Next Article in Journal
Enhancing the Precision of Forest Growing Stock Volume in the Estonian National Forest Inventory with Different Predictive Techniques and Remote Sensing Data
Next Article in Special Issue
From Data to Decision: Interpretable Machine Learning for Predicting Flood Susceptibility in Gdańsk, Poland
Previous Article in Journal
Numerical Weather Prediction of Sea Surface Temperature in South China Sea Using Attention-Based Context Fusion Network
Previous Article in Special Issue
Constructing Rainfall Threshold for Debris Flows of a Defined Hazardous Magnitude
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 16
Issue 20
10.3390/rs16203792
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Disparate Disaster: Spatial Patterns of Building Damage Caused by Hurricane Ian and Associated Socio-Economic Factors

by
Md Zakaria Salim
*,
Yi Qiang
,
Barnali Dixon
and
Jennifer Collins
School of Geosciences, University of South Florida, Tampa, FL 33620, USA
*
Author to whom correspondence should be addressed.
Remote Sens. 2024, 16(20), 3792; https://doi.org/10.3390/rs16203792
Submission received: 20 August 2024 / Revised: 3 October 2024 / Accepted: 8 October 2024 / Published: 12 October 2024
(This article belongs to the Special Issue Advances in GIS and Remote Sensing Applications in Natural Hazards)
Browse Figures
Versions Notes

Abstract

:
The literature shows that communities under different socio-economic conditions suffer different levels of damage in disasters. In addition to the physical intensity of hazards, such differences are also related to the varying abilities of communities to prepare for and respond to disasters. This study analyzes the spatial patterns of building damage in Hurricane Ian in 2022 and investigates the socio-economic disparities related to the damage. Specifically, this study employs NASA’s Damage Proxy Map (DPM2) to analyze spatial patterns of building damage caused by the hurricane. Then, it uses statistical analysis to assess the relationships between building damage and hurricane intensity, building conditions, and socio-economic variables at the building and census tract levels. Furthermore, the study applies geographically weighted regression (GWR) to examine the spatial variation of the damage factors. The results provide valuable insights into the potential factors related to building damage and the spatial variation in the factors. The results also reveal the uneven distribution of building damage among different population groups, implying socio-economic inequalities in disaster adaptation and resilience. Moreover, the study provides actionable information for policymakers, emergency responders, and community leaders in formulating strategies to mitigate the impact of future hurricanes by identifying vulnerable communities and population groups.

Graphical Abstract

1. Introduction

Extreme weather patterns driven by global environmental changes pose significant threats to human communities worldwide [1]. The U.S. has faced substantial consequences from increased flooding and the increasing frequency of hurricane events, resulting in losses of over USD 997.3 billion and over 6500 lives from 1980 to 2020 [2]. Meanwhile, the growth of coastal populations and expansion of urban development has increased the exposure of coastal communities and infrastructure to hurricanes throughout the U.S. [3]. As one of the most vulnerable coastal regions in the U.S., Florida has been affected by some of the most destructive hurricanes in U.S. history and endured the highest amount of hurricane-induced economic loss in the nation [4]. The hurricane threat in Florida was highlighted by Hurricane Ian in 2022. Characterized by severe weather conditions, Hurricane Ian exemplified the vulnerability of areas with urban sprawl and pointed out the significant risks associated with coastal areas in Florida [5]. Hurricane Ian resulted in economic losses exceeding USD 113 billion [6]. The losses included a significant number of homes that were either destroyed or damaged, impacting communities with various socio-economic conditions. The variation in building damage among communities provides an opportunity to investigate the factors contributing to building damage.
Prior research on building damage caused by hurricanes has predominantly centered on the physical intensity of hurricanes and the structural vulnerabilities of buildings [7]. Construction materials, building age, and compliance with hurricane-resistant building codes are often identified as critical determinants of hurricane-induced building damage [8]. In addition to the physical properties of buildings, an increasing number of studies have recognized the importance of socio-economic factors in differentiating hurricane-induced damage [9]. For example, income levels, housing density, and the quality of infrastructure were found to affect damage outcomes most significantly, highlighting the need for targeted mitigation and response strategies in various communities [10,11].
Despite the shifting focus onto socio-economic factors, a notable gap exists in quantitatively measuring how these factors influence building damage. Furthermore, the existing research has yet to fully explore the combined effects of the various factors on building damage and the spatial variation in these effects. Our research aims to bridge these gaps by analyzing the relationship between building damage and a wide range of physical and socio-economic factors during Hurricane Ian at two spatial scales. The main objective of this study is to identify the influential factors that differentiate building damage in a major hurricane. The study also reveals the disparities in levels of building damage in different population groups and geographic locations. Specifically, this study is trying to answer the following research questions: (1) What is the spatial pattern of building damage caused by Hurricane Ian? (2) What influential factors contribute to building damage, and how do these factors change in geographic space? (3) Are there disparities in building damage among different communities and population groups?
To address these questions, statistical analysis will be conducted to quantify the relationships between property damage and socio-economic conditions by controlling for hurricane intensity. The spatial analysis will uncover the spatial patterns of building damage and the spatial variations in the influential factors. The results of these analyses will provide insights into the complex mechanisms of hurricane-induced damage. This, in turn, can lead to more effective and equitable disaster mitigation and response efforts, ensuring that assistance and resources are distributed to address the immediate and long-term needs of those most impacted by hurricanes.

2. Literature Review

2.1. Building Damage in Hurricanes

Hurricanes cause significant damage to buildings, primarily affecting building exteriors, resulting in subsequent interior damage [12]. Building damage is primarily caused by strong wind speeds [13,14], storm surges [15,16], and flood inundation [17]. An analysis of insurance claim records in past hurricanes shows that wind damage to buildings was primarily limited to structures’ outside coverings, such as rooves and external walls [18]. In contrast, storm surges and inundations cause extensive damage to the interiors and foundations of buildings [19]. Compared to individual impacts, buildings are significantly more vulnerable when these impacts converge [20]. Consequently, the most severe building damage in a hurricane is often found near the landfall location, and is affected by the combined effects of wind and storm surges [21]. However, studies have demonstrated that natural barriers such as sand dunes can dissipate storm energy and thus lessen the impacts of storm surges and floods on inland structures [22,23].
The susceptibility of buildings to hurricane damage varies based on factors such as structure, age, construction materials, and the inclusion of hurricane-resistant features [24]. For instance, residential buildings with light-framed wooden structures are particularly vulnerable to hurricane winds and wind-borne debris [25,26]. A comprehensive study on building performance in Florida from 1994 to 2001 indicated that implementing protective measures like window guards, modern roof-to-wall connections, and improved roof-nailing schedules can significantly reduce wind damage [27]. Fronstin and Holtmann (1994) revealed that older homes incurred less damage than new homes in Hurricane Andrew [28]. This counterintuitive trend may be attributed to low-quality construction, faulty designs, and flimsy materials, which are common problems in newer dwellings. In addition to age, the maintenance of buildings also determines their vulnerability to hurricanes. Issues such as cracking, rust, and corrosion can lead to structural instability and severe damage during hurricane events [29]. Moreover, Siegmund’s study (2018) underscores that mobile homes are especially vulnerable in hurricane-affected regions [30]. They have lower structural integrity than site-built homes, making them more susceptible to hurricane damage.
Updates to building codes over recent years have emphasized structural resilience to hurricanes. For instance, the Florida Building Code (FBC) requires stronger roof-to-wall connections, impact-resistant windows, and elevation standards for buildings along coastlines to reduce wind and flood damage [31]. While Florida had some of the strictest building codes in the U.S. for hurricane resistance, the compliance and enforcement of the building codes may vary in different regions or by contractors [32].

