Non-Negative Matrix and Canonical Correlation Method of Field Quantitative Prediction

A short-term climate prediction model(NMF-CCA),based on non-negative matrix factorization and canonical correlation,is designed.This model can predict meteorological elements,through establishing the relation between forecasting object and proper impact factor.By means of preceding winter snow in Tibetan Plateau from NCEP/NCAR,and forecast fields of summer rainfall in Central China from 1971 to 2008,EOF-CCA and NMF-CCA respectively are used to do cross validation in 1999-2008.The results show that two methods obtain good result,but NMF-CCA is better than EOF-CCA,by comparing average scores of ACC,Ps and Ss in ten years.The three different grading results of NMF-CCA are 0.33,76.68 and 0.12,respectively.It indicates that NMF-CCA has certain application value.