Abstract: The paper aims to establish the level of regional meteorological prediction model of rice leaf roller in different plant areas of rice and in different rice growth stages. Based on the data of meteorology and monitoring on rice leaf roller during 2000?-2014 in 103 observation stations, using the principal component analysis and correlation analysis and stepwise regression method by SPSS, the occurrence and development level of meteorological prediction model and the migration level of meteorological prediction model were established in different plant areas of rice and in different rice growth stages.The results showed that, early rice in South China as an example, the key meteorological factors which affected the occurrence and development level of meteorological prediction model in transplanting to tillering stage were cumulative precipitation in the third ten days of March, average temperature in the first and third ten days of April, average temperature and the number days of highest temperature greater than 30℃. The key meteorological factors which affected the migration were average relative humidity of April, average temperature and average relative humidity in the third ten days of March, average hours of sunshine in the first ten days of April. The validation results by the meteorological dates of 2000-2012 return test showed that the forecasting accuracy of the occurrence and development level of meteorological prediction model was above 80% in transplanting to tillering stage and heading to flowering stage in different plant areas of rice; meanwhile the forecasting accuracy of the migration level of meteorological prediction model was above 85%. The meteorological dates of 2013-2014 were used to forecast which showed that the forecasting accuracy of the occurrence and development level of meteorological prediction model was still above 80%. Yet the forecasting accuracy of the migration level of meteorological prediction model was above 78%.The forecasting accuracy was general declined with less number of samples of prediction stations, especially in single season rice in Southwest. The rice prediction models could be put into the operational application in different plant areas of rice in China.