2.2. Socio-Economic Factors Related to Building Damage

Research shows that the economic capacity of individuals and communities is associated with the resilience of buildings to hurricane damage [33]. Larger and more expensive buildings tend to exhibit greater resilience to hurricane damage [9,34]. Higher educational attainment and average income allow households to invest in hurricane-resistant home features [10,35]. Conversely, low-income communities are particularly vulnerable due to the lower construction quality and maintenance standards of their homes [36,37]. These communities often lack resources and knowledge for disaster preparedness and response, causing them to experience more severe damage in hurricanes [38,39,40].
Demographic conditions also significantly affect hurricane impact. Logan et al. (2016) found that urban areas with high population densities, particularly those built using concrete and steel, generally suffer less damage [41]. Moreover, Liu and Fan (2023) demonstrated that urban areas with higher proportions of black and Asian residents are more likely to adopt disaster interventions than areas primarily populated by white residents [42]. Conversely, rural areas often have fewer hurricane-resistant buildings and, therefore, are likely to endure greater damage during hurricanes [43]. Bukvic et al. (2018) indicated that elderly populations also tend to have older housing stock, which is more prone to damage and more challenging to retrofit [44]. In Florida, a popular retirement destination, elderly populations are concentrated in coastal areas, making their residential buildings more susceptible to the adverse effects of hurricanes. Furthermore, building damage varies among racial groups [33]. Owner-occupied homes display a lower likelihood of damage than rental properties, as the latter often lack maintenance and hurricane resistance improvements [45,46]. Despite the wealth of research on individual socio-economic factors, studies quantifying their combined effects on building damage are relatively rare.

3. Study Area and Data

3.1. Hurricane Ian and Its Affected Area

Hurricane Ian in 2022 was the costliest hurricane in Florida and the third-costliest hurricane in the history of the United States [47]. Hurricane Ian originated from a tropical wave that emerged off the west coast of Africa and was upgraded to a tropical storm on 23 September 2022. With its swiftly escalated intensity, Ian grew to a Category 5 Hurricane on 28 September 2022. Ian made landfall near Cayo Costa, Lee County, as a Category 4 Hurricane at 3:05 pm on 28 September 2022, with an average wind speed of 155 mph. As the fifth-strongest hurricane to make landfall in the U.S., Ian extensively impacted the coastal communities, including building and infrastructure damage, casualties, and extensive disruptions to socio-economic activities [48]. Ian directly or indirectly caused the deaths of 161 people [49] and an estimated economic loss of USD 113 billion [6]. Hurricane Ian caused various levels of damage to more than 52,000 houses, ranging from severe structural impairments, such as collapsed roofs and walls, to extensive flooding and foundational issues [6]. Most building damage was concentrated in Charlotte County and Lee County, which were directly near the hurricane’s track and landfall location (Figure 1). Additionally, Hendry County, Glades County, and Collier County also experienced light-to-moderate damage. Thus, the five counties mentioned above, including Charlotte County, Glades County, Lee County, Hendry County, and Collier County, were selected as the study area.

3.2. Datasets

The Advanced Rapid Imaging and Analysis (ARIA) Damage Proxy Map (DPM2) was obtained from NASA’s website on 14 October 2022 to represent building damage [50]. The DPM2 data were derived from synthetic aperture radar (SAR) images captured by the Copernicus Sentinel-1 satellites. The DPM2 dataset covers a rectangular area of approximately 85 by 95 km (53 by 59 miles), including the entire study area at a 30 m resolution. The DPM2 dataset was derived by detecting changes in surface reflectivity before and after the hurricane. The change in pixel value between the two time points indicates a likelihood of surface disturbance based on reductions in radar reflections. In DPM2 rasters, undamaged pixels are coded as NoData, while damaged pixels are assigned an integer pixel value between 128 and 255. However, the actual meaning of the pixel value is unknown in any official sources. Thus, we classified the damage statuses into undamaged (0) and damaged (1) in the following analysis. Next, building footprint polygons from the Microsoft Building Footprint dataset (2011–2020) were then converted into their centroids (points). The building centroids were then overlayed with the DPM2 rasters to determine the damage status of the building footprints.
Building type, building age, roof age, and building size were used to represent building conditions. Building size was calculated from the polygon of building footprints. Using the coordinates of the buildings, the value, year of construction, building type, and year of roof building or improvement were manually acquired from each county’s property appraiser’s website. Due to the time-consuming nature of data acquisition from the appraiser’s website, a random sample of 200 buildings was selected in the most damaged neighborhoods, which were all in Lee County. The sample was balanced equally between damaged and undamaged buildings.
This study uses the speeds of wind gusts and sustained wind, distance to coast, and distance to hurricane track to represent the hurricane intensity (Table 1). A gust of wind, as defined by the National Oceanic and Atmospheric Administration (NOAA), refers to a brief increase in wind speed that is typically measured over a duration of three to five seconds. These gusts can be substantially higher than the average wind speeds [51]. Gusts of wind can potentially cause rapid and severe damage to buildings [52], including direct impacts on vulnerable components such as windows, roofs, and facades [53], and secondary impacts from debris and fallen trees [54]. The gust data in discrete point locations were obtained from the monthly climate report by the NOAA (2022) [55]. Additionally, sustained winds, particularly those exceeding 70 knots, can cause extensive structural damage to buildings [56]. Sustained wind speed data were collected from the NOAA International Best Track Archive for Climate Stewardship [55] and were measured in miles per hour during 1 min intervals at point locations. The Empirical Bayesian Kriging (EBK) method was used to interpolate the gust and sustained wind speeds at point locations into rasters. Then, the “Extract Multi Values to Points” tool in ArcGIS Pro was applied to append values in these rasters to the buildings. The proximity to the coast indicates the potential exposure of buildings to storm surges and coastal flooding caused by hurricanes [57]. Additionally, the proximity to the hurricane track correlates with wind and precipitation, common causes of building damage [58]. The coastline was obtained from the U.S. Census, while the track of Hurricane Ian was obtained from the National Oceanic and Atmospheric Administration [55]. The distances from each building point to the coast and hurricane track were calculated using the “Near” tool in ArcGIS.
Sixteen socio-economic variables were selected in this study (Table 1). The selection of these variables is based on an extensive literature review, as elaborated in Section 3.2 [59,60,61]. The socio-economic variables at the census tract level are acquired from the U.S. Census [62]. The average building size is calculated from all building footprints within the census tracts. A summary of the variables used in this study can be found in Table 1.

4. Research Method

4.1. Calculation of Damage Status and Ratio

Building footprints are overlayed with the DPM2 raster to determine the binary damage statuses. If the centroid of a building footprint overlaps a pixel with a pixel value between 128–255 in the DPM2 raster, the building is classified as damaged. A building is classified as undamaged if its centroid overlaps with a NoData pixel in the DPM2 raster. The damage statuses are coded as binary values (0: undamaged and 1: damaged) and are stored in the attribute table of the building points. To validate the damage classifications derived from the DPM2 data, we compared the damage statuses of 200 random buildings with high-resolution (30 cm) Airbus Earth Observation Satellite Imagery captured on 2 October 2022 (immediately after Ian’s landfall). Using spatial analysis tools, the ratios of damaged buildings to all buildings were calculated in census tracts in association with the socio-economic variables.

