Satellite-Based Flood Resilience Assessment in Data-Scarce Regions: Application to the Downstream Amnok, Taedong, and Tumen River Basins in North Korea

Abstract

Flood hazards are intensifying under climate-driven hydro-meteorological extremes, increasing the need for quantitative assessment of regional flood resilience, especially in areas where ground observations and disaster statistics are unavailable. North Korea represents an example of such data-scarce and inaccessible regions, where repeated large floods along major river basins are reported but their spatial impacts and recovery processes remain poorly quantified. In this context, satellite remote sensing and global geospatial datasets offer a promising alternative for evaluating flood resilience without reliance on in situ measurements. This study develops a satellite-based framework to quantify flood resilience across the downstream basins of North Korea’s three principal rivers: the Amnok, Taedong, and Tumen. Flood vulnerability is estimated by transferring a logistic-regression model trained on South Korean flood records, using topographic, land-surface, climatic, and hydrological predictors. Flood impact is derived as a composite index that integrates population exposure from WorldPop with land-cover–based infrastructure weighting. Recovery is quantified from Sentinel-1 Synthetic Aperture Radar (SAR) time series by detecting the time required for backscatter anomalies to return to a pre-flood baseline. These three components—vulnerability, impact, and recovery—are then integrated into a pixel-level resilience index designed to emphasize the role of recovery where both hazard susceptibility and impact potential are high. The results reveal distinct resilience regimes among the three basins. The Amnok basin exhibits moderate resilience with marked spatial variability driven by delayed recovery in downstream agricultural lowlands. The Taedong basin shows high resilience, as rapid and spatially coherent recovery compensates for structurally high vulnerability and impact in dense urban floodplains. The Tumen basin maintains generally high resilience owing to low hazard and exposure levels, with localized reductions in mountainous agricultural zones where recovery is slower. These findings demonstrate that recovery is the principal driver of resilience contrasts across basins and confirm that the proposed V–I–R framework provides a practical and scalable approach for assessing flood resilience in data-limited regions using satellite-derived indicators alone.