The accuracy of grassland vegetation biomass estimated model based on multi-source remote sensing data ——As a case of experimental area in Sangke grassland in Xiahe County

Based on the remote sensing data of MODIS, Landsat-8 OLI and HJ-1A/1B CCD combined with ground observation data during 2013 and 2014 in the experimental area of Xiahe County, Gansu Province, the alpine grassland biomass inversion models were established and evaluated, and the best inversion models based on different remote sensing vegetation indexes were selected. Meanwhile, the spatial and temporal variations of grassland biomass in experimental area from 2000 to 2013 were also analyzed. The results showed that the best grassland biomass inversion model was logarithmic function of Landsat-8 OLI NDVI data(y=727.54lnx1+495.23, R2=0.772,RMSE=31.333 kg DM·hm-2). With spatial resolution of 30 m and 250 m, estimated biomass based on MODIS NDVI, MODIS EVI, HJ-1A/1B CCD NDVI and Landsat-8 OLI NDVI, were all higher than ground observation biomass. The estimated biomass based on logarithmic function of Landsat-8 OLI NDVI was the most close to observation biomass. The total biomass in the experimental area had great changes from 2000 to 2013. As a whole, the total biomass of grassland has a significant increasing trend (R2 = 0.590 7, P<0.001) from 2000 to 2013, and the average increasing rate reached to 50.57 kg DM ·hm-2·a-1.