Mapping uncertainty and skill in seasonal forecasts and climate data

The skill of seasonal forecasts of temperature and precipitation globally, and over Europe and Colombia were assessed. The forecasts had good skill for temperature in both locations. The skill for precipitation over Europe was low, but was notably better over Colombia. A subset of the EURO-CORDEX regional climate model ensemble was identified. Climate simulations from this subset were compared with the E-OBS gridded dataset for the whole Mediterranean area, and a more in-depth analysis was performed for the MED-GOLD study regions. The analysis focused on key climatic variables (temperature and precipitation), as well as other derived indices (numbers of hot days and nights, numbers of wet and dry days). Overall, the models reproduced the large-scale observed features, but some biases in the simulations were evident. Calibration of the model data significantly reduced the sizes of the biases. All of the climate models projected increases in temperature and numbers of hot days and nights over the MED-GOLD study regions. Reductions in precipitation were projected in spring, summer, and autumn, but some of the changes were not significant until later in the twenty-first century. A small number of simulations from the Central American regional climate model ensemble (CAM-CORDEX) were assessed using the AgMERRA dataset. Large biases in the modelled data were identified, which largely originate from the GCMs used to drive the RCMs. A different method to that used with the EURO-CORDEX data was used to calibrate the climate model data for Colombia. The benefit of the calibration is demonstrated, by comparing projections of temperature and precipitation over Colombia in raw and calibrated data