CCPA Precipitation Analysis and Its Cross Validation
To evaluate the merit of the methodology and the quality of the CCPA data set, cross validation is performed following a data holding approach. Regression parameters are re-estimated with the same sample pool except that the data for a particular one year period (July 1 to June 30) are excluded, and the analysis for the same period is reproduced with these new parameters. The same procedure is repeated for each of the seven years and the data set reproduced is referred to as cross validation analysis. Comparison between the cross validation analysis and the original Stage IV analysis (both aggregated to 0.125 degree resolution and 24 hours accumulation) against the CPC analysis suggests that the former is statistically closer to the CPC analysis and this is true in most individual cases.
In addition to applications in climate study and forecast verifications, the data set provides a proxy of truth for the bias correction and downscaling of NWP forecast, especially ensemble forecast. Development is in progress at EMC/NCEP to bias correct the precipitation forecast from GEFS and NAEFS.