Landscape changes in northern China's arid and semi-arid regions significantly affect ecological security and sustainable development. Understanding vegetation, lakes, and sandy area interactions is essential for landscape management. Using remote sensing data spanning 1980-2024 across 5.6 million km(2), we integrated object-based classification, landscape metrics, Grey Relational Analysis (GRA), and Geodetector to quantify landscape dynamics and driving forces. Four key findings emerged. (1) Landscape evolution comprised three human-dominated phases: agricultural expansion (1980-2024) increased cropland by 7.59% (1980 baseline) while vegetation cover decreased by 2.66%; urbanization (2010-2024) expanded built-up land by 57.65%; a spatial pattern of western desertification and eastern urbanization emerged. (2) Fragmentation increased: patch density (PD) rose 18.70%, Shannon's diversity index (SHDI) declined 9.2%; lakes experienced severe connectivity loss, with PD nearly doubling (+94%). (3) High vegetation cover stabilized sandy areas (r = 0.57, GRA = 0.74) and lakes (r = 0.58, GRA = 0.50); low cover correlated with degradation (r = -0.88 with sandy areas). Over the study period, 42% of small lakes (<1 km(2)) disappeared, and high-cover vegetation declined 34.70%. (4) Human activities - population (q = 0.54) and GDP (q = 0.49) - were the dominant driving forces, with precipitation secondary (q = 0.32). These findings reveal a cascade of agricultural expansion, urbanization, and subsequent ecological constraints under intensifying human-environment interactions, providing a scientific basis for ecological management. The methodology is transferable, and quantified thresholds offer targets for global desertification control.