Statistical downscaling differences strongly alter projected climate damages
Steve J. Miller, Naia Ormaza-Zulueta, Nisha Koppa, Ashley Dancer
Abstract
To project damages from climate change, researchers often pair empirical weather-damage relationships with high-resolution projections of future weather. Weather projections vary across scenarios, climate models, and how climate model output is downscaled to impact-relevant resolutions. Due to the nonlinear and cumulative nature of the relationship between weather and economic output, even small differences among downscaled climate projections could lead to large differences in projected damages. However, damage projections typically employ weather projections derived from a single downscaling approach. Here we examine how the choice of downscaling approach affects future damage projections across the globe, leveraging the historical relationship between mean annual temperature and subnational output. Different downscaling approaches can cause location-specific damage projections to vary by as much as 9.8% of gross regional product, in some locations surpassing uncertainty stemming from climate models and emissions scenarios. These results suggest applied researchers should employ multiple downscaled datasets when projecting climate impacts. Climate change projections can vary significantly based on the choice of downscaling approach, with location-specific projections differing by up to 9.8% of gross regional product, according to an analysis leveraging the historical relationship between mean annual temperature and subnational output across different downscaling methods.