Configuration dependence of atmospheric response to soil moisture perturbations in WRF simulations

This study investigates simulated atmospheric and precipitation sensitivity to soil moisture perturbations using the Weather Research and Forecasting (WRF) model over the Carpathian Basin during the drought-affected 2022 warm season. The anomalously dry soil conditions were replaced with a wetter soil moisture state representative of near-normal conditions (~33% increase over the Carpathian Basin lowlands), while atmospheric forcing was kept unchanged across configurations differing in horizontal resolution, convection representation, and spectral nudging. Enhanced soil moisture led to consistent thermodynamic responses across all configurations: higher latent heat flux, cooler near-surface temperatures (0.8–2.0 °C), and increased lower-tropospheric moisture (dew-point temperature increases of 1.7–2.6 °C). Conversely, precipitation responses depended strongly on model configuration. Simulations with parameterized convection exhibited the largest precipitation increases, primarily in frequency, whereas explicit convection produced smaller and more heterogeneous responses in both frequency and intensity. A two-legged sensitivity framework quantifies this difference: the terrestrial leg (soil moisture to instability) is relatively robust across freely evolving configurations, while the atmospheric leg (instability to precipitation) varies substantially, with the convection-permitting configuration showing the most balanced land–atmosphere coupling. Spectral nudging introduces a distinct coupling response, leading to diverging responses in the atmospheric leg between parameterized and explicit convection, in contrast to their similar sensitivity in freely evolving simulations. While soil moisture and instability are spatially correlated (r ≈ 0.6), the spatial relationship between instability and precipitation is weaker and configuration-dependent (r ≈ −0.3 to 0.1). These findings indicate that thermodynamic adjustments are robust, whereas precipitation responses depend strongly on model formulation, highlighting significant uncertainty in simulated land–atmosphere feedbacks.