Statistical downscaling in the Tropics and Mid-latitudes: a comparative assessment for generating regional information on climate change.

Climate change impact and adaptation studies make use of future climate simulations by Global Climate Models (GCMs). Nevertheless, their coarse resolution makes it necessary to apply some sort of downscaling on them. There are two common approaches for this purpose, dynamical and statistical downscaling (SD), with their particular strengths and limitations. The computational cheapness of SD compared to dynamical downscaling, which allows to explore uncertainties through the generation of large ensembles, as well as its capability to downscale to single point scale, makes this option commonly used for impact and adaptation studies. SD has been extensively evaluated and applied in the extra tropics, but few experiences exist in tropical regions. In this study four state-of-the-art methods belonging to different families (Model Output Statistics, Analog, Transfer Function and Weather Generators) have been evaluated for maximum/minimum daily temperature and daily accumulated precipitation in two regions with very different climates: Spain (Mid-latitudes) and Central America (Tropics). Some key assumptions of SD have been tested: the strength of the predictors/predictand links, the skill of different approaches, the extrapolation capability of each method, the reliability of the GCMs themselves in each region, etc. Although SD has been found to be less skilful in the Tropics, it still adds important value over the raw projections by GCMs. No significant evidence of different reliability by GCMs in both regions has been detected, although this specific question might need a more detailed analysis. Relevant predictors in each region have been found to differ from one to another region, which was expected due to the different climate drivers in both regions. And finally, some methods have been found to behave significantly differently in each region.