Supercomputer-powered climate models are at the heart of our understanding of global climate change – but the climate is a massively interconnected system, and even the most advanced models need to be checked against reality. Now, at the University of Wyoming, researchers are using supercomputers to study how the estimates from computer models stack up against real-world climate data.
Specifically, the researchers looked at the cooling effect of wildfires. While the fires themselves burn hot, the smoke that emerges from them contains both darker particles, which absorb sunlight and trap heat, and whiter particles, which reflect sunlight away from Earth. As a result, the net effect of large amounts of smoke can be difficult to predict.
The researchers aggregated the results of nine “state-of-the-art Earth system models/chemical transport models” that estimated the net atmospheric temperature effects of wildfire smoke. They then compared those models to a dozen observational datasets of the same measurements.
To run this heavy-duty data analysis, the researchers turned to the National Center for Atmospheric Research (NCAR) Wyoming Supercomputing Center, where they used the Cheyenne supercomputer. Cheyenne’s 4,032 Intel Xeon Broadwell-based nodes, 313 TB of memory and Mellanox EDR InfiniBand networking deliver 4.8 Linpack petaflops, placing it 60th on the most recent Top500 list of the world’s most powerful publicly ranked supercomputers.
As for the results?
“When we compare global observations of wildfire smoke to simulated wildfire smoke from a collection of climate models, the vast majority of the models have smoke that is more light absorbing than the observations,” said Hunter Brown, a Ph.D. graduate of the University of Wyoming and lead author of the paper. “This means that more energy from the sun is going toward warming the atmosphere in these models, as opposed to what we see in these field campaigns and laboratory studies, which report less absorbing smoke that has more of a cooling effect by scattering light away from the Earth and back to space.”
The researchers applied their observations and worked to correct an African wildfire smoke model, but the model still overrepresented the warming effects of smoke in that region.
“We were able to trace the disagreement between the model and observations to how the models represented the individual smoke particles, or aerosols, in the model,” Brown said. “This came down to how the model characterized their makeup, their size and the mixtures of different types of biomass burning aerosol. When we changed these variables in one of the models, we saw considerable improvement in the simulated smoke.”