Water level prediction in urban drainage systems using explainable deep learning models

Abstract

Accurate water level prediction in existing urban drainage systems (UDSs) is critical for reliable forecasting of pluvial flooding impacts and reduction of flood-related damages in cities. Conventional physically based urban drainage modelling approaches are constrained by the need for extensive hydro-meteorological, drainage network and surface terrain data and high computational demands. In this research, more computationally efficient Machine Learning based Feedforward Neural Network (FFNN), multi-head Convolutional Neural Network (CNN) and 1D-CNN models were developed and applied to simulate water levels at a bridge crossing downstream of an existing UDS in Kampala City. The study results suggested that the multi-head CNN Deep Learning model resulted in more superior predictive performance (NSE, RMSE and MAE of 0.564, 0.208, and 0.091) when compared to the physically based PCSWMM model (NSE, RMSE, and MAE of 0.505, 0.221 and 0.098). Furthermore, the SHapley Additive exPlanations (SHAP) approach was applied to explain the underlying processes in the developed ML models and to determine the most influential model parameters. The research demonstrates that explainable Deep Learning models can reliably simulate water levels in UDSs, and provide a robust basis for development of real-time pluvial flood early warning systems in data-scarce cities.