Summer season convection inhibition and soil moisture memory in km-scale climate simulations

Soil moisture-atmosphere feedback plays an important role in shaping summer temperature variability and eventually in modulating duration, intensity, and predictability of heat waves. 

Recent studies point out a modulation of summer temperatures introduced by the new generation of km-scale (or convection-permitting, CP) regional climate models (RCMs), compared to convection-parameterized RCMs. Modifications are likely originated from changes in soil moisture-precipitation feedback. This generally turns into an extension of dry spell length (DSL) determining warmer conditions in response to an altered partitioning of surface heat fluxes.

In this study, two potentially relevant factors behind modifications in land-atmosphere interactions at the two resolutions are investigated on a seasonal temporal scale. The first, is the underestimation of summer season convective phenomena, as an outcome of a poor sensitivity to triggering factors and driving longer DSL in km-scale simulations. The second is represented by differences in soil moisture memory between RCM and CPRCM.

We perform simulations with the ECMWF-ERA5 driven WRF-4.2.1 model consisting of a two-step dynamical downscaling at ~15 km (non-CP scale) and ~ 3km (CP scale) respectively. The greater alpine region and extended summer seasons (May to September) represent spatial and temporal domains.

The underestimation of the summer season convection will be explored considering simulations at (i) 15 km with parameterized convection, (ii) 3 km with explicit convection (CPRCM_exp) and (iii) 3 km with parameterized convection (CPRCM_par) according to different numerical schemes. This is to explore whether parameterizing deep convection at CP scale mitigates poor convection-triggering processes sensitivity caused by weak large scale forcing.

The soil moisture memory is assessed through autocorrelation analysis applied to three simulations: one standard and two idealized soil-moisture-perturbed-initialization defining anomalously dry- and wet-initialization experiments. Here, ground-water-aware configuration of Noah-MP land surface model (LSM) will be compared to a more simplified LSM configuration.

Preliminary results for the 2003 summer season show differences in precipitation statistics between the two different resolutions and between CPRCM_exp/CPRCM_par. We observe an increase in wet-hour frequency, an increase and different spatial pattern of precipitation 99th percentile in CPRCM_par.

Concerning soil moisture memory, initial differences are preserved in the two resolutions for the first month run. After that, differences decay in the CPRCM, where all the three simulations converge to similar soil moisture at the end of the run. Differently, RCM preserves a larger difference of soil moisture until the end of the run indicating a longer memory of the initial state, particularly in the wet-initialization experiment.

Several reference products will be considered to evaluate resulting modulations, namely if a km-scale deep convection parameterization can be beneficial and if the shorter soil moisture memory resulting in the km-scale simulations represents an improvement.

Outcomes might offer insights on the km-scale influence on seasonal time-scale prediction of soil-atmosphere-interaction-driven extreme events.