4.2. Building-Level Analysis

Student’s t-test is used in any case where a test is required on the significance of the difference between the means of two samples [63]. In the sample of 200 buildings, Student’s t-test was used to determine whether the means of the variables significantly differed between the damaged and undamaged buildings. Logistic regression is particularly suitable for analyzing the relationship between a binary response variable and multiple explanatory variables [64]. In this study, a logistic regression model was used to analyze the relationships between the damage status (1 = Yes and 0 = No) of buildings and eight independent explanatory variables, including wind gust speed, sustained wind speed, distance to coast, distance to hurricane track, building size, building age, roof age, house value, and building type (single family vs. mobile home). The regression coefficients indicate the effect of the variables on determining damage status.

4.3. Census Tract-Level Analysis

Pearson correlations and hierarchical regression determined the relationships among building damage, hurricane intensity, and the socio-economic variables at the census tract level. Pearson correlation analysis was used to quantify the relationships between each pair of variables. The Pearson correlation coefficients ranged from −1 to +1, where +1 indicates a perfect positive linear correlation, −1 represents a perfect negative linear correlation, and 0 indicates no linear correlation [65]. Hierarchical regression was applied to assess the potential effects of socio-economic variables on building damage under the same hurricane intensity. It allows for assessing the variance in the dependent variable explained by a set of independent variables when controlling for other variables [66]. In the hierarchical regression model, the baseline model quantifies the relationships between the damage ratios and the hurricane intensity variables, including wind speed, distance to the coast, and hurricane track. Then, each step adds a socio-economic variable to the baseline model. Then, the improvement in the model’s goodness-of-fit (R2) indicates the effect of this additional variable on the damage ratio while keeping the hurricane intensity variables constant. Finally, all the socio-economic variables were added to the baseline model to investigate the compound effect of the socio-economic variables on building damage.

4.4. Geographically Weighted Regression

While ordinary least squares (OLS) regression provides a global model for the entire study area, the GWR uses local models to capture the spatial variation between building damage and the selected variables [67]. In this study, the GWR model was first applied to examine the relationships between the damage ratios and the hurricane intensity variables. Then, the GWR model was applied to quantify the effects of the individual socio-economic variables on the damage ratios by controlling the hurricane intensity variables. The neighborhoods were selected using the “Golden Search” method. The “Golden Search” method finds the optimal bandwidth parameter that minimizes the Akaike information criterion (AIC) value in the GWR model, allowing for more accurate and localized spatial predictions [68]. The entire analytical workflow of this study is illustrated in Figure 2.

5. Results

5.1. Accuracy Assessment of Damage Data

The validation shows that the building damage statuses derived from the DPM2 data are highly consistent with the Airbus imageries. About 96% of the 200 samples have the same damage status in the two datasets, indicating the general reliability of the DPM2 data. The confusion matrix in Table 2 shows the numbers and percentages of false negatives, false positives, true negatives, and true positives.

5.2. Spatial Pattern of Building Damage

A hexagon bin map is created to visualize the spatial variation in the damage ratios (Figure 3A). Compared to choropleth maps, hexagon bin maps can eliminate the influence of varying sizes of administrative boundaries and the populations in the boundaries. Unlike other bin shapes (e.g., square bins), hexagonal bins can better reflect natural clustering patterns [69] and minimize edge effects [70]. For these reasons, hexagon bins are often preferred for depicting spatial patterns in heterogenous geographic contexts.
The hexagon bin map in Figure 3A shows that Lee County has the highest damage ratio compared to the other counties. Specifically, the coastal census tracts near the hurricane track suffered the most severe damage. Notably, most damage is concentrated in the areas on the right side of the track, which may reflect the fact that the strongest hurricane winds typically occur on the right side of the hurricane track [71,72]. Census tracts in Collier County, Hendry County, and Glades County, which are further from the coast, endured less damage, possibly due to their distance from the track and weakening hurricane intensity as Ian moved inland. The damage ratios in the census tracts in Figure 3B show a similar pattern to that in Figure 3A despite the varying polygon sizes between the inland and coastal areas. The map in Figure 3B shows the spatial distribution of the dependent variable (damage ratio) used in the subsequent census tract-level analysis.

5.3. Findings from the Building-Level Analysis

5.3.1. Student’s t-Test

The Student’s t-test results show that the means of most of the building-level variables are significantly (p < 0.05) different between the damaged and undamaged buildings (Table 3). Specifically, the damaged buildings experienced significantly higher wind gust speeds (Gust_wind) and sustained wind speeds (Sust_wind) than the undamaged buildings. The damaged buildings were closer to the coastline (Coast_Dist) and the hurricane track (Dist_Track). These differences highlight hurricane intensity as a major factor in building damage. Additionally, the damaged buildings were older than the undamaged buildings (Building_age), suggesting that the newer buildings may be more resilient to hurricanes. The damaged buildings had younger roofs (Roof_age) than the undamaged buildings, which contradicts common sense. Finally, the values of the damaged and undamaged buildings were not significantly different (House_value). The box plots in Figure 4 show the distributions of the variables between the damaged and undamaged samples.

5.3.2. Logistic Regression

Collinearity and outliers are addressed before the regression analysis. The high variance inflation factor (VIF) in Table 4 suggests the strong collinearity between wind gusts and sustained wind, which may affect the interpretability of the regression coefficients. Thus, the wind gust speed was removed from the regression. Eight outliers were also removed from the regression due to their large Cook’s distances [73].
The result of the logistic regression shows a positive correlation between sustained wind and building damage. This relationship confirms wind impact as a major cause of building damage during a hurricane. The negative coefficient of distance to the coast suggests that buildings closer to the coast are more likely to be damaged. Meanwhile, single-family residential buildings are less likely to suffer damage than mobile homes, indicating the higher vulnerability of mobile homes to hurricanes. The building age is positively related to the damage status, suggesting that older buildings are more likely to be damaged. Moreover, roof age is negatively related to damage, which is consistent with the result of the Student’s t-test. This relationship again indicates that the newer rooves were probably more likely to be damaged by this hurricane. Finally, housing value does not show a significant relationship with damage status, which aligns with the t-test result. This model’s ability to explain damage status variation is significant, indicated by a null deviance of 293.97, a residual deviance of 136.32 and an Akaike information criterion (AIC) of 154.32.

5.4. Findings from the Census Tract-Level Analysis

5.4.1. Correlation between Damage Ratios and All Predictors

The correlation among all variables in the census tracts is illustrated in Figure 5, where the divergent color scheme from red to blue indicates the direction (blue: positive; red: negative), and the color intensity represents the strength correlation (i.e., correlation coefficient). As expected, building damage is positively correlated with sustained wind speed and negatively correlated with the coast and hurricane track distances. These correlations confirm that hurricane intensity is an influential factor in building damage. However, the strongest correlations with the damage ratios are found among the socio-economic variables. For example, the damage ratio shows strong positive correlations with the percentage of elderly people and the white population, implying that the elderly and white communities have experienced more damage. In contrast, the damage ratio is negatively correlated with the percentage of children, the Hispanic/Latino population, black population, and non-English speakers, suggesting that communities with a higher ratio of children, Hispanic/Latino populations, black populations, and immigrant populations may have endured less damage.

