Machine learning for postprocessing ensemble streamflow forecasts

Abstract Skillful streamflow forecasts can inform decisions in various areas of water policy and management. We integrate numerical weather prediction ensembles, distributed hydrological model, machine learning to generate ensemble at medium-range lead times (1–7 days). demonstrate the application as postprocessor for improving quality forecasts. Our results show that improve relative low-complexity (e.g., climatological temporal persistence) well standalone hydrometeorological modeling neural network. The gain forecast skill from is generally higher timescales compared shorter times; high flows low–moderate flows, warm season cool ones. Overall, our highlight benefits many aspects both reliability