Cropland carbon stocks driven by soil characteristics, rainfall and elevation

Area of Science:

• Environmental Science
• Soil Science
• Climate Science

Background:

• Soil organic carbon (SOC) plays a crucial role in regulating atmospheric CO2 concentrations and mitigating global climate change.
• Understanding the factors influencing cropland SOC stocks is essential for reducing carbon loss and enhancing carbon sequestration.
• Cropland management strategies require detailed knowledge of environmental drivers impacting soil carbon dynamics.

Purpose of the Study:

• To investigate the influence of 16 environmental variables on SOC stocks and sequestration in croplands.
• To compare the performance of machine learning models (MLR, RF, XGBOOST) for predicting SOC stocks.
• To elucidate the direct and indirect mechanisms driving SOC stocks using a structural equation model (SEM).

Main Methods:

• Utilized 2875 observed soil samples from cropland topsoil in Hunan Province, China (2010).
• Applied three machine learning methods: Multiple Linear Regression (MLR), Random Forest (RF), and Extreme Gradient Boosting (XGBOOST).
• Employed a grid-based Structural Equation Model (SEM) to analyze environmental variable impacts on SOC stocks.

Main Results:

• XGBOOST demonstrated the highest accuracy in predicting SOC stocks, explaining 66% of the variation.
• High SOC stocks were found in low-altitude, water-sufficient cropland areas.
• Elevation was the most influential natural factor, impacting SOC stocks primarily through soil properties; precipitation directly influenced SOC stocks.

Conclusions:

• The combined SEM and machine learning approach effectively explains SOC accumulation mechanisms.
• Environmental factors like elevation and precipitation significantly drive cropland SOC stocks through direct and indirect pathways.
• This modeling approach offers valuable insights for mitigating cropland soil carbon loss and informing climate change mitigation efforts.