Impact of model resolution and input analysis data on convection initiation over flat terrain

The increasing computational capacity has enabled numerical models to reproduce more realistic thunderstorms in the simulations. The Convective and Orographically-induced Precipitation Study (COPS, Wulfmeyer et al., 2011) has focused on the reproducibility of the convective development over a complex terrain, with non-negligible topographic forcing. On the other hand, development of convection over flat terrain is found to be quite sensitive to the choice of the dynamical and microphysical options in the cloud-resolving model, as well as to the large-scale forcing provided through the initial and boundary conditions. Especially in regional-scale studies focusing on convection, it is of critical importance that thunderstorms in question develop at realistic time and location.

This study presents how the interplay between the initial and boundary conditions, model resolution, and the simulation setup, could result in producing thunderstorms in different locations. With our real-data simulations of deep convective clouds in the central U.S., we show that the errors in the location of convection and the amount of accumulated rainfall are likely dependent on the large-scale environment in the initial conditions provided by analysis data, such as the Global Forecast System (GFS), the North American Mesoscale Forecast System (NAM), and the European Center for Medium range Weather Forecasting (ECMWF). In addition, analysis data with a fairly high resolution is actually shown to interfere with the downscaling capability of the cloud-resolving model, leading to a more inaccurate location of the thunderstorm development. This indicates that there seems to be an optimal resolution of the input data to reproduce realistic convection in a simulation, rather than runs with the highest resolution always producing the most realistic thunderstorms.