Evaluation of dynamically downscaled ensemble climate simulations for Vietnam

Abstract

ABSTRACTThis article presents an assessment of the skill of regional climate model PRECIS in simulating seasonal climate over Vietnam. The simulations were conducted at a horizontal resolution of 25 km × 25 km. The model simulations were forced by the ERA‐Interim reanalysis and five members of the Hadley Centre's perturbed physics ensemble (PPE). CRU, APHRODITE, ERA40 datasets and observations recorded at 61 meteorological stations over Vietnam were used to validate the model. The analysis compared seasonal averages of observed and simulated precipitation, temperature, 850 hPa wind speed and direction, as well as the 99th percentile of daily precipitation and the 95th and 5th percentile of daily minimum and maximum temperatures. Annual cycles of temperature and precipitation, and the interannual variability of precipitation were also assessed. The reanalysis‐driven simulation accurately reproduced most of the important characteristics of the observed spatial patterns and annual cycles of circulation rainfall and temperature as well as capturing key characteristics of interannual variability in rainfall and of extremes in precipitation and temperature. Some apparent systematic cool biases were found most likely to be an artefact of inadequacies in the CRU‐gridded temperature observations. The regional model was found to introduce some systematic wet‐biases in rainfall. The five GCM driven simulations demonstrated errors with similar characteristics to the ERA‐Interim‐driven simulations, although with diversity in the magnitude of those errors resulting from the differences in the characteristics of the different members of the HadCM3‐based PPE. By assessing the skill of these models at producing realistic baseline simulations, we gain valuable contextual information to guide the application and interpretation of the future projections over Vietnam generated using these models.