Downscaling simulations

Earth system models (ESMs) simulate climate in grid cells roughly 100 km across, treating everything inside each one as uniform — this is far too coarse to capture the regional and local details that most impact assessments require. Mountain ranges, coastlines, and valleys are far smaller than these grid cells, yet they fundamentally shape how climate change and stratospheric aerosol injection (SAI) will affect specific places.

Agricultural impacts depend on precipitation patterns within a region, not averaged across a grid cell treated as homogeneous; water availability, extreme weather, and ecosystem responses all vary at scales finer than ESMs can resolve. Statistical downscaling (the process of using observed relationships between large-scale climate patterns and local conditions to infer high-resolution regional information) is widely used to bridge this gap.

ESMs also carry systematic biases, such as excess rainfall in certain regions, that must be corrected before impact models can produce realistic results. By providing downscaled, bias-corrected data from the Geoengineering Model Intercomparison Project (GeoMIP) as a public resource, this activity removes a critical barrier that currently prevents many researchers from studying SAI impacts at the scales most relevant to decision-making.

This activity focuses on the provision of downscaled data for impacts analyses, but higher resolution modeling, whether globally, using regional climate models (RCMs) or through nested regions of high resolution in earth system models, can also be an important means of improving mechanistic understanding of the climate dynamics under SAI and providing high resolution output for impacts.

This activity applies downscaling methods to GeoMIP model outputs to produce high-resolution climate data suitable for regional impact analysis. Dynamical downscaling, such as via running regional climate models, may be important for some applications., However, we expect primarily statistical downscaling as part of this activity. Statistical downscaling uses observed relationships between large-scale climate patterns—which ESMs simulate reasonably well—and fine-scale local variations to infer kilometer-scale conditions under different SAI scenarios. The work includes three components:

1. Technique development: developing and selecting appropriate statistical downscaling techniques.
2. Downscaling application: applying those techniques to GeoMIP outputs to generate downscaled data for key climate variables (including temperature and precipitation) at regional scales.
3. Bias correction: implementing bias-correction methods to ensure outputs reflect observed local climate characteristics.

This activity will make downscaled datasets publicly available through a cloud-accessible platform such as the Reflective Cloud Hub, enabling researchers to conduct impact analyses without performing downscaling themselves. The activity may also involve collaboration with specific impact researchers—for example, those modeling water availability or agricultural yields—to ensure that the downscaling approach and output resolution meet their particular needs.