5.4.2. Socio-Economic Factors Influencing Building Damage

The hierarchical regression aimed to quantify the effects of the socio-economic variables under the same hurricane intensity. The baseline model had a low adjusted R2 of 0.072, meaning that the hurricane intensity variables (control variables) could explain only 7.2% of the variance in the building damage ratios. Table 5 shows the adjusted R2 (∆R2) increase by adding the socio-economic variables to the baseline model. In other words, ∆R2 represents the effects of the added variables on the damage ratios with controlled hurricane intensity.
Specifically, the inclusion of the percentage of elderly people, percentage of children, and building size in the model results in notable increases in the adjusted R2, indicating strong effects of elderly people, younger population, and building size on damage under the same hurricane intensity. The positive coefficient of elderly people suggests that communities with higher ratios of elderly people may have experienced more building damage. The negative coefficients of children and Asian populations indicate that communities with higher ratios of children and Asian people suffered less damage. Additionally, the percentages of the population with a bachelor’s degree, Hispanic/Latino population, and population density show negative relationships with the damage ratios, indicating that communities with higher education attainment, higher ratios of Hispanic people, and higher population densities may be more resilient to hurricanes. In contrast, white communities suffered more damage with the same hurricane intensity.
Finally, after adding in all the socio-economic variables, the adjusted R2 increases to 0.361 compared to 0.072 in the baseline model. Such a significant increase implies that socio-economic conditions may be more important determinants of building damage than hurricane intensity. As shown in Table 6, sustained wind (Sust_wind) shows a significant positive correlation with building damage, while the proximity to the coast (Coast_Dist) has a negative correlation. This result implies that under the same socio-economic conditions, buildings in stronger winds and closer to the coast are more likely to be damaged. Additionally, building size (Building_Size) and house value (House_v) are negatively correlated with the damage ratios, suggesting larger and more expensive buildings exhibit comparatively less damage under the same hurricane intensity and socio-economic conditions. Population density (Density) also negatively affects building damage, implying that urban areas with a high population density experienced less building damage than rural areas.

5.4.3. Spatial Variation and the Impact of Physical and Socio-Economic Factors in Building Damages

The results of the GWR model exhibit notable spatial variation in the relations between the damage ratios and the independent variables. Figure 6a shows no significant relationship between sustained wind and building damage in local areas, except for a census tract at the northwest border of Charlotte County, showing a significantly positive relationship. Figure 6b reveals a significantly negative correlation between the distance to the coast and building damage in most of the study area. The negative coefficient increases from the area near the landfall location to distant areas, meaning that the distance to the coast has a stronger effect on building damage near the landfall location. This effect gradually decays when moving away from the landfall location. Figure 6c demonstrates contrasting correlations between building damage and the distance to the hurricane track between the left and right sides of the hurricane track. Specifically, on the right side of the track, building damage is negatively related to the distance to the track, meaning that areas nearer the track have experienced more damage. However, this correlation becomes positive on the left side, implying that areas farther away from the track have experienced more damage. Considering that most building damage occurred on the right side of the track (see Figure 3), the GWR models on the left side may be less reliable due to the smaller sample size of damaged buildings. Moreover, it is worth noting that the distance to the hurricane track has the strongest negative effect on building damage in the far south, while this correlation becomes insignificant (p > 0.05) nearer the track. This trend implies a potential distance threshold beyond which the proximity to the hurricane track becomes a significant factor for building damage.
Next, we conducted a GWR analysis to analyze the effects of socio-economic variables on the damage ratios while controlling for the hurricane intensity variables (control variables). Here, we included the four socio-economic variables that generated the largest increase in R2 (highest ∆R2) in the hierarchical regression (Table 5) GWRs. In Figure 7a, building size negatively affects building damage near the landfall location, indicating that larger buildings are more resistant to hurricane damage in this area. This trend becomes weaker when moving inland. Figure 7b illustrates that the size of the Asian population has the strongest negative effect on building damage around downtown Cape Coral near the coastline, indicating that communities with a higher Asian population suffered less building damage in this area. However, this trend gradually diminishes when moving further from the coast. Figure 7c shows that the elderly population has the strongest positive correlation with building damage near the landfall location, indicating the vulnerability of elderly communities to strong hurricane intensity. But this pattern does not hold strongly further inland. In contrast, the proportion of children in the population exhibits the strongest negative correlation near the landfall location (Figure 7d), implying a higher proportion of children is associated with less damage in this area. These coefficient surfaces of GWR highlight the local areas where the relationship between building damage and the socio-economic variables are most prominent and pinpoint communities where specific measures should be prioritized to reduce the damage risk.

6. Discussion

This study provides a comprehensive analysis of building damage in Hurricane Ian, a major hurricane which struck Florida in 2022. The building-level and census tract-level analyses confirm that the hurricane’s intensity (wind speed, distances to the coast, and hurricane track) were major factors influencing building damage. Moreover, the analyses showed that the building and socio-economic conditions in the communities were also strongly correlated with building damage. As expected, buildings of larger size, higher value, and more recent construction year demonstrated higher resistance to hurricane damage. However, surprisingly, newer roofs appeared to be more susceptible to damage in this hurricane. A possible reason for this finding is the substandard quality and untested wind resistance of some newly built roofs. For example, a study found that many newly built homes with complex roof structures are more likely to be damaged in hurricanes due to uneven wind pressure in certain roof areas [74]. To improve community resilience to hurricanes, stricter structural integrity testing and building code compliance should be applied in high-risk areas.
The hierarchical regression analysis reveals that the socio-economic conditions in the communities further influenced building damage. Under the same hurricane intensity, communities with higher proportions of children, black, Asian, and Hispanic/Latino people in their populations, and with higher education attainment, appeared to experience less building damage. In contrast, elderly and white communities were more susceptible to damage. These findings suggest that minority and marginalized communities may have suffered less damage than the white communities in this hurricane. This trend may be attributed to the lower concentrations of minority and marginalized communities in the most affected areas, particularly in the ocean/waterfront communities, which have less affordable housing. Although further analyses are needed to confirm the causal relations, these findings suggest potential directions for enhancing community resilience to hurricanes. At the census tract level, the hurricane intensity variables can only explain 7.2% of the variance in the building damage ratios. After including the socio-economic variables, the model can explain 36.1% of the variance in the damage ratios. Such a significant increase in the model’s performance highlights the importance of socio-economic factors in the differences in building damage. Finally, the GWR analysis pinpoints areas where the variables have the strongest effect on building damage, providing actionable information for decision-makers to implement specific risk reduction measures in targeted communities.
Despite the interesting findings mentioned above, the study has several limitations that need to be addressed in future research. The damage values in NASA’s DPM2 dataset were reclassified into binary statuses: damaged and undamaged. Future research could enhance our understanding of damage impacts by categorizing damage into finer granularities. The building-level analysis was limited to 200 building samples due to the tedious data acquisition process. Future studies could incorporate larger sample sizes with additional information on physical building attributes, such as construction materials and the existence of hurricane-resistant features, to get deeper insights into the determinants of building damage in hurricanes. Despite the improvement of R2 to 36.1% after adding the socio-economic variables, there is still a need to explore additional factors affecting building damage. Moreover, the analysis of the socio-economic conditions was conducted at the census tract level. Future research may utilize individual or household-level survey data to explore the determinants of building damage at a finer spatial scale.

