Conference Agenda

Commercial microwave links (CMLs) have recently shown great potential in urban drainage modelling due to their ability to provide rainfall-runoff dynamics. Previous studies typically used mechanistic hydrodynamic models driven by quantitative precipitation estimates (QPEs) derived from CML attenuation data. Naturally, some errors related to CML rainfall retrieval are introduced, including uncertainties in wet antenna attention correction and path-averaged QPEs, which result in a loss of valuable information in raw data. To address these issues, we employ raw CML attenuation data without QPE derivation into a Random Forest (RF) model to estimate urban runoff. We find that: (1) RF integrates mean values from CMLs and rain gauges achieved an average NSE of 0.80, PCC of 0.90, RMSE of 38 L/s, and RE of 0.01 across all cross-validation iterations. (2) RF model using raw CML data (RF-CML) yields results comparable to the SWMM model benchmark, especially in high-value discharges. (3) RF-CML enables runoff prediction beyond the catchment's rainfall-runoff concentration time, maintaining PCC > 0.82 for forecasts up to 30 minutes. These results demonstrate that CML raw data can accurately yield runoff dynamics and volumes, and it can be used for short-term runoff predictions.