orleans Parish, LA (66%) are the top three jurisdictions in terms of the lowest percentage of area left as pervious. On the other end of the spectrum, low-populated jurisdictions, such as Kennedy County in TX, where the 2000 U.S. Census estimated a population of only 414 people, has over 99% of its land area calculated as pervious. Results from robust regression analysis (Table 2) indicate that several ecological indicators act to reduce the adverse impacts of flooding, even when controlling for socioeconomic variables. For example, a percent increase in the area of land outside of the 100year floodplain significantly (p < 0.01) reduces insured flood losses from 2001 to 2005. In fact, based on standardized betas, this variable has the strongest effect on observed property damages from floods among all indicators in the model. Increasing levels of soil porosity also has a strong negative effect on flood losses (p < 0.01). In particular, development on well-drained, sandy soils appears to result in the least amount of damage. In contrast, a percent change in wetland area within a jurisdiction significantly increases flood damages (p < 0.01). Wetland losses over time reduce the ability of the hydrological system to naturally attenuate floods, and when replaced with human development, places structures in floodprone areas. Finally, while the effect is negative, large amounts of pervious surfaces alone do not significantly impact the degree of property damage caused by floods along the Gulf coast. Several control variables in the model are also significant predictors of flood loss. As expected, both storm surge and precipitation significantly increase insured flood damage (p < 0.01). Storm surge has the greatest impact among communities directly bordering the
Gulf. Increasing number of housing units and per capita income levels also lead to significantly higher (p < 0.01) amounts of property damage from flooding. More and expensive structures in vulnerable areas amplify the financial impact of each flooding event. Finally, floodplain planning and management mitigation as measured through the CRS does not appear to have a significant effect on flood losses. 5. Discussion Our results show that several ecological indicators can be used to identify areas where flood-related economic impact along the Gulf coast can be high. These findings provide statistical evidence that specific characteristics of the landscape provide a moderating function with respect to property damage resulting from floods. Identifying the importance of key ecological indicators can signal to decision makers where and how to develop communities that are more resilient to flooding over the long term. The indicator with the strongest effect in terms of reducing flood losses is the percentage of area outside of the 100-year floodplain. Specifically, a percent increase in these less vulnerable areas among Gulf coast counties/parishes is equivalent to, on average, $309,540,000 in savings (based on the unstandardized coefficient) from flood damage. This effect translates into an average $89.15 per acre reduction in flood losses. There are several policy implications stemming from this result. First, jurisdictions along the Gulf coast containing larger percentages of land outside the floodplain have more opportunity to become flood resilient and greater flexibility
Fig. 5. Percent pervious surface within Gulf Coastal Communities.
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
499
Table 2 Multiple regression model explaining insured flood losses.
Non-floodplain area Soil permeability Wetland alteration Pervious surface Precipitation Storm surge Number of housing units Per capita income CRS points
Coefficient
Standard error
T-value
Significance
−2.330175 −0.1282054 82.3422 −4.43626 1.663426 0.6577005 0.0000142 0.0001894 −0.0037355
0.7720293 0.0430806 31.17052 4.573353 0.4033117 0.1223042 4.75e−06 0.0000548 0.0478007
−3.02 −2.98 2.64 −0.97 4.12 5.38 2.99 3.46 −0.08
0.003 0.003 0.009 0.334 0.000 0.000 0.003 0.001 0.938
N = 144; F(9, 133) = 28.17; Prob > F = 0.000; Adj. R2 = .3974; VIF = 1.73; AIC = 777.18. Note: dependent variable is the logged insured flood losses from 2001 to 2005.
as to where development takes place. Development in these jurisdictions will also be, on average, less expensive because of lowered requirements for extensive drainage systems and associated infrastructure. Second, jurisdictions containing floodplain area (all but two in the sample) should pursue a development approach that avoids these vulnerable areas. Such an avoidance strategy involves spatially targeting future development in areas well outside of floodplain boundaries. Land use tools, such as cluster development, density bonuses, and capital improvements programming have been shown to be effective in guiding development towards more resilient locations (Beatley, 2009). Naturally occurring wetlands along the Gulf coast appears to be another key ecological indicator of resiliency when dealing with floods. An acre loss of naturally occurring wetlands from 2001 to 2005 increased property damage caused by flooding by an average of $7,457,549, which amounts to approximately $1.5 million per year across the study area. The role of wetlands in attenuating the adverse impacts of floods at the landscape scale should not be overlooked by planners and policy makers as they clearly provide an important protective and economic benefit to coastal communities. These natural features provide not only wildlife habitat and increased biological diversity in coastal areas, but also a critical buffer to storms which, as our model demonstrates, translates into significant dollar savings. Land acquisition techniques, coastal buffers, and critical area setbacks are a few of the tools available to planners that, when implemented, can help protect wetland functions. One of the more overlooked landscape characteristics in terms of reducing flood impacts is soil porosity. Soil structure may be considered at the site level, but rarely does this factor enter the minds of planners and policy makers as they consider future development patterns. However, our results show that well-drained soils are critical to reducing storm-water run-off and associated flood damages. Based on our model, a percent increase in soil permeability translates into a $3,403,600 reduction in property damage from floods per year. Coastal planners, policy makers, and developers should consider in their decisions soils alongside floodplains and any other indicator of vulnerability that could impact property and lives. Conservation overlay zones, subdivision ordinances, site plan requirements, and public expenditure guidelines are among the more effective techniques that can help reduce development pressure on impermeable soils. Overall, using ecological indicators to devise a program that combines structural and non-structural mitigation techniques can help reduce the adverse impacts of floods along the Gulf coast and increase the resiliency of local communities over the long term. While our flood mitigation measure was not significant in the statistical model, the techniques listed above can be effective in reducing flood losses. Previous studies have shown that communities that participate in FEMA’s CRS experience significantly less property damage caused by flood events (Brody et al., 2007a,b, 2008).
