Examining Tropical Convection Features at Storm‐Resolving Scales Over the Maritime Continent Region
This study uses a regionally refined mesh centered on the Maritime Continent to investigate how tropical convection manifests with climate model physics at storm-resolving resolutions.
This study uses a regionally refined configuration of the Community Atmosphere Model that has storm-resolving resolution over the Maritime Continent. Analysis of the season-long simulation shows that adding resolution shows clear regional improvements. Precipitation is better represented in terms of the distribution of intensities, including extremes, as well as an improvement in the diurnal cycle. Tropical cyclones and other organized convective storms form and better match observations. Outside the high-resolution region, however, the deficiencies of the traditional parameterizations become apparent. In particular, the deep convection parameterization is unable to remove moist instabilities effectively, leading to significant biases, and highlighting the need for scale-aware physics for climate studies using regionally refined configurations.
There is increasing interest in and use of global storm-resolving models with a horizontal grid spacing of less than 5 km. These models remain extremely computationally expensive, making it difficult to conduct experiments. This study uses a numerical mesh that has 3 km grid spacing
only over a small region of the globe, stretching to 60 km for most of the world. This allows analysis of regional features at ultra-high resolution at a fraction of the cost of global storm-resolving models. Running the model for several months gives reasonable statistics for an
assessment of convective features compared to a control simulation using a uniform 60 km mesh. Within the high-resolution region, several improvements are apparent compared to observations. These improvements arise as convective features become more realistic, aided by
much more realistic topography and associated mesoscale circulations. Looking outside the refined region showed biases compared to the control 60 km simulation, which occurs as the model time step becomes small to account for the high-resolution region. This is a consequence of parameterized physics sensitivities, highlighting the importance of scale-aware physics for multi-scale simulations.