Downscaling

Industry needs transparency and measures of uncertainty to make use of downscaled future projections of the risk from extreme weather events. Most downscaled estimates of risk are deterministic, i.e., not expressed in a framework of uncertainty. This work will create a model of 1-kilometer extreme weather employing a hybrid physics and machine learning approach to climate modeling.

Team: Kinter (GMU), Burls (GMU), Cash (GMU)

Extreme heat is the clearest signal we have for a changing climate and can impact worker productivity, cause power outages, and increase energy costs. The cost is projected to be severe, with heat-related losses estimated at over $2T annually by 2035, with a significant share potentially covered by insurance. However, there is a lack of granular air temperature data to inform action at the scales at which humans are exposed to heat. This project will use Machine Learning to quantify the effectiveness of urban heat island (UHI) mitigation strategies—including cool roofs, green infrastructure, and urban planning interventions—at actionable scales to enhance urban resilience against extreme heat events using a combination of satellite imagery and ground based observations.

Team: Azarderakhsh (CUNY), Ortiz (GMU)