Remote Sensing Methods For Monitoring Grassland Production In Inner Mongolia

As one of the country's five major pastoral regions, grassland in Inner Mongolia is a natural ecological barrier for Northern China, and is also China's important base for stockbreeding. To be accurately and timely get forage production over different time and space in this region is the key to achieve a reasonable, efficient, sustainable use of grassland resources. Therefore the use of remote sensing technology for Inner Mongolia grassland dynamic monitoring is of great significance. Secondly, the types of grassland in Inner Mongolia are typical, and plenty of samples were gathered. The remote sensing research of grass output based on grassland in Inner Mongolia is very useful to other similar studies.Currently, statistical models are more widely used in dynamic monitoring of grassland production. But different researchers get different results with different numbers of ground samples, and validation is also not performed in many cases. Only several remote sensing indicators are compared with the determination coefficient. The reliability of the results is difficult to guarantee. Therefore, it is necessary that production dynamic monitoring model is established in Inner Mongolia to provide a basis for management, through comparisons of a variety of remote sensing indicators and methods on the same dataset obtained in 2006.This paper based on TERRA/MODIS data, AQUA/MODIS data and a lot of ground data collected from the Inner Mongolia Autonomous Region in 2006, single-band, multi-band, vegetation index model and Multi-band model of TERRA/AQUA binaries are established. Vegetation indices include the nine vegetation indices already widely used and the twelve new vegetation indices including shortwave infrared band. A total of 450 regression equations are obtained. Through statistical judgment, sensitivity analysis and model validation, a dynamic partitioning approach is proposed which hold an advantage that in the areas with high vegetation cover, models also guarantee that vegetation index is notably sensitive to the change of production. By the combination of sensitivity analysis and the field validation, the quantitative standards of dynamic partitioning are determined. Finally the optimal model including two VIs is established, which guarantees accurate estimation to every partitioning, and overcomes the disadvantage of selecting the optimal model only through determination coefficient in the past. When production is lower than 1800 kg/hm~2, RRMSE equals to 0.25 and RRMSE equals to 264.7 kg/hm~2. When production is higher than 1800 kg/hm~2, RRMSE equals to 0.22 and RRMSE equals to 684.5 kg/hm~2. This model covers four bands and has a small amount of calculation, thus it can easily come true and has enough precision to meet real-time monitoring grassland production on a large scale.At last through the application, total grassland production in Inner Mongolia is calculated by the end of July to the beginning of August 2006.