The Effects of Space–Time Dynamics of Precipitation on the Shape and Timing of Streamflow Event Hydrographs

To decipher the effects of space–time dynamics of precipitation on the resulting streamflow hydrographs, we herein analyse the controls of timing and shape of the 85 863 hourly streamflow events observed in 180 small German catchments. Using rainfall radar observations, spatially distributed snowmelt, soil moisture data and landscape properties we derive a comprehensive set of potential dynamic controls that apart from standard catchment- and event-averaged precipitation and wetness (i.e., lumped) characteristics represent: the space–time structure and the location of precipitation events within catchments; interaction of precipitation with surface (land use) and subsurface (soil) properties; and interaction of precipitation with antecedent wetness conditions. Interpretable machine learning based on random forest and accumulated local effects shows that among considered spatially and temporally differentiated controls, particularly the characteristics describing the location of precipitation events relative to catchment outlet and stream network, as well as the interaction of precipitation with the dynamic soil moisture and static soil characteristics have a strong effect on the timing of hydrographs. Instead, spatial and temporal structure (i.e., its uniformity or variability in space and time) affects their shapes. We also find that lumped precipitation and wetness characteristics are less relevant for large streamflow events (i.e., magnitudes larger than the 95th percentile). Instead, the space–time interaction of precipitation events with antecedent soil moisture is crucial for accurately predicting the timing and shape of large events. Their importance highlights the need to account for these aspects to improve the accuracy of flood simulations.