Statistical Downscaling of CMIP5 data to predict future dry day frequency in the El Yunque National Forest

The El Yunque National Forest (EYNF), situated in the Luquillo Mountains of east Puerto Rico, is home to a wide range of climate-sensitive ecosystems and forest types. In particular, these ecosystems are highly sensitive to changes in the hydroclimate, even on short time scales. Any systematic reductions in annual or seasonal precipitation pose serious implications for the tropical rainforest ecosystem which relies on the region's abundant rainfall. Even in the absence of large annual or seasonal precipitation deficits, future changes in short-duration dry spells represent an important disturbance to the ecological and biogeochemical processes in EYNF. Current global climate models predict coarse-scale reductions in precipitation across the Caribbean prompting the need to investigate future fine-scale hydroclimate variability in the EYNF.

This study analyzes future rainfall variability and dry day frequency during the early rainfall season (AMJJ) at a high elevation site in the EYNF. A better understanding of future changes during the early rainfall season is important as it has been shown to drive much of the annual precipitation variability in eastern Puerto Rico, and it also serves as the recovery period for the region after the dry season spanning December - March. A binary output artificial neural network (ANN) is trained using ERA-Interim Reanalysis data to predict wet and dry days. The ANNs are trained using 10 atmospheric predictor variables which includes the Galvez-Davison Index (GDI) and its components as well as low- and mid-tropospheric wind fields. The ANNs are run using a CMIP5 global climate model ensemble to analyze future changes in precipitation variability. Additionally, changes in future convective potential are examined by using shifts in GDI regimes.