Applications of flexible spatial and temporal discretization techniques to a numerical weather prediction model

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

“…The determination of zones and their boundaries is performed with a minimization of computational cost in parallel computing. This hierarchical time‐stepping (HTS) technique essentially avoids cells with large grid spacing from using small timesteps unnecessarily and enables MPAS‐A to be used on meshes of very large resolution variation ranges with affordable computational resources (Cheung et al., 2022; Ng et al., 2019). For the 200 m‐resolution mesh (details in Section 2.1), there are 9 HTS levels, with timestep 0.781 s for the finest resolution zone (with grid spacing range 0.171–0.513 km), timestep 1.563 s for the second highest HTS level zone (with grid spacing range 0.349–1.16 km), so on and so forth to 200.0 s for the coarsest resolution zone (with grid spacing 42.9–59.7 km).…”

“…Moreover, a technique based on the OLAM (Cheung et al., 2022; Walko & Avissar, 2011) for Customizable Unstructured Mesh Generation (CUMG) enables local mesh refinement in arbitrary shape using user‐defined target horizontal resolution in any desired locations. It generates SCVTs that satisfy the original formulation of MPAS‐A dynamic core.…”

Deep Learning Improves GFS Wintertime Precipitation Forecast Over Southeastern China

“…Many efforts have been devoted to improving the forecast performance of precipitation. These efforts include the resolution refinement of the numerical weather prediction (NWP) models (Ashrit et al., 2020; Cheung et al., 2022), the improvements in the physical parameterization schemes (Stoelinga et al., 2003; Thompson et al., 2008; Y.‐M. Yang et al., 2020), the improvements in the observational network (Ceccherini et al., 2015; Kidd et al., 2012), the improvements of the data assimilation methods (Dance, 2004; Lee et al., 2022; Županski & Mesinger, 1995) and so on.…”