Evaluation of a Customized Variable‐Resolution Global Model and its Application for High‐Resolution Weather Forecasts in East Asia

Abstract

The performance of a variable-resolution global model, based on the Model for Prediction Across Scales-Atmosphere (MPAS-A) framework and with customized 160-to-1 km resolution grid mesh, was tested by simulating idealized flow fields as well as forecasting the evolution of actual weather systems. The mesh contains five levels of refinement, with 20, 15, 9, 3, and 1 km resolution covering central to East Asia, central to southern China, southeastern China, and Greater Bay Area/Hong Kong, respectively. Using a shallow-water solver and MPAS-A's solver, the mesh was evaluated against standard circularly refined meshes in simulating idealized steady-state flows. Conservation properties and error growths in the 160-to-1 km and counterpart simulations were fairly comparable. By perturbing the steady flow, realistic baroclinic wave evolutions could be captured. Initialized by Global Forecast System (GFS), parallel experiments were further conducted with new-Tiedtke (nTDK), Kain-Fritsch (KF), and Tiedtke (TDK) cumulus schemes and a convection-permitting suite (CP). Experiments showed that the model can give reasonable 5-day outlooks and evolution of synoptic-scale weather systems typically found in East Asia in various seasons. In particular, it can reproduce the mesoscale precipitation related to the Meiyu rainband (cold fronts) in summer (winter). When compared with station data, promising skills in predicting local temperature, humidity, and wind changes were found. It also performed slightly better using nTDK, KF, and TDK schemes, than adopting CP. Overall, by capturing multiscale features concurrently, these experiments gave reasonable global, regional, and local weather predictions, thereby demonstrating the practicality of using customized variable-resolution meshes for high-resolution short-range weather forecasts under MPAS-A framework. Plain Language SummaryWeather forecasting is usually carried out by dividing the atmosphere into many polygonal grid cells and solving physical equations in each cell. In general, the smaller the grid size, the more accurate is the prediction. However, computation becomes extremely expensive if the whole globe is covered by equally small grid cells. A numerical model, designed to accurately predict regional weather patterns, has been set up by using small grids over a target region, but with larger grids elsewhere, so as to minimize the computational cost. It was shown that such model can capture the general behavior of the atmosphere circulation. The model's skill was further assessed, by conducting retrospective predictions of historical weather events over East Asia. The model was capable of capturing the evolution of weather from the global to local scale, thereby demonstrating the practicality of using this model for fine-resolution short-range weather forecasts. Furthermore, it was also found that certain model configurations can give best performance in forecasting East Asian weather in various seasons.

Numerical Weather Prediction at 200 m Local Resolution Based on an Unstructured Grid Global Model

“…We compare the results of the above mesh with large resolution variation and HTS (HK200 m‐100 km with 11 HTS levels), three cases with mesh generated by Lloyd‐based algorithm and time‐integrated without HTS (Lui et al., 2020) (160‐1 km‐no‐HTS, 60‐3 km‐no‐HTS and 92‐25 km‐no‐HTS), three global near‐uniform meshes with HTS (2 HTS levels) (48 , 64 and 96 km) and a mesh NH24‐SH48, having 24 km on Northern Hemisphere and 48 km on Southern Hemisphere, with HTS (3 HTS levels).…”

Performance of MPAS-A and WRF in predicting and simulating western North Pacific tropical cyclone tracks and intensities

“…Voronoi cell vertices) were generated by a modified Frontal-Delaunay algorithm (Engwirda 2017) and bisection techniques. The positions of the points were then further optimized by Lloyd's method (Du et al 1999;Ringler et al 2008;Lui et al 2020) with respect to a mesh density function (which depends on the mesh specification) until they converged to a SCVT. The 160-to-2-km mesh was customized specifically for simulating WNP TCs that make landfall over southern China.…”

“…By using such unstructured spherical centroidal Voronoi tessellations (SCVTs), local refinement is achieved via variable-resolution mesh such that the horizontal resolution would exhibit gradual transition from coarser region to finer region (e.g. Lui et al 2020). As a global NWP model, MPAS-A can be integrated forward in time, given the initial conditions, without any lateral boundary forcing.…”

Impact of parameterization schemes on temperature simulation over China using the variable-resolution MPAS-A

“…However, there are some problems in the nesting between GCMs and RCMs (Warner et al 1997;Inatsu et al 2012). The RCMs need to constant updated the lateral boundary conditions which leads some issues on boundary-related conditions (Lui et al 2020a), because of the difference of dynamical cores and 4 parameterizations between GCMs and RCMs (Giorgi and Mearn 1991;Wang et al 2004). For the oneway nesting, the regional signals could not be feedback to the GCMs and the major errors from driving data cannot be corrected (Giorgi 2006).…”