6. Conclusion Our results demonstrate that several ecological characteristics act to mitigate the economic impacts of flooding along the Gulf of Mexico coast. These indicators significantly influence the degree to which storms impact property within coastal communities. Accordingly, they should be strongly considered by planners and policy makers as they decide where and how much development should take place in the future. While our study provides insights on the moderating effect of landscape characteristics and the interaction between social and ecological systems, it should be considered only a starting point for more thorough research on the topic. First, our study examines only four indicators of resiliency. Future research should consider a wider range of factors, including different wetland types, forestlands, topography, etc. Second, the county/parish unit of analysis provides only a broad assessment of indicator effectiveness. Addition work should be done at smaller scales, preferably at the neighborhood level where there may be reduced spatial error and stronger statistical validity when making conclusions. Third, we analyzed data over only a fiveyear period, which offers a small window for assessment. Future research should look over larger time horizons with multiple storm events to increase the ability to make generalizations about the results. Fourth, the analysis is based on a cross-section research design that does not take into consideration changes within the study period. Future research should consider panel or time series models that account for temporal changes from year to year. Finally, we examine only one region in the U.S. A comparative study evaluating multiple coastlines would shed additional light on the notion of ecological resiliency and the relative effectiveness of landscape features in moderating the impacts of storm events. Acknowledgements This paper is based on research supported by NOAA Award No. NA07NOS4730147. The findings and opinions reported are those of the authors and are not necessarily endorsed by the funding organizations or those who provided assistance with various aspects of the study. References Arnold, C.L., Gibbons, J.C., 1996. Impervious surface coverage: the emergence of a key environmental indicator. Journal of the American Planning Association 62 (2), 243–258. Beatley, T., 2009. Planning for Coastal Resilience: Best Practices for Calamitous Times. Island Press, Washington, DC. Berkes, Colding, Folke (Eds.), 2003. Navigating Social Ecological Systems: Building Resilience for Complexity and Change. Cambridge University Press, Cambridge. Brezonik, P.L., Stadelman, T.H., 2002. Analysis and predictive models of stormwater runoff volumes loads and pollutant concentrations from watersheds in the twin cities metropolitan area, Minnesota, USA. Water Resources 36, 1743–1757. Brody, S.D., Highfield, W.E., Kang, J.E., 2011. Rising Waters: The Causes and Consequences of Flooding in the United States. Cambridge University Press, Cambridge, UK.
500
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
Brody, S.D., Zahran, S., Maghelal, P., Grover, H., Highfield, W., 2007a. The rising costs of floods: examining the impact of planning and development decisions on property damage in Florida. Journal of the American Planning Association 73 (3), 330–345. Brody, S.D., Highfield, W., Ryu, H.C., Spanel-Weber, L., 2007b. Examining the relationship between wetland alteration and watershed flooding in Texas and Florida. Natural Hazards 40 (2), 413–428. Brody, S.D., Zahran, S., Highfield Grover, H., Vedlitz, A., 2008. Identifying the impact of the built environment on flood damage in Texas. Disasters 32 (1), 1–18. Bullock, A., Acreman, M., 2003. The role of wetlands in the hydrological cycle. Hydrology and Earth System Sciences 7 (3), 358–389. Chang, H., Franczyk, J., 2008. Climate change land-use change, and floods. Toward an integrated assessment. Geography Compass 2 (5), 1549–1579. Costanza, R., Pérez-Maqueo, O., Luisa, M., Sutton, P., Anderson, S.J., Mulder, K., 2008. The value of coastal wetlands for hurricane protection. Ambio 37 (4), 241–247. Crossett, K., Culliton, T.J., Wiley, P., Goodspeed, T.R., 2004. Population trends along the coastal United States, 1980–2008. National Oceanic and Atmospheric, Administration Coastal Trends Report Series. Cutter, S.L., 1996. Vulnerability to natural hazards. Progress in Human Geography 20 (4), 529–539. Folke, C., Hahn, T., Olsson, P., Norberg, J., 2005. Adaptive governance of social–ecological systems. Annual Review of Environmental Resources 30, 441–473. Godschalk, D., 2003. Urban hazard mitigation: creating resilient cities. Natural Hazards Review 4 (3), 136–142. Hey, D.L., 2002. Modern drainage design: the pros, the cons, and the future. Hydrological Science and Technology 18 (14), 89–99. Hirsch, R.M., Walker, J.F., Day, J.C., Kallio, R., 1990. The influence of main on hydrological systems. In: Wolman, M.G., Riggs, H.C. (Eds.), Surface Water Hydrology. Geological Society of America, Boulder, CO, pp. 329–359.
Hsu, M.H., Chen, S.H., Chang, T.J., 2000. Inundation simulation for urban drainage basin with storm sewer system. Journal of Hydrology 234, 21–37. Lebel, L., Anderies, J.M., Campbell, B., Folke, C., Hatfield-Dodds, S., Hughes, T.P., Wilson, J., 2006. Governance and the capacity to manage resilience in regional social–ecological systems. Ecology and Society 19, 1–15. Lewis, W.M., 2001. Wetlands Explained: Wetland Science, Policy, and Politics in America. Oxford University Press, New York, NY. Longhurst, R., 1995. The assessment of community vulnerability in hazard prone areas. Disasters 19, 269–270. Mileti, D.S., 1999. Disaster by Design: A Reassessment of Natural Hazards in the United States. Joseph Henry Press, Washington, DC. Mitch, W.J., Gosselink, J.G., 2000. Wetlands, 3rd ed. John Wiley & Sons, New York. National Research Council (NRC), 2000. Risk Analysis and Uncertainty in Flood Damage Reduction Studies. National Academy Press, Washington, DC. Paton, D., Johnston, D. (Eds.), 2006. Disaster Resilience: An Integrated Approach. Charles C. Thomas, Springfield, IL. Paul, M.J., Meyer, J.L., 2001. Streams in the urban landscape. Annual Review of Ecological Systems 32, 333–365. Rasmussen, P.P., Perry, C.A., 2000. Estimation of Peak Streamflows for Unregulated Rural Streams in Kansas. U.S. Geological Survey, Water-Resources Investigations, Report 00–4079. Saxton, K.E., Shiau, S.Y., 1990. Surface waters of North America; influence of land and vegetation on streamflow. In: Wolman, M.G., Riggs, H.C. (Eds.), The Geology of North America, vol. 0–1, Surface Water Hydrology. The Geologic Society of America, Washington, DC, pp. 55–80. Smith and Petley Environmental Hazards: Assessing Risk and Reducing Disaster. Routledge, New York, 2009. Tollan, A., 2002. Land-use change and floods: what do we need most, research or management? Water Science and Technology 45 (8), 183–190. Tourbier, J.T., Westmacott, R., 1981. Water Resources Protection Technology: A Handbook of Measures to Protect Water Resources in Land Development. The Urban Land Institute, Washington, DC.
