China's arid and semi-arid regions are facing severe land degradation. To combat desertification, China has launched large-scale ecological engineering, such as the Three-North Shelterbelt Program (TNSP). This study aims to uncover the feedback mechanisms of ecological spatial networks on ecological engineering effectiveness and analyze the spatial patterns of ecological sources and their ecosystem service responses. It also optimizes techniques for sand fixation, carbon sequestration, and desertification prevention. This study uses multi-source remote sensing and vector data (1986-2021) and complex network theory to extract ecological spatial networks and analyze TNSP's spatiotemporal trends. The network motif discovery algorithm identifies spatial patterns and explores the relationships between pattern structure, distribution, and topological properties. Finally, it analyzes the response mechanisms between spatial patterns and ecosystem functions. The results show that since the TNSP's implementation, the number of ecological sources (up by 653, with an area increase of 78,014 km2) and corridors (up by 1,579, with a length increase of 29,591 km) have increased. Spatiotemporal changes in ecological networks reflect ecological engineering effectiveness and guide future optimization. The star, corelinked loop, and triangle patterns can enhance network stability. The star pattern exhibits the highest degree (4.66) and betweenness centrality (53,086), while the triangle and core-linked loop patterns have higher clustering coefficients of 0.315 and 0.250, respectively. Forests demonstrate the strongest sand fixation and carbon sequestration capacities across different patterns. Shrub sources under linear and star patterns significantly enhance ecological value and network stability.