Cumulative catchments: A novel approach for catchment-based regional flood susceptibility

Abstract

Abstract Catchments are a fundamental unit of hydrological and geomorphological systems. Although floods occur in rivers, they also result from processes within catchments. For this reason, catchment morphometry and parameters such as climatic, soil, and hydrological are widely used to define flood susceptibility in regions where data is scarce. However, floods mostly occur at inter-catchments. An inter-catchment is a transitional area between adjacent drainage basins where water flow is not confined to a single channel and may contribute to multiple basins. However, morphometric calculations often neglect spatial gradients between sub-catchments and inter-catchments in assessing flood susceptibility. In this study, we therefore developed the cumulative catchment approach, which delineates small sub-catchments that reflect the upstream drainage contribution while not neglecting the spatial gradients within sub-catchments. Based on this approach, many morphometric parameters have been calculated. Additionally, parameters related to climatic and hydrological conditions, as well as land-use/soil types, have been assigned to the catchments using zonal statistics. Although the aim of the study was not to compare machine learning methods, the models that performed best in flood prediction were compared, and flood susceptibility was obtained on a cumulative catchment basis using machine learning. The algorithms such as XGBoost, random forest, logistic regression, and support vector machine were used, and XGBoost was determined to be the best model for defining flood susceptibility based on the ROC curve using the inventory belong to FlooDOT (FlooD invetory Of Türkiye) dataset. Furthermore, a bivariate map was constructed between model parameters and susceptibility values to understand the impact of covariates. To the best of our knowledge, this study provides the pioneer flood susceptibility analysis regionally for Türkiye, based on cumulative catchments. Model results present flood probabilities in a more realistic and high-resolution manner, based on the cumulative characteristics of the catchments, in the form of sub-catchments (small catchments). This approach also offers an opportunity in terms of regional applicability, particularly in data-scarce areas.