Most climate models for the Caribbean predict drying trends, increased temperature, and precipitation variability. We used regression analyses to evaluate how global predictions translated into local temporal patterns of climatic variation within the Luquillo Mountains in Puerto Rico, and which variables were more useful for expressing precipitation and temperature variability trends there. We used cross-correlation analyses to evaluate how local precipitation and temperature data related to regional atmosphere-ocean teleconnections indices (ENSO, NAO, TNA, and AMO). Contrary to regional climate predictions, we found an increase in mean annual precipitation during the dry season, and no trends in annual and wet season precipitation data sets were observed. Precipitation variance increased after 1995, but only when considering dry season data, which showed a decrease in the total number and consecutive days without rain after 1995. In agreement with regional predictions, maximum and minimum annual mean temperatures increased between 1975 and 2015. We found strong positive associations between maximum and minimum temperatures and rainfall distribution (i.e. more wet days, less dry intervals). Precipitation and temperature were more strongly associated to changes in the Tropical Atlantic than to changes in the North Atlantic or the Eastern Pacific. Our downscaling analysis (one station data) shows the relevance of evaluating global climate predictions in specific areas where landscape features can modify the patterns observed at bigger scales. The combination of downscaling analysis and the inclusion of long-term climate trends can facilitate the development of better future scenarios of climate variability and evaluate their importance to local forest ecosystems.