Revealing the dynamic evolution of the sandy ecosystem service value and its stability network pattern has important practical guiding significance for the regional management of degraded desert and the adjustment of the agricultural， forestry and animal husbandry industry structure. Based on the land use change data since 1986， this paper statistically analyzes the temporal and spatial differentiation of the ecosystem service value （ESV） of the Mu Us Sandy Land. It takes the maximization of ecological service value as the goal and strategy， and uses the least resistance model （MCR） to discussing the evolution law of the ecological network pattern in the Mu Us Sandy Land. The results show that： （1） In the past 34 years， with the progress of the Mu Us Sandy Land oasis process， woodland has replaced desert to become the dominant land type in the Mu Us Sandy Land， and the forest and grass coverage rate reach 71.38% in 2020. （2） From 1986 to 2020， the overall ecological service value of the Mu Us Sandy Land has increased by 19.876 billion yuan. During the process of oasis， the growth rate of desert ecological service value from 1986 to 2006 and 2006 to 2020 was 251 million yuan·yr-1 and 1.062 billion yuan·yr-1 respectively； from a spatial point of view， extremely high ecological service value areas were concentrated in Henan Township and Maugaitu Sumu， and has gradually expanded to Sumitusum， Gartu Town and other areas since 2016. （3） The ecological network pattern is obviously different， namely the northwest area where the ecological network is sparse， mainly sand reclamation land， mobile sand dunes and semi-mobile sand dunes； and the southeast area with dense ecological network is covered by forests and grasses. High fixed desert and areas with rich water resources in rivers and lakes are the main areas， and with the migration of high value points of ecological service value， the stability network structure shifts to the northeast direction year by year.

Keywords： land use ; ecological service value ; ecological network pattern ; dynamic evolution ; Mu Us Sandy Land