The meridional extent and complex orography of the South American continent contributes to a wide diversity of climate regimes ranging from hyper-arid deserts to tropical rainforests to sub-polar highland regions. In addition, South American meteorology and climate are also made further complicated by ENSO, a powerful coupled ocean-atmosphere phenomenon. Modelling studies in this region have typically resorted to either atmospheric mesoscale or atmosphere-ocean coupled global climate models. The latter offers full physics and high spatial resolution, but it is computationally inefficient typically lack an interactive ocean, whereas the former offers high computational efficiency and ocean-atmosphere coupling, but it lacks adequate spatial and temporal resolution to adequate resolve the complex orography and explicitly simulate precipitation. Explicit simulation of precipitation is vital in the Central Andes where rainfall rates are light (0.5-5 mm hr-1), there is strong seasonality, and most precipitation is associated with weak mesoscale-organized convection. Recent increases in both computational power and model development have led to the advent of coupled ocean-atmosphere mesoscale models for both weather and climate study applications. These modelling systems, while computationally expensive, include two-way ocean-atmosphere coupling, high resolution, and explicit simulation of precipitation. In this study, we use the Coupled Ocean-Atmosphere-Wave-Sediment Transport (COAWST), a fully-coupled mesoscale atmosphere-ocean modeling system. Previous work using COAWST has demonstrated its ability to reproduce the current climate’s seasonal cycle (Oct 2003 – Oct 2004) in the Central Andes using historical CMIP5 model data as boundary conditions on a 27-/9-km grid configuration (Outer grid extent: 60.4°S to 17.7°N and 118.6°W to 17.4°W).
Our investigation of the mesoscale features of future precipitation regimes (spatial distribution, precipitation rates, and diurnal cycle) of the Central Andes involved boundary conditions from three CMIP5 models (MIROC5, GFDL-ESM2M, and CCSM). Prior to running simulation with these models, we first ran year-long (Oct 2031 – Oct 2032) COAWST simulations with MIROC5 data from three regional climate pathways (RCP) (4.5, 6.0, and 8.5) on its boundaries. All three simulations show similar cumulative distribution curves of precipitation, however the curves from RCP 6.0 and 8.5 were shifted increasingly rightward mainly in response to an enhanced South Atlantic Convergence Zone. This rightward shift motivated our choice of the moderate RCP 6.0 scenario for future climate simulations. Because running multi-year simulations with a mesoscale model is expensive computationally, we generated October to October “snapshots” in the years 2031, 2059, and 2087 and then used CMIP5 model output to analyze large-scale trends between these yearly “snapshots” with hourly precipitation output. We show a strong dependence between the simulated future climate and the CMIP5 model boundary conditions with differences up to 120 precipitation days (>5 mm/day) and 400 mm less rainfall annually in the Central Andes between simulations. These results are likely associated with differences in simulated diurnal convection in the western Amazon which is dynamically linked to the intensity of the Bolivian High (a key moisture transport mechanism for the Central Andes). Regardless of these differences, COAWST simulations show little to no change in the timing of diurnal precipitation, yet the magnitude of diurnal precipitation, the total annual precipitation, and the number of precipitation days tended to decrease in future years in the Central Andes. Specifically, in 2087, annual precipitation is simulated to have fallen 100 mm and there were 30 fewer precipitation days when compared to 2003. Cumulative distribution functions of regional precipitation show increased variability in annual precipitation where both low and high-end precipitation are seen to increase, yet locations with moderate annual precipitation decrease.