Large-eddy simulations of land-atmosphere interactions and mid-latitude storms: from conceptual models to realistic cases

An ability to predict the weather depends on an ability to explicitly represent atmospheric processes in Numerical Weather Prediction (NWP) models. Limited computational resources still force many weather operational centers to perform simulations with a grid spacing of O(10 km), thus requiring to parametrize many subgrid-scale processes, and in particular moist convection. On the other hand, Large Eddy Simulation (LES) models, characterized by a grid spacing of O(100 m), are usually employed for research purposes in highly idealized setups. This dissertation aims at bridging together these two approaches by using the newly developed Icosahedral Non-hydrostatic Large-Eddy Model (ICON-LEM), which finally allows running LES mod- els over large scales, thereby capturing individual convective clouds, their associated mesoscale circulation and the large-scale environment in which they are embedded. The main idea is to exploit this new ability to represent convection in order to reconsider old unsolved questions and to focus on new problems which were not affordable with older approaches. In the first part of the dissertation the main objective is to develop conceptual models to explain different aspects of the coupling be- tween soil moisture and convective precipitation. I firstly focus on initial homogeneous conditions and on the precipitation response to increase/decrease of the initial soil moisture. Although convection can be triggered earlier over dry soils than over wet soils under certain atmospheric conditions, total precipitation is found to always decrease over dry soils. A simple conceptual model shows that this can be explained by the fact that precipitation intensity strongly correlates with surface latent heat flux, hence implying more rain over wet soils. Secondly, the focus is on heterogeneous conditions as resulting from heterogeneous surface soil moisture. Such heterogeneous conditions are known to trigger thermally-indirect circulations that advect moisture from wet to dry patches. Given that also local evaporation contributes to precipitation, my aim is to disentangle the effect of both sources over the course of a diurnal cycle. A simple conceptual model is derived using the results of LES simulations performed over a surface with varying degrees of soil moisture heterogeneity. The precipitation is expressed as a sum of advection and evaporation, each weighted by its own efficiency. By expressing advection and evaporation as functions of soil moisture I isolate the main parameters that control the variations of precipitation over a spatially drier patch. It is found that these changes surprisingly do not depend on soil moisture itself but instead solely on the initial atmospheric state. This is so because the propagation of the front associated with the mesoscale circulation is mostly driven by cold pools which effectively remove the dependency on the surface state. In the second part of the dissertation ICON-LEM is employed to simulate more realistic atmospheric cases including a convective diurnal cycle over Germany and a Tropical-Like Cyclone (TLC) over the Mediterranean Sea. The first scenario is used to test the conceptual model capacity to link precipitation to advection and evaporation weighted by different efficiencies. Two days from the exceptional period of severe weather that interested Germany in June 2016 are chosen due to their distinct diurnal cycle. It is found that, depending on the day and on the area analyzed, the model predicts different values for the efficiencies and that the one related to advection is not usually larger than the one associated with evaporation. Moreover, the model using two different efficiencies exhibits a greater skill than the one using a single efficiency in explaining precipitation variability over different regions of the domain. A different conceptual model, which attempts to reproduce the de- pendency of latent heat release by deep convection on the model resolution, is tested against simulations performed with ICON-LEM in the last chapter of the dissertation. The rare case of a TLC that affected the central Mediterranean between 7 and 9 November 2014 is simulated. The goal is to determine whether ICON-LEM can reproduce the internal dynamic and evolution of this cyclone, which was poorly predicted by most General Circulation Models (GCMs). It is found that simulations performed with grid spacing larger than 2.5 km completely miss the intensification phase of the cyclone due to poorly resolved convective processes. The simulations characterized by a smaller grid spacing, instead, show the ability of the model to reproduce the internal structure of the cyclone, characterized by a warm core, and its evolution over time. An interpretation of this abrupt change in the model skill is given based on a thermodynamic argument and on Potential Vorticity (PV)-thinking. Overall this dissertation shows how the interaction between convection and the land surface or the large-scale circulation can now be well captured with convection-explicit models like ICON-LEM which allow us to resolve the full spectrum of scales on which convection acts