You are here:
HPC Aspects of Variable-Resolution Global Climate Modeling using a Multi-scale Convection Parameterization
Citation:
Alapaty, Kiran, L. Fowler, J. Herwehe, AND B. Skamarock. HPC Aspects of Variable-Resolution Global Climate Modeling using a Multi-scale Convection Parameterization. Presented at IUSSTF Workshop, Bangalore, INDIA, December 16 - 18, 2013.
Impact/Purpose:
The National Exposure Research Laboratory (NERL) Atmospheric Modeling and Analysis Division (AMAD) conducts research in support of EPA mission to protect human health and the environment. AMAD research program is engaged in developing and evaluating predictive atmospheric models on all spatial and temporal scales for forecasting the air quality and for assessing changes in air quality and air pollutant exposures, as affected by changes in ecosystem management and regulatory decisions. AMAD is responsible for providing a sound scientific and technical basis for regulatory policies based on air quality models to improve ambient air quality. The models developed by AMAD are being used by EPA, NOAA, and the air pollution community in understanding and forecasting not only the magnitude of the air pollution problem, but also in developing emission control policies and regulations for air quality improvements.
Description:
High performance computing (HPC) requirements for the new generation variable grid resolution (VGR) global climate models differ from that of traditional global models. A VGR global model with 15 km grids over the CONUS stretching to 60 km grids elsewhere will have about ~2.5 times more grid columns than that of a typical ~60 km grid global model. Also, at these ultra-high resolutions a VGR global model requires a dynamical non-hydrostatic atmospheric representation, state-of-the art and complex physical and chemical processes suitable for these ultra-high resolution grids as well as special visualization software. If these data were to be used for studying hurricanes and associated storm surge or any other extreme events, climate change impacts on ecosystem and human health for decision making—then, hourly-daily resolution data need to be saved from these global models. Also, it’s a challenge to analyze these big data sets quickly. Another critical hindrance for global models to use ultra-high resolution (~10km) grids is the lack of suitable physical formulations applicable at these ultra-high resolutions. For example, present convection parameterizations won’t work at these finer resolutions or seamlessly work across various spatial scales. Also, incorporating a convective cloud microphysics into a convection parameterization increases computational requirements. All together, HPC requirements are quite huge for the ultra-high resolution global climate modeling. Recently, we have developed a multi–scale Kain-Fritsch (KF) scheme that can be used seamlessly across all spatial scales down to convection-resolving scales. The new science updates include: convective cloud-radiation interactions and scale-dependent dynamic adjustment timescale and entrainment formulations. In addition, we have also included impacts of convective updrafts on the grid-scale vertical velocity and a diagnostic convective cloud microphysics to facilitate aerosol–convection interactions. These new science updates are tested in a regional climate model, the Weather Research and Forecasting (WRF) model and were implemented into the next generation climate model, the Model for Prediction Across Scales (MPAS). We have performed two sets of simulations with the MPAS using the standard KF scheme and the multi-scale KF scheme to study the improvements resulting from the latter scheme. First, the MPAS global simulations were performed for one month using a quasi-uniform 60 km Voronoi cells. Then, we conducted MPAS simulations using variable resolution Voronoi cells of about ~15 km over the continental U.S. and gradually stretching to about ~60 km for the rest of global domain. Preliminary results obtained from as well as HPC requirements for using the MPAS will be presented.