Estimation Of Fractional Cover Of Non-photosynthetic Vegetation And Its Spatial-temporal Variations In Xilingol Grassland Based On MODIS Data

Non-photosynthetic vegetation?NPV?is a primary component of vegetation productivity in grasslands,and they not only maintain the structure,function and dynamic stability of grassland ecosystem,but also affect the circulation of carbon and water,energy flow in the ecosystem.The timely and accurate acquisition of information on grassland fractional cover of NPV(f_NPV)and its spatio-temporal variations is of great significance for monitoring and conserving of grassland ecosystems,managing grassland resources,assessing soil erosion and grassland fire risk,and preserving the grassland ecological environment.In this study,we selected Xilingol Grassland as the study area.Firstly,using the MODIS band obtained from field ground quadrat spectral simulations and MODIS image data,we evaluated the applicability and effectiveness of eight selected non-photosynthetic vegetation indices?NPVIs??NDI5,NDI7,NDTI,STI,NDSVI,MSACRI,SWIR32,and DFI?to estimate f_NPV,so as to determine the best NPVI for f_NPV estimation.Then,based on the determined NPVI,we established a remote sensing estimation f_NPV model using MODIS image data and measured f_NPV data.Finally,we used this model to estimate the f_NPV information of Xilingol grassland during the non-growth season from 2000 to 2016,and its temporal and spatial dynamic variations were analyzed.The main conclusions are summarized as follows:?1?The MODIS-based DFI has a good applicability in estimating f _NPV when PV,NPV,and BS coexisted.By analyzing several regression models of NPVIs,the correlation between the DFI and f_NPV is significant,with high R²?0.68?and low RMSECV?0.1390?;the soil-line-adjusted MSACRI?R²=0.53?has a slightly better correlation with f_NPV than the NDTI?R²=0.51?,which is not adjusted using the soil line;while the correlation between the NDI5,NDI7,and NDSVI with f _NPV are low.Compared with the NPVIs simulated by the measured spectrum,the applicability of the MODIS image-based DFI to estimate f_NPV has reduced to a certain extent,R² and RMSECV are 0.59 and 0.1081,respectively;the applicability of SWIR32,NDTI,MSACRI,STI and NDI7 are all significantly lower than the DFI;while the applicability of the NDI5 and NDSVI is very low,R² are 0.19 and 0.13,and RMSECV are 0.1443 and 0.1522,respectively.?2?The DFI-f_NPV linear regression model established using the MODIS image-based DFI and the measured f_NPV data is the best remote sensing model for estimating f_NPV in the Xilingol grassland,with an R² and RMSECV are 0.60 and0.1574,respectively,which can effectively estimate f_NPV information in the Xilingol grassland.?3?Spatial analysis shows that the spatial distribution of f_NPV in the Xilingol grassland is obviously heterogeneous,showing a pattern of high in the northeast and low in the southwest.The spatial distribution of f_NPV is also affected by the grassland type.The f_NPV values gradually decreases from the meadow grassland?0.65?,to the typical grassland?0.42?,to the sandy grassland?0.24?,to the desert grassland?0.23?.The dynamic changes in f_NPV in the Xilingol grassland from 2000 to 2016 shows that the overall trend of f_NPV increased in a fluctuating manner.An increase in f_NPV occurred in 78.57%of Xilingol grassland,of which 20.90%increased significantly,mainly distributed in the central and southern typical grasslands;the area without significant change of f_NPV accounted for about 9.22%of Xilingol grassland,which distributed sporadically;while the decrease area in f_NPVPV was found in a relatively small proportion?12.20%?of Xilingol grassland,these areas were mainly concentrated in the western desert grassland,the northeast of the typical grassland and a small part of the meadow grassland.