ArticlePDF Available

Disparities of population exposed to flood hazards in the United States


Abstract and Figures

This study integrates publicly available datasets to provide a county-based assessment of socio-economic disparities of population exposure to flood hazards in the United States. Statistical methods were applied to reveal the national trends and local deviations from the trends. Results show that approximately 21.8 million (6.9% of) U.S. population are living in flood zones in 2015. The flood exposure varies greatly across the space. Communities near water bodies are more responsive to potential flood hazards by avoiding residence in flood zones than inland communities. At the national scale, economically disadvantaged population are more likely to reside in flood zones than outside. At the national scale, economically disadvantaged population tend to reside in flood zones in inland areas, while coastal flood zones are more occupied by wealthier and old people. These findings point to an alarming situation of inland communities where people are generally less responsive to flood hazards and people in flood zones have a lower economic condition. Using “hot spot” analysis, local clusters of disadvantaged population groups with high flood exposure were identified. Overall, this study provides important baseline information for policy-making of different levels of administration and pinpoints local deviations where diversified and targeted strategies are needed to mitigate flood risk in communities with different socio-economic characteristics. The valuable insights offered in this study advance understanding about the dynamic interactions between flood exposure and human factors.
Content may be subject to copyright.
Page 1
Disparities of Population Exposed to Flood Hazards 1 in the United States 2 3
Keywords: flood hazard; population exposure; disadvantaged population; vulnerability; socio-4
economic disparities; environmental justice 5
1 Introduction 6
Floods are the most common and costliest natural hazards in the United States in terms of lives 7
and property losses (FEMA 2004). In addition to the changing climate and rising sea level, the risk 8
of flood for human societies is intensified by population growth and demographic transformation 9
in coastal and inland floodplains (McGranahan, et al. 2007; Neumann et al. 2015). Flood risk can 10
be generally considered as a function of the flood hazard, flood exposure and vulnerability (IPCC 11
2012; Koks et al. 2015). The impact of a flood hazard is greatly dependent on the level of 12
vulnerability and exposure of human communities to the hazard. Vulnerability and exposure are 13
varying across space and time, and dependent on economic, social, geographic, demographic, 14
cultural, institutional, governance, and environmental conditions (Cutter et al. 2010; Koks et al. 15
2015; De Moel et al. 2011). Flood exposure can be mitigated by human interventions such as land 16
use control, population relocation and building levees along rivers and coasts (Wheater and Evans 17
2009; Pottier et al. 2005). Adaptation and mitigation practices will be more successful if the 18
dynamic nature of vulnerability and exposure is taken into account. In contrast, high vulnerability 19
and exposure are usually the product of socio-economic disparities and unsustainable development 20
such as environmental mismanagement, inappropriate urban planning, and failed governance. 21
Page 2
A spatial assessment of flood risk can be conducted by superimposing the spatial assessments of 22
its components (i.e. hazard, exposure and vulnerability). The locality of flood hazards is usually 23
estimated by hydrologic and hydraulic models that take into account topography, frequency of 24
extreme rainfall and run-offs, and human structures (such as levees) (Wing et al. 2017, Sampson 25
et al. 2015). For instance, the flood maps of U.S. Federal Emergency Management Agency (FEMA) 26
have delineated flood zones with the 100-year return period in most of the inhabited territory of 27
the U.S. The assessments of vulnerability are usually based on an index approach that aggregates 28
a variety of socio-economic and environmental variables into an index describing vulnerability at 29
different geographic scales (Cutter et al. 2003; Yusuf and Francisco 2009; Nelson et al. 2010). 30
Similar approaches have been applied to assess a closely-related concept, resilience, which is often 31
considered the opposite of vulnerability (Adger et al. 2005; Cutter et al. 2010; Lam et al. 2016). 32
As the focus of this study, flood exposure is usually assessed by intersecting the distributions of 33
flood hazard and population (e.g. Thieken et al. 2016; B. Jongman et al. 2014, Wing et al. 2018). 34
Thus, an extensive spatial assessment of flood exposure require large-scale population and flood 35
hazard data derived by standardized approaches. At the national scale, Qiang et al. (2017) have 36
conducted a county-level assessment of population exposure to flood hazards for the contiguous 37
United States by intersecting the FEMA flood maps and population data. Using a similar approach, 38
Wing et al. (2018) has applied a different flood model to estimate the population and GDP 39
exposure to flood hazard in the contiguous U.S. At the global scale, Jongman et al. (2012) provided 40
country-based assessment of urban and population exposure to flood hazards by combining 41
multiple flood databases. 42
Theoritically, urban and population development in flood-prone areas should be avoided or at least 43
minimized in order to reduce flood exposure. However, urban and population growth continues at 44
Page 3
flood-prone areas (De Moel et al. 2011, Jongman et al. 2012, Collenteur et al. 2014), where 45
political, cultural and economic factors often cause disproportionate exposure of some ratio/ethnic 46
minorities and disadvantaged population groups to flood hazard. For instance, poor people may be 47
disproportionately exposed to flood hazards due to the amenities (e.g. employment, education, and 48
transportation) and low property prices in flood-prone areas (Winsemius et al. 2018, Bin and 49
Landry 2013, Beltrán et al. 2018). Meanwhile, population in flood zones have a higher odd of 50
being affected by flood hazards to fall into poverty or be trapped in poverty (Masozera et al. 2007). 51
Such disproportionate exposure to negative environmental impacts has been widely discussed in 52
literature of environmental justice (e.g. Cutter 2012, Chakraborty et al. 2011, Morello-Frosch et 53
al. 2001). Empirical evidence of environmental injustice associated with different hazards have 54
been discovered in local areas. For instance, Ueland and Warf (2006) examined the altitudinal 55
residential segregation in 146 cities in the southern U.S. and found that blacks are 56
disproportionately concentrated in lower-altitude (flood-prone) areas in the inland cities and an 57
inverse trend near the coast, where whites dominate higher-valued coastal properties. By 58
intersecting demographic data with FEMA flood maps, Montgomery and Chakraborty (2015) 59
revealed that some ethnic minority groups are inequitably exposed to flood risks in Miami, Florida. 60
Additionally, Maantay and Maroko (2009) applied a dasymetric method, which is a population 61
mapping technique (Mennis 2015), to conducted an environmental justice assessment of people 62
exposed to flood risk in New York City. 63
Beyond the previous studies on local areas, this study provides a nationwide county-based 64
assessment of population exposure to flood hazards and socio-economic disparities of the exposed 65
population for the United States. By intersecting the spatial distributions of population and flood 66
hazards, the exposure of population to flood hazards was estimated. In this study, the spatial 67
Page 4
distribution of flood hazards was represented by the 100-year-flood (also known as flood of more 68
1 percent annual chance) zones defined in the Federal Emergency Management Agency (FEMA) 69
flood maps. The population distribution was downscaled from demographic data at a block group 70
level onto 30m-resolution land cover data. Finally, flood exposure was quantified as the count and 71
ratio of population located in 100-year-flood zones for each county. In addition to total population, 72
a number of disadvantaged population groups that are considered vulnerable to natural hazards in 73
literature were studied respectively. The major contributions of this study can be summarized as 74
follows. First, this study provides a county-level assessment of population exposure to flood 75
hazards for the entire United States using updated data and a refined population downscaling 76
approach. Second, this study is the first quantitative assessment of the disparities of population 77
exposed to flood hazards in the United States. The assessment results uncover the general trends 78
of flood exposure of the total population and the disadvantaged population groups. The spatial 79
analysis reveal local deviations from the general trends. This study provides empirical evidence of 80
socio-economic disparities and environmental injustice associated with flood exposure in the U.S. 81
and offers valuable insights to the underlying factors. 82
2 Data Acquisition and Processing 83
2.1 Flood Zone Determination 84
The spatial distribution of flood hazards was represented by the 100-year-flood zone in the FEMA 85
flood maps, which is a national standard used by FEMA and all federal agencies for the purposes 86
of requiring and rating flood insurance and regulating new development in floodplains. The FEMA 87
flood maps are stored as polygons in the ESRI shapefile format, which can be freely downloaded 88
from FEMA Flood Map Service Center ( The FEMA flood maps 89
Page 5
were then converted into a 30m-resolution raster to be overlaid with the population data. At the 90
moment of the study, the FEMA flood maps have not covered the entire territory of the United 91
States, but it is continuously updating with newly published maps and appealed revisions. The 92
database includes effective and preliminary flood maps. The former is officially published, while 93
the latter is not official and in the public appeals period during which relevant stakeholders can 94
appeal information contained in the preliminary maps (FEMA 2017b). Despite the unofficial status, 95
the preliminary maps present the best information available at the current time and provide the 96
public an early look at their home or community’s projected risk to flood hazards. To create more 97
extensive coverage for the United States, both the effective and preliminary flood maps were used 98
for analysis in this study. 99
The flood maps used in this study (acquired in September 2017) covers 57.3% of the territory of 100
