Sparse Local Implicit Image Function for sub-km Weather Downscaling (Papers Track)
Yago del Valle Inclan Redondo (Recursive); Enrique Arriaga-Varela (Recursive); Dmitry Lyamzin (Recursive); Pablo Cervantes (Recursive); Tiago Ramalho (Recursive)
Abstract
We introduce SpLIIF to generate implicit neural representations and enable arbitrary downscaling of weather variables. We train a model from sparse weather stations and topography over Japan and evaluate in- and out-of-distribution accuracy predicting temperature and wind, comparing it to both an interpolation baseline and CorrDiff. We find the model to be up to 50% better than both CorrDiff and the baseline at downscaling temperature, and around 10-20% better for wind.