Predicting surface soil pH spatial distribution based on three machine learning methods: a case study of Heilongjiang...

Area of Science:

• Soil Science
• Environmental Science
• Geospatial Analysis

Background:

• Accurate surface soil pH data is vital for monitoring soil degradation and informing agricultural practices.
• Understanding spatial pH distribution aids in sustainable land management and crop yield optimization.
• Heilongjiang Province, a key agricultural region, requires precise soil information for effective resource management.

Purpose of the Study:

• To comprehensively map the spatial distribution of surface soil pH in Heilongjiang Province.
• To compare the performance of Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) for soil pH prediction.
• To identify key environmental covariates influencing soil pH and determine the optimal modeling approach.

Main Methods:

• Utilized data from 125 soil survey sampling points in Heilongjiang Province.
• Employed Pearson correlation analysis and recursive feature elimination (RFE) to select key environmental covariates.
• Applied and compared three machine learning models: SVM, RF, and XGBoost for surface soil pH prediction.

Main Results:

• Mean monthly temperature maximum, mean monthly precipitation minimum, mean annual precipitation, drought index, and mean monthly wind speed maximum were identified as the most significant covariates.
• XGBoost demonstrated the highest prediction performance (R=0.705), followed by RF (R=0.688), while SVM showed instability.
• The optimal XGBoost model indicated that black soils in Heilongjiang Province generally exhibit weak acidity (average pH 6.42), with spatial trends and localized acidification in eastern/northeastern areas.

Conclusions:

• Climate variables are effective predictors of soil pH, especially in large, flat agricultural regions.
• XGBoost is the optimal machine learning model for accurate surface soil pH mapping in this study area.
• Findings highlight the need for controlled nitrogen fertilizer application and improved soil acid-base buffering capacity in affected regions of Heilongjiang Province.