Contents lists available at SciVerse ScienceDirect
Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind
Ecological indicators of flood risk along the Gulf of Mexico Samuel D. Brody a,∗ , Walter Gillis Peacock b , Joshua Gunn b a Texas A&M University at Galveston, Department of Marine Sciences, Ocean and Coastal Studies Building, Building 3029, Room # 366, 200 Seawolf Parkway, Galveston, TX 77553, USA b Texas A&M University at College Station, Department of Landscape Architecture and Urban Planning, TAMU 3731, Texas A&M University, College Station, TX 77843-3137, USA
a r t i c l e
i n f o
Article history: Received 3 July 2011 Received in revised form 2 January 2012 Accepted 4 January 2012 Keywords: Ecological indicators Gulf of Mexico Flood damage Social–ecological systems
a b s t r a c t Despite mounting economic losses from both acute and chronic flood events in coastal areas of the U.S., little empirical research has been conducted on the importance of existing landscape-level ecological components in mitigating the economic impacts to vulnerable coastal communities over the long term. In recognition of this lack of knowledge base, we examine several ecological indicators across 144 counties bordering the Gulf of Mexico. Specifically, we identify and measure the following four indicators: floodplain area, soil porosity, naturally occurring wetlands, and pervious surfaces. We then statistically test the degree to which these indicators reduce insured flood losses observed across the study area over a five-year period from 2001 to 2005. Results based on multiple regression models controlling for various environmental and socioeconomic characteristics support the notion that certain features of the natural environment help mitigate the negative economic consequences that arise from floods. The findings provide guidance to local and regional policy makers on where to guide future development. Reducing the amount of flood losses helps building more flood-resilient human communities along the Gulf coast not only in terms of economic savings but also to reduce human loss. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction The Gulf of Mexico coastal margin has long been one of the regions in the U.S. most vulnerable to the adverse impacts of storms and associated flooding. Rapid population growth and development in low-lying areas have only exacerbated flood-related losses. For example, from 2000 to 2005 Gulf coast counties incurred over $52 billion in overall flood damage with $19 billion reported through the National Flood Insurance Program (NFIP). With mounting losses from both acute and chronic flood events, decision makers are beginning to focus on how to construct coastal communities that are more resilient to economic loss over the long term. Despite the increasing impacts of floods on localities, little research has been conducted on defining, measuring, and testing the effectiveness of ecological indicators related to flood risk. We address this issue by examining ecological conditions indicative of flood risk across 144 counties bordering the Gulf of Mexico. Specifically, we identify and measure four ecological indicators critical to protecting local communities against the adverse impacts of floods, then statistically test the degree to which they reduce flood losses observed across the study area over a five-year
∗ Corresponding author. Tel.: +1 409 740 4939; fax: +1 409 740 4787. E-mail addresses: [email protected] (S.D. Brody), [email protected] (W.G. Peacock), [email protected] (J. Gunn). 1470-160X/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.ecolind.2012.01.004
period. Study findings identify certain features of the natural environment that work to moderate the effects of floods and provide a foundation for building less flood-prone communities over the long term. Local and regional policy makers can use this information to develop more flood-resilient communities and reduce the amount of flood-related economic and human losses accruing along the Gulf coast. 2. Deriving ecological indicators of flood risk Earlier work in the field of disaster reduction and hazard mitigation recognizes community vulnerability as a place-based concept (Longhurst, 1995). In this sense, susceptibility to the adverse effects of natural hazards is based on the interaction of biophysical risk and socioeconomic conditions within a specific geographic domain (Cutter, 1996). When conditions associated with ecological risk coincide with human settlements, social and economic functions are disrupted (Smith and Petley, 2009). Recognizing that the social component critically interacts with the ecological component in social–ecological systems provides the basis for building more hazard-resilient communities (Berkes et al., 2003; Folke et al., 2005). The concept of an integrated socio-ecological system provides insights into how to facilitate community development while minimizing risk (Lebel et al., 2006; Paton and Johnston, 2006). It is important for planners and decision makers to
494
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
reduce pre-existing vulnerabilities by reducing exposure to hazards, such as flooding and storm surge (Mileti, 1999; Godschalk, 2003). Ecological components essentially moderate the impacts of natural hazards on human populations and the built environment (Beatley, 2009). Precisely where human communities persist within ecological and geophysical landscapes thus become a central aspect in determining the degree of risk and vulnerability. Jurisdictions that locate or build in risk-prone areas will be more likely to incur adverse impacts associated with natural hazards unless they can adopt appropriate coping strategies. While scholars recognize the importance of location-based factors when seeking to develop communities in vulnerable areas, they often stop short when it comes to identifying and measuring specific features of the natural environment that provide a foundation for minimizing flood hazard impacts. In response to this lack of study, we propose four baseline ecological indicators that serve as essential moderators between flood events and their degree of economic impact on local communities. Decision makers along the Gulf coast can use these measures to better understand their level vulnerability to flooding and help guide future development that is less damage-prone over the long term.
