We implemented a hybrid downscaling model using classification and regression trees and support vector regression with evolutionary strategies to statistically downscale precipitation occurrences and amounts from 16 points across North America. All the selected points belong to different climate regions. In addition, to evaluate the downscaling model’s historical and future performances we used daily precipitation outputs (from a high resolution ~25km grid spacing global atmospheric model) as predictands, and coarsened versions of the same high-resolution outputs from the nearest nine gridpoints (interpolated to a ~100km grid), as predictors. This experimental setup, known as “Big-Brother” allows us to use the high-resolution (historical and future) model outputs as pseudo-observations so we can validate the downscaled values against them. The downscaled precipitation occurrences were evaluated in terms of historical and future Peirce Skill Score (PSS), while the downscaled precipitation amounts were evaluated in terms of mean absolute error skill score (MAE SS) to assess if the skills were time-invariant. Our results show that from the selected 16 points, those located west of the Rocky Mountains obtained the highest positive historical and future MAE SS; while the points near the northern part of the east coast also obtained positive scores but their scores were smaller in magnitude. These results contrast the negative or null skill scores obtained when downscaling to points located between the Rockies and the east coast. We also found that the CART model under predicted the total number of rainy days and might not be the most appropriate model for obtaining precipitation occurrences with this set of predictors. Future implementations will expand the predictor set aiming to improve the overall MAE SS.
Gaitan, Carlos F.; Dixon, Keith W.; Balaji, Venkatramani; and McPherson, Renee, "Statistically Downscaled North American Precipitation Using Support Vector Regression And The Big Brother Approach." (2014). CUNY Academic Works.