Using Statistics to Select Atmospheric Predictor Variables

Contributed by Kristen Sanfilippo

An exploratory analysis found that atmospheric predictor variables selected through statistical methods can lead to more accurate rainfall projections when compared to the previous predictor set used in statistical downscaling for rainfall projections in Hawaiʻi. This figure shows the skill of a linear model in projecting rainfall for 1980 to 2007 using the previous predictor set (blue bars) and using a predictor set selected through our predictor selection process (orange bars). Zero represents no skill and 1 represents perfect skill. A shift towards increased skill is apparent for the selected predictor set.

Leave a Reply

Your email address will not be published. Required fields are marked *