Abstract: Land degradation (desertification) in arid region, as one of the global ecological challenges, is crucial to food security, environmental quality and regional natural resource management. Fortunately, the development of remote sensing technology greatly improves the ability of land degradation information extraction and assessment, which can effectively reveal conditions of land degradation in arid region and provide the scientific basis for land degradation trend prediction and formulating the corresponding preventing measures. In fact, the essence of land degradation is the decrease of ecological service following the change of type, quantity, structure and function of land use/cover which emerges from interactions of the coupled human-environment (H-E) factors at multiple spatio-temporal scales, leading the difficulty of classification and assessment. This prompts an attempt to reduce the complexity of the assessment by identifying a limited suite number of processes and variables which makes the problem tractable at particular scale like the interaction process of soil and vegetation, the core of the degradation. Based the standard spectral endmember space of GF-1 satellite, this paper innovatively established three levels evaluation system of land degradation state in arid areas, including land use/cover structure, degradation type and degradation degree. Secondly, we realized dryland multi-temporary linear spectral mixture analysis. The remote sensing images of GF-1/WFV endmember fractions time series were applied to characterize the quality attributes of the underlying surface and organize the classification knowledge, and then to complete the fine land cover/use classification of Minqin county in 2015. Based on land use/cover mapping in arid region, the time series EMs including vegetation (GV), sand land (SL), saline land (SA), dark surface (DA) were combined to obtain and express the variables reflecting the ecosystem function, and then, were organized with decision tree (DT) for degradation state mapping. Finally, the results of land degradation state were assessed using photos of the site landscape and the measured data of the sampling sites. The results showed that: 1) The MESMA method could decompose four stable EMs types, which were consistent with the physical components of the surface cover in study area. The model could effectively simulate the spectral information of study area. More importantly, it solved the problem that the bands number of GF-1 WFV was few and the information mining was insufficient. And which provides a basis for establishing knowledge for decision tree classification and assessment. 2) The time series EMs in standard spectral endmember space could perfectly highlight the temporal interaction characteristics details between vegetation types and habitat and the ability of remote sensing to identify land degradation types and degree, with corresponding accuracy of 87.5% and 78.7%, respectively. 3) For Minqin, a typical dryland system, the sandification process and the sand-salinization process were the dominant land degradation processes. Light sandification and moderate sandification were the dominant degradation degrees. Through the relative evaluation results of land degradation types and degree, it helps divide the key control areas, ecological conservation areas and the use of decompression areas. The research result contributes to the sustainable development of regional management and control of ecological fragile areas, land space development and ecological environment protection. General remote-sensing technical framework for land degradation assessment was confirmed that it have the potential to be applied to the study of land degradation assessment in arid regions across time and regions.