This dataset comprises output from global atmospheric simulations performed with the Model for Prediction Across Scales - Atmosphere (MPAS-A) as part of the DYnamics of the Atmospheric general circulation Modeled On Non-hydrostatic Domains (DYAMOND) intercomparison project (Phases 1 and 2). Simulations were conducted over two 40-day periods: Phase 1 (boreal summer: 1 August to 10 September 2016) and Phase 2 (boreal winter: 20 January to 1 March 2020). Phase 1 features horizontal mesh spacings ranging from 480 km to 3.75 km, alongside a suite of sensitivity experiments designed to isolate the effects of convective parameterization and changes to the microphysics. Phase 2 focuses on higher-resolution configurations, featuring mesh spacings of 30 km, 15 km, 7.5 km, and 3.75 km. Output for both phases consists of two-dimensional diagnostic fields (15-minute frequency) and three-dimensional history files (3-hourly frequency) on MPAS unstructured Voronoi meshes. The dataset supports research into atmospheric phenomena across all scales - from individual convective clouds to global circulation - including tropical waves, tropical cyclone activity, precipitation, and the sensitivity of numerical models to physical parameterization choices under different seasonal regimes.

Background and Objectives: DYAMOND (Stevens et al. 2019) is a community intercomparison framework for global storm-resolving models that explicitly represent deep convection without convective parameterization at kilometer-scale grid spacings. This dataset supports analysis of MPAS-A across a broad range of resolutions and physics configurations, enabling study of resolution dependence, parameterization sensitivity, and seasonal variability in a global convection-permitting context.

Model and Simulations: MPAS-A was configured with the "convection_permitting" physics suite, which includes Thompson microphysics, MYNN boundary and surface layers, the Noah land surface model, RRTMG radiation, and a scale-aware Tiedtke cumulus parameterization. Phase 1 utilized the default suite from MPAS v6.1, while Phase 2 employed an experimental, custom-built suite that was not adopted as a standard release. Simulations were initialized on 1 August 2016 (Phase 1) and 20 January 2020 (Phase 2) from ECMWF analyses, with observed SSTs updated throughout the respective integrations. Variables use standard MPAS names, which can be referenced here.