2.1. Floodplain area The 100-year floodplain (where there is a 1% change of flooding every year) is a longstanding marker for determining the possibility of an area being inundated by a rainfall event. Communities developing within designated floodplain boundaries are at a greater risk of flood impacts unless mitigation measures are in place. In fact, over the course of a 30-year mortgage, there is a 26% chance that a home in the floodplain will experience flooding (NRC, 2000). Filling-in areas of the floodplain or raising sites above flood elevation may be a protective measure for a specific development, but flooding risks are often increased downstream. The cumulative impacts of seemingly minor alterations can compromise the ability of hydrologic systems to store runoff and literally alter the boundary of a floodplain. For example, Brody et al. (2007a) found that, on average, increasing areas of floodplain in Florida were correlated with larger amounts of property damage from floods. We expect that jurisdictions along the Gulf of Mexico coast with larger percentages of floodplain area will experience significantly greater amounts of flood losses. We statistically test the hypothesis: localities with larger areas outside of the floodplain will experience lower amount of losses from flooding events.
2.2. Soil porosity Soil porosity is another important indicator of vulnerability to flooding because it helps determine the rate of surface water infiltration (Tollan, 2002; Chang and Franczyk, 2008). The amount of water that any given soil will infiltrate and retain depends primarily upon its texture and current moisture condition (Saxton and Shiau, 1990). Porous soils, such as those with high sand content drain much more quickly than low porosity soils, making them a potentially more suitable substrate for development. The potential for higher peak and mean annual flows from basins with low soil permeability is greater than that for basins with higher permeability soils, as higher permeability allows greater infiltration, more storage, and less runoff (Rasmussen and Perry, 2000). We test the hypothesis that: counties or parishes containing soils with higher levels of porosity will incur significantly lower amounts of property damage from floods.
2.3. Naturally occurring wetlands Naturally occurring wetlands are also key ecological indicators when considering the risk to flood damage because of their ability to attenuate floods caused by precipitation and storm surge events (Bullock and Acreman, 2003; Costanza et al., 2008). Both anecdotal and empirical research suggest that wetlands may reduce or slow flooding (Mitch and Gosselink, 2000; Lewis, 2001; Bullock and Acreman, 2003). For example, empirical research in Texas and Florida demonstrates the value of naturally occurring wetlands in reducing the adverse impacts of floods. Brody et al. (2007b) found that the development of wetlands significantly increased the number of exceedances in stream-flow across 85 watersheds in Texas and Florida. When controlling for multiple socioeconomic and geophysical contextual characteristics, Brody et al. (2008) found that the loss of wetlands across 37 coastal counties in Texas from 1997 to 2001 significantly increased observed amount of property damage from floods. Each permit granted by the USACE to alter a naturally occurring wetland under Section 404 of the Clean Water Act translated into an average of $211.88 in added property damage per flood over the five-year period. On average, wetland alteration added over $38,000 in property damage per flood (Brody et al., 2008). A companion analysis for every county in Florida showed similar results (Brody et al., 2007a). In this case, the alteration of wetlands increased the average property damage per flood by over $400,000. Given the evidence, communities that can maintain naturally occurring wetlands should be less vulnerable to flood disasters. Accordingly, we test the hypothesis that: wetland loss will exacerbate property damage from floods even over larger study areas, such as the Gulf of Mexico coast. 2.4. Pervious surfaces Another key ecological indicator moderating flood impacts is the amount of area within a jurisdiction not covered by impervious surfaces, such as roads, rooftops, and parking lots (Arnold and Gibbons, 1996). Pervious surfaces (e.g. green space, protected areas, and coastal prairies) serve important hydrological functions because they absorb, store, and slowly release water (Tourbier and Westmacott, 1981). Conversely, large areas of impervious surface coverage correspond to a decrease in rainfall infiltration and an increase in surface runoff (Paul and Meyer, 2001). Impervious surfaces are especially implicated in increased peak discharge (Brezonik and Stadelman, 2002). Under compromised hydrological conditions, the lag time between the center of precipitation volume and runoff volume is compressed so that floods peak more rapidly (Hirsch et al., 1990). This reduced lag time occurs because runoff reaches water bodies more quickly when rainfall is unable to infiltrate into the soil (Hsu et al., 2000; Hey, 2002). Overall, there is a substantial body of evidence to support the notion that development-based impervious surface increases runoff volume, peak discharges, and associated flood magnitudes. These conditions set the stage for more regular flood events and increasing human losses. For example, Brody et al. (2007b) found that an increase in impervious surfaces (as measured through remote sensing imagery) correlated with a significant increase in stream flow exceedances over a twelve-year period across 85 coastal watersheds in Texas and Florida. A follow-up study in coastal Texas examined the influence of the built environment from 1997 to 2001. The authors observed that over 37 counties, every square meter of additional impervious surface translated into approximately $3602 of added property damage caused by floods per year (Brody et al., 2008). Based on this evidence, undeveloped or pervious areas indicate a certain level of resiliency because they act to moderate flooding events and the degree of impact in terms
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
of property damage and human casualties. Therefore, we specifically test the hypothesis that: increasing areas of pervious surface will result in decreasing amounts of flood loss.