the 50 United States, including 98.1% effective and 1.9% preliminary flood maps. The coverage 101
of flood maps varies from county to county. In general, most counties with a moderate population 102
density are covered by flood maps. Large blank areas of flood maps are distributed in Alaska and 103
the middle and western areas of the contiguous U.S. where the population density is low and the 104
demand for flood maps is less pressing. Some small blanks in coastal areas (such as Mississippi 105
Delta) can be a result of local conflicts in flood zone delineation (Linskey 2013). In this study, 106
counties with >5% of area covered by flood maps were included for analysis, leading to 2,351 107
qualified counties out of the 3,142 counties (74.8%) in the United States (see Figure 1). Most of 108
the counties with a partial flood map coverage are located in sparsely populated areas, where flood 109
maps are only available in the population clusters. Thus, the assessment made in the partial 110
coverage can still reflect the situation of the county. The 2,351 qualified counties contain 93.6% 111
Page 6
of the U.S. population. Thus, the analyses conducted with these counties generally reflect the 112
national trends. 113
Figure 1: The coverage of FEMA flood maps in counties of the United States. 115
The flood maps classify geographic areas into three general categories according to the annual 116
chance of flood inundation. First, high flood risk zones are defined as areas that have equal to or 117
more than 1 percent chance of being inundated by flood in any given year (FEMA 2017a). The 1 118
percent chance flood is also termed base flood or 100-year flood. FEMA defines the 100-year-119
flood zones as Special Flood Hazard Area (SFHA) in which floodplain management regulations 120
must be enforced and purchase of flood insurance is mandatory (FEMA 1986). Second, moderate121
low flood risk zones are defined as areas that have less than 1 percent annual flood chance. Third, 122
undetermined flood zones are areas where flood chance is possible but undetermined. In this study, 123
the locality of flood hazards was represented by the 100-year-flood zones, which was denoted as 124
flood zones for simplicity in the remaining of this article. The moderate-low flood risk zones were 125
referred to as non-flood zones. The undetermined flood zones were excluded from the analyses. 126
Page 7
2.2 Population Downscaling 127
Current nationwide population datasets, such as LandScan (Bright et al. 2013) and Gridded 128
Population of the World (CIESIN 2015), are presented at a ~1km resolution, which are too coarse 129
to be compared with flood zones at the household level to determine flood exposure. To derive the 130
population distribution at a finer resolution, the population data in census block groups were 131
downscaled to land cover data at a 30m or finer resolution. The block group data associated with 132
per capita income, social and demographic variables were acquired from the website of U.S. 133
Census Bureau (i.e. 2012-2016 American Community Survey 5-year Estimates). The 2011 land 134
cover data at 30m resolution of the Contiguous U. S. and Alaska were acquired from the National 135
Land Cover Database ( The land cover data of Hawaii were acquired from 136
NOAA C-CAP database (, which were created 137
between 2010 – 2011 at a 2.4m resolution. Both the NLCD and C-CAP are based on the Anderson 138
Land Cover Classification System (Anderson et al. 1976), in which the class of developed land 139
can represent man-made structures in both urban and rural areas. 140
The downscaling of population data is based on three assumptions: (1) population (same as 141
households) are only distributed in pixels classified as developed land in the land cover data; (2) 142
population density within a census block group is even; (3) socio-economic and demographic 143
characteristics within a census block group are even. Based on the first and second assumption, 144
population per developed pixel can be derived as the quotient of the total population and number 145
of developed pixels in a block group. Based on the third assumption, population in a demographic 146
group per developed pixel is the quotient of the total population in that group and number of 147
developed pixels in a block group. Per capita income of all pixels in a block group is the same. 148
Finally, total population, population of a particular group, per capita income can be estimated for 149
Page 8
each developed pixel. To offset the local biases of the assumptions, these quantities in pixels were 150
aggregated into counties after their flood exposure (in or out of flood zones) was determined. 151
In this study, flood exposure was calculated for the total population and a number of disadvantaged 152
population groups. These disadvantaged groups are that are commonly used as indicators in social 153
vulnerability and resilience assessments (e.g. Cutter et al. 2003; Burton 2010, Lam 2016) and are 154
available in U.S. Census block group data. The disadvantaged groups including population above 155
75 (ELDERLY), population under 5 (CHILD), population above 25 with no schooling completed 156
(NO_SCHOOL), population above 16 unemployed (UNEMPLOYED), female householder with 157
no husband present (SINGLE_FEMALE), female householder with no husband present and with 158
children under 6 (SINGLE_MOM), household with limited English ability (LIMITED_EN), 159
household with an income below poverty level (POVERTY), population without health insurance 160
3 Analysis 162
To analyze the total population and socio-economic disparities of population exposed to flood 163
hazards, four analyses were carried out in this study. 164
First, exposure of total population to flood hazards was estimated by intersecting the population 165
distribution and flood zones for each county. Total population in flood zones (denoted as P) was 166
the sum of population of all developed pixels located in flood zones. Then, the ratio of population 167
in flood zones (R) was the quotient of population in flood zones and total population covered by 168
flood maps in the county. The first quantity represents the total population and associated socio-169
economic resources exposed to flood hazards. The second quantity standardizes the total quantities 170
by the density of population so that counties that are less populated but have a high ratio of flood 171
Page 9
exposure can receive the same attention as the populated counties. The exposed population (P) and 172
exposure ratios (R) of the 791 counties not covered by FEMA flood maps were estimated using 173
ordinal kriging interpolation. To perform kriging interpolation, the county polygons were first 174
converted to centroid points. Then, the ratios of population in flood zones in the counties 175
(represented as points) without flood maps were predicted from the counties within flood map 176
coverage. Kriging interpolation was applied separately for the contiguous U.S. and Alaska (Hawaii 177
has full flood map coverage). Finally, multiplying the exposure ratios by the populations of all the 178
U.S. counties, the total population exposed to 100-year-flood was estimated. 179
Second, the difference between the ratio of population in flood zone and ratio of land area in flood 180
zones (denoted as ) was compared for each county (Equation 1). The significance of the 181
difference of all counties are tested using Student’s t-test, with a null hypothesis that the two ratios 182
are equal (i.e., the difference is zero). The land area is the county’s total area excluding 183
undevelopable areas such as water bodies (from the land cover data), military sites (U.S. Census 184
data), wildlife refuge (U.S. Fish and Wildlife Service), federal land (USGS), and national parks 185
(National Park Service). Assuming a community is not concerned with the distribution of potential 186
flood hazards, the ratio of population in flood zones is expected to be equal to the ratio of land in 187
flood zones (i.e. the difference is zero). A deviations of the difference from zero reflects the degree 188
to which people are aware of, attach importance to (as a trade-off decision between flood risk and 189
other amenities), and mitigate and adapt to flood hazards. For ease of discussion, we use the term 190
responsiveness to flood hazards in this article to represent the implications of the deviations. A 191
negative deviation indicates can be interpreted as less population located in flood zones than 192
expected, further suggesting that the community is more responsive to flood hazards. Conversely, 193
a positive deviation would imply the community is less responsive to flood hazards and do not 194
Page 10
avoid or even favor flood zones for residence. Hotspot analysis (Getis-Ord Gi statistic) was used 195
to detect the local clusters of deviations. 196
=   
     
 
Equation 1
Third, the difference between per capita incomes in and out of flood zones () was compared for 197
each county (Equation 2). The significance of the difference was also tested by Student’s t-test. 198
Total income in flood zones is the summation of income of all developed pixels in flood zones. 199
Then, per capita income in (or out of) flood zones is the quotient of the total income and population 200
in (or out of) flood zones. Assuming the per capita incomes in and out of flood zones are equal, 201
the expected value of the difference between the two ratios should be zero. A positive deviation 202
(i.e. > 0) would indicate a higher per capita income of people in flood zones than those outside, 203
while a negative deviation means the opposite. In addition to the t-test for all the counties, the 204
Getis-Ord Gi* statistic (Getis and Ord 1992) is applied to detect local clusters that are significantly 205
deviated from the mean difference. 206
= .     .    
Equation 2
Fourth, the difference between the ratios of disadvantaged population in and out of flood zones 207
() was compared (Equation 3). Again, a positive deviation of the difference from the zero 208
implies a higher ratio of disadvantaged population located in flood zones than outside, and a 209
negative deviation indicates the opposite. Due to overlapped population representations, the ratios 210
of the nine disadvantaged groups may be correlated among each other. For instance, people in a 211
poor economic condition may belong to POVERTY, UNEMPLOYED and NOT_INSURED at the 212
same time. To reduce the redundancy and distinguish the non-overlapped population groups, 213
principal component analysis (PCA) was used to aggregate the nine disadvantaged groups into a 214
Page 11
fewer number of groups. The spatial patterns of the deviations of the aggregated groups were 215
analyzed respectively. Analogous to the third analysis, the Getis-Ord Gi* statistic was applied to 216
detect local clusters of . 217
 =.   
     .    
