Ejin Oasis, located in the lower reaches of the Heihe River Basin (HRB), has experienced severe ecosystem decline between the 1960s and 1990s. In response, the Chinese Government implemented the Ecological Water Diversion Project (EWDP) in 2000. To evaluate the effects of the EWDP, this study monitored changes in land use and vegetation in the Ejin Oasis since 2000 and examined driving factors behind such changes. Results demonstrated that the Ejin Oasis ecosystem generally improved between 2000 and 2011. Water body area significantly increased. Lake area of once dried-up Sogo Nuur increased to 45 km². Accordingly, vegetation cover restoration has also significantly increased. For example, the Seasonally Integrated Normalized Difference Vegetation Index (SINDVI) has shown that 31.18% of the entire study area experienced an increase in vegetation area. On the other hand, even though the EWDP has been successful in driving vegetation recovery and lake restoration, farmland reclamation has counteracted such restoration initiatives. Farmland area almost doubled between 2000 and 2011. Thus, farmland expansion management is necessary for the full restoration of the Ejin Oasis ecosystems as well as HRB sustainable development. The results of this study can provide a reference for the management of the HRB.

Land use and land cover change (LUCC) is an important issue in global environmental change and sustainable development, yet spatial simulation of LUCC remains challenging due to the land use system complexity. The cellular automata (CA) model plays a crucial role in simulating LUCC processes due to its powerful spatial computing power; however, the majority of current LUCC CA models are binary-state models that cannot provide more general information about the overall spatial pattern of LUCC. Moreover, the current LUCC CA models rarely consider background artificial irrigation in arid regions. Here, a multiple logistic-regression-based Markov cellular automata (MLRMCA) model and a multiple artificial-neural-network-based Markov cellular automata (MANNMCA) model were developed and applied to simulate complex land use evolutionary processes in an arid region oasis (Zhangye Oasis), constrained by water resources and environmental policy change, during the period 2000–2011. Results indicated that the MANNMCA model was superior to the MLRMCA model in simulated accuracy. Furthermore, combining the artificial neural network with CA more effectively captured the complex relationships between LUCC and a set of spatial driving variables. Although the MLRMCA model also showed some advantages, the MANNMCA model was more appropriate for simulating complex land use dynamics. The two integrated models were reliable, and could reflect the spatial evolution of regional LUCC. These models also have potential implications for land use planning and sustainable development in arid regions.