Jinwon Kim and Norman Miller,
Lawrence Berkeley National Laboratory
Chris Ding and Andreas Mueller, NERSC,
Lawrence Berkeley National Laboratory

Research Objectives

The purpose of this project is to generate physically and dynamically consistent information on regional climate variability due to natural and anthropogenic factors and to assess its impacts on various sectors (e.g., water resources, agricultural production, ecosystems, etc.). An instrumental goal is to convert the present serial Regional Climate System Model (RCSM) to a high-performance parallel code.

Computational Approach

The core of RCSM is a mesoscale atmospheric model that solves the stratified, quasi-2D flow equations for 3D advection and physical parameterizations with a third-order finite difference scheme. The parallel RCSM code uses a 2D domain decomposition with message passing for distributed memory architectures. It takes 15 minutes to simulate one model day on 14 nodes of the Cray T3E, with about the same performance on a Cray T90. Ongoing research includes improving scalability and load balancing as well as cross-platform portability and benchmarking.

As part of a three-day weather forecast for the western U.S., RCSM produced this wind speed and direction forecast for six-hour intervals from 10 pm October 27 to 4 pm October 28, 1999.

Accomplishments

We have produced a multi-year "hindcast" of the hydroclimate of the western U.S., which we are using to validate and improve our model and to analyze the observational data. The hindcast included a successful simulation of the streamflow in northern California's Russian River basin over a three-month period. The model has also been applied to eastern Asia. In addition, we completed a preliminary study of doubled CO₂ impacts on the western U.S. based on the HadCM2 ensemble projections.

Significance

Generating information on regional-scale precipitation patterns and other aspects of climate variability requires high-resolution regional climate models. Computational requirements increase to the third power as the resolution increases (e.g., doubling the spatial resolution requires about 8 times the computational load). Hence, high-performance computing is a crucial component to successful improvements in regional climate research. The high performance RCSM will become a tool to generate science-based information to improve climate change impact assessments, as envisioned by the DOE Accelerated Climate Prediction Initiative (ACPI).

Publications

J. Kim, N. Miller, J. Farrara, D. Cayan, and K. Mo, "Winter-season hydroclimate study for the western US using the RCSM," in Proc. 11th Conf. on Appl. Climatology (Dallas, Texas, 1999), p. 350.

N. Miller, J. Kim, J. Farrara, K. Mo, and D. Cayan, "Short-term and seasonal streamflow predictions for a California coastal basin during the 97-98 winter," in Proc. 14th Conf. on Hydrology (Dallas, Texas, 1999), p. 255.

N. Miller, J. Kim, and R. Hartman, "Downscaled climate and streamflow study of the southwestern United States," J. Amer. Water Res. Assoc. (submitted).