     
Equation 3
4 Results 218
Results from the four analyses are organized as follows. Section 4.1 presents the results of the first 219
analysis, which estimates the total population and ratio of population in flood zones per county. 220
Section 4.2 includes the result of the second analysis, analyzing responsiveness of population to 221
flood hazards. Section 4.3 includes the results of the third and fourth analysis, which compare per 222
capita incomes and ratios of disadvantaged population in and out of flood zones. All analyses are 223
conducted at both the national and county levels, reflecting the general trend and local deviations 224
from the trends. 225
4.1 Exposure of Population to Flood Hazard 226
As expected, population in flood zones are concentrated in metropolitan areas along the coast, 227
including New York City, Miami, Naples, Tampa, Houston, New Orleans, Los Angeles, and San 228
Francisco (Figure 2). These areas are highly populated and have large low-laying areas subject to 229
coastal flooding. As shown in Table 1 (left), seven of the top ten counties ranked by total 230
population in flood zone are in southern Florida. The remaining three are near Houston (TX), New 231
Orleans (LA) and Los Angeles (CA). Meanwhile, several inland areas with high flood exposure 232
are noticeable in Figure 2, such as counties around Phoenix (AZ), Salt Lake City (UT), and Dallas 233
(TX), which are inland cities with a large population exposed to riverine flood. 234
Page 12
Figure 2: Total population in flood zone per county. 236
The ratio of population located in flood zones presents a different spatial pattern (Figure 3). In 237
addition to the coastal counties, many inland counties with high ratios of population in flood zones 238
stand out, including counties along the Lower Mississippi River, the western hillside of 239
Appalachian Mountains, and some counties scattered in the western mountainous region. In Table 240
1 (right), it is noticeable that none of the top ten counties of percentage of population in flood zone 241
are in large coastal cities. Instead, three inland counties, including Nobel (Oklahoma), Lincoln 242
(Louisiana), and Issaquena (Mississippi), pop up in the list. The remaining seven are less populated 243
coastal communities, including three counties around Pamlico Sound in North Carolina, Monroe 244
County (the Key West) and Collier County (Nápoles) in Florida, Cameron County (Lake Charles) 245
in Louisiana, and Poquoson County in Virginia. 246
In the 2351 counties covered by flood map coverage, the total ratio of population in flood zone is 247
6.84%. To obtain a national estimation, the exposure ratios of the counties not covered by flood 248
Page 13
maps were estimated using kriging interpolation. The result shows that in total 21.8 million people 249
(6.87% of total population) in the U.S. are exposed to 100-year-flood zones. 250
Figure 3: Ratio of population in flood zone per county. 252
Table 1: Top 10 counties ranked by total population in flood zone (left) and percentage of population in flood zone (right). 253
% of population
in FZ
Palm Beach
Page 14
4.2 Responsiveness of Population to Flood Hazard 255
The result of t-test shows that the ratio of population in flood zones is significantly (p < 0.001) 256
lower than the ratio of land in flood zones, meaning that people in the U.S. are generally responsive 257
to flood hazards by avoiding development in flood zones. However, the difference (Dp) between 258
the two ratios varies over the space, which presents two opposite trends (Figure 4). Counties near 259
water bodies, including those along the Gulf Coast, East Coast, and the middle-lower Mississippi 260
River, have lower Dp values. These areas are historically flood-prone, but communities there are 261
more responsive to flood hazards by avoiding residence in flood zones. The area around Miami 262
(FL) is a noticeable exception in the East Coast, where people appear not responsive to flood 263
hazards. In contrast, counties in the western mountainous region and the eastern inland region have 264
higher Dp values. In these areas, flood hazard could be considered less important compared with 265
other factors for choosing locations for population placement. 266
Figure 4: Difference between the ratio of population in flood zone and ratio of land area in flood zone (). SD denotes standard 268
deviation(s) from the mean. 269
Page 15
4.3 Disparities of Population Exposed to Flood Hazards 270
4.3.1 Income 271
The result from t-test shows no significant difference (p = 0.198) between the per capita incomes 272
in and out of flood zones over the country (Table 2). However, the spatial pattern of the difference 273
(Di) is uneven, showing local pockets with high positive or negative deviations from zero (Figure 274
5). Using the Getis-Ord Gi* analysis, clusters with a positive deviation are detected as “hot spot”, 275
in which counties with a high positive deviation are surrounded by counties with a high positive 276
deviation. Conversely, clusters of negative deviations are denoted as “cold spot”. In this study, 277
counties that share a common boundary or vertex are defined as neighbors. Due to the isolation of 278
Hawaiian and Alaska counties (no adjacent counties), these two states are excluded from the Getis-279
Ord Gi* analysis. 280
Figure 5: Difference between per capita incomes in and out of flood zones (Di). SD denotes standard deviation(s) from the mean. 282
As shown in Figure 6, most “hot spots” of per capita income are located along the East Coast and 283
Gulf Coast, including counties around New York City, Delmarva Peninsula (Virginia and 284
Page 16
Maryland), Charleston (South Carolina) and Wilmington (Georgia), and Mobile and Escambia 285
County (Alabama). In these “hot spots”, per capita income of people in flood zones is higher than 286
those outside. To the contrary, most “cold spots” of per capita income are located in inland areas 287
besides the coastal counties in California. In the “cold spots”, per capita income in flood zones is 288
lower than outside. 289
Figure 6: Significant clusters of the difference between per capita incomes in and out of flood zones (Di). 291
Table 2: T-test results of the differences between per capita income and ratios of disadvantaged populations in and out of flood 292
zones. Significant differences (p<0.05) are in bold font and underlined. 293
Population group
Average per capita income
Ratio of population above 75
Ratio of population under 5
Ratio of household with an income below poverty level
Page 17
Ratio of population above 16 unemployed
Ratio of female householder with no husband presented
Ratio of female householder with no husband and with children under 6
Ratio of population above 25 with no schooling completed
Ratio of household with limited English ability
Ratio of population with no health insurance
4.3.2 Ratios of Disadvantaged Population 295
The results from t-test analysis show that the null-hypothesis should be rejected for ELDERLY, 296
POVERTY, UNEMPLOYED, SINGLE_MOM, and NOT_INSURED (see Table 2). The ratios of 297
ELDERLY in flood zones are significantly (p<0.001) higher that the ratios out of flood zones, 298
indicating that elderly people are generally less likely to reside in flood zones in the U.S. The ratios 299
of POVERTY, UNEMPLOYED, SINGLE_MOM, and NOT_INSURED in flood zones are higher 300
than those out of flood zones, reflecting these disadvantaged population groups are more likely to 301
reside in flood zones in the U.S. 302
Using principal component analysis (PCA), the ratios of the nine disadvantaged groups were 303
aggregated into three principal components. The first component (PC1) occupies 61.6% of the 304
total variance, in which POVERTY, UNEMPLOYED, SINGLE_FEMALE, and NOT_INSURED 305
have the highest loading (Table 3). These variables all represent certain aspects of economic 306
condition. Thus, we use the first component (PC1) to represent the general group of the 307
economically disadvantaged people. LIMITED_EN and ELDERLY are dominant variables with 308
outstanding loadings in the second (PC2) and third component (PC3) respectively, indicating 309
LIMITED_EN and ELDERLY are two groups of people that do not overlap with the economically 310
disadvantaged (PC1). Due to the dominant loadings, LIMITED_EN and ELDERLY were analyzed 311
independently rather than being aggregated into components. Again, the Getis-Ord Gi* statistic 312
was used to detect local deviations from the mean difference between the ratios of the 313
Page 18
disadvantaged population in and out of flood zones (). “Hot spots” denote local clusters where 314
the ratio of the disadvantaged population in flood zones is higher than outside, while “cold spots” 315
are counties with a lower ratio of disadvantaged people in flood zones. 316
Table 3: Top three components and loadings of variables from the principal component analysis. 317
Principal components (PC)
Proportion of variance explained
As shown in Figure 7(a) “hot spots” of the economically disadvantaged are mostly located in 319
inland areas, except counties near Pamlico Sound in North Carolina and coastal area in Mississippi, 320
where the economically disadvantaged are more likely to reside in flood zones than outside. To 321
the contrary, most “cold spots” are detected in coastal and riverine areas, such as Gulf Coast, East 322
Coast, and counties along Mississippi River, where a low ratio of the economically disadvantaged 323
people are in flood zone. This pattern is generally in line with the result of the third analysis that 324
people in the coastal flood zones are in a better economically condition than people outside. 325
The two largest “hot spots” of LIMITED_EN are located in central California and the area between 326
Nevada and Utah (Figure 7(b)). At the meantime, other smaller “hot spots” are scattered in the 327
inland areas. In the “hot spots”, people with limited English ability are more likely to reside in 328
flood zone than outside. To the contrary, large “cold spots” can be found in southern California, 329
Page 19
the cross-boundary area between Arizona and New Mexico, Tampa in Florida and New York City. 330
The two largest “hot spots” of ELDERLY in Florida and the shores of Chesapeake Bay (Maryland 331
and Virginia) are most prominent (Figure 7(c)), where old people are more likely to live in coastal 332
flood zones possibly due to the aesthetical and restorative values of the coasts. The smaller “hot 333
spot” in Matagorda (Texas) may fall to the same category. Additionally, other “hot spots” can be 334
found in the inland areas such as western Mississippi, the areas near Reno (Nevada) and Santa Fe 335
(New Mexico). In these areas, the underlying factors that cause the old people to be crowded in 336
flood zones need further investigations. 337
Figure 7: Clusters of differences between ratios of disadvantaged population in and out of flood zones (). (a) economically 339
disadvantaged; (b) people with limited English ability (LIMITED_EN); (c) people above 75 years old (ELDERLY). 340
Page 20
5 Discussion 341
This study provides a county-level assessment of population exposure to flood hazards for the 342
entire United States. This assessment approach has improved from the previous study (Qiang et al. 343
2017) by using more updated population data (i.e. 2015 census data) at a finer spatial resolution 344
(i.e. the block group level), extending the assessment to the entire United States, and investigating 345
socio-economic disparities in flood zones. The assessment approach utilizes publicly available 346
databases and thus is transferable to other regions where hazard maps are available. Based on this 347
assessment approach, the study has analyzed four general types of quantities including (1) 348
population exposure (total and ratio) to flood zones, (2) responsiveness of population to flood 349
hazards, (3) difference of per capita incomes in and out of flood zones, and (4) differences of ratios 350
of the disadvantaged groups in and out of flood zones. The national trends and local deviations 351
discovered in this study provide important policy implications. 352
At the national scale, it was estimated that 21.8 million (6.87%) of the U.S. population are exposed 353
to 100-year-flood. According to the 1% annual inundation chance in the 100-year-flood zones, 354
0.218 million (6.87%) U.S. population will be affected by a certain level of flood hazards annually. 355
