Afforestation and reforestation are often advocated as effective strategy for increasing terrestrial carbon sequestration and mitigating climate change. However, the estimation of their carbon sink remains uncertain and subject to considerable controversy due to the scarcity of holistic evaluations based on large-scale sampling data. In particular, understanding the interaction between biomass and soil organic carbon dynamics following forestation remains elusive. While it is widely acknowledged that forestation contributes to the expansion of the biomass carbon pool, how increasing biomass acts on soil organic carbon dynamics is still elusive. On one hand, the growth of biomass can enhance the accumulation of soil organic carbon via increased litter production. On the other hand, however, an increase in biomass may also trigger the loss of soil organic carbon due to accelerated decomposition of soil organic matter caused by plants’ nutrient acquisition. We conducted an extensive pairwise field survey in China’s green great wall—the Three-North Shelterbelt Development Program, to resolve the complex relationship between biomass and soil organic carbon dynamics following forestation and estimated the carbon sink due to forestation in northern China via a machine learning approach.

• An extensive field survey campaign to decode the secrets of forestation's carbon sequestration

We conducted a large-scale survey encompassing 163 control plots and 614 forested plots to provide a comprehensive evaluation of forestation-induced carbon sequestration in northern China. During the field campaign, 25304 trees were surveyed and 11700 soil samples were collected to measure various physicochemical properties. Leveraging this extensive dataset, we estimated the carbon sink associated with a large-scale forestation project in northern China using machine learning models and further explored the intricate relationship between biomass and soil organic carbon dynamics.

• The size of carbon sink induced by forestation in northern China

Forestation in northern China has gifted us with a nonnegligible carbon sequestration capacity. The study unveils a considerable carbon sink of 913.19±47.58 Tg C, with 74% from biomass while only 26% from soil organic carbon. When averaged over the planted area, the intensity of forest-induced carbon sink for biomass, soil organic carbon and their combination were 5.51±0.31, 1.91±0.08 and 7.42±0.39 kg C m^-2, respectively. From the first implementation of the Three-North Shelterbelt Development Program in 1978 to the sampling year 2012, the mean annual carbon sequestration due to forestation in northern China has reached 26.86 Tg C year^-1. Such a value equates to 10.3% of the terrestrial carbon sink in China and 14.9% to 19.2% of the forest carbon sink in China, showcasing the power of forestation.

• Asymmetric dynamics of biomass and soil organic carbon

The study unravels the interplay between biomass and soil organic carbon dynamics following forestation. The dynamics of biomass soil organic carbon following forestation were found to be asymmetric. Within this intricate performance, a key player emerges—the level of soil total nitrogen. Forestation resulted in more significant increases of biomass density within regions characterized by abundant soil nitrogen and older stands. In contrast, areas with limited soil nitrogen and young forest experienced only marginal enhancements in biomass density, and in some cases, even a decrease was observed. The change in soil organic carbon density was much smaller compared to the alterations in biomass density, exhibiting diverse patterns along the soil nitrogen gradient. Specifically, forestation generally decreased soil organic carbon density in soils with high nitrogen content, while it increased soil organic carbon density in soils with low nitrogen content, especially in older stands.

Interestingly, we found a trade-off between soil organic carbon and biomass carbon dynamics among different planted tree species. For tree species characterized by a fast increase in biomass carbon sequestration (e.g. L. gmelinii and P. sylvestris var. mongholica), soil organic carbon sequestration showed a large decrease along the soil nitrogen gradient, and even had a considerable carbon loss in nitrogen-rich soils. In contrast, for species with a slower increase rate of biomass carbon sequestration (e.g. P. tabuliformis), soil organic carbon only showed a minor loss in nitrogen-rich soils. All these results suggest that promotion of plant biomass growth may lead to soil carbon loss because of nutrient acquisition.

Figure 1. Dependency of changes in carbon densities induced by forestation on background soil total nitrogen density (STND). a. Relationship between changes in carbon densities of biomass and soil organic carbon (SOC) with STND. b. Relationship between changes in total organic carbon density and STND. Lines in a and b were fitted based on linear mixed model. c. Comparison of changes in carbon densities in groups with different STND values. Independent sample t-tests with correction for false discovery rates were conducted to compare the data of each group with 0. *, ** and *** indicate that the null hypothesis can be rejected at p < 0.05, 0.01 and 0.001, respectively. Error bars indicate standard errors. This figure is based on field sampling data at plot level. d. Trade-off between biomass and soil carbon dynamics among tree species. Increase rate of Δ(Biomass density) with stand age refers to the slope between Δ(Biomass density) and stand age. Change rate of ΔSOCD with STND refers to the regression slope between ΔSOCD and STND.

• The implications of our findings

The introduction of forestation contributes to an initial increase in biomass carbon sink. However, our findings demonstrate that this sink eventually decreases as soil nitrogen levels rise. Particularly, there is a substantial reduction in soil organic carbon observed in nitrogen-rich soils, implying a trade-off between biomass and soil organic carbon dynamics that is closely tied to plant nitrogen acquisition. These results have significant implications. Firstly, relying on a fixed ratio between biomass and soil organic carbon may overestimate the carbon sink potential associated with forestation. Furthermore, the assumption commonly employed in most ESMs, which assumes a positive relationship between biomass and soil organic carbon, proves to be unreliable based on our findings. Therefore, we argue that more experimental results and ecological theories should be used to improve C-N schemes within ESMs. In addition, our study also highlights the importance of careful site and species choices when aiming to maximize forestation-induced C sequestration.