3. Research methods 3.1. Study area We selected as our study area 144 coastal counties and parishes along the Gulf of Mexico as identified by the National Oceanic and Atmospheric Administration (NOAA) (Crossett et al., 2004). This region extends from the Florida Keys westward to the Southern tip of Texas, and includes counties from the following six states: Florida, Georgia, Alabama, Mississippi, Louisiana, and Texas (Fig. 1). The NOAA special projects office defines a county (or parish if in Louisiana) as coastal if one of the following two criteria is met: (1) at a minimum, 15% of the county’s total land area is located within a coastal watershed or, (2) a portion of, or an entire county accounts for at least 15% of a coastal cataloging unit. Coastal counties associated with the Gulf of Mexico provide an ideal area in which to study ecological risk and flooding for several reasons. First, this low-lying coastal margin is extremely vulnerable to the adverse effects of rainfall and surge-based flooding events, resulting in a long history of losses to examine. For example, data on insured losses from 1996 to 2007 show that jurisdictions bordering the Gulf coast experienced the largest amount of property damage in the U.S. (Brody et al., 2011). Louisiana, which suffered extensively from Hurricane Katrina in 2005, reported the highest amount of property damage with over $13.8 billion over a twelve-year period, followed by Mississippi (over $2 billion), Florida (over $2 billion), Texas (over $1.8 billion), and Alabama (over $793 million). Second, this study area contains a wide range of ecological conditions that may act as moderators between storm events and impacts to human communities. For example, some of the largest areas of naturally occurring wetlands and coastal floodplain are located within the study area, providing an ideal test bed for examining their influence on flood losses. Third, these ecological features have been heavily impacted by human settlement. Localities fringing the Gulf of Mexico have a legacy of rapid population growth and a diversity of development patterns in flood-prone areas. For example, a NOAA survey of land use change along U.S. coasts from 1996 to 2001 revealed that 53% of new development occurred in the southeast between Texas and North Carolina. Finally, the study area and unit of analysis is a principle target for NOAA policy and planning for coastal resiliency and hazard reduction. The results of our study can therefore be directly used by decision makers as they work to reduce the adverse impacts of floods and other natural hazards in the Gulf coast.
3.2. Concept measurement Flood loss, the dependant variable for the study, was measured using National Flood Insurance Program (NFIP) property damage claims aggregated from 2001-2005. The NFIP was established in 1968 and is administered through FEMA to provide flood insurance to residents and businesses. In participating communities, residential structures within the 100-year floodplain with a mortgage are required to purchase flood insurance. For properties outside the floodplain, insurance coverage is optional. However, throughout the study area, we found that approximately 30% of claims are from properties outside the 100-year floodplain. The dollar value of individual insured claims was aggregated to the county or parish level and log-transformed to better approximate a normal distribution (see Table 1).
495
Ecological indicators of flood risk were measured using a variety of data sources. Non-floodplain area was derived by first calculating the percentage of a county or parish occupied by FEMA-defined 100-year floodplains. Information was acquired from FEMA Q3 data. On average, 35% of jurisdictions in the study area are within the 100-year floodplain. We then subtracted this percentage from 1 to calculate the percent area for a locality outside of the floodplain, indicating areas least at risk. Indicators of porosity and potential run-off were measured using the State Soil Geographic Database (STATSGO). Average permeability, based inches per hour of water infiltration, was generated for each jurisdiction. Higher averages of porosity indicate a greater degree of soil infiltration for rainwater. The amount of naturally occurring wetlands across the study area was measured as a percent change score over a five-year time period. We chose this approach because it is easier to detect a statistical signal representing the effect of wetlands on flood damages as these areas change over time (loss of ecological services) rather than examining only one time period. Wetland alteration was measured by calculating the percent change in wetland landcover classified by NOAA C-CAP using Landsat Thematic Mapper data at 30 m resolution from 2001 to 2005. This change measure represents loss (or in some cases gains) of wetlands within a county/parish over the five-year study period. While our measurement approach captures the amount of wetland change relative to the total area of wetland cover, other measures, such as timeaverages could better capture the impact of change. We also do not examine the specific type of wetland cover for each jurisdiction. Future work should address these measurement issues. Pervious surfaces were measured by subtracting the percent of impervious surface within a jurisdiction from 1. Impervious surfaces were calculated using NOAA C-CAP landcover classifications derived from Landsat Thematic Mapper data at 30 m resolution. Percentages of land cover were calculated in a GIS by using a shapefile of the 144 counties in the study area to compute zonal statistics. The area of all impervious surface types (low, medium, and high intensity) were aggregated and divided by the area of each jurisdiction to derive a percentage. We then measured the area within a county/parish not classified as impervious. Several contextual control variables were also measured and included in the explanatory model. Precipitation is an essential control variable as it is usually considered the most important factor contributing to local flooding and associated property damage. Generally, the more rainfall, the greater the likelihood streams and rivers will overflow due to excessive runoff. We calculated precipitation based on annual rainfall amounts from the PRISM dataset. Annual precipitation data for the five-year period was mapped at a scale of 30 arc second normals and then averaged annually in millimeters for each jurisdiction. To capture precipitation events most likely to result in flooding, we totaled the number of times per year in a jurisdiction rainfall amounts exceeded the 75th percentile for the study period. This variable was thus measured as a count variable, where higher values represent a greater number of extreme rainfall events. Surge-based flooding events also affected the study area from 2001 to 2005. During this time period, there were several waveinduced surge events, most notable the one caused by Hurricane Katrina in 2005. As with precipitation, surge inundates the landscape, but more from overland flow, which can cause sudden and catastrophic damage. We measured the control variable storm surge by summing the number of storm surge events over the study period as reported by the Spatial Hazard Events and Losses Database for the United States (SHELDUS version 8.0). This count variable ranges from 0 to 7 surge events per jurisdiction during the study period. Two additional socioeconomic controls were included in the final model. First, we measured the number of housing units in each
496
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
Fig. 1. Gulf of Mexico Coastal Study Area.
county/parish based on 2000 U.S. Census data. We expect that jurisdictions with more housing units are more likely to report higher amounts of flood-related property loss. Second, we controlled for relative wealth of the population using per capita income levels for each county or parish in the sample based on 2000 U.S. Census data. We expect that higher incomes are associated with the presence of more expensive structures and resulting flood damage. Finally, we measured a flood mitigation variable based on the FEMA Community Rating System (CRS) activity for floodplain management and planning (activity 510) for the year 2000. This variable was constructed based on the number of points a community receives by participating in the CRS and receiving credit for the specific activity. Jurisdictions received a 0 if they do not participate or receive credit for the activity. The highest scoring jurisdiction in the sample received 24 points, although the majority was coded as 0.
