Experimental study of stony debris flow and its feature importance with varying coarse grain and water content

Abstract

The Western Himalayas in India have witnessed increased geohazards, notably debris flows, due to increased precipitation and subsequent rapid landslides. These flows threaten flat landscapes, particularly through the deposition fans they form. The increase in debris flow hazards makes it essential to understand the changes in runout deposits with varying water content and coarser particles to better capture solid–liquid interactions at a small scale. Additionally, there is a need for prediction models to analyze key features such as coarse-grained particles and water content in shaping deposits. This study offers an experimental exploration of debris flow deposition kinematics in the Western Indian Himalayas context. Utilizing reconstituted debris material from the region, experiments were conducted using a flume setup to simulate debris flow. Subsequent machine learning and Particle Image Velocimetry (PIV) provided insights into flow dynamics and helped analyze sediment accumulation patterns. Extreme gradient boosting (XGBoost) analysis revealed the significant role of stony particles in influencing mobility, with compositions between 8 and 12% showing pronounced effects of increasing deposit thickness and width. XGBoost demonstrated high predictive accuracy, with an impressive correlation between predicted and actual values for length (r² = 0.95), thickness (r² = 0.91), and width (r² = 0.94) of deposit fans. Water content was found to negatively impact the thickness of the deposits, with a greater reduction in thickness at higher water content. However, it positively influenced the overall mobility of the debris flow. The study underscores the importance of understanding debris flow mechanisms to mitigate the associated geohazard risks.