A framework for identifying calcium accumulation problem in cropland: Integrating field surveys, legacy soil map, and machine learning models.

The calcium accumulation problem (CAP) in cinnamon soil regions of northern China significantly impacts crop yields. Identifying and mitigating CAP is crucial for improving soil quality and agricultural productivity. This study, based on field research in Aohan Banner, Chifeng City, utilizes legacy soil maps to construct a CAP dataset and evaluates the predictive performance of several machine learning models. The influence of topography on CAP is also analyzed. Key findings include: (1) In the study area, CAP predominantly manifests as block formations in dry land. Of the surveyed farmers, 58% report CAP in their cropland, with 84% noting reduced yields, though 76% have not implemented any specific mitigation measures. (2) Evaluation of machine learning models shows that tree-based models (BRT and XGBoost) outperform others in predicting CAP, with BRT demonstrating superior mapping capabilities. (3) Spatial analysis reveals that CAP is more common in the eastern and central regions of Aohan Banner, particularly in terrains such as slopes, ridges, and peaks. Additionally, the cold-to-hot zone ratio increases significantly as terrain transitions from dry to humid. (4) Regression analysis shows a strong negative correlation between terrain variables (e.g., MRVBF and GEO) and the likelihood of CAP. A further analysis indicates that CAP is more likely to occur in areas with higher soil erosion risk. These findings provide valuable insights for identifying CAP in regional soil mapping and for guiding future research in this area.