3.3. Data analysis Data on ecological indicators of flood risk are analyzed in two stages. First, we map and describe the spatial pattern for each indicator across the study area. Visual and descriptive analysis provides insights on the variation in landscape conditions and where the most resilient communities are situated. Second, we use robust regression analysis to explain the variation in insured flood losses across the study area. While we pursue a cross-sectional research design with which to analyze the data, we temporally position independent variables to reduce validity threats. For example, independent variables, such as landcover, and socioeconomic characteristics were measured at the beginning of the flood damage record (the year 2000 or 2001) to reinforce a statistical interpretation that these factors predict flood impacts. As with any study of this type, it is not possible to eliminate
Table 1 Concept measurement dependent and independent variables. Variable
Measurement
Source
Flood loss
Logged NFIP paid claims for flood losses from 2001 to 2005. Proportion area outside of the FEMA-defined 100-year floodplain. Average soil permeability in inches per hour.
FEMA
Non-floodplain area Soil permeability
Wetland alteration
Pervious surface
Precipitation
Storm surge Number of housing units Per capita household income Floodplain management & planning
Proportion area change in wetland cover from 2001 to 2005. Based on summing 30 m2 pixels based from Landsat Thematic Mapper remote sensing imagery. Proportion area not covered by impervious surfaces based on summing 30 m2 pixels from Landsat Thematic Mapper remote sensing imagery. Number of times precipitation exceeded the 75th percentile for during the study period. Number of storm surge events per jurisdiction during the study period. U.S. Census 2000 estimates for the number of housing units in each jurisdiction. U.S. Census 2000 estimates for the median household income for each jurisdiction. Number of credit points received through the FEMA CRS.
FEMA
Range 0–22.6 0.03–1
State Soil Geographic Database NOAA, Coastal Change & Analysis Program
0.7–13
NOAA, Coastal Change & Analysis Program
0.40–0.99
−0.038 to 0.002
Mean
Standard deviation
13.66
4.47
0.658
0.245
5.35
4.33
−0.004
0.005
0.943
0.082
PRISM Climate Group
0–5
1.65
1.44
SHELDUS v8.0
0–7
0.68
1.62
U.S. Census
U.S. Census
FEMA
281–1,298,130
7069–31,195
0–24
54,396
126,431
16,194
3,593
1.11
4.21
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
497
Fig. 2. Percent of Gulf Coastal Communities within FEMA’s 100-year floodplain.
history threats, but temporally ordering variables better justifies the stated x–y relationships. Diagnostics for model misspecification and AIC analysis, variance inflation/multicollinearity, and spatial-autocorrelation did not yield any major violations. We did, however, find heteroskedacticity in the data, leading us to estimate the model with robust standard errors. No other violations of OLS regression assumptions were detected. 4. Results Descriptive and visual analysis illustrates extensive variation in the pattern of ecological flood indicators across the study area. The amount of area outside of the 100-year floodplain varies widely across the study area, ranging 3 to over 99% of a jurisdiction. As shown in Fig. 2, the highest proportion of floodplain occurs in the central part of the study area in and adjacent to New Orleans. Areas along the Florida Panhandle (Gulf and Franklin Counties) and upper Texas coast (Jefferson and Brazoria Counties) also contain well-above average (35%) proportions of floodplain area. Generally, communities situated away from the coastline at higher elevations contain a lower percentage of floodplain area, making it easier to develop without the threat of flooding during severe storms. For example, inland counties in Texas and parts of Florida contain the least amount of floodplain area, making them ideal locations for resilient development projects. As indicated in Table 1, the average level of soil permeability for the study area is 5 inches per hour, with a range from 0 to 13 inches per hour. Overall, this indicates poorly drained soils
indicative of coastal plain geography associated with the Gulf of Mexico Coast. Fig. 3 illustrates a well-defined spatial trend for soil permeability across the 144-county study area. Generally, counties in Florida with high sand content are the most permeable while the clay dominated counties in the upper Texas coast are the least permeable. In fact, Galveston and Brazoria counties in Texas have the least permeable soils in the study area while Gulf and Hendry counties in western Florida have the highest levels of porosity. Thus, from an ecological perspective, localities in Florida are preferable for human settlement because surface runoff will, on average, drain much more quickly. The mean calculated proportion of wetlands per jurisdiction in 2001 was approximately 3%, with a wide range from 0.04 to 90%. Monroe and Collier counties contain the most wetland area due to their association with the Everglades Ecosystem in south Florida. Several parishes in Louisiana, including Iberville, St. Martin, and Terrebonne contain over 60% wetlands within their boundaries. In contrast, large percentages of wetlands are scarce among Texas inland counties. As shown in Fig. 4, the loss of wetlands from 2001 to 2005 occurred in rapidly expanding urban and suburban coastal areas. Specifically, wetland areas surrounding New Orleans, LA, Sarasota, FL, and the Houston-Galveston region in TX were converted for development during the study time period. While there exist large population centers, very little of the total study area is developed. The average percent of pervious surface across the entire study area is over 94% (Fig. 5). As expected, areas containing large urban and sprawling suburban development patterns have the least amount of pervious land. Pinellas County, FL
Fig. 3. Soil permeability for Gulf Coastal Communities.
498
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
Fig. 4. Wetland losses and gains in Gulf Coastal Communities.
