Geospatial analysis of Flood susceptibility through a combined bivariate statistical model and Sentinel-1 SAR data in Qarqan River basin, China

<h2>Abstract</h2> Floods pose significant threats due to their escalating impacts on infrastructure and socio-economic sustainability. Therefore, flood susceptibility mapping is crucial for informed flood risk management. The Qarqan River Basin, prone to recurrent flood events that disrupt local livelihoods and infrastructure, presents a unique challenge due to the lack of measured data and studies, highlighting the need for comprehensive flood susceptibility assessments in the region. This study maps flood susceptibility in the Qarqan River Basin using Sentinel-1 SAR data and a Bivariate Statistical Model. Based on historical flood timing reports, multi-temporal Sentinel-1 SAR data (2017–2022) were processed in GEE to extract flooded pixels using Otsu thresholding, demonstrated to have high accuracy. A fishnet analysis was applied to derive flood inventories, which were partitioned into a 75 % training set for developing the Frequency Ratio model and a 25 % validation set for assessing the accuracy of the resulting flood susceptibility map. A Frequency Ratio model was developed using 13 conditioning factors, with assigned weights using bivariate statistical analysis. Results show that the southern plain has more extensive flood-susceptible areas than the northern plain of The Qarqan River Basin. Drainage density was identified as the most influential predictor with the highest average weight, followed by Curvature, HAND and Land Cover Land Use (LCLU). The flood susceptibility map, categorized into five classes, reveals 6 % and 22 % of the basin as very high susceptibility and high susceptibility, respectively, covering primarily low-lying areas, while moderate susceptibility is predominant (35 %), with low (29 %) and very low susceptibility (8 %) observed in elevated or well-drained regions. Spatial pattern analysis shows a clustered pattern, and the AUC validation yields 0.804, indicating approximately 80 % accuracy. These findings will contribute to flood risk assessment in the Qarqan River Basin and areas of similar geography, supporting informed decision-making by local authorities and policymakers.