Identifying sensitivity of factor cluster based gully erosion susceptibility models.

Area of Science:

• Earth Science
• Environmental Science
• Geomorphology

Background:

• Gully erosion poses a significant threat to agricultural land and food security.
• Understanding erosion susceptibility is vital for effective land management and conservation efforts.

Purpose of the Study:

• To map gully erosion susceptibility in the Mayurakshi river basin.
• To identify the sensitivity of different factor clusters (erodibility, erosivity, resistance, topography) to gully erosion.
• To determine the dominant contributing factors to gully erosion susceptibility.

Main Methods:

• Integration of 18 parameters into four factor clusters: erodibility, erosivity, resistance, and topographical.
• Application of four machine learning models: Random Forest (RF), Gradient Boost (GBM), Extreme Gradient Boost (XGB), and Support Vector Machine (SVM).
• Sensitivity analysis to identify dominant contributing factors.

Main Results:

• Approximately 20% and 25% of the upper catchment fall into extreme and high gully erosion susceptibility zones, respectively.
• Random Forest (RF) emerged as the best-performing machine learning model.
• Erosivity and erodibility clusters showed the highest spatial association with the final susceptibility model.
• Geology (unclassified granite gneiss) and soil texture were identified as dominant factors.

Conclusions:

• The study provides a detailed gully erosion susceptibility map for the Mayurakshi river basin.
• Geology and soil texture are critical factors driving gully erosion in the region.
• Findings are essential for developing targeted gully erosion control measures and ensuring agricultural sustainability and food security.