LAUSR

Abstract This study assesses the impact of an improved soil moisture (SM) initialization using direct insertion methodology for convection-resolving modelling of heavy precipitation events (HPEs). State-of-the-art 1 km SM data from the Soil Moisture and Ocean Salinity (SMOS) mission, SMOS-BEC L4 version 3 are used for this purpose. A strategy is developed to prepare the SMOS-L4 surface soil moisture (SSM) product for the COnsortium for Small-scale MOdelling (COSMO) model initialization by applying a cumulative density function (CDF)-matching bias-correction and the exponential filter method to calculate corresponding SM profiles (L4-Expo). The processed satellite-derived product is validated with 38 observing sites from three in-situ SM networks REMEDHUS (19), SMOSMANIA (11) and VAS (8). All networks measure at a soil depth of 5 cm, only at the SMOSMANIA network additional measurements at 10, 20 and 30 cm are available. Four HPEs are selected to evaluate the impact of the high-resolution realistic initialization. The results show a high agreement index (AI = 0.91) and a low root-mean-square deviation (0.03 m3/m3) of the high-resolution, bias-corrected SMOS-L4 SSM product compared to in-situ observations. Moreover, the derived L4-Expo SM profile field agrees with ground-based observations and successfully removes the wet bias of the original COSMO simulated SM profile. Enhanced SM initialization with the SMOS-L4 SM derived profiles improves precipitation modelling in all selected HPEs as a consequence of improved near-surface 2 m-temperature and induced changes in the pressure field (−0.5 hPa), atmospheric humidity distribution (dqs = 5–15%) as well as wind circulations (dw700hPa = 25%), thus convergence/divergence fields. The sensitivity study, applying the same methodology for SM initialization for a SMOS-L3 (~25 km), shows a weaker improvement of the precipitation forecast of an analysed HPE than the 1 km SMOS-L4 product. Our results highlight the benefit of high-resolution SSM remote sensing satellite data for scientific disciplines like meteorology in overcoming present limitations such as the uncertainty associated to SM initialization in models.