Estimation of Grass Yield Based on MODIS Data in Sichuan Province, China

Grassland in Sichuan Province is one of the five most important pastoral areas in China.Available natural grassland accounts for 85% of the total grassland area of Sichuan Province.Accurate grass yield is significant for management of grassland and local economics.In this study,data of MODIS-NDVI/EVI with different resolution(250m,500m,1km) in July of 2011 was utilized to model the relationship between grass yield of the four main types(i.e.,alpine meadow type,alpine shrub grassland,alpine swamp grassland and mountain drain grassland) and RS data,respectively.Vegetation coverage combined with NDVI/EVI was used to reduce the effect of mixed pixel.The main conclusions are as follows: NDVI was better than EVI to establish the relationship.Data of 250m was ideal among data of different resolution tested.Model for each main type of grassland was better than model for all the four grassland types.Exponential model was better to simulate the relationship compared with other models for the four main types of grassland except for alpine swamp grassland.And for alpine swamp grassland,power function model was better.Optimization model for each of the four main type grassland was obtained.Accuracy of each optimization model was more than 70% while R2was larger than 0.75.Then these optimization models were utilized to estimate grass yield of Sichuan Province in 2011 and the accuracy of estimation was about 90%.This suggests that these optimization models are feasible and the accuracy is satisfied in practice.