Based on North American Multi Model Ensemble(NMME)and monthly average soil moisture return data of five models in Xinjiang region during 1982-2010,using ERA-interim reanalysis soil moisture data for verification,the a...Based on North American Multi Model Ensemble(NMME)and monthly average soil moisture return data of five models in Xinjiang region during 1982-2010,using ERA-interim reanalysis soil moisture data for verification,the abilities of five models and simple ensemble mean to predict winter soil moisture in Xinjiang region were evaluated.The results showed that forecasting skill of CanCM4 in single model was the best,while forecasting skill of CESM1 was the lowest.With the increase of lead time,the forecasting skills of all models gradually decreased,containing NMME.But NMME gradually reflects the advantages in longer lead time,and its forecasting skills in Lead1-Season,Lead2-Season and Lead3-Season were higher than all single model,illustrating that multi model simple ensemble mean is more useful for predicting soil moisture in longer lead time,which is helpful for improving prediction skills.In spatial distribution,correlation of all models in winter in southwest Xinjiang(Tarim Basin)and northwest region was worse.CanCM3 and CanCM4 models had higher correlation in northern Xinjiang.Meanwhile,root mean square error was also smaller in corresponding region and larger in southwest region.Forecasting skills of GFDL-CM2p1 and CESM1 in eastern area of central Xinjiang were higher,and root mean square error in this area was also smaller.This paper also could lay the foundation for further studying prediction of soil drought.展开更多
文摘Based on North American Multi Model Ensemble(NMME)and monthly average soil moisture return data of five models in Xinjiang region during 1982-2010,using ERA-interim reanalysis soil moisture data for verification,the abilities of five models and simple ensemble mean to predict winter soil moisture in Xinjiang region were evaluated.The results showed that forecasting skill of CanCM4 in single model was the best,while forecasting skill of CESM1 was the lowest.With the increase of lead time,the forecasting skills of all models gradually decreased,containing NMME.But NMME gradually reflects the advantages in longer lead time,and its forecasting skills in Lead1-Season,Lead2-Season and Lead3-Season were higher than all single model,illustrating that multi model simple ensemble mean is more useful for predicting soil moisture in longer lead time,which is helpful for improving prediction skills.In spatial distribution,correlation of all models in winter in southwest Xinjiang(Tarim Basin)and northwest region was worse.CanCM3 and CanCM4 models had higher correlation in northern Xinjiang.Meanwhile,root mean square error was also smaller in corresponding region and larger in southwest region.Forecasting skills of GFDL-CM2p1 and CESM1 in eastern area of central Xinjiang were higher,and root mean square error in this area was also smaller.This paper also could lay the foundation for further studying prediction of soil drought.
