Land use and cover change accelerated China's land carbon sinks limits soil carbon

The study investigates how land use and cover change (LUCC) has shaped China’s terrestrial carbon dynamics, particularly soil organic carbon (SOC), gross primary productivity (GPP), and net ecosystem productivity (NEP). Human-driven LUCC contributes substantially to global carbon emissions and alters land-atmosphere interactions, yet its impacts on regional carbon budgets remain uncertain due to methodological inconsistencies and limited long-term LUCC datasets. SOC is a critical carbon pool for ecosystem functioning and climate mitigation, and LUCC can either deplete or enhance SOC depending on conversion types and management. China has experienced extensive LUCC since 1900, including major afforestation and ecological restoration initiatives (e.g., Grain for Green, Three-North Shelterbelt, Natural Forest Conservation), but the precise contribution of LUCC to China’s carbon budget is unresolved. The objective is to quantify LUCC impacts on SOC and land carbon fluxes in China from 1979–2014 by integrating a three-level meta-analysis of SOC responses with process-based Earth system modeling, and to disentangle LUCC-driven effects from those driven by climate, CO₂ fertilization, and nitrogen deposition.

The paper reviews approaches to estimate LUCC effects on the carbon cycle, including IPCC Tier 1 list methods, remote sensing, and process-based ecosystem models such as Dynamic Global Vegetation Models used in the Global Carbon Budget. The Land-Use Harmonization (LUH2) dataset, built from FAO reports and HYDE, provides historical land-use trajectories but carries uncertainties. Prior studies show large model spread in LUCC-attributed land carbon fluxes due to input and structural uncertainties. Meta-analyses have been effective in synthesizing SOC responses to LUCC globally, revealing consistent SOC losses following conversions to cropland and gains following restoration, but with climate and soil context dependencies. For China, previous assessments were limited by oversimplified forest dynamics, inconsistent methods, and insufficient consideration of management and forest types, motivating a combined meta-analysis and modeling framework to improve regional estimates.

The study combines a three-level meta-analysis of SOC responses to LUCC in China with process-based simulations using CESM2/CLM5. - Meta-analysis: The authors systematically searched Web of Science, Scopus, CNKI, and Google Scholar (Nov 9, 2023; updated Jan 1, 2024) for meta-analyses on SOC stocks or concentrations related to LUCC, retaining 132 publications contributing 1248 paired observations across Chinese ecosystems. Inclusion criteria: analysis of bulk SOC stocks/concentrations; consistent stratified sampling to at least 10 cm depth; reported precision (SE or CI). SOM values were converted to SOC using SOC = SOM × 0.58. - Three-level random-effects model: Accounts for sampling variance (level 1), within-study variance (level 2), and between-study variance (level 3). Effect sizes were weighted by inverse variance, with additional quality-based weighting (quality effects) and a variance–covariance matrix to handle non-independence due to shared primary studies (pseudo-correlation based on overlap). Model selection used AIC. Publication bias was assessed via funnel plots, Tandem Procedure, fail-safe N, Egger’s and Begg’s tests (all p > 0.05). Analyses used metafor (rma.mv) in R. - Statistical analyses of moderators: Simple linear regression tested relationships between SOC response ratio (RRsoc) and mean annual temperature (MAT), mean annual precipitation (MAP), elevation, bulk density (BD), soil pH, and duration since conversion. A meta-forest (random forest adapted for meta-analysis) evaluated variable importance (IncMSE%) among seven predictors (climate, soil properties including BD and initial pH, and duration), with missing data imputed via missForest. Correlation analyses and structural equation modeling (SEM; lavaan) quantified direct and indirect effects of climate, soil, latitude, elevation, and duration on RRsoc; model fit indices were evaluated. - Modeling: Two CLM5 experiments (with LUCC vs no-LUCC) were run at 0.25° × 0.25° over 1979–2014. Forcing: China Meteorological Forcing Dataset (CMFD). LUCC trajectories: LUH2 (including secondary forests, wood harvest, and agricultural expansion). Vegetation represented by 16 plant functional types. Outputs included annual GPP, NEP, and SOC stocks at 0–20 cm and 0–100 cm. - Validation: Simulated GPP was compared against nine ChinaFlux tower sites (2003–2010), yielding R² = 0.8 (p < 0.001). Inventory-based forest carbon stock changes at national and site levels further supported model performance. - Data processing and trends: Annual means of GPP, NEP, and SOC were computed for LUCC and no-LUCC scenarios. Trends were assessed with the Mann–Kendall test and Sen’s slope. Additional analyses quantified fluxes from wood product pools to litter and fossil fuel-related pools associated with LUCC.

