Improving NASA GPM Data Exploitation through NASA Cloud-Resolving Model Empowered with Parallel AsyncIO Management

Abstract:

The effects of aerosols have the largest uncertainty when it comes to predicting the anthropogenic impact with current weather and climate models. The NASA Goddard Cumulus Ensemble model (GCE) with advanced spectral bin microphysics has been used to study the impact of aerosols on deep convective precipitating systems, which contribute more than 50% of the total surface rainfall in the Tropics and for floods and are associated with severe weather events in mid-latitudes. Global Precipitation Measurement (GPM) has released its data to the public. High-resolution GCE simulations can be combined with incoming NASA GPM data to provide direct assistance to both retrieval algorithm developers as well as developers of model physics for coarse resolution (i.e., hydrological, regional and global/climate) simulations with more limited computational resources. However, the spectral bin microphysical scheme is very different from previous bulk schemes. It explicitly resolves the particle size distributions using 33 (or more if needed) size bins for each cloud species. This introduces ~400 more prognostic variables. In addition, bin microphysical schemes need to resolve the interactions between these 400 additional variables, making it highly compute as well as I/O intensive. To address those challenges, we have been improving GCE scalability as well as developing a parallel AsyncIO management system consisting of MPI IO, data burst caching and data compression. In this paper, we will present the results of ultra-high resolution (up to 0.125 x 0.125 degree global-model-equivalent resolution) GCE simulations empowered with the parallel AsyncIO management system.