(40%), Harris County, TX (50%), and Orleans Parish, LA (66%) are the top three jurisdictions in terms of the lowest percentage of area left as pervious. On the other end of the spectrum, low-populated jurisdictions, such as Kennedy County in TX, where the 2000 U.S. Census estimated a population of only 414 people, has over 99% of its land area calculated as pervious. Results from robust regression analysis (Table 2) indicate that several ecological indicators act to reduce the adverse impacts of flooding, even when controlling for socioeconomic variables. For example, a percent increase in the area of land outside of the 100year floodplain significantly (p < 0.01) reduces insured flood losses from 2001 to 2005. In fact, based on standardized betas, this variable has the strongest effect on observed property damages from floods among all indicators in the model. Increasing levels of soil porosity also has a strong negative effect on flood losses (p < 0.01). In particular, development on well-drained, sandy soils appears to result in the least amount of damage. In contrast, a percent change in wetland area within a jurisdiction significantly increases flood damages (p < 0.01). Wetland losses over time reduce the ability of the hydrological system to naturally attenuate floods, and when replaced with human development, places structures in floodprone areas. Finally, while the effect is negative, large amounts of pervious surfaces alone do not significantly impact the degree of property damage caused by floods along the Gulf coast. Several control variables in the model are also significant predictors of flood loss. As expected, both storm surge and precipitation significantly increase insured flood damage (p < 0.01). Storm surge has the greatest impact among communities directly bordering the
Gulf. Increasing number of housing units and per capita income levels also lead to significantly higher (p < 0.01) amounts of property damage from flooding. More and expensive structures in vulnerable areas amplify the financial impact of each flooding event. Finally, floodplain planning and management mitigation as measured through the CRS does not appear to have a significant effect on flood losses. 5. Discussion Our results show that several ecological indicators can be used to identify areas where flood-related economic impact along the Gulf coast can be high. These findings provide statistical evidence that specific characteristics of the landscape provide a moderating function with respect to property damage resulting from floods. Identifying the importance of key ecological indicators can signal to decision makers where and how to develop communities that are more resilient to flooding over the long term. The indicator with the strongest effect in terms of reducing flood losses is the percentage of area outside of the 100-year floodplain. Specifically, a percent increase in these less vulnerable areas among Gulf coast counties/parishes is equivalent to, on average, $309,540,000 in savings (based on the unstandardized coefficient) from flood damage. This effect translates into an average $89.15 per acre reduction in flood losses. There are several policy implications stemming from this result. First, jurisdictions along the Gulf coast containing larger percentages of land outside the floodplain have more opportunity to become flood resilient and greater flexibility
Fig. 5. Percent pervious surface within Gulf Coastal Communities.
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
499
Table 2 Multiple regression model explaining insured flood losses.
Non-floodplain area Soil permeability Wetland alteration Pervious surface Precipitation Storm surge Number of housing units Per capita income CRS points
Coefficient
Standard error
T-value
Significance
−2.330175 −0.1282054 82.3422 −4.43626 1.663426 0.6577005 0.0000142 0.0001894 −0.0037355
0.7720293 0.0430806 31.17052 4.573353 0.4033117 0.1223042 4.75e−06 0.0000548 0.0478007
−3.02 −2.98 2.64 −0.97 4.12 5.38 2.99 3.46 −0.08
0.003 0.003 0.009 0.334 0.000 0.000 0.003 0.001 0.938
N = 144; F(9, 133) = 28.17; Prob > F = 0.000; Adj. R2 = .3974; VIF = 1.73; AIC = 777.18. Note: dependent variable is the logged insured flood losses from 2001 to 2005.
as to where development takes place. Development in these jurisdictions will also be, on average, less expensive because of lowered requirements for extensive drainage systems and associated infrastructure. Second, jurisdictions containing floodplain area (all but two in the sample) should pursue a development approach that avoids these vulnerable areas. Such an avoidance strategy involves spatially targeting future development in areas well outside of floodplain boundaries. Land use tools, such as cluster development, density bonuses, and capital improvements programming have been shown to be effective in guiding development towards more resilient locations (Beatley, 2009). Naturally occurring wetlands along the Gulf coast appears to be another key ecological indicator of resiliency when dealing with floods. An acre loss of naturally occurring wetlands from 2001 to 2005 increased property damage caused by flooding by an average of $7,457,549, which amounts to approximately $1.5 million per year across the study area. The role of wetlands in attenuating the adverse impacts of floods at the landscape scale should not be overlooked by planners and policy makers as they clearly provide an important protective and economic benefit to coastal communities. These natural features provide not only wildlife habitat and increased biological diversity in coastal areas, but also a critical buffer to storms which, as our model demonstrates, translates into significant dollar savings. Land acquisition techniques, coastal buffers, and critical area setbacks are a few of the tools available to planners that, when implemented, can help protect wetland functions. One of the more overlooked landscape characteristics in terms of reducing flood impacts is soil porosity. Soil structure may be considered at the site level, but rarely does this factor enter the minds of planners and policy makers as they consider future development patterns. However, our results show that well-drained soils are critical to reducing storm-water run-off and associated flood damages. Based on our model, a percent increase in soil permeability translates into a $3,403,600 reduction in property damage from floods per year. Coastal planners, policy makers, and developers should consider in their decisions soils alongside floodplains and any other indicator of vulnerability that could impact property and lives. Conservation overlay zones, subdivision ordinances, site plan requirements, and public expenditure guidelines are among the more effective techniques that can help reduce development pressure on impermeable soils. Overall, using ecological indicators to devise a program that combines structural and non-structural mitigation techniques can help reduce the adverse impacts of floods along the Gulf coast and increase the resiliency of local communities over the long term. While our flood mitigation measure was not significant in the statistical model, the techniques listed above can be effective in reducing flood losses. Previous studies have shown that communities that participate in FEMA’s CRS experience significantly less property damage caused by flood events (Brody et al., 2007a,b, 2008).
