Monitoring Grassland Biomass Based On Multi-source Remote Sensing Data In Gannan Pastoral Area

Gannan Tibetan autonomous prefecture, located in the eastern edge of Tibetan Plateau, is a crucial recharge area of the Yellow River and is an important livestock supply region of Gansu province. Grassland, as the fundamental element for animal husbandry and water maintenance, plays an important role in the ecological regulation and economic development. Grassland biomass is one of the most important factors in grassland resource remote sensing research, Monitoring the dynamic change is important for the grassland degradation and forage and livestock balance research since it is able to accurately reveal the grassland growing conditions.Based on the Vegetation Indices (VI) from various remote sensing sources and grassland dry biomass data surveyed in field, we developed the retrieval models of grassland biomass, assessed the accuracy of these models, and found the optimal VI and retrieval model for Gannan. After then, we calibrated the biomass retrieval model based on MODIS (Moderate Resolution Imaging Spectroradiometer) data by using retrieved biomass from high resolution image WorldView-2as ground truth, and analyzed the grassland growing condition. Furthermore, we developed an empirical model between MODIS NDVI (Normalized Difference Vegetation Index) and AVHRR (Advanced Very High Resolution Radiometer) NDVI, calculated the MODIS NDVI from1982to1999based on the empirical model, then calculated the maximum annual grassland biomass of Gannan from1982to2011, and analyzed the grassland biomass change rate for the three periods of1982-1999,2000-2011, and1982-2011. The results are as following:1) NDVI and RVI (Ratio Vegetation Index) of MODIS, Landsat TM, WorldView-2, and HIS (HyperSpectral Image) data were well related to the grassland dry biomass. For MODIS, Landsat TM, and HIS data, the accuracy of retrieved biomass data from RVI was much better than that based on NDVI, while WorldView-2data had an opposite result.2) After comparison of the performances of21models that based on the four kinds of remotely sensed data, we found that the optimal biomass retrieval model for Gannan was the one based on the NDVI of WorldView-2and the equation is y=14129x-7801.4, and R2=0.6193. 3) From1982to2011, the area with a decreasing trend of grassland biomass is larger than that with an increasing trend. The overall change trend of Gannan grassland was degradating over the30years, in which grassland condition was seriously deteriated from1982to1999while slightly improved in some regions from2000to2011. The various grassland types had different change trend. For example, the lowland meadow was stable, the conditions of temperate grassland type and temperate meadow type were improved; meanwhile, the situations of warm hassock, marsh, alpine shrub and alpine meadow were seriously damaged.4) The calibrated biomass retrieval model of MODIS NDVI was y=1.7494x-2019, where, x is the retrieved biomass based on MODIS data.. Based on MODIS NDVI data we calculated the grassland biomass of the year2012. Overall, the Gannan grassland condition was good in2012as68.72%of grassland were level1, level2, and level3, and the grassland biomass increased from north to south.5) Comparing of the Gannan grassland biomass in the last10years (2002to2011) and the year2012, we found that the average annual dry grassland biomass in2012was higher than the average value of the last10years. The grassland condition in Lintan and Zhouqu counties was best in the eight counties, relatively good in Hezuo, and seriously bad in other counties.