Spatial heterogeneity in social vulnerability to flood exposure

Indicators of social vulnerability are frequently analyzed using methods that assume spatial stationarity, meaning that the relationships between these indicators and the outcome of interest are presumed to remain consistent across space. However, this assumption can obscure important variations if spatial heterogeneity is present, that is if the relationships between social vulnerability indicators and the outcome vary across different geographic locations. Failing to account for spatial heterogeneity may lead to mischaracterizations of where socially vulnerable populations are most at risk when assessing vulnerability to specific hazards like flood exposure. This study investigates the spatial dynamics of the relationships between social vulnerability indicators and flood exposure. First, a systematic literature review assesses whether spatial heterogeneity is evident in existing studies. Next, using Texas as a case study, we apply Multiscale Geographically Weighted Regression (MGWR) to directly assess spatial heterogeneity and then compare these results to those from Ordinary Least Squares (OLS) regression, which assumes spatial stationarity. The systematic literature review highlights significant variability across study findings, with no indicator consistently demonstrating the same relationship with flood exposure. In the MGWR, we find that only six of twenty indicators exhibit stationary relationships to flood exposure, while the majority demonstrate spatial heterogeneity, with localized variations in strength, direction, and significance. Only four indicators show complete consistency between OLS and MGWR, underscoring how accounting for spatial heterogeneity unveils critical localized patterns masked by the assumption of spatial stationarity. These findings highlight the importance of spatially nuanced approaches for assessing social vulnerability.