These estimates provide base-line information for flood preparation and mitigation for the federal 356
level decision-making. As expected, large metropolitan areas along the coasts have high 357
concentrations of population, economy and associated assets exposed to flood zones. However, 358
small communities (both inland and coastal) have the highest ratios of population in flood zones. 359
Compared with the large coastal cities where assistance resources and public attention are 360
concentrated, the small communities with a high ratio of flood exposure may be oversighted in the 361
efforts of hazard mitigation and disaster relief. 362
Page 21
Population exposure to flood hazards can be a result of lack of awareness of potential hazard 363
(awareness), being able to cope with and adapt to the adverse impacts (coping and adaptive 364
capacity), a trade-off decision between flood risk and amenities in flood zones (trade-off), and 365
governmental and instructional factors. Changes of flood exposure in space and time can be driven 366
by any of these factors. In this article, the term responsiveness has been used to generalize the 367
combined effects of these factors. The national trend indicates that people in the U.S. are generally 368
responsive to flood hazards by avoiding living in flood zones. This trend can be intervened by 369
policy and institutional levers such as the enforcement of floodplain development regulations at 370
the federal scale. Thus, by monitoring the trend over time, the effectiveness of federal level 371
interventions to the reduction of flood exposure can be monitored. At the local scale, deviations 372
from the general trends reflect varying conditions of individuals’ awareness, local governance, 373
dependence on water resource, and other socio-economic factors in different places. Possibly due 374
to the higher public awareness and more governmental interventions, communities near coasts and 375
rivers, which are historically flood-prone, are more responsive to flood hazards than the inland 376
communities (shown in Figure 4). The exception of Miami (a coastal city with low responsiveness) 377
could be caused by the attraction of amenities in the flood zones. Conversely, the low 378
responsiveness of inland communities to flood hazards may reflect the negative situation (e.g. lack 379
of awareness and adaptive governance). With the changing climate and precipitation pattern, the 380
low responsiveness of inland communities can potentially amplify the adverse impact of flood 381
hazards, which is the first alarm to the inland communities raised in this study. 382
The choice of living in flood zone or outside is also influenced by individuals socio-economic 383
conditions. At the national level, a higher ratio of economically disadvantaged people (including 384
POVERTY, UNEMPLOYED, SINGLE_FEMALE, and NOT_INSURED) choose to live in flood 385
Page 22
zones than outside. This trend is potentially related to the lower property prices in flood zones, 386
which were discussed in a number of studies (e.g. Speyrer and Ragas 1991, Okmyung and Stephen, 387
2004, Bin and Landry 2013). This tendency is more prominent in the inland areas than the coasts. 388
Most clusters of low per capita income (the third analysis) and high ratios of economically 389
disadvantaged people (the fourth analysis) are located in the inland areas. In contrast, the opposite 390
clusters are mostly coastal. For instance, southern Florida is the largest “hot spot” of per capita 391
income (higher income in flood zones) and “cold spot” of economically disadvantaged people 392
(lower ratio in flood zones). This inland-coastal contract confirmed the empirical findings in 393
previous studies focused on local areas (e.g. Ueland and Warf 2006, Montgomery and Chakraborty 394
2013), which revealed that minorities and disadvantaged groups are segregated in flood-prone 395
areas in inland cities, whereas the higher-valued coastal and waterfront properties are occupied by 396
middle and upper-classes. Since a lower economic condition can limit one’s abilities to mitigate, 397
cope with and recover from the negative impacts of hazards, the disproportionate exposure of 398
economically disadvantaged population in flood zones is the second alarm posed to the inland 399
communities in this study. 400
LIMITED_EN and ELDERLY represent different groups of people from the economically 401
disadvantaged population. The largest “hot spot” of LIMITED_EN is located in California, which 402
is one of the most ethnically diversified state in the U.S. The second largest “hot spot” is in the 403
Great Basin between Nevada and Utah, which is historically inhabited by indigenous American 404
tribes speaking Washo and Numic languages. Due to the arid to semi-arid environment, the 405
livelihood and culture of the indigenous people heavily rely on ecosystem services provided by 406
limited water resources, resulting large overlaps between their residence and potential flood 407
hazards. In these “hot spots”, limited English ability and cultural barrier of ethnical minorities and 408
Page 23
new immigrants may cause difficulties in accessing hazard information, leading to lower 409
awareness of flood risk and limited knowledge about climate change. Additionally, cultural 410
disadvantages can impose obstacles in communication and acquisition of assistance resources 411
during and after hazard (Cutter et al. 2010). Ideally, hazard education and information 412
dissemination in non-English languages should be improved in these areas to prompt the awareness 413
of flood hazard and reduce vulnerability. When flood hazards strike, special assistance with 414
language support should be offered to help the people with a limited English ability withstand and 415
recover from the adverse impacts of flood hazards. 416
“Hot spots” of ELDERLY were found in southern Florida, Chesapeake Bay, and Matagorda in 417
Texas, which are all popular retirement destinations in the U.S. The high density of ELDERLY in 418
these areas could be explained by the recreational and restorative effects of the oceanic blue spaces. 419
Although the generally higher economic condition would benefit elderly people in coping with 420
and adapting to flood hazards, mobility constraints and social isolation will increase their 421
difficulties in evacuation and seeking support during hazard events (Siagian et al. 2014; Walker 422
and Burningham 2011). Besides the coastal “hot spots”, further investigations are needed to 423
understand the causes of the inland “hot spots” of ELDERLY. Special measures should be taken 424
to mitigate the impact of potential flood hazards to the elderly communities. 425
The analyses of the study are limited in the following aspects. First, the population distribution 426
was downscaled from the block group data into 30m resolution land cover data, assuming that the 427
population density and socio-economic conditions are even within block groups. The spatial 428
variability of population within block groups have not been taken into account. The exposure ratios 429
(P) were validated against the ratios estimated using the 2010 block level data in the 2,351 counties 430
with flood maps. The overall exposure ratio of the block-level data is 6.75%, compared to 6.84% 431
Page 24
obtained in this study. The Mean Absolute Error (MAE) and Root-Mean-Square Error (RMSE) of 432
the validation per county are 0.016 and 0.010. Given the different year of the validation data and 433
potential errors in the downgrading process, the uncertainty of the assessment need to be further 434
evaluated with ground truth data. Second, despite the FEMA flood maps have covered the majority 435
(~93.6%) of U.S. population, the interpolated values in the unmapped areas can be a source of 436
uncertainty. Also, the estimated exposure is based on residential population. Further studies should 437
consider the dynamics of population such as people in travel, as evidence shows that the majority 438
of fatalities in flood events occur when people attempt to drive or walk in floodwaters (Kellar, 439
2010; Arrighi et al., 2017). Third, only the100-year-flood was used in the assessment. A 440
comprehensive assessment should include more frequent floods (such as 30 and 50-year-flood) 441
which may also cause impacts to human communities. Fourth, in spite of being a national standard, 442
FEMA flood maps are often criticized for the varying age and levels of quality. For instance, using 443
a newly-developed flood model, Wing et al. (2018) estimated that 40.8 million people (13.3% of 444
the population) in the contiguous U.S. are exposed to 100-year-flood, which nearly doubles the 445
estimations (21.7 million and 6.87%) derived in this study. This difference can possibly be 446
attributed to the incomplete coverage of the FEMA flood maps over the U.S. and different flood 447
zones projected by the two models. Wing et al. (2017 and 2018) claimed that their flood model 448
can identify flood zones in small catchments that are often missed by FEMA flood maps. In the 449
future work, the uncertainty of the assessment needs to be further evaluated against ground-truth 450
data. 451
6 Conclusion 452
This study provides a county-based assessment of population exposure to flood hazards and socio-453
economic disparities in the exposed population in the United States. Instead of developing an 454
Page 25
overall index, this study aimed to gain new insights to the interrelations between flood exposure 455
and human factors by analyzing socio-economic disparities of population exposed to flood hazards. 456
The general trends derived at the national scale provide important baseline information for the 457
federal level policy-making. The local deviations from the general trends pinpoint areas that are 458
potentially more vulnerable to flood hazards than the average. The analyses of the disadvantaged 459
population uncovered potential environmental injustice of flood exposure confronted by different 460
population groups. The identified ‘hot spots’ can inform decision-makers to develop diversified 461
and targeted strategies to mitigate flood risk in communities with skewed socio-economic 462
structures. Major findings derived from this study include: (1) Approximately 21.8 million (6.87%) 463
U.S. population are located in 100-year-flood zones. Although population exposed to flood hazards 464
are concentrated in large coastal cities, small communities (both inland and coastal) have the 465
highest ratios of population in flood zones. (2) Communities near water bodies (i.e. coasts and 466
rivers) were more responsive to flood hazards and tended to avoid residence in flood zones. 467
Conversely, inland communities are less responsive to flood hazards and do not avoid flood zones 468
for residence. (3) There are socio-economic disparities between population in and out of flood 469
zones. At the national level, the economically disadvantaged groups (including POVERTY, 470
UNEMPLOYED, SINGLE_MOM, and NOT_INSURED) generally tend to reside in flood zones 471
than outside. At local scales, coastal flood zones are more crowded by richer and old people, while 472
inland flood zones are more occupied by poorer people. The second and third finding both point 473
to an alarming situation of the inland communities where people are generally less responsive to 474
flood hazards and people in flood zones have a lower economic condition. 475
The analyses of socio-economic disparities of population exposed to flood hazards have advanced 476
our understanding of the dynamic interactions among exposure, vulnerability and resilience. The 477
Page 26
trends and deviations quantified in this study have important policy implications on flood risk 478
management and environmental justice for different levels of decision-makers. The assessment 479
method integrates publicly available datasets, and thus is reproducible and transferable to other 480
countries where hazard maps are available. The assessment can be reproduced with historical or 481
updated datasets to monitor the dynamics of flood exposure to evaluate the effectiveness of 482
mitigation policies. The assessment and analysis results are available in a web-based GIS 483
( for public users to freely access to increase 484
awareness of flood hazard and inform decision-making. 485
References 486
1. Adger, N., Hughes, T., Folke, C., Carpenter, S., and Rockström, J. 2005. “Social-Ecological 487
Resilience to Coastal Disasters.” Science 309 (5737):1036–39.