6. Conclusion Our results demonstrate that several ecological characteristics act to mitigate the economic impacts of flooding along the Gulf of Mexico coast. These indicators significantly influence the degree to which storms impact property within coastal communities. Accordingly, they should be strongly considered by planners and policy makers as they decide where and how much development should take place in the future. While our study provides insights on the moderating effect of landscape characteristics and the interaction between social and ecological systems, it should be considered only a starting point for more thorough research on the topic. First, our study examines only four indicators of resiliency. Future research should consider a wider range of factors, including different wetland types, forestlands, topography, etc. Second, the county/parish unit of analysis provides only a broad assessment of indicator effectiveness. Addition work should be done at smaller scales, preferably at the neighborhood level where there may be reduced spatial error and stronger statistical validity when making conclusions. Third, we analyzed data over only a fiveyear period, which offers a small window for assessment. Future research should look over larger time horizons with multiple storm events to increase the ability to make generalizations about the results. Fourth, the analysis is based on a cross-section research design that does not take into consideration changes within the study period. Future research should consider panel or time series models that account for temporal changes from year to year. Finally, we examine only one region in the U.S. A comparative study evaluating multiple coastlines would shed additional light on the notion of ecological resiliency and the relative effectiveness of landscape features in moderating the impacts of storm events. Acknowledgements This paper is based on research supported by NOAA Award No. NA07NOS4730147. The findings and opinions reported are those of the authors and are not necessarily endorsed by the funding organizations or those who provided assistance with various aspects of the study. References Arnold, C.L., Gibbons, J.C., 1996. Impervious surface coverage: the emergence of a key environmental indicator. Journal of the American Planning Association 62 (2), 243–258. Beatley, T., 2009. Planning for Coastal Resilience: Best Practices for Calamitous Times. Island Press, Washington, DC. Berkes, Colding, Folke (Eds.), 2003. Navigating Social Ecological Systems: Building Resilience for Complexity and Change. Cambridge University Press, Cambridge. Brezonik, P.L., Stadelman, T.H., 2002. Analysis and predictive models of stormwater runoff volumes loads and pollutant concentrations from watersheds in the twin cities metropolitan area, Minnesota, USA. Water Resources 36, 1743–1757. Brody, S.D., Highfield, W.E., Kang, J.E., 2011. Rising Waters: The Causes and Consequences of Flooding in the United States. Cambridge University Press, Cambridge, UK.
500
S.D. Brody et al. / Ecological Indicators 18 (2012) 493–500
Brody, S.D., Zahran, S., Maghelal, P., Grover, H., Highfield, W., 2007a. The rising costs of floods: examining the impact of planning and development decisions on property damage in Florida. Journal of the American Planning Association 73 (3), 330–345. Brody, S.D., Highfield, W., Ryu, H.C., Spanel-Weber, L., 2007b. Examining the relationship between wetland alteration and watershed flooding in Texas and Florida. Natural Hazards 40 (2), 413–428. Brody, S.D., Zahran, S., Highfield Grover, H., Vedlitz, A., 2008. Identifying the impact of the built environment on flood damage in Texas. Disasters 32 (1), 1–18. Bullock, A., Acreman, M., 2003. The role of wetlands in the hydrological cycle. Hydrology and Earth System Sciences 7 (3), 358–389. Chang, H., Franczyk, J., 2008. Climate change land-use change, and floods. Toward an integrated assessment. Geography Compass 2 (5), 1549–1579. Costanza, R., Pérez-Maqueo, O., Luisa, M., Sutton, P., Anderson, S.J., Mulder, K., 2008. The value of coastal wetlands for hurricane protection. Ambio 37 (4), 241–247. Crossett, K., Culliton, T.J., Wiley, P., Goodspeed, T.R., 2004. Population trends along the coastal United States, 1980–2008. National Oceanic and Atmospheric, Administration Coastal Trends Report Series. Cutter, S.L., 1996. Vulnerability to natural hazards. Progress in Human Geography 20 (4), 529–539. Folke, C., Hahn, T., Olsson, P., Norberg, J., 2005. Adaptive governance of social–ecological systems. Annual Review of Environmental Resources 30, 441–473. Godschalk, D., 2003. Urban hazard mitigation: creating resilient cities. Natural Hazards Review 4 (3), 136–142. Hey, D.L., 2002. Modern drainage design: the pros, the cons, and the future. Hydrological Science and Technology 18 (14), 89–99. Hirsch, R.M., Walker, J.F., Day, J.C., Kallio, R., 1990. The influence of main on hydrological systems. In: Wolman, M.G., Riggs, H.C. (Eds.), Surface Water Hydrology. Geological Society of America, Boulder, CO, pp. 329–359.
Hsu, M.H., Chen, S.H., Chang, T.J., 2000. Inundation simulation for urban drainage basin with storm sewer system. Journal of Hydrology 234, 21–37. Lebel, L., Anderies, J.M., Campbell, B., Folke, C., Hatfield-Dodds, S., Hughes, T.P., Wilson, J., 2006. Governance and the capacity to manage resilience in regional social–ecological systems. Ecology and Society 19, 1–15. Lewis, W.M., 2001. Wetlands Explained: Wetland Science, Policy, and Politics in America. Oxford University Press, New York, NY. Longhurst, R., 1995. The assessment of community vulnerability in hazard prone areas. Disasters 19, 269–270. Mileti, D.S., 1999. Disaster by Design: A Reassessment of Natural Hazards in the United States. Joseph Henry Press, Washington, DC. Mitch, W.J., Gosselink, J.G., 2000. Wetlands, 3rd ed. John Wiley & Sons, New York. National Research Council (NRC), 2000. Risk Analysis and Uncertainty in Flood Damage Reduction Studies. National Academy Press, Washington, DC. Paton, D., Johnston, D. (Eds.), 2006. Disaster Resilience: An Integrated Approach. Charles C. Thomas, Springfield, IL. Paul, M.J., Meyer, J.L., 2001. Streams in the urban landscape. Annual Review of Ecological Systems 32, 333–365. Rasmussen, P.P., Perry, C.A., 2000. Estimation of Peak Streamflows for Unregulated Rural Streams in Kansas. U.S. Geological Survey, Water-Resources Investigations, Report 00–4079. Saxton, K.E., Shiau, S.Y., 1990. Surface waters of North America; influence of land and vegetation on streamflow. In: Wolman, M.G., Riggs, H.C. (Eds.), The Geology of North America, vol. 0–1, Surface Water Hydrology. The Geologic Society of America, Washington, DC, pp. 55–80. Smith and Petley Environmental Hazards: Assessing Risk and Reducing Disaster. Routledge, New York, 2009. Tollan, A., 2002. Land-use change and floods: what do we need most, research or management? Water Science and Technology 45 (8), 183–190. Tourbier, J.T., Westmacott, R., 1981. Water Resources Protection Technology: A Handbook of Measures to Protect Water Resources in Land Development. The Urban Land Institute, Washington, DC.