Enhancing the digital mapping accuracy of farmland soil organic carbon in arid areas using agricultural land use history

Abstract

Accurate digital mapping of soil organic carbon (SOC) over cultivated land is significant for estimating potential soil carbon sequestration and mitigating future climate changes. Large-scale land and water resource development in arid Northwest China has led to significant changes in SOC. Therefore, agricultural land use history, including reclamation source (RS) and cultivation year (CY), has significantly influenced SOC. However, when this information is ignored in digital mapping, biases arise. To solve this problem, this study applied RS and CY to SOC mapping and examined its impact on the results. The cultivated land of Qitai County was selected as the study area. Land use from 1980 to 2018 was superimposed to identify the agricultural land use history using our proposed methods. RS and CY were incorporated as environmental covariates and combined with other natural variables and field soil samples to predict the spatial distribution of SOC using the random forest (RF) model. The results showed that the SOC contents of plots reclaimed from high-coverage grassland and bare land were higher and lower, respectively, than other plots. The SOC increased in the short-term farming, but declined after reaching mid-and-long term. RS and CY were important environmental covariates for predicting cultivated land SOC. Incorporating RS and CY increased the mapping accuracy of SOC compared to only using natural variables. Adding RS and CY into the model resulted in an R2 increase from 0.381 to 0.469, as well as an decrease in statistical errors. In addition, RS and CY provided more spatial detail attributed to land reclamation when mapping SOC. Overall, this study provided a new and improved method for integrating human activities into digital soil mapping.