Uncertainties in discharge predictions based on microwave link rainfall estimates in a small urban catchment

Abstract

A wide range of applications in the field of urban hydrology requires rainfall data with high spatio-temporal resolutions. Commercial microwave links (CMLs) densely cover urban areas and can provide high-resolution path-averaged quantitative precipitation estimates (QPEs). This study aims to reduce systematic errors in CML QPEs using rainfall and discharge observations commonly available in urban areas and to assess the potential of such precipitation estimates for discharge predictions in small urban catchments. CML QPEs are optimized using flow data observed at the catchment outlet and using hourly rain rates from rain gauges located at different distances from the catchment. Both optimized CML QPEs and traditional rain gauge data are propagated through a rainfall-runoff model and evaluated against observed discharges. To quantify uncertainties of runoff predictions, the deterministic hydrodynamic model is extended by a stochastic error model explicitly accounting for model bias. Resulting runoff prediction intervals, namely their width and reliability, show that optimized CML QPEs predict discharges only slightly worse than those based on benchmark rain gauge data (1 gauge / 0.5–1 km2), especially for rainfall events with high spatial variability. Unbiased CML QPEs obtained in this study represent high-quality rainfall data suitable for many urban hydrology tasks, including assessment and real-time control of urban stormwater systems.