An Improved Statistical Downscaling Method Based On Copula And Its Application

This research includes three parts:First, basing on the flexibility of Copula in constructing joint distribution for any marginal distributions, the joint distribution of Global Climate Models (GCM) simulation and observation is established with Copula. Second, a new statistical downscaling method which utilizes Copula is suggested-Copula-Bias Correction and Spatial Downscaling (C-BCSD). This method corrects bias in GCM by constructing the conditional probability distribution with the established joint distribution of GCM simulation and observation and obtaining the maximum conditional probability. Third, C-BCSD and BCSD are compared by simulating the precipitation of a GCM (GFDL-CM3) cell in the Columbia River basin. The simulation period is January (wet season) and July (dry season) from1979to2003. The model performances are evaluated with metrics MRE, NSE, MAE and RMSE in GCM bias correction, spatial downscaling, and temporal disaggregation. All the metrics show that simulated precipitation from C-BCSD is closer to observation data in each step than simulation from traditional BCSD.