Study on statistic models of drought prediction for grassland in Inner Mongolia

Using evapotranspiration and precipitation as drought index, statistic models of growth season from April to September were founded based on 34 years daily data from 1971 to 2004 from 20 weather stations and other factors related to soil moisture and pasture yield in Inner Mongolia, 30 yearsreturn test and 4 years trial forecast were also conducted. The results showed that daily statistical prediction model was able to better simulate trends of difference between evapotranspiration and precipitation. And by return test and trial forecast, the mean relative errors of April to August model (R20.96) were all less than 10% and the mean accuracy were all above 90%. September daily statistical correlation coefficient was lower, which might be related to lack of consideration of the relationships between abnormal weather events, the atmospheric circulation, the context of future climate, biological and nonbiological factors and grassland drought.