Mapping dynamics of soil organic matter in croplands with MODIS data and machine learning algorithms

Area of Science:

• Soil Science
• Environmental Science
• Remote Sensing

Background:

• Soil organic matter (SOM) is a key indicator of soil quality, impacting land productivity and ecosystem health.
• Accurate regional SOM mapping is vital for sustainable agriculture and land management.
• Few studies have addressed multi-year SOM mapping and temporal dynamics.

Purpose of the Study:

• To compare four machine learning algorithms (DT, BDT, RF, GBRT) for mapping SOM in Hubei, China.
• To analyze the temporal dynamics of SOM over an 18-year period (2000-2017).
• To identify spatial patterns and driving factors of SOM changes.

Main Methods:

• Utilized soil sampling data and four machine learning algorithms: Decision Tree (DT), Bagging Decision Tree (BDT), Random Forest (RF), and Gradient Boosting Regression Trees (GBRT).
• Compared algorithm performance using metrics like R², ME, RMSE, and LCCC.
• Employed the best-performing model (GBRT) with spatially explicit explanatory variables (climate, terrain, remote sensing) to predict SOM across 500m x 500m grid cells.

Main Results:

• GBRT demonstrated superior performance in mapping SOM distribution compared to RF.
• SOM content in Hubei ranged from 0.89 to 58.86 g/kg, averaging 20.52 g/kg.
• A province-wide increasing trend in cropland SOM was observed (0.26 g/kg increase, 1.28% growth rate) from 2000-2017.
• Spatial analysis revealed SOM increases in southern Hubei and decreases in central and northern regions.
• Decreasing SOM in northern Hubei was linked to reclaimed cropland, while urbanization impacted high-quality cropland in the east.

Conclusions:

• Gradient Boosting Regression Trees (GBRT) is a highly effective algorithm for regional SOM mapping and temporal dynamic analysis.
• SOM content exhibits significant spatial and temporal variability influenced by agricultural practices and land use changes.
• Findings highlight the need for targeted management strategies to mitigate SOM decline in specific regions and preserve soil health.