Susceptibility modeling of hydro-morphological processes considered river topology

Hydro-morphological processes (HMP, any natural phenomenon contained within the spectrum defined between debris flows and flash floods) are most likely to occur in small catchments, especially buffer zones along or near rivers. However, previous studies on HMP prediction lacked consideration of the physical interactions between catchments, resulting in insufficient predictive and explanatory capabilities of the models. In this work, we fully considered the role played by river topology and developed a Topology-based HMP susceptibility Modeling (Topo-HMPSM) to simulate the dynamic interactions between catchments and predict the susceptibility of HMPs for the Yangtze River Basin during 1985-2015. Results confirmed that our proposed model outperforms four selected baseline models (RF, GBDT, GRU, and LSTM) with the best F1-score and relatively lower uncertainties. This work is a new attempt to incorporate physical mechanisms into deep learning models. A graph-based deep neural network improves the predictive and interpretability of HMP susceptibility modeling using embedding learning techniques. Our findings highlight the consideration of river topology to predict HMP to support hazard mitigation.