Climate Model Simulation of Present and Future Extreme Events in Latin America and the Caribbean: What Spatial Resolution is Required?

Abstract:

Latin America and the Caribbean are at risk of extreme climate events, including flooding rains, damaging winds, drought, heat waves, and in high elevation mountainous regions, excessive snowfalls. The causes of these events are numerous – flooding rains and damaging winds are often associated with tropical cyclones, but also can occur, either separately or in tandem, due to smaller, more localized storms. Similarly, heat waves and droughts can be large scale or localized, and frequently occur together (as excessive drying can lead to enhanced heating, while enhanced heating in turn promotes additional drying). Even in the tropics, extreme snow and ice events can have severe consequences due to avalanches, and also impact water resources. Understanding and modeling the climate controls behind these extreme events requires consideration of a range of time and space scales. A common strategy is to use a global climate model (GCM) to simulate the large-scale (~100km) daily atmospheric controls on extreme events. A limited area, high resolution regional climate model (RCM) is then employed to dynamically downscale the results, so as to better incorporate the influence of topography and, secondarily, the nature of the land cover. But what resolution is required to provide the necessary results, i.e., minimize biases due to improper resolution? In conjunction with our partners from participating Latin American and Caribbean nations, we have made an extensive series of simulations, both region-wide and for individual countries, using the WRF regional climate model to downscale output from a variety of GCMs, as well as Reanalyses (as a proxy for observations). The simulations driven by the Reanalyses are used for robust model verification against actual weather station observations. The simulations driven by GCMs are designed to provide projections of future climate, including importantly how the nature and number of extreme events may change through coming decades. Our results suggest that for proper simulation of both mean climate, and importantly extreme events, a spatial resolution of 4km is required in regions of complex mountainous topography. A somewhat coarser resolution of 12km is adequate in regions without much topographic relief, and where differing land covers account for most of the spatial heterogeneity.