/science.1112122. 489
2. Arrighi, C, Oumeraci, H., Castelli, F. 2017. Hydrodynamics of pedestrians' instability in 490
floodwaters”. Hydrology and Earth System Sciences, vol. 21, p. 515-531, ISSN: 1027-5606, 491
doi: 10.5194/hess-21-515-2017 492
3. Anderson, J., Ernest, H., John, R., and Richard, W., 1976. A Land Use and Land Cover 493
Classification System for Use with Remote Sensor Data. Vol. 964. US Government Printing 494
Office. 495
4. Beltrán, Allan, David Maddison, and Robert J R Elliott. “Is Flood Risk Capitalised Into 496
Property Values?” Ecological Economics 146 (April 1, 2018): 668–85. 497 498
Page 27
5. Bin, Okmyung, and Craig E. Landry. “Changes in Implicit Flood Risk Premiums: Empirical 499
Evidence from the Housing Market.” Journal of Environmental Economics and Management 500
65, no. 3 (May 1, 2013): 361–76. 501
6. Bright, E. A., Rose, A. N., and Urban, M. L. 2013. “LandScan 2012.” Oak Ridge National 502
Laboratory. 503
7. Burton C. G. 2010. “Social Vulnerability and Hurricane Impact Modeling.” Natural Hazards 504
Review 11 (2): 58–68. 505
8. Center for International Earth Science Information Network, Columnbia University (CIESIN). 506
2015. “Gridded Population of the World (GPW), V4, UN-Adjusted Population Count for 2015. 507
Center for International Earth Science Information Network.” Columbia University. 508
population-count-adjusted-to-2015-unwpp-country-totals/data-download. 510
9. Chakraborty, J., Maantay, J. A. and Brender, J. D. 2011. “Disproportionate Proximity to 511
Environmental Health Hazards: Methods, Models, and Measurement.” American Journal of 512
Public Health 101 (S1): S27–36. 513
10. Collenteur, R. A., De Moel, H., Jongman, B., and Di Baldassarre, G. 2015. “The Failed-Levee 514
Effect: Do Societies Learn from Flood Disasters?” Natural Hazards 76, no. 1: 373–88. 515 516
11. Cutter, S. L. 2012. Hazards Vulnerability and Environmental Justice. Routledge. 517
12. Cutter, S. L., Boruff, B.J., and Shirley, W. L. 2003. “Social Vulnerability to Environmental 518
Hazards.” Social Science Quarterly 84 (2):242–261.
6237.8402002. 520
Page 28
13. Cutter, S. L., Burton, C. G., and Emrich, C. T., 2010. “Disaster Resilience Indicators for 521
Benchmarking Baseline Conditions.” Journal of Homeland Security and Emergency 522
Management 7 (1):14. 523
14. Federal Emergency Management Agency (FEMA). 2017a. “Flood Zones.” 2017. 524 525
15. Federal Emergency Management Agency (FEMA). 2017b. “View Your Community’s 526
Preliminary Flood Hazard Data.” 2017.
preliminary-flood-hazard-data-0. 528
16. Federal Emergency Management Agency. 2016. NFIP Flood Insurance Manual, Washington, 529
D. C. Available at
14365ad154954e4dc69/changepackage_508_apr16b.pdf. 531
17. Federal Emergency Management Agency (FEMA). 2004. “Flooding: America’s #1 Natural 532
Hazard!” 2004.
hazard. 534
18. Getis, A., and Ord, J. K. 1992. “The Analysis of Spatial Association by Use of Distance 535
Statistics.” Geographical Analysis 24 (3):189–206.
4632.1992.tb00261.x. 537
19. Lam, N.S., Reams, M., Li, K., Li, C., and Mata, L.P. 2016. “Measuring Community Resilience 538
to Coastal Hazards along the Northern Gulf of Mexico.” Natural Hazards Review 17 (1). 539 540
20. Mennis, J. 2015. “Increasing the Accuracy of Urban Population Analysis With Dasymetric 541
Mapping.” Cityscape 17 (1): 115–26. 542
Page 29
21. McGranahan, G., Balk, D., and Anderson, B. 2007. “The Rising Tide: Assessing the Risks of 543
Climate Change and Human Settlements in Low Elevation Coastal Zones.” Environment and 544
Urbanization 19 (1):17–37. 545
22. Morello-Frosch, R., Pastor, M., and Sadd, J. 2001. “Environmental Justice and Southern 546
California’s ‘Riskscape’: The Distribution of Air Toxics Exposures and Health Risks among 547
Diverse Communities.” Urban Affairs Review 36 (4): 551–78. 548 549
23. IPCC. 2012. Chapter 3 Changes in Climate Extremes and Their Impacts on the Natural 550
Physical Environment. 551
24. Jongman, B., Koks, E. E., Husby, T. G., and Ward, P. J. 2014. “Increasing Flood Exposure in 552
the Netherlands: Implications for Risk Financing.” Nat. Hazards Earth Syst. Sci. 14 (5):1245553
55. 554
25. Jongman, B., Ward, P.J., and Aerts, C. J. 2012. “Global Exposure to River and Coastal 555
Flooding: Long Term Trends and Changes.” Global Environmental Change 22 (4):823–35. 556 557
26. Kellar, D. M. M. The examination of vehicle related flood fatalities in the United States, 558
Commonwealth of Puerto Rico and the U.S. Territories of the Virgin Island and Guam: 1995-559
2005”, A thesis submitted to Kent State University, available at:
!etd.send_file?accession=kent1290537007&disposition=inline 561
27. Koks, E. E., Jongman, B., Husby, T. G., and Botzen, W. J. W. 2015. “Combining Hazard, 562
Exposure and Social Vulnerability to Provide Lessons for Flood Risk Management.” 563
Environmental Science & Policy 47 (Supplement C):42–52.
/j.envsci.2014.10.013. 565
Page 30
28. Lam, N. S., Reams, M., Li, K., Li, C., and Mata, L. P. 2016. “Measuring Community Resilience 566
to Coastal Hazards along the Northern Gulf of Mexico.” Natural Hazards Review 17 (1):e193567
e193. 568
29. Lindsay, J. R. 2003. “The Determinants of Disaster Vulnerability: Achieving Sustainable 569
Mitigation through Population Health.” Natural Hazards 28 (2–3):291–304. https://doi.org570
/10.1023/A:1022969705867. 571
30. Linskey, A. 2013. “FEMA’s New Flood Maps Pressure Homeowners to Raise Their Houses.” 572
Bloomberg. 2013.
floodmaps-pressure-homeowners-to-raise-their-houses. 574
31. De Moel, H., Aerts, C.J., and Koomen, E. 2011. “Development of Flood Exposure in the 575
Netherlands during the 20th and 21st Century.” Global Environmental Change, Special Issue 576
on The Politics and Policy of Carbon Capture and Storage, 21 (2):620–27. 577 578
32. Masozera, Michel, Melissa Bailey, and Charles Kerchner. “Distribution of Impacts of Natural 579
Disasters across Income Groups: A Case Study of New Orleans.Ecological Economics, 580
Ecological Economics of Coastal Disasters, 63, no. 2 (August 1, 2007): 299–306. 581 582
33. Nelson, R., Kokic, P., Crimp, S., Meinke, H., and Howden. S. M. 2010. “The Vulnerability of 583
Australian Rural Communities to Climate Variability and Change: Part I—Conceptualising 584
and Measuring Vulnerability.” Environmental Science & Policy 13 (1):8–17. 585 586
Page 31
34. Neumann, B., Vafeidis, A.T., Zimmermann, J., and Nicholls, R.J., 2015. “Future Coastal 587
Population Growth and Exposure to Sea-Level Rise and Coastal Flooding - A Global 588
Assessment.” PLOS ONE 10 (3):e0118571. 589
35. Pottier, N., Penning-Rowsell, E., Tunstall, S., and Hubert, G., 2005. “Land Use and Flood 590
Protection: Contrasting Approaches and Outcomes in France and in England and Wales.” 591
Applied Geography 25 (1):1–27. 592
36. Qiang, Yi, Nina S. N. Lam, Heng Cai, and Lei Zou. 2017. “Changes in Exposure to Flood 593
Hazards in the United States.” Annals of the American Association of Geographers 107 (6): 594
1332–50. 595
37. Siagian, T. H., Purhadi, P., Suhartono, S., and Ritonga, H., 2014. “Social Vulnerability to 596
Natural Hazards in Indonesia: Driving Factors and Policy Implications.” Natural Hazards 70 597
(2):1603–17. 598
38. Sampson, Christopher C., Andrew M. Smith, Paul D. Bates, Jeffrey C. Neal, Lorenzo Alfieri, 599
and Jim E. Freer. “A High-Resolution Global Flood Hazard Model.” Water Resources 600
Research 51, no. 9: 7358–81. 601
39. Thieken, A. H., Cammerer, H., Dobler, C., Lammel, J. and Schöberl, F., 2016. “Estimating 602
Changes in Flood Risks and Benefits of Non-Structural Adaptation Strategies - a Case Study 603
from Tyrol, Austria.” Mitigation and Adaptation Strategies for Global Change 21 (3):343–76. 604 605
40. Walker, G., and Burningham, K. 2011. “Flood Risk, Vulnerability and Environmental Justice: 606
Evidence and Evaluation of Inequality in a UK Context.” Critical Social Policy 31 (2):216607
40. 608
Page 32
41. Winsemius, H.C., Jongman, B., Veldkamp, T. I. E., Hallegatte, S., Bangalore, M., and Ward., 609
P. J., 2018. “Disaster Risk, Climate Change, and Poverty: Assessing the Global Exposure of 610
Poor People to Floods and Droughts.” Environment and Development Economics 23 (3): 328611
48. 612
42. Wing, Oliver E. J., Paul D. Bates, Christopher C. Sampson, Andrew M. Smith, Kris A. Johnson, 613
and Tyler A. Erickson. “Validation of a 30 m Resolution Flood Hazard Model of the 614
Conterminous United States.” Water Resources Research 53, no. 9 (2017): 7968–86. 615 616
43. Wing, Oliver E. J., Paul D. Bates, Andrew M. Smith, Christopher C. Sampson, Kris A. Johnson, 617
Joseph Fargione, and Philip Morefield. (2018). “Estimates of Present and Future Flood Risk 618
in the Conterminous United States.” Environmental Research Letters 13 (3): 034023. 619 620
44. Wheater, H., and Evans, E., 2009. “Land Use, Water Management and Future Flood Risk.” 621
Land Use Policy, Land Use Futures, 26 (Supplement 1): S251–64.
