Asynchronous recovery of grassland ecosystem health in northern China: A three-dimensional assessment of structure, function, and environmental quality

Assessing ecosystem health under global change remains a central challenge in dryland ecology, as most existing approaches rely on single indices or aggregated metrics that obscure fundamental differences among ecosystem components. Here, we introduce a three-dimensional framework that explicitly separates grassland ecosystem health (GEH) into structural integrity, functional performance, and environmental quality, allowing us to diagnose not only overall trends but also hidden divergences among ecosystem dimensions. Applying this framework to the arid and semi-arid grasslands of northern China over the period 2000–2023, and combining long-term remote sensing data, ecosystem service modeling, and interpretable machine learning, we uncover a previously underappreciated pattern of ecosystem change. We show that GEH has improved overall, but through strongly asynchronous pathways: structural health and environmental quality increased consistently, whereas functional health exhibited a slight but persistent decline. This result challenges the common assumption that vegetation greening and environmental improvement reflect holistic ecosystem recovery. Instead, our findings demonstrate that ecosystem components respond at different rates and through distinct mechanisms, leading to a decoupling between structure, environment, and function. Precipitation emerged as the dominant driver, interacting nonlinearly with topography, and grazing intensity to regulate ecosystem responses. In particular, climate-driven gains in vegetation structure are partially offset by grazing pressure and soil constraints, which limit the recovery of ecosystem services. By explicitly revealing this multidimensional decoupling, our study advances ecosystem health assessment beyond conventional composite indices and identifies asynchronous recovery as a fundamental feature of dryland ecosystems under global change. Critically, this asynchronous decoupling creates a misleading signal of overall ecosystem stability while substantially lowering long-term grassland resilience against extreme drought and intensified human disturbance, generating hidden ecological vulnerability despite widespread vegetation greening. Our results highlight the need to move from structure-based restoration targets toward function-oriented management strategies. The proposed framework provides a transferable tool for diagnosing ecosystem condition and guiding adaptive management across dryland regions worldwide.