资源环境科技发展态势分析平台

Many climate models from the fifth phase of the Coupled Model Intercomparison Project predict decreases in precipitation in Central America and northern South America by the year 2100 for the Representative Concentration Pathway 8.5 scenario. Here we show that the fifth phase of the Coupled Model Intercomparison Project models able to more accurately simulate warm North Atlantic sea surface temperatures (SSTs) for the present climate (strong Atlantic meridional overturning circulation models) are more likely to project a larger precipitation decrease. Drought amplification from the slowdown of Atlantic meridional overturning circulation is more significant during the wet season in the Northern Hemisphere, with an SST-constrained model estimate yielding a 73% larger decrease in precipitation (-1.11mm/day) than the multimodel mean (-0.64mm/day). Since most Earth system models underestimate contemporary SSTs in the North Atlantic, the use of the multimodel mean for impact analysis likely underestimates drought stress and the vulnerability of neotropical forests to increasing drought from climate change.

Plain Language Summary Many climate models predict that in response to global warming, rainfall will decrease in Central America and northern South America. However, there is large disagreement regarding the magnitude of this drying in different models. We separate these models into different groups based on their ability to accurately simulate sea surface temperatures in the North Atlantic Ocean. One group of models simulates higher present-day sea surface temperatures, which agree better with present-day air temperature and ocean current observations. We find that these models predict more future drying in Central America and northern South America, particularly during the wet season, relative to the full model archive. By using sea surface temperature observations as a constraint, we create an optimal prediction and estimate that the drying will be intensified by an additional 73% by 2100 compared to the multimodel mean. Our work suggests that future drought stress and the vulnerability of tropical forests to climate change may be potentially larger than previous estimates.