Work

Digital mapping of soil organic carbon in a plain area based on time-series features

Year: 2025

Type: article

Abstract: Enhancing the accuracy of digital soil organic carbon (SOC) mapping in plain Areas is crucial for agricultural development and environmental protection. It is important to obtain spatial variability information on SOC efficiently and comprehensively. Digital soil mapping (DSM) is a vital method for monitoring the SOC. However, the use of traditional landscape co-variates to estimate soil properties is not suitable for plain areas due to the high variability of soil properties in similar topographic and vegetation conditions. This study explored whether time-series environmental factors can effectively improve the accuracy of SOC mapping. Using Fengqiu County as the study area, this study employed topographical, climatic, and ecological variables alongside the time-series features of multiple ecological factors as environmental covariates. SOC prediction models were constructed using the random forest (RF), backpropagation neural network (BP), and support vector machine (SVM) methods to test the effectiveness of time-series data in enhancing SOC mapping accuracy. Among the environmental variables, ecological factors such as the normalized vegetation, drought, and humidity indices, along with the distance from rivers, play a dominant role in SOC mapping. The relative importance of the time-series features of the ecological factors was higher than that of single-time-point vegetation indices. Introducing the time-series features of ecological factors resulted in a decrease in the mean error (ME) and root mean square error (RMSE), whereas the coefficient of determination (R2) and concordance correlation coefficient (CCC) showed increasing trends across different models. A comparative analysis of variable selection methods (stepwise variance regression, the Boruta algorithm, and the SPA algorithm) revealed that the Boruta algorithm selected the optimal variable set, resulting in the greatest improvement in accuracy for various models. The RF time-series model (RFSTB)showed the most significant improvement, with R2 increasing by 65.45% and the RMSE decreasing by 47.12%. Introducing time-series ecological factor variables reduced the prediction errors and increased the model reliability to varying degrees. This study introduces new environmental variables for SOC mapping and provides new insights into digital SOC mapping in the plain areas.

Source: Ecological Indicators

DOI: 10.1016/j.ecolind.2025.113215

Authors: Kun Yan, Decai Wang, Yongkang Feng, Siyu Hou, Yamei Zhang, Huimin Yang

Institution: Henan Agricultural University

Language: English

Topic: Soil Geostatistics and Mapping

Subfield: Environmental Engineering

Field: Environmental Science

Domain: Physical Sciences

SDGs: Life on land, Climate action

Open Access status: gold

Estimated APC paid: $3,650