Multimodel output statistical downscaling prediction of precipitation in the Philippines and Thailand

摘要

Six dynamical seasonal model outputs, which are currently used in the APEC Climate Center Multimodel Ensemble (MME) prediction system, are employed for statistical downscaling prediction of station-scale precipitation in the Philippines and Thailand. Correlation analysis and Singular Value Decomposition Analysis are used to reveal atmosphere dynamic linkage based on the observed data other than model data. The observed linkage provides a robust basis for the choice of predictor and its range in predicted fields. In order to avoid spatial shift of predicted field away from observed climate, a movable window is set to select the most sensible area within the range of predictor for downscaling. The downscaled MME prediction is verified against observed station precipitation in a cross-validation manner, and the prediction skill is apparently improved compared with the simple composite of raw model predictions for most of the stations.