landusepol.2009.08.019. 623
45. Yusuf, A. A., and Francisco, H., 2009. “Climate Change Vulnerability Mapping for Southeast 624
Asia.” 625
... Often, people with the least capacity to prepare, respond, and recover from flooding events also tend to live where their exposure to such events is greater [23,24]. Nevertheless, in other contexts, socioeconomically-advantaged populations may experience greater exposure to flooding [25], mainly in coastal regions, where locational and environmental benefits such as the attraction of waterfront amenities [26] are contrasted with high-impact weather events like hurricanes that often produce significant coastal flooding. ...
... Differentiating floods by type, as shown in this study, may be another way to address currently conflicting findings in the scientific literature. For example, floods can manifest as large regional 4 In defining social vulnerability throughout this study, we use the following, well-established definition: 'the characteristics of a person or group and their situation that influence their capacity to anticipate, cope with, resist, and recover from the impact of a natural hazard' [26,33,[39][40][41][42]. 5 river floods over long periods of time, local shortduration flash floods, coastal storm surges, and urban drainage overflow [28]. ...
... In addition, we selected socioeconomic variables (table 2) that indicate different household deprivation domains, such as income, employment, housing, household characteristics, transportation, and demographics [19]. These variables are commonly used in distributive environmental justice research to evaluate socioeconomic variability in health outcomes and social vulnerability to natural hazards [26,33,[40][41][42][43]. ...
Full-text available
Previous studies have drawn attention to racial and socioeconomic disparities in exposures associated with flood events at varying spatial scales, but most of these studies have not differentiated flood risk. Assessing flood risk without differentiating floods by their characteristics (e.g., duration and intensity of precipitation leading to flooding) may lead to less accurate estimates of the most vulnerable locations and populations. In this study, we compare the spatial patterning of social vulnerability, types of housing, and housing tenure (i.e., rented vs. owned) between two specific flood types used operationally by the National Weather Service—flash floods and slow-rise floods— in the floodplains across the contiguous United States. We synthesized several datasets, including established distributions of flood hazards and flooding characteristics, indicators of socioeconomic status, social vulnerability, and housing characteristics, and used generalized estimating equations to examine the proportion of socially vulnerable populations and housing types and tenure residing in the flash and slow-rise flood extents. Our statistical findings show that the proportion of the slow-rise flooded area in the floodplains is significantly greater in tracts characterized by higher percentages of socially vulnerable. However, the results could not confirm the hypothesis that they are exposed considerably more than less vulnerable in the flash flooded floodplains. Considering housing-occupancy vulnerability, the percentage of renter-occupancies are greater in the flash flood floodplains compared to slow-rise, especially in areas with high rainfall accumulation producing storms (e.g., in the Southeast). This assessment contributes insights into how specific flood types could impact different populations and housing tenure across the CONUS and informs strategies to support urban and rural community resilience and planning at local and state levels.
... To measure people's exposure to flooding, we then calculated the relative area of habitability across scales using land cover data. Population exposure disaggregated based on land cover is one approach to measuring flood exposure (Debbage, 2019;Qiang, 2019a;Tate et al., 2021); however, there are several alternative methods of quantifying exposure that vary in accuracy and resolution, see (Crowell et al., 2010;Huang & Wang, 2020;Yager & Rosoff, 2017). ...
... Dasymetric mapping is a geospatial technique that can utilize land cover to distribute population data more accurately across a geographic boundary, such as BGs, tracts, and counties (US EPA, O., 2015). This technique has been applied in other flood exposure and risk analyses see (Debbage, 2019;Flores et al., 2023;Maantay & Maroko, 2009;Montgomery & Chakraborty, 2013;Qiang, 2019a;Tate et al., 2021;Wing et al., 2018). ...
Full-text available
Flooding is a natural hazard that touches nearly all facets of the globe and is expected to become more frequent and intensified due to climate and land-use change. However, flooding does not impact all individuals equally. Therefore, understanding how flooding impacts distribute across populations of different socioeconomic and demographic backgrounds is vital. One approach to reducing flood risk on people is using indicators, such as social vulnerability indices and flood exposure metrics, to inform decision-making for flood risk management. However, such indicators can face the scale and zonal effect produced by the Modifiable Areal Unit Problem (MAUP). This study investigates how the U.S. Census block group, tract, and county scale selection impacts social vulnerability and flood exposure outcomes within coastal Virginia, USA. Here we show how (1) scale selection can obstruct our understanding of drivers of vulnerability, (2) increasingly aggregated scales significantly undercount highly vulnerable populations, and (3) hotspot clusters of social vulnerability and flood exposure can identify variable priority areas for current and future flood risk reduction. Study results present considerations about using such indicators, given the real-life consequences that can occur due to the MAUP. The results of this work warrant understanding the implications of scale selection on research methodological approaches and what this means for practitioners and policymakers that utilize such information to help guide flood mitigation strategies.
... FEMA flood maps cover 57% of the territory of the 50 US states, but 93.6% of the population 150 (Qiang, 2019). In our analysis, we additionally filter for counties in the contiguous US where more 151 than 50% of residential properties are accounted for within flood mapped areas, leading to a final 152 sample of 2,392 counties out of 3,108 total counties, accounting for 94% of the CONUS population 153 (Fig. S3). ...
Despite increasing exposure to flooding and associated financial damages, estimates suggest more than two-thirds of flood-exposed properties are currently uninsured. This low adoption rate could undermine the climate resilience of communities and weaken the financial solvency of the United States National Flood Insurance Program (NFIP). We study whether repeated exposure to flood events, especially disaster-scale floods expected to become more frequent in a warming climate, could spur insurance adoption. Using improved estimates of residential insurance take-up in locations where such insurance is voluntary, and exploiting variation in the frequency and severity of flood events over time, we quantify how flood events impact local insurance demand. We find that a flood disaster declaration in a given year increases the take-up rate of insurance by 7% in the following year, but the effect diminishes in subsequent years and is gone after five years. This effect is more short-lived in counties in inland states that do not border the Gulf and Atlantic coasts. The effect of a flood on takeup is substantially larger if there was also a flood in the previous year. We also find that recent disasters are more salient for homeowners whose primary residences are exposed to a disaster declaration compared to non-primary residences. Our results provide a more comprehensive understanding of the salience effect of flooding on insurance demand compared to previous studies. Overall, these findings suggest that relying on households to self-adapt to increasing flood risks in a changing climate is insufficient for closing the insurance protection gap.
... Similarly, studies on flood exposure and social vulnerability have typically used a spatial analysis approach in which flood risk layers, typically 100-year flood zones, are superimposed on social vulnerability indicators and compared to identify spatial clusters of vulnerability (e.g., Emrich & Cutter. 2011;Finch et al., 2010;Qiang, 2019;Tate et al., 2021;Wing et al., 2022). ...
Full-text available
Infrastructure equity is an immediate concern with levees, constituting the backbone of the U.S. protection against flooding. Flooding patterns are exacerbated by anthropogenic climate change in several regions, posing a significant risk to the economy, safety, and well‐being of the nation. The evolving risk of flooding is shown to disproportionately affect historically underserved and socially vulnerable communities (HUSVCs). Here we compare the sociodemographic and socioeconomic composition of leveed and non‐leveed U.S. communities and show a substantial overrepresentation of HUSVCs in leveed areas at the state, regional, and national levels. Further, we analyze the proportion of communities designated as “disadvantaged” in leveed versus non‐leveed areas, revealing a substantially larger population of disadvantaged communities residing behind levees. Our analyses show that nationally, Hispanic are the most overrepresented population in leveed areas yielding a disparity percentage of 39.9%, followed by Native American (18.7%), Asian (17.7%), and Black (16.1%) communities. Communities characterized by low education, poverty, and disability exhibit a disproportionately higher presentation of 27.8%, 20.4%, and 5.4% in leveed areas across the U.S. In 43 states, disadvantaged communities are overrepresented behind levees, with a national disparity percentage of 40.6%. At the regional level, the highest disparity was observed in the Northeast (57.3%), followed by the West (51.3%), Southeast (38%), Midwest (29.2%), and Southwest (25%). The findings can enable decision‐ and policy‐makers to identify hotspots within HUSVCs that need to be prioritized for enhancing the integrity and climate adaptation of their levee systems.
