October 19, 2019
Highlight

A Faster Way to Explore Earth System Uncertainty

A new Earth system model emulator provides internal variability and space-time correlation

Earth system

Scientists at Pacific Northwest National Laboratory have developed an Earth system model emulator (ESM) with internal variability and space-time correlation.

The Science
Earth system models (ESMs) provide the best predictions of future climate trajectories by simulating all of the physical processes that drive the climate system. However, they require a large amount of computing time, memory, and power. As a result, researchers are generally limited to running a small selection of scenarios, so they cannot study the full range of possible climate outcomes that an ESM could provide. Scientists at the U.S. Department of Energy’s Pacific Northwest National Laboratory (PNNL) have developed an emulator called fldgen, which allows researchers to estimate what an ESM would have produced had they been able to run a larger suite of calculations.

The Impact
The ability to run a wide range of climate scenarios is crucial to studying the risk from climate related extremes. Events such as droughts, floods, or heat waves are rare and might not be well sampled in the limited ESM runs that are available. Emulators can fill in the gaps by providing enough runs to sample these rare extremes, but only if they display the variability seen in ESM runs. Most emulators are deterministic, producing a single answer that can be interpreted as the average result. Real climate is variable and if the model always produces the average temperature, a lot of important climate phenomena is missed. Fldgen captures this variability in its results, including all of its spatial correlation (locations near each other have similar weather) and time correlation (it produces cyclic phenomena like El Niño). This allows researchers to explore the risks faced by human systems due to an uncertain future climate in a way that has not previously been possible.

Summary
In the past, ESM emulators have focused on estimating the average response of an ESM. To study risk, however, researchers must capture the variability and uncertainty in future climate. Each time an ESM is run with a different initial condition, it produces a different result. Emulators that display the same patterns of variability over space and time as the ESMs are needed. When fldgen v. 1.0 is run in analysis mode, it analyzes the available ESM outputs for space and time correlation. Fldgen can then be run in generator mode to produce global temperature outputs with these same correlations, as many times as desired. Fldgen runs in minutes on an ordinary scientific workstation, instead of the months of supercomputer time ESMs require. Therefore, fldgen is a tool uniquely suited to generating the large collections of data needed to study risk to human systems due to uncertain temperature in the future. Although version 1.0 of fldgen only covers temperature, subsequent versions—already developed and under scientific review—include additional variables such as precipitation, making it possible to study a wider range of extreme event risks. The code for fldgen can be found here: https://github.com/JGCRI/fldgen. The version 1.0 release, specifically, is here: https://github.com/JGCRI/fldgen/releases/tag/v1.0.0. 

PI Contact
Mohamad Hejazi
Pacific Northwest National Laboratory
Mohamad.Hejazi@pnnl.gov

Funding
This research is based on work supported by the U.S. Department of Energy, Office of Science, as part of research in the MultiSector Dynamics, Earth and Environmental System Modeling Program. This research was supported in part by the Indiana University Environmental Resilience Institute and the Prepared for Environmental Change grand challenge initiative.  

Revised: October 19, 2019 | Published: July 2, 2019

Publication
Link R, A Snyder, C Lynch, C Hartin, B Kravitz, and B Bond-Lamberty. 2019. “Fldgen v1.0: an emulator with internal variability and space–time correlation for Earth system models.” Geoscientific Model Development., 12:1477–1489, https://doi.org/10.5194/gmd-12-1477-2019.