How well do high resolution limited area models simulate sub-tropical precipitation? A comparison of High Resolution Regional Climate Model output for Puerto Rico

Regional climate modeling of precipitation processes in tropical / subtropical regions is subject to structural uncertainty from a number of sources. This includes the ability to represent convective processes and moisture availability in these regions and the choice of grid resolution in a regional climate model. Both of these sources influence the ability to capture precipitation in tropical / subtropical regions with different regional climate models (RCMs). The complex topography and background easterly flow leads to unique patterns for precipitation (both convective and non-convective) in Puerto Rico. The sharp gradient in precipitation across the island leads to a high diversity of habitats and species that could be vulnerable to anthropogenic climate change (ACC). Currently, global climate models (GCMs) do not capture local climate gradients given their coarse (>25 km) resolution, limiting their utility for decision makers who wish to plan for and adapt to ACC. Dynamic downscaling is an approach for resolving, using RCMs, the effects of local-scale forcings that are not resolved by GCMs and which can mediate the local response to global ACC. However, the uncertainty associated with the differences between RCMs, specifically their representation of convective processes, leads to the need to compare RCMs ability to capture precipitation in subtropical and tropical regions. Here we evaluate output from two reanalysis-forced RCMs centered over Puerto Rico, the Weather Research and Forecasting (WRF) model and a dual configuration of the Non-Hydrostatic Model and Regional Spectral Model (NHM / RSM). Both models are run at a maximum horizontal resolution of 2km, and the results emphasize the potential differences associated with the triggering of convection and moisture availability. WRF and NHM / RSM are comparable in capturing the annual cycle of precipitation and share a similar tendency to underestimate precipitation. In addition, results indicate that while WRF produces clouds and precipitation, there is a strong tendency to underestimate the frequency of heavy rain events. NHM / RSM underestimate precipitation, but also do not capture the frequency of rain events. The former problem with WRF suggests potential challenges regarding triggering convection with appropriate frequency in simulation. The latter problem with NHM / RSM suggests a lack of available moisture in simulation, causing total precipitation to be underestimated in the region. For studies in Puerto Rico and similar regions, WRF is recommended for the overall region, but for mountainous regions only the NHM / RSM configuration is recommended. The challenges with convection triggering in WRF also suggest that this may limit the ability of WRF to assess the impact of intense precipitation changes on Puerto Rico. However, future research regarding the causes for both regional climate model configurations' underestimate of precipitation is needed.