... The standard approach to flood exposure assessment focuses on place of residence as the location from which the extent and disparities of hazard exposure are quantified. This approach is based on overlaying flood hazard map or floodplain areas with population distribution information to estimate flood exposure (Boulange, Hanasaki, Yamazaki, & Pokhrel, 2021;Chakraborty et al., 2022;Mohanty & Simonovic, 2021;Qiang, 2019). Studies focusing primarily on residential locations assume that the impact of floods on people is solely based on damage to their residence. ...
Full-text available
Current characterization of flood exposure is largely based on residential location of populations; however, location of residence only partially captures the extent to which populations are exposed to flood. An important, though yet under-recognized aspect of flood exposure is associated with human mobility patterns and population visitation to places located in flood prone areas. This study analyzed large-scale, high-resolution location-intelligence data to characterize human mobility patterns and the resulting flood exposure in counties of the United States. We developed the metric of mobility-based exposure based on dwell time in places located in the 100-year floodplain. The results of examining the extent of mobility-based flood exposure reveal a significant disparity across race, income , and education level groups. Black and Asian, economically disadvantaged, and undereducated populations are disproportionally exposed to flood due to their daily mobility activities, indicating a pattern contrary to residential flood exposure. The results suggest that mobility behaviors play an important role in extending flood exposure reach disproportionally among socio-demographic groups. Mobility-based flood exposure provides a new perspective regarding the extent to which floods could disrupt people's life activities and enables a better characterization of disparity in populations' exposure to flood hazards beyond their place of residence.
... The risk of property flooding is not distributed equally across all areas and communities in a city [20,21]. Studies have shown that low-income communities and communities of color are often more vulnerable to flooding than wealthier, white communities [22][23][24][25]. Spatial inequality of property flood risk is in part influenced by the patterns of growth and development in cities [26]. ...
Full-text available
Understanding the relationship between urban form and structure and spatial variation of property flood risk has been a longstanding challenge in urban planning and city flood risk management. Yet limited data-driven insights exist regarding the extent to which variation in spatial inequality of property flood risk in cities can be explained by heterogenous features of urban form and structure. In this study, we explore eight key features (i.e., population density, point of interest density, road density, minority segregation, income segregation, urban centrality index, gross domestic product, and human mobility index) related to urban form and structure to explain variability in spatial inequality of property flood risk among 2567 US counties. Using rich datasets related to property flood risk, we quantify spatial inequality in property flood risk and delineate features of urban form and structure using high-resolution human mobility and facility distribution data. We identify significant variation in spatial inequality of property flood risk among US counties with coastline and metropolitan counties having the greatest spatial inequality of property flood risk. The results also reveal variations in spatial inequality of property flood risk can be effectively explained based on principal components of development density, economic activity, and centrality and segregation. Using a classification and regression tree model, we demonstrate how these principal components interact and form pathways that explain levels of spatial inequality in property flood risk in US counties. The findings offer important insights for the understanding of the complex interplay between urban form and structure and spatial inequality of property flood risk and have important implications for integrated urban design strategies to address property flood risk as cities continue to expand and develop.
... It provides powerful post processing and result representation features ideal for hazard mapping. The modelling involves extensive datasets depicting topographical (e.g., digital elevation model (DEM), land use land cover (LULC)), hydraulic (e.g., river discharge, river water level), and meteorological (e.g., rainfall) inputs (Thorndahl and Willems, 2008;Qiang, 2019;Mohanty et al., 2020b). In the context of Surat City, previous researchers (Timbadiya et al., 2014(Timbadiya et al., , 2015Patel et al., 2017) developed hydrodynamic models to study the behavior of past floods across coastal Surat City, although flood hazard analysis was not performed. ...
In the current study, flood risk assessment of densely populated coastal urban Surat City, on the bank of the lower Tapi River in India, was conducted by combining the hydrodynamic model-based flood hazard and often neglected socioeconomic vulnerability. A two-dimensional (2D) hydrodynamic (HD) model was developed using physically surveyed topographic data and the existing land use land cover (LULC) of the study area (5248 km 2). The satisfactory performance of the developed model was ascertained by comparing the observed and simulated water levels/depths across the river and floodplain. The 2D HD model outputs with geographic information system (GIS) applications were further used to develop probabilistic multiparameter flood hazard maps for coastal urban city. During a 100-year return period flood (Peak discharge = 34,459 m 3 /s), 86.5% of Surat City and its outskirt area was submerged, with 37% under the high hazard category. The north and west zones are the worst affected areas in Surat City. The socioeconomic sensitivity and adaptive capacity indicators were selected at the city's lowest administrative (ward) level. The socioeconomic vulnerability was evaluated by employing the robust data envelopment analysis (DEA) technique. Fifty-five of 89 wards in Surat City, covering 60% of the area under the jurisdiction of the Municipal Corporation, are highly vulnerable. Finally, the flood risk assessment of the city was conducted using a bivariate technique describing the distinctive contribution of flood hazard and socioeconomic vulnerability to risk. The wards adjoining the river and creek are at high flood risk, with an equal contribution of hazard and vulnerability. The ward-level hazard, vulnerability, and risk assessment of the city will help local and disaster management authorities to priorities high risk areas while planning flood management and mitigation strategies.
... That Sapelo's Black Geechee population is at high risk to flooding is likely no surprise to many readers. It is well-documented within academic scholarship across the United States that lower elevation, higher flood risk zones are often disproportionately inhabited by economically-disadvantaged people (Qiang 2019) and/or people of color (Ueland and Warf 2006;Bullard and Wright 2009;Knighton et al. 2021). Yet, while the broader pattern is not consistent (see Maantay and Maroko 2009), it generally holds that inland flood zones are disproportionately inhabited by people of color and/or low income households while coastal flood zones are typically inhabited by affluent white people with access to resources (Ueland and Warf 2006;Collins, Grineski, and Chakraborty 2018). ...
Decades of environmental justice research has focused on identifying existing patterns of disproportionate burdens to environmental harms across social difference. However, relatively few studies examine the “legacy effect” of historical patterns. In flood risk studies specifically, several scholars have highlighted the role of systemic processes in historically shaping and producing observed disparities in flood risk patterns. These studies reveal that such relations are tied to histories of racialized land struggles and territorial dispossessions. In this paper, I argue that scholars need to do more than quantify today’s disproportionate burdens across social difference or explain the systemic processes causing those disparities. I suggest that “legacy vulnerability” helps identify how the potential for harm from flood risk to marginalized groups may reside in events of the past that have imprinted a spatially hidden, but spatiotemporally revealed unjust pattern upon today’s landscape. In a flood risk assessment of Sapelo Island, the initial results suggest that when comparing contemporary flood risk of Sapelo’s Geechee descendant (Black and mostly low-to-middle income) to non-descendant newcomer owners (mostly white and affluent) an environmental justice disparity in proportional flood risk burden does not exist. However, results of a counterfactual flood risk assessment show that approximately one-third of historically owned, Geechee property is located outside the contemporary 100-year flood zone compared to zero percent outside of it today. In other words, roughly one-third of Geechee property’s flood risk today is a legacy vulnerability directly tied to racialized land dispossessions that unfolded in the middle twentieth century.
The Hindu-Kush-Himalaya is abode to numerous severely flood-prone mountainous stretches that distress vulnerable communities and cause massive destruction to physical entities such as hydropower projects. Adopting commercial flood models for replicating the dynamics of flood wave propagation over such regions is a major constraint due to the financial economics threaded to flood management. For the first instance, the present study attempts to investigate whether advanced open-source models are skillful in quantifying flood hazards and population exposure over mountainous terrains. While doing so, the performance of 1D-2D coupled HEC-RAS v6.3 (the most recent version developed by the U.S. Army Corps of Engineers) is reconnoitred for the first time in flood management literature. The most flood-prone region in Bhutan, the Chamkhar Chhu River Basin, housing large groups of communities and airports near its floodplains, is considered. HEC-RAS v6.3 setups are corroborated by comparing them with 2010 flood imagery derived from MODIS through performance metrics. The results indicate a sizable portion of the central part of the basin experiences very-high flood hazards with depth and velocities exceeding 3 m, and 1.6 m/s, respectively, during 50, 100, and 200-year return periods of floods. To affirm HEC-RAS, the flood hazards are compared with TUFLOW at 1D and 1D-2D coupled levels. The hydrological similarity within the channel is reflected at river cross-sections (NSE and KGE > 0.98), while overland inundation and hazard statistics differ, however, very less significant (<10 %). Later, flood hazards extracted from HEC-RAS are fused with the World-Pop population to estimate the degree of population exposure. The study ascertains that HEC-RAS v6.3 is an efficacious option for flood risk mapping over geographically arduous regions and can be preferred in resource-constrained environments ensuring a minimal degree of anomaly.
Full-text available
Climate change and natural hazards pose great threats to road transport systems which are ‘lifelines’ of human society. However, there is generally a lack of empirical data and approaches for assessing resilience of road networks in real hazard events. This study introduces an empirical approach to evaluate road network resilience using crowdsourced traffic data in Google Maps. Based on the conceptualization of resilience and the Hansen accessibility index, resilience of road network is measured from accumulated accessibility reduction over time during a hazard. The utility of this approach is demonstrated in a case study of the Cleveland metropolitan area (Ohio) in Winter Storm Harper. The results reveal strong spatial variations of the disturbance and recovery rate of road network performance during the hazard. The major findings of the case study are: (1) longer distance travels have higher increasing ratios of travel time during the hazard; (2) communities with low accessibility at the normal condition have lower road network resilience; (3) spatial clusters of low resilience are identified, including communities with low socio-economic capacities. The introduced approach provides ground-truth validation for existing quantitative models and supports disaster management and transportation planning to reduce hazard impacts on road network.
