Understanding the mechanisms of soil organic carbon (SOC) sequestration in forests is vital to ecosystem carbon budgeting, and helps gain insight in the functioning and sustainable management of world forests. An explicit knowledge of the mechanisms driving global SOC sequestration in forests is still lacking because of the complex interplays between climate, soil and forest type in influencing SOC pool size and stability. Based on a synthesis of 1179 observations from 292 studies across global forests, we quantified the relative importance of climate, soil property and forest type on total SOC content and the specific contents of physical (particulate vs. mineral-associated SOC) and chemical (labile vs. recalcitrant SOC) pools in upper 10 cm mineral soils, as well as SOC stock in the O horizons. The variability in the total SOC content of the mineral soils was better explained by climate (47~60%) and soil factors (26%~50%) than by NPP (10~20%). The total SOC content and contents of particulate (POC) and recalcitrant SOC (ROC) of the mineral soils all decreased with increasing mean annual temperature because SOC decomposition overrides the C replenishment under warmer climate. The content of mineral-associated organic carbon (MAOC) was influenced by temperature, which directly affected microbial activity. Additionally, the presence of clay and iron oxides physically protected SOC by forming MAOC. The SOC stock in the O horizons was larger in the temperate zone and Mediterranean regions than in the boreal and sub/tropical zones. Mixed forests had 64% larger SOC pools than either broadleaf or coniferous forests, because of i) higher productivity, and ii) litter input from different tree species resulting in diversification of molecular composition of SOC and microbial community. While climate, soil and forest type jointly determine the formation and stability of SOC, climate predominantly controls the global patterns of SOC pools in forest ecosystems.

We used the Google Scholar (https://scholar.google.com) and China National Knowledge Infrastructure (CNKI, https://www.cnki.net) to search for published papers on studies of SOC in forest ecosystems globally. The searched keywords were “(soil organic matter OR soil carbon OR soil organic carbon physical fractions OR chemical fractions OR particulate OR mineral‐associated OR labile OR recalcitrant) AND (forest OR forest ecosystems)”. Specifically, the searched keywords were in Chinese terminology when using CNKI. The search yielded more than 1000 studies matching those keywords. To avoid bias in selection of publications and to increase the comparability of data, studies that satisfied the following criteria were incorporated into the final dataset for synthesis: (1) field studies conducted with natural and mature forests; (2) studies exclusively concerned with separation of SOC into physical (determined by density or particle size into POC and MAOC, respectively) or chemical (determined using the potassium permanganate [KMnO₄] oxidation or acid [H₂SO₄] hydrolysis into LOC and ROC) pools (Table S1); and (3) information available on forest type (i.e., broadleaf, coniferous and mixed forest), total SOC content, and at least one of the variables related to forest and environmental conditions (Table S2). The resulting dataset contains the values of total SOC content and the contents of specific SOC pools in the mineral soil at varying depths down to 30 cm, as well as SOC content and stock for the O horizon.