RESEARCH PROJECT

Machine learning for high-resolution downscaling in the Hawaiian Islands

Climate change could significantly alter precipitation patterns in tropical islands like the state of Hawaiʻi with significant consequences for water resources. Unfortunately, models of global climate change have insufficient resolution to model the processes that generate most rainfall in Hawaiʻi. Specialized models are needed to predict how global changes will affect local climate, including statistical models trained on historical data.

This project will develop machine learning methods to predict precipitation at locations where no measurement data is available, using contemporaneous rainfall measurements (or downscaled projections of future rainfall) at other locations. Compared to traditional methods that rely on interpolation between stations, this project will use machine learning to leverage information about the orography to inform these predictions. With improved models for spatial interpolation, the project will produce historical rainfall maps at unprecedented resolution and accuracy. The project will also evaluate this as a method for projecting future rainfall and compare against existing statistical downscaling models. These historical and future rainfall maps will be shared via the Hawaiʻi Climate Data Portal to help stakeholders better understand the effect of climate change on water resources in the state of Hawaiʻi.