Phenology-based Residual Trend Analysis Of MODIS-NDVI Time Series For Assessinghuman-inducedland Degradation

With the acceleration of urbanization and the rapid change of climate,Land degradation has developed into a major global environmental problem.It poses a serious threat to the sustainable development of human beings and nature.An accurate monitoring method is necessary for governance of land degradation.The main object of this study is MODIS—NDVI time series.And we assessed the land degradation of Songnen grassland by considering phenological information of grassland.Land degradation is a long evolutionary process,and scope of occurrence is usually contagious.MODIS—NDVI(Normalized Difference Vegetation Index)time series provide basic data for land degradation monitoring.But time series cannot directly reflect the relevant information of land degradation.In fact,both human activities and climate can changes of vegetation productivity in arid and semi-arid areas.Therefore,removing the influence of climate on NDVI change is one of the important challenges in land degradation monitoring by remote sensing.RESTREND is one of the most widely used land degradation monitoring models.However,the traditional RESTREND algorithm does not consider the spatial and temporal differences of vegetation growth season length in the study area.Moreover,the influence of pre-growing season precipitation on NDVI was not considered in this method.In this study,we investigated the growth of grassland vegetation by using the NDWI(Normalized Difference Water Index)and the technology of grassland phenological signal detection.The phenological information and RESTREND model was combined to improve the RESTREND algorithm.This study selected the available MODIS09A1 data sets in the study area from 2000 to 2015 to calculate NDWI value and structure NDWI time series;then we structured the smooth and continuous NDWI time series by using smoothing-interpolation coupling method.In this paper,the current mainstream methods of phenological extraction were selected to detect the phenological information of grassland vegetation.We extract the growing season information of grassland vegetation by setting threshold.The growing season information was used to quantify the two important parameter of RESTREND: accumulated NDVI and accumulated precipitation in growth season.Considering the lagging effect of precipitation on vegetation growth,we also quantified the precipitation in pre-season.The correlation between rainfall and NDVI was determined by analyzing the regression relationship among the three parameters.Then we removed the influence of precipitation by this regression relationship and structured the residual series.The interannual variation of vegetation productivity was analyzed by trend analysis method to assess land degradation in Songnen grassland.In this paper,the RESTREND algorithm coupled with phenological information was compared with the traditional RESTREND algorithm.The results show that: The RESTREND algorithm coupled with phenological have better performance in removing the influence of precipitation.Simultaneously,this algorithm can detect the degradation area which is not monitored by the traditional RESTREND algorithm which improves the accuracy of land degradation monitoring.