Land-use/land-cover change (LULC) fundamentally alters regional ecosystem service provisioning, necessitating multi-scenario simulations coupled with multi-objective optimization (MOP) to inform territorial spatial governance. This study develops an integrated PLUS-InVEST-MOP framework to simulate land-use dynamics and their ecological consequences in the Haibei Tibetan Autonomous Prefecture, a critical ecological barrier on the northeastern Qinghai–Tibet Plateau. Utilizing GlobeLand30 data (2000–2020), we projected 2030 land-use patterns under four scenarios (natural development, ecological conservation, economic development, and climate change) and quantified spatiotemporal variations in carbon storage and habitat quality. Results showed that grassland degradation accelerated during 2010–2020, with 1202.43 km² converted primarily to bare ground, contrasting with restoration gains observed in 2000–2010. Elevation emerged as the predominant natural driver of glacial and forestland expansion, whereas demographic and economic factors governed land encroachment on artificial surfaces. Under the climate change scenario, habitat quality declined by 3.78% (to 0.596), and carbon storage decreased by 70.04 Tg relative to 2020, underscoring the vulnerability of alpine ecosystems to warming. Conversely, the MOP-optimized scenario achieved synergistic improvements in carbon sequestration (+1.8% vs. 2020) and habitat quality (+2.3% vs. 2020) while adhering to arable land preservation targets (≥54,800 ha) and constraints on artificial surfaces. These findings provide quantitative decision support for balancing ecological conservation and sustainable development in high-altitude cold regions.