An early warning signal for grassland degradation on the Qinghai-Tibetan Plateau

The Qinghai-Tibetan Plateau (QTP) comprises the world’s largest alpine grassland system, supporting major ecosystem services and the livelihoods of millions of pastoralists. Rapid warming and changing precipitation, together with intensified human activities and increased livestock numbers, have raised concerns about grassland degradation characterized by declines in plant and soil quality and ecosystem functioning. Traditional small-herd nomadic grazing was generally sustainable, but recent overgrazing is a key driver of degradation. To provide an early warning and management tool, the study proposes a stocking rate threshold—defined via grassland net primary productivity (NPP)—that would trigger extreme degradation. The research asks: where current grazing is sustainable relative to this threshold; how quickly degradation may occur under various grazing intensities; and how climate change and elevated CO2 may modify risks. The goal is to guide sustainable grazing patterns regionally across the QTP, complementing national restoration programs with process-based, grid-scale diagnostics.

Past work on the QTP has linked grazing to changes in vegetation cover, species diversity, aboveground productivity, soil carbon, respiration, and nutrient dynamics at local scales, but extrapolation to regional management thresholds is uncertain. Process-based ecosystem models, which integrate biogeophysics, biogeochemistry, vegetation dynamics, and land–atmosphere exchanges, have shown promise for capturing productivity responses to climate and management. National conservation programs (e.g., Grain to Green; Retire Livestock and Restore Pastures; Return Grazing to Grassland) have reduced pressures in some areas, yet a robust, regional early-warning threshold for sustainable stocking remains lacking. Literature also indicates that aridity and overgrazing interact to accelerate degradation, and that climate warming and humidification trends could enhance productivity and carrying capacity in some contexts, though compensatory growth and CO2 responses remain uncertain.

Model: The TRIPLEX-GHG dynamic global vegetation model, building on IBIS modules, was used to simulate vegetation dynamics, land surface processes, soil biogeochemistry, and greenhouse gas cycles. Vegetation is represented by plant functional types including C3 and C4 grasses. Canopy photosynthesis follows Farquhar-type formulations; nitrogen constraints are included via soil N availability modifiers and dynamic C:N ratios. Indicator: Net primary productivity (NPP) was used as a proxy for ecosystem functioning and degradation. NPP equals gross primary productivity minus plant maintenance and growth respiration, with photosynthesis limited by light, Rubisco capacity, and CO2 (for C4 also CO2 limitation), and modified by nitrogen availability. Grazing module: Grazing seasons were defined as summer (May–October) and winter (November–April). Biomass consumption depends on stocking rate (in sheep units per hectare per year), livestock weight, intake rate (2% body weight day−1), and digestibility (0.65 live, 0.45 dead). Summer consumption draws primarily from live biomass (95% live, 5% dead), using an exponential function to vary uptake with available aboveground biomass. Winter consumption draws from standing dead/litter, with livestock assumed to lose 25% body weight over winter, and daily intake offsets maintenance respiration minus prescribed weight loss. Trampling reduces aboveground biomass proportionally to stocking rate (0.8% effect coefficient). Livestock body weight is updated daily based on intake, digestibility, and maintenance costs. Stocking rate mapping: County-level annual stocking rates (1980–2017) were derived from provincial/prefecture statistical yearbooks, converting livestock numbers to sheep units. Stocking rate per county accounts for edible grassland fraction and seasonal pasture fractions, with dynamic grassland area from ESA CCI Land Cover time series. Spatial resolution for modeling and drivers was ~0.08333° (~10 km). Forcing data: Daily climate (precipitation, temperatures, humidity, radiation, wind) was generated by spline interpolation from 2400 stations for China (1960–2017) and subset for the QTP. Soil texture came from high-resolution national maps, and initial grassland distribution from the 1:1,000,000 vegetation map. A 400-year spin-up with 1960–1990 average climate equilibrated carbon pools. Transient simulations used daily climate 1960–2017, with grazing applied from 1980 onward. Threshold detection: For each grassland grid cell, 31 simulations were run with stocking rates from 0.0 to 15.0 sheep units ha−1 yr−1 in 0.5 increments, over 120 years using mean 1980–2017 climate and CO2. The stocking rate threshold is defined as the rate at which NPP declines to 1% of the pre-grazing baseline, indicating extreme degradation. From these runs, maps of thresholds and time until degradation at specified stocking levels were produced. Future thresholds were similarly computed under RCP2.6, 4.5, and 8.5 using 2020–2100 average climate and CO2 (from 24 GCMs) and a scenario with climate and CO2 fixed at 2020 levels. Model evaluation: Simulated thresholds were compared with heavy grazing levels from 38 field sites (slope 0.923, R2 = 0.77, p < 0.01). Simulated and observed aboveground biomass across 15 sites showed good agreement (R2 = 0.74, p < 0.01). Simulated GPP matched eddy covariance at three flux sites (R2 > 0.74, p < 0.001). Simulated multi-year NPP correlated with AVHRR (R2 = 0.62) and MODIS (R2 = 0.47), and with their mean (R2 ~ 0.6).

