As highlighted in a recent issue of DOE Pulse, which features science and technology news from the DOE National Labs, scientists from Pacific Northwest National Laboratory (PNNL) have developed a more efficient way to perform computationally expensive global climate simulations.

As climate models become more precise and detailed, the simulation takes more time, putting pressure on already limited supercomputing resources. Since getting more cycles isn’t always an option, scientists are encouraged to make their codes as efficient as possible. Researchers at PNNL took this message to heart and the result was a novel computational approach that promises equally reliable results using only a fraction of the conventional computational load. In simple terms, it works by replacing a single long computer run with multiple short runs.

Current high-resolution simulations takes days to months to complete even on the most powerful supercomputers. The goal is to produce a reliable signal distinct from the noise that is inherent in such highly complex system, but this is very time-consuming. The PNNL researchers examined statistics from a number of short simulations rather than from a single, multi-year simulation. Using the Community Atmosphere Model (CAM), they initialized the short simulations with different weather conditions, creating independent runs that could be carried out simultaneously. By transforming a single long process into multiple short tasks, researchers made better use of leadership-class supercomputer systems, and had their results more quickly too.

The researchers conclude that they were able to reduce computational time by a factor of about 15, and shorten turnaround time by a factor of several hundred. They expect that this dramatic improvement in efficiency will benefit the scope and depth of other climate investigations and similar projects.

“Using short ensembles instead of traditional multi-year climate simulations, sensitivity studies can be carried out more efficiently, benefiting from a substantial reduction of the total CPU time spent on numerical integration, as well as a much faster turnaround in the investigation because the independent ensemble members introduce an additional dimension of parallelism that can be exploited with current flagship supercomputers,” the team writes.

“The efficiency of the method makes it particularly useful for the development of high-resolution, costly, and complex climate models,” they explain.

They do caution that the technique has its limitations, noting “it cannot be used as formulated here to investigate modes of climate variability or feedback mechanisms that operate on time scales of months to years, thus could not replace long-term simulations when long time scales are important.”

The next step for the project is to improve how climate process interactions are represented in the CAM5 model.

Read the DOE Pulse article here.