Spatiotemporal changes of soil organic carbon in intensive croplands over three decades: Emerging role of farmland utilization shifts

Abstract

Understanding spatiotemporal variation characteristics and key driving factors of soil organic carbon (SOC) is crucial for refined managements of farmland quality and carbon emissions. Nevertheless, the impact of human-induced farmland utilization activities on regional SOC dynamics remains unclear. In this study, based on 1577 farmland topsoil (0–20 cm) samples, we developed a two-tiered stratification-contextualized framework within digital soil mapping paradigm for determining the drivers for spatiotemporal SOC dynamics by partitioning farmland and constructing individual machine learning method within farmland units (FUs) in East China during the 1980s-2010s, and then estimated spatiotemporal patterns of SOC as well as evaluated drivers on SOC changes within FUs using random forest models. Our results showed that the temporal changes in topsoil SOC stocks exhibited high spatial heterogeneity across three decades. The average SOC densities for the 1980s, 2000s, and 2010s were 41.4 ± 9.3 C ha⁻¹, 47.4 ± 8.6 C ha⁻¹ and 39.7 ± 14.4 C ha⁻¹, respectively, with SOC densities initially increasing and then decreasing in our intensively cultivated region. Climatic changes accounted for > 75 % of the relative importance (RI) to SOC dynamics over the past 30 years. Farmland utilization shifts accelerated temporal SOC changes, with the effects coefficient increasing from 2.6 % (95 % CI: 1.7∼3.1 %) to 6.4 % (95 % CI: 3.9∼7.4 %). Induced by farmland utilization shifts, the overall changes of SOC stock increased by 0.10 Mt C during the 1980s-2000s with minimal SOC changes in FU3 and parts of FU1 (only 1 %), while decreased by 0.33 Mt C during the 2000s-2010s with approximately 12.9 % of regions in FU3, FU5, FU1 and FU6 exhibiting changes over 3 %. This work enhanced the understanding of spatiotemporal SOC variability induced by farmland utilization using machine learning method based on determined FUs, which also provided valuable guidance for soil monitoring and carbon accounting management for intensively cultivated farmland.