Full-text available
Hurricanes, as one of the most devastating natural hazards, have posed a great threat to people in coastal areas. A better understanding of the spatiotemporal dynamics of human settlement in hurricane-prone areas largely benefits sustainable development. This study uses the nighttime light (NTL) data from the Defense Meteorological Satellite Program's Operational Linescan System (DMSP/OLS) to examine human settlement development in areas with different levels of hurricane proneness from 1992 to 2013. The DMSP/OLS NTL data from six satellites were intercalibrated and desaturated with the Advanced Very High Resolution Radiometer (AVHRR) and Moderate Resolution Imaging Spectroradiometer (MODIS) optical imagery to derive the Vegetation Adjusted NTL Urban Index (VANUI), a popular index that quantifies human settlement intensity. The derived VANUI time series was examined with the Mann–Kendall test and Theil–Sen test to identify significant spatiotemporal trends. To link the VANUI product to hurricane impacts, four hurricane-prone zones were extracted to represent different levels of hurricane proneness. Aside from geographic division, a wind-speed-weighted track density function was developed and applied to historical storm tracks which originated in the North Atlantic Basin to better categorize the four levels of hurricane proneness. Spatiotemporal patterns of human settlement in the four zones were finally analyzed. The results clearly exhibit a north–south and inland–coastal discrepancy of human settlement dynamics. This study also reveals that both the zonal extent and zonal increase rate of human settlement positively correlate with hurricane proneness levels. The intensified human settlement in high hurricane-exposure zones deserves further attention for coastal resilience.
Full-text available
Preserving coral reef resilience is a major challenge in the Anthropocene, yet recent studies demonstrate failures of reef recovery from disturbance, globally. The wide and vigorous outer-reef system of French Polynesia presents a rare opportunity to assess ecosystem resilience to disturbances at a large-scale equivalent to the size of Europe. In this purpose, we analysed long-term data on coral community dynamics and combine the mixed-effects regression framework with a set of functional response models to evaluate coral recovery trajectories. Analyses of 14 years data across 17 reefs allowed estimating impacts of a cyclone, bleaching event and crown-of-thorns starfish outbreak, which generated divergence and asynchrony in coral community trajectory. We evaluated reef resilience by quantifying levels of exposure, degrees of vulnerability, and descriptors of recovery of coral communities in the face of disturbances. Our results show an outstanding rate of coral recovery, with a systematic return to the pre-disturbance state within only 5 to 10 years. Differences in the impacts of disturbances among reefs and in the levels of vulnerability of coral taxa to these events resulted in diverse recovery patterns. The consistent recovery of coral communities, and convergence toward pre-disturbance community structures, reveals that the processes that regulate ecosystem recovery still prevail in French Polynesia.
Full-text available
Whereas monthly and annual nighttime light (NTL) composite datasets are being increasingly used to estimate socioeconomic status, use of the National Polar-orbiting Partnership Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) Day/Night Band (DNB) daily data has been limited for detecting and assessing the impact of short-term disastrous events. This study explores the application of daily NPP-VIIRS DNB data in assessing the impact of three types of natural disasters: earthquakes, floods, and storms. Daily DNB images one month prior to and 10 days after a disastrous event were collected and a Percent of Normal Light (PNL) image was produced as the ratio of the mean DNB radiance of the pre- and post-disaster images. Areas with a PNL value lower than one were considered as being affected by the event. The results were compared with the damaged proxy map and the flood proxy map generated using synthetic aperture radar data as well as the reported power outage rates. Our analyses show that overall NPP-VIIRS DNB daily data are useful for detecting damages and power outages caused by earthquake, storm, and flood events. Cloud coverage was identified as a major limitation in using the DNB daily data; rescue activities, traffic, and socioeconomic status of the areas also affect the use of DNB daily data in assessing the impact of natural disasters. Our findings offer new insight into the use of the daily DNB data and provide a practical guide for researchers and practitioners who may consider using such data in different situations or regions.
Full-text available
Time series monitoring of earthquake-stricken areas is significant in evaluating post-disaster reconstruction and recovery. The time series of nighttime light (NTL) data collected by the defense meteorological satellite program-operational linescan system (DMSP/OLS) sensors provides a unique and valuable resource to study changes in human activity (HA) because of the long period of available data. In this paper, the DMSP/OLS NTL images’ digital number (DN) is used as a proxy for the intensity of HA since there is a high correlation between them. The purpose of this study is to develop a methodology to analyze the changes of intensity and distribution of HA in different areas affected by a 2008 earthquake in Wenchuan, China. In order to compare the trends of HA before and after the earthquake, the DMSP/OLS NTL images from 2003 to 2013 were processed and analyzed. However, their analysis capability is greatly limited owing to a lack of in-flight calibration. To improve the continuity and comparability of DMSP/OLS NTL images, this study developed an automatic intercalibration method to systematically correct NTL data. The results reveal that: (1) compared with the HA before the earthquake, the reconstruction and recovery of the Wenchuan earthquake have led to a significant increase of HA in earthquake-stricken areas within three years after the earthquake; (2) the fluctuation of HA in a severely-affected area is greater than that in a less-affected area; (3) recovery efforts increase development in the most affected areas to levels that exceeded the rates in similar areas which experienced less damage; and (4) areas alongside roads and close to reconstruction projects exhibited increased development in regions with otherwise low human activity.
Full-text available
Past attempts to estimate rainfall-driven flood risk across the US either have incomplete coverage, coarse resolution or use overly simplified models of the flooding process. In this paper, we use a new 30 m resolution model of the entire conterminous US with a 2D representation of flood physics to produce estimates of flood hazard, which match to within 90% accuracy the skill of local models built with detailed data. These flood depths are combined with exposure datasets of commensurate resolution to calculate current and future flood risk. Our data show that the total US population exposed to serious flooding is 2.6–3.1 times higher than previous estimates, and that nearly 41 million Americans live within the 1% annual exceedance probability floodplain (compared to only 13 million when calculated using FEMA flood maps). We find that population and GDP growth alone are expected to lead to significant future increases in exposure, and this change may be exacerbated in the future by climate change.
Social media such as Twitter is increasingly being used as an effective platform to observe human behaviors in disastrous events. However, uneven social media use among different groups of population in different regions could lead to biased consequences and affect disaster resilience. This paper studies the Twitter use during 2017 Hurricane Harvey in 76 counties in Texas and Louisiana. We seek to answer a fundamental question: did social-geographical disparities of Twitter use exist during the three phases of emergency management (preparedness, response, recovery)? We employed a Twitter data mining framework to process the data and calculate two indexes: Ratio and Sentiment. Regression analyses between the Ratio indexes and the social-geographical characteristics of the counties at the three phrases reveal significant social and geographical disparities in Twitter use during Hurricane Harvey. Communities with higher disaster-related Twitter use in Harvey generally were communities having better social and geographical conditions. These results of Twitter use patterns can be used to compare with future similar studies to see whether the Twitter use disparities have increased or decreased. Future research is also needed to examine the effects of Twitter use disparities on disaster resilience and to test whether Twitter use can predict community resilience.
Economic theory suggests that, other things being equal, properties located within a floodplain should suffer a price discount. A survey of the existing evidence nonetheless reveals that this price discount lies anywhere between − 75.5% to a + 61.0% price premium. In this paper we summarise and explore the wide variation in the results to obtain ‘best’ estimates with which to guide policy. Results from our meta-analysis comprising 37 published works and 364 point estimates indicate marked differences between studies according to when and where they were conducted. For coastal regions the results show that properties located in the floodplain command higher prices; this finding is however likely to be caused by a high correlation between omitted coastal amenities and flood risk. There is moreover, evidence that publication bias affects the coastal flooding literature. Results from meta-regression analyses intended to uncover sources of heterogeneity confirm that controlling for time elapsed since the most recent flood is especially important. For inland flooding the price discount associated with location in the 100-year floodplain is − 4.6%. Although other estimates are defensible, we suggest this figure be used as a rule of thumb to determine the benefits of flood relief projects to households.
Studies on how variables of community resilience to natural hazards interact as a system that affects the final resilience (i.e., their dynamical linkages) have rarely been conducted. Bayesian network (BN), which represents the interdependencies among variables in a graph while expressing the uncertainty in the form of probability distributions, offers an effective way to investigate the interactions among different resilience components and addresses the natural–human system as a whole. This article employs a BN to study the interdependencies of ten resilience variables and population change in the Lower Mississippi River Basin (LMRB) at the census block group scale. A genetic algorithm was used to identify an optimal BN where population change, a cumulative resilience indicator, was the target variable. The genetic algorithm yielded an optimized BN model with a cross-validation accuracy of 67 percent over a period of 906 generations. Six variables were found to have direct impacts on population change, including level of threat from coastal hazards, hazard damage, distance to coastline, employment rate, percentage of housing units built before 1970, and percentage of households with a female householder. The remaining four variables were indirect variables, including percentage agriculture land, percentage flood zone area, percentage owner-occupied house units, and population density. Each variable has a conditional probability table so that its impacts on the probability of population change can be evaluated as it propagates through the network. These probabilities could be used for scenario modeling to help inform policies to reduce vulnerability and enhance disaster resilience.
People living in poverty are particularly vulnerable to shocks, including those caused by natural disasters such as floods and droughts. This paper analyses household survey data and hydrological riverine flood and drought data for 52 countries to find out whether poor people are disproportionally exposed to floods and droughts, and how this exposure may change in a future climate. We find that poor people are often disproportionally exposed to droughts and floods, particularly in urban areas. This pattern does not change significantly under future climate scenarios, although the absolute number of people potentially exposed to floods or droughts can increase or decrease significantly, depending on the scenario and region. In particular, many countries in Africa show a disproportionally high exposure of poor people to floods and droughts. For these hotspots, implementing risk-sensitive land-use and development policies that protect poor people should be a priority.