This dataset contains output from a suite of 10-day, IFS-initialized MPAS-Atmosphere forecast experiments designed to diagnose the sources of convectively coupled Kelvin wave (CCKW) forecast error in an accompanying publication. Six configurations vary horizontal resolution (30 km and 15 km global meshes, plus a variable-resolution mesh with 3 km convection-permitting spacing between 10S and 30N), convective parameterization (scale-aware Tiedtke vs. Grell-Freitas), and initial CCKW amplitude (Kelvin-filtered wind, temperature, and moisture fields added to or subtracted from the initial state). Forecasts are initialized at 00 UTC every three days during 2018, matching an existing suite of operational forecast runs (Different Models, Same Initial Conditions; Magnusson et al., 2022), for a total of 118 forecasts (1,180 forecast days). This dataset contains both the native-mesh output and 1 by 1 degree post-process fields for each of these model runs, with the exception of the variable-resolution 15-3 km tropical mesh simulation. For those simulations, only the post-processed data are provided.

Details: Accurately predicting convectively coupled Kelvin waves (CCKWs) remains a major challenge for global models, despite their importance in modulating tropical weather and extreme events. This dataset was produced to investigate the sources of those errors by varying horizontal resolution, convective parameterization, and initial CCKW amplitude one factor at a time from a common baseline. All experiments use the global configuration of the Model for Prediction Across Scales - Atmosphere (MPAS-A) version 8.2, a fully compressible, non-hydrostatic model on an unstructured Voronoi mesh. Common settings across every experiment, including: 55 vertical levels, RRTMG radiation, Thompson microphysics, and the MYNN boundary-layer scheme. Forecasts are 10 days long, initialized at 00 UTC from ECMWF IFS analysis fields every three days. There are six configurations:

• MPAS 30km: Baseline. 30 km quasi-uniform global mesh, scale-aware Tiedtke convection. 50 forecasts, 2018-06-06 through 2018-10-31.
• MPAS 15km: Same as baseline but 15 km quasi-uniform global mesh. 16 forecasts: 2018-06-30, and 2018-09-01 through 2018-10-13.
• MPAS 3km: Variable-resolution mesh: 3 km grid spacing in a continuous band from 10S to 30N at all longitudes, 15 km elsewhere. 6 forecasts: 2018-06-30, and 2018-09-16, -19, -22, -25, -28.
• MPAS GF: Same as baseline but with scale-aware Grell-Freitas convection replacing scale-aware Tiedtke. 16 forecasts, same dates as MPAS 15km.
• MPAS Damp.: Same as baseline, but Kelvin-wave-filtered wind, temperature, and moisture fields are subtracted from the initial state to weaken the initial CCKW. 15 forecasts, 2018-09-01 through 2018-10-13.
• MPAS Ampl. Same as MPAS Damp., but the filtered fields are added rather than subtracted, strengthening the initial CCKW. 15 forecasts, same dates.

Total: 118 forecasts, 1,180 forecast days. The initial-amplitude perturbation method follows Lawton et al. (2024). Every run writes 6-hourly instantaneous output every six hours. Post-processed products were generated by conservative remapping (CDO remapcon) from the native unstructured mesh to a regular 1 by 1 degree latitude-longitude grid. Derived quantities computed at this stage include apparent heating (Q1), divergence and vorticity, and tropical-belt rainfall subsets. Several diagnostic and PV-budget fields are interpolated to 28 isobaric levels between 1000 hPa and 100 hPa, spaced 25 hPa apart. Files are organized as experiment/YYYYMMDDHH/, one directory per initialization, with post-processed products in a post_process/ subdirectory alongside the native-mesh output. Note that the MPAS 30km baseline directories sit at the top level rather than under an experiment-named subdirectory. Several post-processed files hold identical data under different coordinate-variable naming conventions, distinguished by the merged_, fixed_merged_, and fixed_for_MSE_merged_ filename prefixes.

Model Simulation

UCAR/NCAR  |  NIL-U/EAE

0.0 MB

HDF5/NetCDF4

This work is licensed under a Creative Commons Attribution 4.0 International License.