7. Conclusions

This study examines the relationship between building damage and various hurricane intensity, building condition, and socio-economic variables to reveal factors contributing to building damage in Hurricane Ian. Specifically, this study utilizes NASA’s Damage Proxy Map (DPM2) to examine the spatial pattern of building damage resulting from the hurricane. It applies statistical analysis to explore the relationship between building damage and the various physical and socio-economic factors at the building and census tract levels. Additionally, the study employs geographically weighted regression (GWR) to investigate how the effect of these factors varies across different locations. The analyses offer crucial insights into the factors contributing to building damage and how these factors differ spatially. They also highlight disparities in building damage across different socio-economic and demographic groups, uncovering disparities in disaster adaptation and resilience in different communities. Furthermore, this research provides policymakers, emergency responders, and community leaders with actionable data to develop strategies to reduce the effects of future hurricanes, particularly by identifying vulnerable population groups.

Author Contributions

Lead author, data collection, processing, and analysis, M.Z.S.; advising the overall research direction, Y.Q.; manuscript writing, M.Z.S.; manuscript revision, Y.Q.; manuscript reviewing, commenting, and proofreading, B.D. and J.C. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Science Foundation (Grant No. 2325631).

Data Availability Statement

The datasets of damage in hurricane Ian analyzed during the study mainly include The Advanced Rapid Imaging and Analysis (ARIA) Damage Proxy Map (DPM2) (https://aria-share.jpl.nasa.gov/202209-Hurricane_Ian_USA/DPM/, accessed on 9 August 2024). Building footprints were sourced from the Microsoft Building Footprint database (2011–2020) (https://github.com/Microsoft/USBuildingFootprints?tab=readme-ov-file, accessed on 9 August 2024). All socio-economic indicators are available at https://data.census.gov/table (accessed on 9 August 2024). Information on the buildings of Lee County is available at https://www.leepa.org/search/propertysearch.aspx (accessed on 9 August 2024). Information on buildings of Charlotte County is available at https://www.ccappraiser.com/RPSearchEnter.asp? (accessed on 9 August 2024). Information on the buildings of Collier County is available at https://www.collierappraiser.com/ (accessed on 9 August 2024).