- SOC response to LUCC: - Overall, LUCC resulted in a 39.2% loss in SOC change across China. Conversions to cropland caused the largest losses: grassland→cropland (−56%) exceeded forest→cropland (−37%). - Restoration increased SOC: cropland→forest (+11%), cropland→grassland (+13%), cropland→shrubland (+22%). SOC losses were smaller in tropical climates than in temperate and cold plateau regions. LUCC improved SOC in acidic and neutral soils (InRR = 0.38, p < 0.01; 0.47, p < 0.05). - Loess Plateau showed significant positive SOC response to LUCC, aligned with large-scale ecological restoration policies (e.g., GFGP). - Moderator effects: MAP correlated positively with RRsoc; MAT negatively. Higher elevation associated with lower SOC change. BD positively correlated with SOC response (R² ≈ 0.03, p < 0.001), but LUCC effects on SOC were negative when BD > 1.6 g cm⁻³. Longer duration since conversion correlated with lower RRsoc (negative recovery effect over time). - Meta-forest and SEM: Soil bulk density was the most influential factor; duration was second. Climate impacted SOC mainly indirectly via environmental gradients (latitude, elevation). Soil properties exerted direct positive effects on SOC RRs after LUCC. - Carbon fluxes and sinks (1979–2014): - GPP: LUCC increased GPP, with a trend of ~0.02 Pg C yr⁻¹, about double the no-LUCC trend (~0.01 Pg C yr⁻¹). LUCC led to substantial increases in GPP, with high values (>1500 g C m⁻² yr⁻¹) in southeast China; reductions occurred in rapidly urbanizing regions (central China, Yangtze and Pearl River deltas). - NEP: LUCC-induced annual NEP averaged 2.53 g C m⁻² yr⁻¹, contributing to 0.02 ± 0.12 Pg C yr⁻¹ of national carbon uptake. In contrast, no-LUCC NEP was −0.06 ± 0.16 Pg C yr⁻¹. About 80% of China’s area functioned as a carbon sink under LUCC, with strongest sinks in southwest China and mountainous regions (Daxing’anling, Xiaoxing’anling, Changbai Mountains). - SOC stocks: Topsoil (0–20 cm) SOC was 73.33 ± 0.45 Pg C (no LUCC) vs 71.73 ± 0.56 Pg C (with LUCC), indicating a 1.6 Pg C loss attributable to LUCC; no significant difference in LUCC effects between 0–20 cm and 20–100 cm layers. Spatial decreases occurred in parts of Jilin, Xinjiang, and Shanghai, while increases appeared in Tianjin, Jiangsu, Henan, and Shandong. - Product pools and litter/CWD fluxes: LUCC increased carbon fluxes to wood product pools, with a cumulative addition of ~40.16 Pg C yr⁻¹ (cumulative over 1979–2014; regional maxima noted, e.g., Hainan). - Overall, LUCC has significantly enhanced China’s terrestrial carbon sink despite concomitant SOC losses, underscoring the role of afforestation and reforestation.

The study demonstrates that LUCC has been a dominant driver of China’s land carbon sink since 1979, boosting GPP and NEP beyond what would be expected from climate, CO₂ fertilization, and nitrogen deposition alone. However, this enhancement in atmospheric carbon uptake co-occurred with significant SOC losses (notably when ecosystems were converted to croplands), indicating a trade-off between aboveground carbon gains and belowground carbon depletion. Afforestation and ecological restoration (e.g., Grain for Green) effectively increased SOC in targeted regions such as the Loess Plateau and contributed to land carbon uptake, partially offsetting urbanization-induced declines in central and coastal regions. The analyses highlight soil bulk density and conversion duration as critical mediators of SOC responses, with climate effects acting largely indirectly via environmental gradients. The integration of meta-analysis with process-based modeling reduces uncertainty and aligns simulated outcomes with empirical evidence, providing a robust depiction of spatially heterogeneous LUCC impacts on China’s carbon cycle.

By integrating a three-level meta-analysis with CESM2/CLM5 simulations, the study provides comprehensive evidence that LUCC has accelerated China’s land carbon sinks from 1979–2014, while simultaneously limiting soil carbon through SOC losses. Afforestation and reforestation, especially under national programs, have been effective in enhancing SOC and strengthening carbon sinks, though responses vary by region, climate, soil properties, and land-use history. The findings emphasize the importance of sustained ecological restoration and conservation to maximize carbon sequestration benefits and mitigate climate change. Future research should refine model representations of plant carbon allocation, tissue turnover and mortality, and soil carbon–nutrient cycling, and improve historical LUCC inputs to better partition above- versus below-ground carbon sinks.

Key limitations stem from uncertainties in LUH2 historical land-use data, model initial conditions, and structural/process representations in CESM2/CLM5. The partitioning of carbon sinks between above- and below-ground pools remains uncertain due to simplified internal cycling mechanisms. Although validated against flux towers and inventories and constrained by 1248 observed SOC effect sizes, residual uncertainties persist in parameterization, forest management representation, and regional LUCC histories. The study recommends improving input data accuracy and advancing process-based representations of plant allocation, tissue lifespan/mortality, and soil carbon–nutrient dynamics.