Generative multi-objective optimization of national park boundary growth within coupled human and natural systems: Evidence from the candidate Altai Mountains National Park.

National park boundary delineation is both a conservation planning task and a land-use governance issue, requiring coordination among ecological priorities, human activities, and existing protected-area systems. Existing approaches have primarily focused on priority-area identification, while translating these priorities into spatially explicit and implementable boundaries remains challenging in coupled human and natural systems. To address this gap, we developed a generative boundary optimization framework guided by the Non-dominated Sorting Genetic Algorithm III (NSGA-III). Within this framework, NSGA-III optimizes seed configurations and growth-preference parameters that govern a multi-seed boundary-growth process, which is implemented through region-growing decoding to progressively generate candidate boundaries under fixed-area constraints. Using the candidate Altai Mountains National Park as a case study, we generated alternative boundaries across area scenarios ranging from 10% to 90% and evaluated trade-offs among ecological value, human activity intensity, and spatial aggregation. The optimized boundaries exhibited a staged expansion trajectory from core ecological spaces to mountain ecological belts and subsequently to peripheral areas. Increasing area produced diminishing ecological gains and greater human-use conflicts, whereas spatial coherence and protected-area integration improved most under intermediate scenarios. Multi-dimensional indicator evaluation identified a Conservation-priority Scenario (20%) focused on high-value ecological cores and a Balanced Management Scenario (50%) that better balanced ecological representativeness, spatial integration, and implementation feasibility. This study advances national park boundary delineation from static priority mapping toward generative optimization of boundary-growth processes and provides an adaptable decision-support framework for conservation planning in coupled human and natural systems.