Acknowledgments

The authors would like to thank Dominic Del Pino for reviewing the paper.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Trenberth, K.E. Framing the way to relate climate extremes to climate change. Clim. Chang. 2012, 115, 283–290. [Google Scholar] [CrossRef]
  2. NOAA. Hurricane Costs. Available online: https://coast.noaa.gov/states/fast-facts/hurricane-costs.html (accessed on 17 July 2023).
  3. Klotzbach, P.P.J.; Bowen, S.G.; PielKe, R.G.R.; Bell, M. Continental U.S. Hurricane landfall frequency and associated damage: Observations and future risks. Bull. Am. Meteorol. Soc. 2018, 99, 1359–1376. [Google Scholar] [CrossRef]
  4. Malmstadt, J.; Scheitlin, K.; Elsner, J. Florida hurricanes and damage costs. Southeast. Geogr. 2009, 49, 108–131. [Google Scholar] [CrossRef]
  5. Larsen, L. Reimagining sustainable cities: Strategies for designing greener, healthier, more equitable communities. J. Am. Plan. Assoc. 2023, 89, 148–149. [Google Scholar] [CrossRef]
  6. NOAA. Monthly National Climate Report for September 2022. Available online: https://www.nhc.noaa.gov/data/tcr/ (accessed on 25 July 2023).
  7. Mazumder, R.K.; Enderami, S.A.; Sutley, E.J. A novel framework to study community-level social and physical impacts of hurricane-induced winds through synthetic scenario analysis. Front. Built Env. 2023, 9, 1005264. [Google Scholar] [CrossRef]
  8. Alduse, B.; Pang, W.; Tadinada, S.K.; Khan, S. A framework to model the wind-induced losses in buildings during hurricanes. Wind 2022, 2, 87–112. [Google Scholar] [CrossRef]
  9. Burton, C. Social vulnerability and hurricane impact modeling. Nat. Hazards Rev. 2010, 11, 58–68. [Google Scholar] [CrossRef]
  10. Milch, K.; Broad, K.; Orlove, B.; Meyer, R. Decision science perspectives on hurricane vulnerability: Evidence from the 2010-2012 Atlantic hurricane seasons. Atmosphere 2018, 9, 32. [Google Scholar] [CrossRef]
  11. Sutley, E.J.; Hamideh, S. Post disaster housing stages: A Markov chain approach to model sequences and duration based on social vulnerability. Risk Anal. 2020, 40, 2675–2695. [Google Scholar] [CrossRef]
  12. FEMA Building Performance: Hurricane Andrew in Florida. Available online: https://www.fema.gov/sites/default/files/2020-08/fia22_complete.pdf (accessed on 28 August 2022).
  13. Arch, G.; Hajra, B.; Moravej, M.; Zisis, I.; Irwin, P.; Chowdhury, A.G.; Suaris, W. An experimental study to assess the effect of soffit louvered vents on wind loads and wind driven rain intrusion on low rise buildings. Sustain. Cities Soc. 2017, 34, 43–55. [Google Scholar] [CrossRef]
  14. Paleo-Torres, A.; Gurley, K.; Pinelli, J.P.; Baradaranshoraka, M.; Zhao, M.; Suppasri, A.; Peng, X. Vulnerability of Florida residential structures to hurricane induced coastal flood. Eng. Struct. 2020, 220, 111004. [Google Scholar] [CrossRef]
  15. Li, Y.; van de Lindt, J.W.; Dao, T.; Bjarnadottir, S.; Ahuja, A. Loss analysis for combined wind and surge in hurricanes. Nat. Hazards Rev. 2012, 13, 1–10. [Google Scholar] [CrossRef]
  16. Mudd, L.; Wang, Y.; Letchford, C.; Rosowsky, D. Assessing climate change impact on the U.S. east coast hurricane hazard: Temperature, frequency, and track. Nat. Hazards Rev. 2014, 15, 04014001. [Google Scholar] [CrossRef]
  17. Xing, Z.; Yang, S.; Zan, X.; Dong, X.; Yao, Y.; Liu, Z.; Zhang, X. Flood vulnerability assessment of urban buildings based on integrating high-resolution remote sensing and street view images. Sustain. Cities Soc. 2023, 92, 104467. [Google Scholar] [CrossRef]
  18. Ayscue, J.K. Hurricane Damage to Residential Structures: Risk and Mitigation; The Johns Hopkins University: Baltimore, MD, USA, 1996; Available online: https://hazards.colorado.edu/research/working-papers/94 (accessed on 25 July 2022).
  19. Svenningsen, L.S.; Bay, L.; Doemgaard, M.L.; Halsnaes, K.; Larsen, M.D. Beyond the stage-damage function: Estimating the economic damage on residential buildings from storm surges. Nat. Hazards Earth Syst. Sci. Discuss. 2020. [preprint]. [Google Scholar] [CrossRef]
  20. Nofal, O.M.; Amini, K.; Padgett, J.E.; van de Lindt, J.W.; Rosenheim, N.; Darestani, Y.M.; Enderami, A.; Sutley, E.J.; Hamideh, S.; Duenas-Osorio, L. Multi-Hazard socio-physical resilience assessment of hurricane-induced hazards on coastal communities. Resilient Cities Struct. 2023, 2, 67–81. [Google Scholar] [CrossRef]
  21. Xian, S.; Feng, K.; Lin, N.; Marsooli, R.; Chavas, D.; Chen, J.; Hatzikyriakou, A. Brief communication: Rapid assessment of damaged residential buildings in the Florida keys after Hurricane Irma. Nat. Hazards Earth Syst. Sci. 2018, 18, 2041–2045. [Google Scholar] [CrossRef]
  22. van Wiechen, P.P.J.; de Vries, S.; Reniers, A.J.H.M.; Aarninkhof, S.G.J. Dune erosion during storm surges: A review of the observations, physics and modelling of the collision regime. Coast. Eng. 2023, 186, 104383. [Google Scholar] [CrossRef]
  23. Shepard, C.C.; Crain, C.M.; Beck, M.W. The protective role of coastal marshes: A systematic review and meta-analysis. PLoS ONE 2011, 6, e27374. [Google Scholar] [CrossRef]
  24. Blendon, R.J.; John Benson, S.M.; Catherine DesRoches, M.M.; Katherine Lyon-Daniel, D.; Mitchell, E.W.; Pollard, W.E. The public’s preparedness for hurricanes in four affected regions. Public. Health Rep. 2007, 122, 167–176. [Google Scholar] [CrossRef]
  25. Abdelhady, A.U.; Spence, S.M.J.; McCormick, J. Risk and fragility assessment of residential wooden buildings subject to hurricane winds. Struct. Saf. 2022, 94, 102137. [Google Scholar] [CrossRef]
  26. Masters, F.J.; Gurley, K.R.; Shah, N.; Fernandez, G. The vulnerability of residential window glass to lightweight windborne debris. Eng. Struct. 2010, 32, 911–921. [Google Scholar] [CrossRef]
  27. Gurley, K.; Masters, F. Post-2004 hurricane field survey of residential building performance. Nat. Hazards Rev. 2011, 12, 177–183. [Google Scholar] [CrossRef]
  28. Fronstin, P.; Holtmann, A.G. The determinants of residential property damage caused by Hurricane Andrew. South. Econ. J. 1994, 61, 387–397. [Google Scholar] [CrossRef]
  29. Maalej, M.; Chhoa, C.Y.; Quek, S.T. Effect of cracking, corrosion and repair on the frequency response of RC beams. Constr. Build. Mater. 2010, 24, 719–731. [Google Scholar] [CrossRef]
  30. Siegmund, S. Protecting against hurricane damage with mathematics. Not. Am. Math. Soc. 2018, 65, 967–968. [Google Scholar] [CrossRef]
  31. FBC. Florida Building Code—Existing Building, 8th ed.; International Code Council: Nappanee, Indiana, 2023; ISBN 978-1-960701-22-0. [Google Scholar]
  32. Prevatt, D.O.; Roueche, D.B.; Van De Lindt, J.W.; Pei, S.; Dao, T.; Coulbourne, W.; Graettinger, A.J.; Gupta, R.; Grau, D. Building Damage Observations and EF classifications from the Tuscaloosa, AL, and Joplin, MO, tornadoes. ASCE 2012, 35, 999–1010. [Google Scholar] [CrossRef]
  33. Pinelli, J.; Weekes, J.; Pita, G.; Hamid, S.; Gurley, K.; Subramanian, C. Vulnerability of mid-high rise commercial-residential buildings in the Florida public hurricane loss model. Environ. Eng. Sci. 2009, 1140–1149. [Google Scholar] [CrossRef]
  34. Peacock, W.G.; Van Zandt, S.; Zhang, Y.; Highfield, W.E. Inequities in long-term housing recovery after disasters. J. Am. Plan. Assoc. 2014, 80, 356–371. [Google Scholar] [CrossRef]
  35. Miljkovic, T.; Miljkovic, D. Modeling impact of hurricane damages on income distribution in the coastal U.S. Int. J. Disaster Risk Sci. 2014, 5, 265–273. [Google Scholar] [CrossRef]
  36. Cutter, S.L.; Finch, C. Temporal and spatial changes in social vulnerability to natural hazards. Proc. Natl. Acad. Sci. USA 2008, 105, 2301–2306. [Google Scholar] [CrossRef] [PubMed]
