Assessing the Credibility of Downscaled Climate Projections

Overview

This year, NASA started a project to assess the credibility of dynamically downscaled climate projections—taking results from global climate models (GCMs) to drive regional climate models (RCMs). These projections have been used for climate assessments, impact studies, and resource and management decision supports. However, since the credibility of downscaled climate projections has not been scientifically well established, the application and decision support communities are using the projection data with significant risk of relying on incomprehensive information.

Our goal is to establish a NASA framework for assessing the credibility of dynamically downscaled climate projections for the continental United States. Research will be done in collaboration with our interagency and international partners using NASA’s unique high-resolution climate modeling, supercomputing, and big data analytic capabilities.

Project Details

The project has six components: (1) acquisition of a high-performance computing (HPC) system and storage devices, (2) high-resolution GCM simulations, (3) RCM downscaled simulations, (4) satellite data reanalysis products, (5) an RCM evaluation system, and (6) metrics development.

Evaluation focus areas will include West Coast wintertime atmospheric rivers, Southern/Central summertime convection and precipitation, and Northeast winter storms. Possible assessment criteria are: precipitation distribution; snow cover/elevation; local orographic flow (for example, barrier jets); atmospheric moisture profiles/characteristics; and cloud, radiation, temperature, and storm track statistics.

Results and Impact

The project will help NASA scientists to answer the following questions:

• Under ideal forcing conditions (for example, high-quality reanalyses), how good are RCMs at replicating important weather/climate features?
• Under what conditions does downscaling (RCMs driven by GCMs) give valid results?
• Do high-resolution GCMs (10 kilometers or finer) offer anything that can’t be obtained via downscaling?

Why HPC Matters

Long-term climate projections typically require at least 100-year model integration. Because of significant computing resource requirements, it was not possible in the past to run long integrations of climate models at high-enough resolution for decision support. The current generation of HPC systems, with millions of cores, is on the cusp of being able to provide the necessary resources. In addition, high-resolution climate simulations will generate multiple petabytes of data, so a big data management and data analytic system will be necessary for data analysis. This is a very timely problem that the HPC community can help solve, and NASA provides a unique environment to make this possible.