Abstract. Thunderstorm forecasting remains challenging despite advances in numerical weather prediction (NWP) systems. The microphysics scheme that represents clouds in the model partly contributes to the introduction of uncertainties in the simulations. To better understand the discrepancies, synthetic radar data simulated by a radar forward operator (applied to model outputs) are usually compared to dual-polarization radar observations, as they provide insight into the microphysical structure of clouds. However, the modeling of polarimetric values and radar signatures such as the ZDR column (ZDRC) remains a complex issue, despite the diversity of microphysics schemes and forward operators, especially above the freezing level where values that are too low are often found. The aim of this work is to assess the ability of the AROME NWP convective model, when coupled with two distinct microphysics schemes (ICE3 one-moment and LIMA partially two-moment; LIMA: Liquid Ice Multiple Aerosols), to accurately reproduce thunderstorm characteristics. A statistical evaluation is conducted on 34 convective days of 2022 using both a global and an object-oriented approach, and a ZDRC detection algorithm is implemented. Simulations performed with LIMA microphysics showed good agreement with observed ZH, ZDR, and KDP below the melting layer in convective cores. Moreover, it demonstrated a remarkable capacity to generate a realistic number of ZDRCs, as well as a distribution of (1) the ZDRC area and (2) the first ZDRC occurrence, very close to the observations. Enhancements in the forward operator have also been suggested to improve the simulations in the mixed phase and cold phase regions. These findings are highly encouraging in the context of data assimilation, where one could leverage the combination of advanced microphysics schemes and improved forward operators to improve storm forecasts.
David et al. (Mon,) studied this question.