Background National forest parks represent natural ecosystems with demonstrated therapeutic potential for public mental health. As multifunctional wellness spaces, these protected areas contribute to emotional regulation, stress mitigation, and psychological wellbeing enhancement. Current research provides essential foundations for optimizing health landscape configurations through quantitative analysis of forest environmental factors and mental health outcomes. Methods This study employed structural equation modeling to investigate mediating pathways through which national forest park environments influence mental health. Grounded in Stress Recovery Theory and Attention Restoration Theory, the analysis used a 2025 survey dataset comprising 618 valid responses from Lanzhou residents. Key environmental perception dimensions were operationalized as exogenous variables to examine their direct and indirect effects on mental health indicators. Results (1) Environmental perception dimensions significantly predicted psychological wellbeing, with naturalness perception showing the strongest effect (β = −0.487, p 0.001), followed by health facility (β = −0.296, p 0.001), service quality (β = −0.124, p 0.05), and tranquility perceptions (β = −0.108, p 0.10); (2) forest therapy participation served as a significant mediator with indirect effects ranging from 0.144 to 0.243 (all p 0.001); and (3) high-stress individuals demonstrated substantially stronger therapeutic benefits, with indirect effects 5.7–6.3 times larger than low-stress counterparts and significantly enhanced participation-wellbeing relationships ( p 0.001). Conclusion The findings indicate substantial mental health benefits from national forest park environments. Strategic planning should prioritize: (1) enhancing naturalness through ecological restoration; (2) optimizing landscape configuration for sensory tranquility; and (3) expanding health-promoting infrastructure. These interventions may enhance ecosystem services in addressing urban mental health challenges.
Meng et al. (Thu,) studied this question.