He Yina, Dirk Pﬂugmachera, Ang Lib, Zhengguo Lic, Patrick Hosterta,d
Remote Sensing of Environment
Land cover/use change
Land use and land cover change Deforestation Aﬀorestation Cropland retirement MODIS time series Trajectory-based approach Inner Mongolia Machine learning Random forest Probability scores
During thepastdecades,overuseoflandresourceshasincreasingly contributedtoenvironmental crisesinChina. To mitigate wide-spreadland degradation, actionshavebeentaken tomaintain andrestoreecologically valuable landscapes such as natural forests. However, the eﬀects of the various vegetation protection policies that have been implemented in China since the late 1990′s still remain largely unknown. In this paper, we therefore focus on mapping land use and land cover change (LULCC) in Inner Mongolia, one of the key regions targeted by Chinese ecological restoration programs. We used 250-m MODIS time series and a random forest classiﬁcation approach to generate annual probabilities for each land cover class between 2000 and 2014. We then applied a trajectory-based changedetection approachbasedonamodiﬁedversionoftheLandsat-based detection oftrends in disturbance and recovery (LandTrendr) algorithm to the probability time series and mapped land cover change trajectories. We found that our trajectory-based approach achieved high accuracies (overall accuracy 0.95 ± 0.02). It provides spatial-temporal land change maps that allow a land-use related interpretation of changepatterns.Ourchangemapsshowthati)forestlossdecreasedrapidlyafter2000(from15,717 ± 1770 ha in 2001 to 1313 ± 165 ha in 2014) and forest gain (190,645 ± 28,352 ha during 2001–2014) occurred in the ecological program zones, leading to a net forest increase in Inner Mongolia, and ii) cropland retirement (212,979 ± 54,939 ha during 2001–2014) mostly occurred at the early stage of ecological programs and mainly concentrated in drier environments and steep terrain. Overall, land cover mapping and trajectory-based land use analyses allowed a consistent characterization of LULCC over large areas, which is crucial for gaining a better understanding of environmental changes in the light of rapidly changing environmental policies and governance regimes in China.