Disentangling climate and fencing effects reveals aridity-dependent time lags in alpine grassland multifunctionality

Quantifying the true ecological outcomes of fencing in climate-sensitive alpine grasslands remains challenging because conventional before-after comparisons often fail to separate management effects from concurrent climatic trends , which leads to evaluation biases where regional climatic greening can inflate perceived management success. To resolve this casual attribution gap, we constructed a counterfactual attribution framework across the Changtang Plateau (2001−2020) to effectively decouple the independent effects of grazing exclusion from climate variability across five core ecosystem functions. Our findings reveal three complementary mechanisms underlying ecosystem functional recovery following grazing exclusion. First, counterfactual attribution demonstrated that grazing exclusion, rather than climate variability, predominantly controlled changes in primary productivity ( PP ) and carbon sequestration ( CS ), although its relative contribution progressively declined along the aridity gradient. Specifically, in relatively humid alpine meadows, the stable-effect stage of short-term grazing exclusion (2004–2012) additionally increased PP and CS by approximately 13.18 (95% CI [11.20, 15.16]) and 13.24 (95% CI [11.19, 15.29]) g C km −2 inside fences, respectively, with changes in PP being almost entirely attributable to grazing exclusion (102.45%). Grazing exclusion also dominated short-term recovery of soil retention ( SR ) in alpine meadows and alpine steppes. In contrast, climate exerted stronger control over SR recovery in desert steppes during the early stage, with fencing becoming the primary driver only after prolonged implementation. By comparison, the responses of water conservation ( WC ) and sand prevention ( SP ) remained highly variable, reflecting their greater sensitivity to climatic regulation. Second, the ecological benefits of grazing exclusion were governed by an aridity-dependent time-lag mechanism. While humid meadows reached functional saturation, a plateau in net functional gains occurring after the rapid recovery of photosynthetic capacity following the removal of grazing pressure, within approximately three years, and maintained stable enhanced functioning for nearly eight years. By contrast, resource-limited arid steppes required prolonged accumulation (>10 years) to overcome intrinsic restoration thresholds, with significant improvements in PP , CS , and SR emerging only during the long-term grazing exclusion stage. Third, the intensity and spatial dominance of grazing exclusion exhibited a pronounced spatial decoupling, with mean local Shannon entropy (MLSE) ranging from 0.19 to 0.25, indicating fragmented control regimes. Although grazing exclusion-dominated areas occupied a larger proportion of the landscape, the proportion of these areas progressively declined with increasing aridity, indicating that the regional influence of grazing exclusion is constrained by macroclimatic conditions. By explicitly separating management effects from climatic noise, the proposed counterfactual attribution framework advances conservation evaluation from conventional before–after comparisons toward causal attribution. Our results necessitate a paradigm shift from “one-size-fits-all” policies toward aridity-specific strategies, i.e., rotational grazing for humid meadows after functional saturation (∼eight years) and sustained exclusion over 10 years for arid steppes, providing a robust tool for optimizing dryland protected-area networks.