  37. Mitsova, D.; Esnard, A.M.; Sapat, A.; Lai, B.S. Socioeconomic vulnerability and electric power restoration timelines in Florida: The case of Hurricane Irma. Nat. Hazards 2018, 94, 689–709. [Google Scholar] [CrossRef]
  38. Emrich, C.T.; Tate, E.; Larson, S.E.; Zhou, Y. Measuring social equity in flood recovery funding. Environ. Hazards 2020, 19, 228–250. [Google Scholar] [CrossRef]
  39. SAMHSA Disaster Technical Assistance Center Supplemental Research Bulletin Greater Impact: How Disasters Affect People of Low Socioeconomic Status. Available online: https://www.samhsa.gov/sites/default/files/dtac/srb-low-ses_2.pdf (accessed on 17 June 2022).
  40. Zamboni, L.M.; Martin, E.G. Association of US households’ disaster preparedness with socioeconomic characteristics, composition, and region. JAMA Netw. Open 2020, 3, e206881. [Google Scholar] [CrossRef] [PubMed]
  41. Logan, J.R.; Issar, S.; Xu, Z. Trapped in place? segmented resilience to hurricanes in the Gulf coast, 1970–2005. Demography 2016, 53, 1511–1534. [Google Scholar] [CrossRef] [PubMed]
  42. Liu, T.; Fan, C. Impacts of disaster exposure on climate adaptation injustice across U.S. cities. Sustain. Cities Soc. 2023, 89, 104371. [Google Scholar] [CrossRef]
  43. Eaton, K.J. Low-income housing and hurricanes. Wind. Eng. 1980, 1, 7–21. [Google Scholar] [CrossRef]
  44. Bukvic, A.; Gohlke, J.; Borate, A.; Suggs, J. Aging in flood-prone coastal areas: Discerning the health and well-being risk for older residents. Int. J. Environ. Res. Public. Health 2018, 15, 2900. [Google Scholar] [CrossRef]
  45. Acierno, R.; Ruggiero, K.J.; Galea, S.; Resnick, H.S.; Koenen, K.; Roitzsch, J.; de Arellano, M.; Boyle, J.; Kilpatrick, D.G. Psychological sequelae resulting from the 2004 Florida hurricanes: Implications for Post disaster intervention. Am. J. Public Health 2007, 97 (Suppl. S1), S103–S108. [Google Scholar] [CrossRef]
  46. Estrada, F.; Botzen, W.J.W.; Tol, R.S.J. Economic losses from U.S. hurricanes consistent with an influence from climate change. Nat. Geosci. 2015, 8, 880–884. [Google Scholar] [CrossRef]
  47. NOAA. Costliest U.S. Tropical Cyclones Tables Update. Available online: https://www.ncei.noaa.gov/access/billions/dcmi.pdf (accessed on 17 September 2023).
  48. Karimiziarani, M.; Moradkhani, H. Social response and disaster management: Insights from twitter data assimilation on Hurricane Ian. Int. J. Disaster Risk Reduct. 2023, 95, 103865. [Google Scholar] [CrossRef]
  49. Florida Department of Law Enforcement. Florida Medical Examiners Commission Hurricane Ian Deaths. Available online: https://www.fdle.state.fl.us/News/2023/January/Update-Florida-Medical-Examiners-Commission-Hu-(2) (accessed on 10 December 2022).
  50. ARIA. Hurricane Ian USA. Available online: https://aria-share.jpl.nasa.gov/202209-Hurricane_Ian_USA/DPM/ (accessed on 13 February 2023).
  51. Kristensen, L.; Casanova, M.; Courtney, M.; Troen, I. In search of a gust Definition. Bound. Layer. Meteorol. 1991, 55, 91–107. [Google Scholar] [CrossRef]
  52. Konthesingha, K.M.C.; Stewart, M.G.; Ryan, P.; Ginger, J.; Henderson, D. Reliability based vulnerability modelling of metal-clad industrial buildings to extreme wind loading for cyclonic regions. J. Wind. Eng. Ind. Aerodyn. 2015, 147, 176–185. [Google Scholar] [CrossRef]
  53. Wills, J.A.B.; Lee, B.E.; Wyatt, T.A. A model of wind-borne debris damage. J. Wind. Eng. Ind. Aerodyn. 2002, 90, 555–565. [Google Scholar] [CrossRef]
  54. Cugnoni, H.; Whitworth, I. Injuries related to wind speed. Ann. R. Coll. Surg. Engl. 1992, 74, 294–296. [Google Scholar]
  55. NOAA. NHC Data in GIS Formats. Available online: https://www.nhc.noaa.gov/gis/ (accessed on 25 June 2023).
  56. Townsend, B.W.; Ohanehi, D.; Dillard, D.; Austin, S.R.; Salmon, F.; Gagnon, D. Developing a simple damage model for the long-term durability of acrylic foam structural glazing tape subject to sustained wind loading. J. Archit. Eng. 2012, 18, 214–222. [Google Scholar] [CrossRef]
  57. Lin, N.; Emanuel, K.; Oppenheimer, M.; Vanmarcke, E. Physically based assessment of hurricane surge threat under climate change. Nat. Clim. Chang. 2012, 2, 462–467. [Google Scholar] [CrossRef]
  58. Emanuel, K.; Ravela, S.; Vivant, E.; Risi, C. Supplement to a statistical deterministic approach to hurricane risk assessment. Bull. Am. Meteorol. Soc. 2006, 87, S1–S5. [Google Scholar] [CrossRef]
  59. Ye, M.; Wu, J.; Liu, W.; He, X.; Wang, C. Dependence of tropical cyclone damage on maximum wind speed and socioeconomic factors. Environ. Res. Lett. 2020, 15, 094061. [Google Scholar] [CrossRef]
  60. Liew, H.-P.; Eidem, N. Examining the potential impacts of social vulnerability on damage levels in areas affected by Hurricane Harvey. J. Homel. Secur. Emerg. Manag. 2022, 19, 51–66. [Google Scholar] [CrossRef]
  61. Spekkers, M.H.; Kok, M.; Clemens, F.H.L.R.; Ten Veldhuis, J.A.E. Decision-tree analysis of factors influencing rainfall-related building structure and content damage. Nat. Hazards Earth Syst. Sci. 2014, 14, 2531–2547. [Google Scholar] [CrossRef]
  62. U.S. Census Bureau. American Community Survey, ACS 5-Year Estimates Subject Tables. Available online: https://www.census.gov/programs-surveys/acs (accessed on 5 March 2023).
  63. Katz, R.W. Stochastic modeling of hurricane damage. J. Appl. Meteorol. Clim. 2002, 41, 754–762. [Google Scholar] [CrossRef]
  64. Mitsova, D.; Escaleras, M.; Sapat, A.; Esnard, A.M.; Lamadrid, A.J. The effects of infrastructure service disruptions and socio-economic vulnerability on hurricane recovery. Sustainability 2019, 11, 516. [Google Scholar] [CrossRef]
  65. Benesty, J.; Chen, J.; Huang, Y. On the Importance of the Pearson Correlation Coefficient in Noise Reduction. IEEE Trans. Audio Speech Lang. Process 2008, 16, 757–765. [Google Scholar] [CrossRef]
  66. Cohen, J.; Cohen, P.; West, S.G.; Aiken, L.S. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, 3rd ed.; Routledge: London, UK, 2023. [Google Scholar] [CrossRef]
  67. Fotheringham, A.S.; Charlton, M.E. Geographically weighted regression: A natural evolution of the expansion method for spatial data analysis. Environ. Plan. A Econ. Space 1998, 30, 1905–1927. [Google Scholar] [CrossRef]
  68. Fotheringham, A.S.; Brunsdon, C.; Charlton, M. Geographically Weighted Regression: The Analysis of Spatially Varying Relationships; Wiley: Hoboken, NJ, USA, 2002; ISBN 978-0-471-49616-8. [Google Scholar]
  69. Birch, C.P.D.; Oom, S.P.; Beecham, J.A. Rectangular and hexagonal grids used for observation, experiment and simulation in ecology. Ecol. Model. 2007, 206, 347–359. [Google Scholar] [CrossRef]
  70. Carr, D.; Olsen, A.; White, D. Hexagon mosaic maps for display of univariate and bivariate geographical data. Cartogr. Geogr. Inf. Syst. 2013, 19, 228–236. [Google Scholar] [CrossRef]
  71. Boose, E.R.; Foster, D.R.; Fluet, M. Hurricane impacts to tropical and temperate forest landscapes. Ecol. Monogr. 1994, 64, 369–400. [Google Scholar] [CrossRef]
  72. Coch, N.K. Hurricane Hazards along the Northeastern Atlantic coast of the United States. J. Coast. Res. 1994, 115–147. Available online: http://www.jstor.org/stable/25735594 (accessed on 10 August 2023).
  73. Sun, H.; Cui, Y.; Wang, H.; Liu, H.; Wang, T. Comparison of methods for the detection of outliers and associated biomarkers in mislabeled omics data. BMC Bioinform. 2020, 21, 357. [Google Scholar] [CrossRef]
  74. Meloy, N.; Sen, R.; Pai, N.; Mullins, G. Roof Damage in New Homes Caused by Hurricane Charley. J. Perform. Constr. Facil. 2007, 21, 97–107. [Google Scholar] [CrossRef]
Remotesensing 16 03792 g001
Figure 1. Study area.
Figure 1. Study area.
Remotesensing 16 03792 g001
Remotesensing 16 03792 g002
Figure 2. The analytical workflow.
Figure 2. The analytical workflow.
