Assessment of CMIP6 in Representing Precipitation Patterns in Peru

Climate change is affecting precipitation patterns around the world, especially in Andean tropical countries such as Peru, causing extreme hydrological events that put at risk ecosystems and communities. The aim of this study is to assess the simulation of precipitation from CMIP6 models in Peru and its three drainage regions during the historical period from 1995 to 2014. We assessed the performance of 10 CMIP6 models considering monthly and mean annual rainfall during the reference period for each region. This assessment was conducted through the comparison of monthly series to determine correlation, standard deviation, and RMSE values and through a relative error mapping based on the inverse distance weighting method for an annual scale. It was determined that Ec-Earth3 is the best representative model, especially in the total area of Peru and the Titicaca drainage region for obtaining a high correlation of 0.87, low RMSE, and a standard deviation close to the observed series. This model also obtained the highest correlation for the Pacific region but a great overestimation determined by its elevated RMSE value, while in the Atlantic, Ec-Earth3 presented a strong correlation yet not the best. It was found that models performed best over the Atlantic region, with correlation values in a range from 0.78 to 0.87, higher than in the other areas. Future work will focus on assessing downscaling techniques, using novel approaches such as machine learning, to reduce the relative error and bias of the simulations of the most representative CMIP6 models for Peru found in this research.