Abstract. Machine learning provides a great potential to model hydrological variables at a spatial resolution beyond the capabilities of traditional physically-based modelling. This study features an application of Random Forests (RF) to model the depth to the shallow water table, for a wintertime minimum event, at 50 m resolution over a 15,000 km2 large domain in Denmark. In Denmark, the shallow groundwater poses severe risks of groundwater induced flood events affecting both, urban and agricultural areas. The risk is especially critical in wintertime, when the shallow groundwater is close to terrain. In order to advance modelling capabilities of the shallow groundwater system and to provide estimates at scales required for decision making, this study introduces a simple method to unify RF and physically-based modelling. Results from the national water resources model in Denmark (DK-model) at 500 m resolution are employed as covariate in the RF model. Thereby, RF ensures physical consistency at coarse scale and fully exhausts high-resolution information from readily available environmental variables. The vertical distance to the nearest waterbody was rated the most important covariate in the trained RF model followed by the DK-model. The validation test of the trained RF model was very satisfying with a mean absolute error of 79 cm and a coefficient of determination of 0.55. The resulting map underlines the severity of groundwater flooding risk in Denmark, as the average depth to the shallow groundwater is 1.9 m and approximately 29 % of the area is characterised with a depth less than 1 m during a typical wintertime minimum event. This study brings forward a novel method to assess the spatial patterns of covariate importance of the RF predictions which contributes to an increased interpretability of the RF model. Quantifying uncertainty of RF models is still rare for hydrological applications. Two approaches, namely Random Forests Regression Kriging (RFRK) and Quantile Regression Forests (QRF) were tested to estimate uncertainties related to the predicted groundwater levels. This study argues that the uncertainty sources captured by RFRK and QRF can be considered independent and hence, they can be combined to a total variance through simple uncertainty propagation.
Physically‐Based Modelling of the Post‐Fire Runoff Response of a Forest Catchment in Central Portugal: Using Field versus Remote Sensing Based Estimates of Vegetation Recovery