Remotesensing 16 03792 g002
Remotesensing 16 03792 g003
Figure 3. Spatial pattern of damage ratios in census tracts.
Figure 3. Spatial pattern of damage ratios in census tracts.
Remotesensing 16 03792 g003
Remotesensing 16 03792 g004
Figure 4. Comparative box plot of standardized factors affecting building damage.
Figure 4. Comparative box plot of standardized factors affecting building damage.
Remotesensing 16 03792 g004
Remotesensing 16 03792 g005
Figure 5. Correlation between damage ratios and all variables.
Figure 5. Correlation between damage ratios and all variables.
Remotesensing 16 03792 g005
Remotesensing 16 03792 g006
Figure 6. Coefficient surfaces of the three hurricane intensity variables: (a) sustained wind speed, (b) distance to the coast, and (c) distance to hurricane track in the GWR models. The colored polygons are census tracts that show a significant relationship (p < 0.05) in the neighborhoods. The color intensity represents the regression coefficient.
Figure 6. Coefficient surfaces of the three hurricane intensity variables: (a) sustained wind speed, (b) distance to the coast, and (c) distance to hurricane track in the GWR models. The colored polygons are census tracts that show a significant relationship (p < 0.05) in the neighborhoods. The color intensity represents the regression coefficient.
Remotesensing 16 03792 g006
Remotesensing 16 03792 g007
Figure 7. Coefficient surfaces of four independent socio-economic variables: (a) Building_Size; (b) Asian population; (c) elderly population; and (d) children. The colored polygons are census tracts that show a significant relationship (p < 0.05) in the neighborhoods. The color intensity represents the regression coefficient.
Figure 7. Coefficient surfaces of four independent socio-economic variables: (a) Building_Size; (b) Asian population; (c) elderly population; and (d) children. The colored polygons are census tracts that show a significant relationship (p < 0.05) in the neighborhoods. The color intensity represents the regression coefficient.
Remotesensing 16 03792 g007
Table 1. Data type and sources of variables used in this study.
Table 1. Data type and sources of variables used in this study.
ThemeVariableVariable Abbr.Source
Dependent
Variable		Building
Damage		Damage RatioDRNASA
Independent variableHurricane IntensitySpeed of Wind Gust (miles per hour)Gust_windNOAA
Speed of Sustained Wind (miles per hour) Sust_wind
Distance to Hurricane Track (meter)Dist_Track
Distance to Coast (meter)Coast_DistU.S. Census
Building ConditionsAge of Building (year)Building_ageCounty Property Appraiser
Age of Roof (year)Roof_age
Building TypeBuilding_Type
Average Building Size (square meters)Building_SizeMicrosoft Building Footprint (2011–2020)
Socio-economicAverage House Value
(dollars)		House_vU.S. Census (2022)
Number of Population with Bachelor’s DegreeBach
Household Average Income (dollars)Income
Percentage of Owner-occupied Homes (%)Own_h
Unemployment Rate (%)Unemployment
Percentage of Individuals with Income under Poverty Level (%)Poverty
Population Density (1000/sqrt. km)Density
Median Rent Paid by Households (dollars)Rent
Population with Limited English SpeakerN_EngU.S. Census (2022)
DemographicPercentage of Hispanic/Latino Population (%)Hispanic
Percentage of White Population (%)White
Percentage of Black Population (%)Black
Percentage of Asian Population (%)Asian
Percentage of Population aged 65 and Above (%)Elderly
Percentage of Population under 5 Years Old (%)Children
Table 2. Confusion matrix for accuracy assessment of damage classification.
Table 2. Confusion matrix for accuracy assessment of damage classification.
Damage Status
Sample = 200 PositiveNegative
Airbus Imagery
(Validation Data)		Positive97 (48.50%)5 (2.50%)
Negative3 (1.50%)95 (47.50%)
Table 3. Student’s t-test results (larger mean values are highlighted with a grey background in the table).
Table 3. Student’s t-test results (larger mean values are highlighted with a grey background in the table).
VariablesMean of Damaged BuildingsMean of Undamaged Buildingsp-Value
Gust_wind (mph)55.740.4<0.001
Sust_wind (mph)38.928.2<0.001
Coast_Dist (m)10,975.214,401.80.020
Dist_Track (m)14,102.426,879.3<0.001
Building_Size (sq.m.)230.7305.6<0.001
Building_age (year)3527<0.001
Roof_age (year)1923<0.001
House_value (dollars)763,939901,8690.470
Table 4. Summary of logistic regression results (bold font: p < 0.05).
Table 4. Summary of logistic regression results (bold font: p < 0.05).
Independent VariableVIFCoefficientp-Value
Gust_wind (removed due to collinearity)17.34N/AN/A
Sust_wind17.111.628<0.001
Coast_Dist1.83−2.363<0.001
Dist_Track2.11−0.4200.138
Building_Size1.310.1750.501
Building_type: Single-Family Residential1.54−5.785<0.001
Building_age1.550.9210.016
Roof_age1.58−0.6810.049
House_value1.230.0180.945
Table 5. Comparison between the baseline model and other models (bold font: p < 0.05).
Table 5. Comparison between the baseline model and other models (bold font: p < 0.05).
ModelCoefficientAdjusted R2∆R2p-Value
Baseline modelIn Table 60.072 <0.001
Baseline model + Children−0.3690.1670.10<0.001
Baseline model + Elderly0.3380.1640.09<0.001
Baseline model + Asian−0.2910.1480.08<0.001
Baseline model + Building_Size−0.2790.1370.07<0.001
Baseline model + Bach−0.2120.1120.04<0.001
Baseline model + Density−0.2100.1110.04<0.001
Baseline model + White0.2010.1020.030.001
Baseline model + Hispanic−0.1930.0950.030.005
Baseline model + Black−0.1430.0870.020.019
Baseline model + House_v−0.1390.0830.010.037
Baseline model + Income−0.1260.0820.010.048
Baseline model + Own_h−0.1060.0810.010.061
Baseline model + Rent−0.1040.0780.010.099
Baseline model + Poverty0.0730.0740.0020.245
Baseline model + N_Eng−0.0880.0760.0040.150
Baseline model + Unemployment−0.0010.069<0.0010.984
Baseline model + all other variablesIn Table 60.3610.29<0.001
Table 6. Coefficients of variables in hierarchical regression model at census tract level (bold font: p < 0.05).
Table 6. Coefficients of variables in hierarchical regression model at census tract level (bold font: p < 0.05).
VariablesCoefficientStd. Errorp-Value
Sust_wind0.3000.060<0.001
Coast_Dist−0.3350.068<0.001
Dist_Track0.0710.0610.248
Building_Size−0.2220.0640.001
House_v−0.2210.1090.043
Bach−0.1820.1130.109
Income−0.1700.1300.192
Own_h0.0610.1070.568
Unemployment−0.0890.0550.103
Poverty0.1340.0730.066
Density−0.1640.0600.006
Rent0.0580.0710.417
N_Eng0.0420.0850.619
Hispanic0.0200.1780.912
White0.2480.1780.165
Black0.0670.1200.575
Asian−0.0870.0620.156
Elderly0.2870.1560.067
Children−0.1370.1900.469
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Salim, M.Z.; Qiang, Y.; Dixon, B.; Collins, J. A Disparate Disaster: Spatial Patterns of Building Damage Caused by Hurricane Ian and Associated Socio-Economic Factors. Remote Sens. 2024, 16, 3792. https://doi.org/10.3390/rs16203792

AMA Style

Salim MZ, Qiang Y, Dixon B, Collins J. A Disparate Disaster: Spatial Patterns of Building Damage Caused by Hurricane Ian and Associated Socio-Economic Factors. Remote Sensing. 2024; 16(20):3792. https://doi.org/10.3390/rs16203792

Chicago/Turabian Style

Salim, Md Zakaria, Yi Qiang, Barnali Dixon, and Jennifer Collins. 2024. "A Disparate Disaster: Spatial Patterns of Building Damage Caused by Hurricane Ian and Associated Socio-Economic Factors" Remote Sensing 16, no. 20: 3792. https://doi.org/10.3390/rs16203792

APA Style

Salim, M. Z., Qiang, Y., Dixon, B., & Collins, J. (2024). A Disparate Disaster: Spatial Patterns of Building Damage Caused by Hurricane Ian and Associated Socio-Economic Factors. Remote Sensing, 16(20), 3792. https://doi.org/10.3390/rs16203792

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop