On Grass Yield Remote Sensing Estimation Models of China's Northern Farming-Pastoral Ecotone

摘要

On the basis of grassland zoning, using the NASA MODIS data and the 668 ground sample data from mid July to September 2005, this paper took China's northern farming-pastoral ecotone as its subject of study and built the linear, nonlinear models and BP neural network models by using 5 vegetation indexes (NDVI, EVI, MSAVI, OSAVI and SAVI) and thereby proposed a whole set of feasible methods to estimate the grass yields in China's northern farming-pastoral ecotone. Some conclusions are drawn: (i) The zoned models are superior to the non-zoned models for reflecting the actual grass yield condition better in China's northern farming-pastoral ecotone; (ii) The grass yield estimation models based on BP neural network are superior to the linear and nonlinear models, and more accurate and most suitable for estimation the grass yields of China's northern farming-pastoral ecotone; (iii) NDVI and SAVI have the highest precision of fitting with the sample biomass and thereby, they are the vegetation indexes suited most to be applied in grass yield remote sensing estimation of China's northern farming-pastoral ecotone.