- Spatial thresholds: Stocking rate thresholds increase from northwest to southeast; lowest in the northwest (<1.5 SU ha−1 yr−1) and highest in the east (>9 SU ha−1 yr−1). Most of the Plateau falls between 1 and 4 SU ha−1 yr−1. - Current status vs threshold: Approximately 80% of QTP grasslands have current stocking rates below the local threshold, especially centrally, but in about 55% of these areas the stocking rate already exceeds half of the threshold. Overgrazed areas (actual > threshold) cluster in the northwest, south, and northeast. - Degradation timing: If grazing is maintained at the threshold, degradation occurs within <20 years in the northwest, >80 years in the south and southeast, and 40–80 years across much of the Plateau. At 10% or 30% above threshold, most grasslands degrade within 20–50 years or 10–20 years, respectively, accelerating degradation by ~25 or ~40 years compared with threshold-level grazing. In currently overgrazed areas, 83.0% would have degraded within 20 years and 96.2% within 40 years under actual rates (grazing applied since 1980), implying many have already degraded; cutting to threshold extends time to >60 years in the south and northeast but remains <10 years in the northwest. - Climate and CO2 offsets: Positive effects of climate change and elevated CO2 offset grazing losses in 68.3% of grasslands, nearly all under normal grazing (98.5% of offset areas; 83.6% of normally grazed area). Only 5.3% of overgrazed areas experience offsets. In offset areas, 69.2% have a stocking-to-threshold ratio <0.6 and 83.0% <0.7. In non-offset areas (31.7% of grasslands), 58.5% are overgrazed with ratios 2–6; among normally grazed non-offset areas, most ratios exceed 0.5–0.6. - Management window: Setting the stocking-to-threshold ratio between 0.5 and 0.7 preserves 70–80% of normally grazed areas that benefit from climate/CO2 offsets and maximizes the area that can be converted from non-offset to offset. An additional simulation enforcing ratio ≤0.6 where it exceeded 0.6 increased offset areas to 89.1%, converting ~70% of prior non-offset areas (from 78% of overgrazed and 58% of normally grazed non-offset areas). - Future thresholds: Under RCP2.6, 4.5, and 8.5, thresholds increase relative to current climate, with the greatest expansion of high-threshold areas under RCP8.5; areas with thresholds >4 SU ha−1 yr−1 and >7 SU ha−1 yr−1 expand notably, especially in the south and southeast. - Policy-relevant recommendation: A stocking rate not exceeding about 60% of the modeled threshold (range 50–70%) is suggested to balance human demand with grassland protection, leveraging positive climate and CO2 effects only where below-threshold grazing is maintained.

The study introduces a regional, process-based stocking rate threshold grounded in NPP dynamics to identify where grazing is sustainable and to forecast time until extreme degradation. Unlike carrying capacity, which aims for sustainable yield, this threshold marks an unsustainable boundary beyond which extreme degradation (NPP ~1% of baseline) is expected. Spatial gradients in thresholds reflect climatic productivity differences between alpine meadow (southeast, wetter, higher productivity) and alpine steppe (northwest, drier, lower productivity), consistent with observed aridity–overgrazing interactions. Results indicate most grasslands are currently below the threshold, but many operate near it (>50% of threshold), implying vulnerability to modest increases. Overgrazed zones, especially in the northwest, likely already experienced degradation given historical grazing trajectories. Climate change and CO2 fertilization can buffer grazing impacts only where stocking remains below threshold; they cannot compensate overgrazing. Thus, managing stocking at 50–70% of the local threshold optimizes resilience and harnesses beneficial climate/CO2 effects, offering a practical, spatially explicit early warning and planning tool. The threshold maps and time-to-degradation diagnostics can inform interventions (e.g., stocking reductions, temporal grazing adjustments, targeted fencing) and prioritize regions for intensified management.

By integrating a grazing module into a process-based vegetation model, the study quantifies a productivity-based stocking rate threshold for each grid cell across the Qinghai-Tibetan Plateau and estimates time until extreme degradation under varying grazing intensities. It shows that although most areas remain below thresholds, many are near critical levels; overgrazed areas are concentrated in the northwest, south, and northeast. Climate change and elevated CO2 can partially offset grazing impacts only when stocking is below threshold. Management aligning stocking at roughly 50–70% of the threshold (about 60% on average) can balance livelihood needs with ecosystem protection, serving as an early warning framework under current and future climates. Future research should refine thresholds using multiple degradation indicators, incorporate additional human disturbances, fencing effects, trait-based species responses, compensatory growth, nutrient constraints, and expanded field validation to improve accuracy and applicability.

- Thresholds consider only grazing; other human disturbances (urbanization, infrastructure, reclamation, medicinal plant harvesting) are excluded, likely causing overestimation of true thresholds and underestimation of degradation risk. - Compensatory growth under light to moderate grazing is not modeled; evidence is mixed and context-dependent. - Fencing distribution and temporal effects are not incorporated; localized overgrazing around fences is unmodeled. - Uncertainties in grass responses to elevated CO2 and nutrient limitations (N, P) could alter productivity responses and thresholds. - Vegetation is simplified to two grass PFTs; species-specific traits and palatability/grazing tolerance are not represented. - Scale mismatches exist between model grids (~10 km) and field plots (hectares) and flux footprints; heavy stocking in experiments may not directly correspond to modeled thresholds. - Model relies on gridded climate interpolations and land cover datasets with inherent uncertainties.