Answers are grounded in this paper and cite their sources.
Habitat quality and carbon density showed contrasting environmental associations, with several SHAP zero crossings recurring across three boosting models. Background and Objectives: Nonlinear ecosystem-service relationships can reveal spatial heterogeneity relevant to conservation, but their sensitivity to model choice remains an important source of uncertainty. Materials and Methods: We combined the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model, 12 machine-learning algorithms, and Shapley additive explanations (SHAP). Model fitting and SHAP interpretation used five historical snapshots from 2000 to 2020; extrapolated estimates for 2025 were used only for illustrative temporal comparison. Results: Both indicators were generally higher in the southeastern than northwestern Qilian Mountains. Mean annual temperature and elevation ranked highest for habitat quality, with zero crossings near −4 °C and 3.84 km. Normalized difference vegetation index ranked highest for carbon density, with zero crossings at 0.16–0.17; evapotranspiration crossings occurred around 147–150 mm. Precipitation boundaries varied more among models. Conclusions: The contrasting associations identify candidate environmental gradients for monitoring and restoration assessment. Agreement among boosting models supports consistency within this model family, while the crossings remain statistical descriptors of model-derived indicators rather than demonstrated ecological tipping points.
No takes yet. Share an insight, caveat, or question.
An et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: