Uncertainties of soil organic carbon stock estimation caused by paleoclimate and human footprint on the Qinghai Plateau

Background

Quantifying the stock of soil organic carbon (SOC) and evaluating its potential impact factors is important to evaluating global climate change. Human disturbances and past climate are known to influence the rates of carbon fixation, soil physiochemical properties, soil microbial diversity and plant functional traits, which ultimately affect the current SOC storage. However, whether and how the paleoclimate and human disturbances affect the distribution of SOC storage on the high-altitude Tibetan Plateau remain largely unknown. Here, we took the Qinghai Plateau, the main component of the Tibetan Plateau, as our study region and applied three machine learning models (random forest, gradient boosting machine and support vector machine) to estimate the spatial and vertical distributions of the SOC stock and then evaluated the effects of the paleoclimate during the Last Glacial Maximum and the mid-Holocene periods as well as the human footprint on SOC stock at 0 to 200 cm depth by synthesizing 827 soil observations and 71 environmental factors.

Results

Our results indicate that the vegetation and modern climate are the determinant factors of SOC stocks, while paleoclimate (i.e., paleotemperature and paleoprecipitation) is more important than modern temperature, modern precipitation and the human footprint in shaping current SOC stock distributions. Specifically, the SOC stock was deeply underestimated in near natural ecosystems and overestimated in the strongly human disturbance ecosystems if the model did not consider the paleoclimate. Overall, the total SOC stock of the Qinghai Plateau was underestimated by 4.69%, 12.25% and 6.67% at depths of 0 to 100 cm, 100 to 200 cm and 0 to 200 cm, respectively. In addition, the human footprint had a weak influence on the distributions of the SOC stock. We finally estimated that the total and mean SOC stock at 200 cm depth by including the paleoclimate effects was 11.36 Pg C and 16.31 kg C m⁻², respectively, and nearly 40% SOC was distributed in the top 30 cm.

Conclusion

The paleoclimate is relatively important for the accurate modeling of current SOC stocks. Overall, our study provides a benchmark for predicting SOC stock patterns at depth and emphasizes that terrestrial carbon cycle models should incorporate information on how the paleoclimate has influenced SOC stocks.