Soil carbon sequestration by agroforestry systems in China: A meta-analysis

China has a rich historical heritage of agroforestry, but a quantitative analysis of the potential of agroforestry systems (AFS) for soil organic carbon (SOC) sequestration is missing. A comprehensive meta-analysis of soil C sequestration rates derived from 43 studies was undertaken to determine its most influential parameters. Soil C sequestration rates were calculated for topsoils (0–20 cm, 97 sites) and at two subsoil layers (20–40 cm, 73 sites; 40–60 cm, 54 sites). The results showed highest C sequestration rates for the AFS-type shelterbelt in topsoils (0.92 Mg ha−1 yr−1), upper subsoils (0.72 Mg ha−1 yr−1) and lower subsoils (0.52 Mg ha−1 yr−1), followed by agrosilvicultural systems (0.70, 0.48 and 0.43 Mg ha−1 yr−1, respectively) and silvopastoral systems (0.23, 0.08 and 0.02 Mg ha−1 yr−1, respectively). We tested potential effects of different predictor variables (soil class, AFS-type, land use of the control site, system age, initial SOC stock, tree components, legumes and climatic properties) on soil C sequestration rates using a Random Forest regression model. We found changes in the conditional importance of the predictors for different soil layers. For both top- and subsoils, the AFS-type, initial SOC and soil class were most influential, followed by age. Other factors such as land use of the control, climate factors (climate zone, mean annual temperature, mean annual precipitation), leguminous species and tree components were of minor importance. We conclude that besides the AFS-type and the initial SOC, soil type plays a decisive role for the efficiency of soil C sequestration by agroforestry. Our meta-analysis provided evidence that existing AFS in China, particularly shelterbelts and agrosilvicultural systems, are effective practices to increase SOC stocks, both in top- and subsoils and especially in the subtropical climate zone.