Abstract: The frangibility of eco-environment and interference of socio-economy in arid and semi-arid areas often causes high tendency to land degradation and distinct variations in regional landscape between seasons. Moreover, constrained to the spatial resolution of remotely sensed data, high spatial heterogeneity of landscape indicating desertification status often exists in the form of mixture pixels with remote sensing assessment. Hence, traditional methods including pixel-based classification have their limitations. This paper taking Minqin as an example expanded the linear mixture analysis based on multi-seasonal data for extracting seasonal characteristics of desertification heterogeneity including determination of types of end-members, selection of multi-seasonal representative end-members and optimization of end-member spectra. The results indicated that the multi-seasonal linear mixture model allowing for the comparability between seasons and physical meaning of decomposition results could be interpreted physically and effectively simulate the distribution of landscape elements in Minqin oasis and the estimated end-member fractions of seasons could describe seasonal variations of oasis landscape, which could offer support for land desertification dynamic assessment.