# Multi-scale effects of soil and water conservation on runoff and sediment transport in a Chinese loess plateau basin

## TL;DR

This study examines how soil and water conservation measures affect runoff and sediment in a Chinese basin over multiple time scales, revealing seasonal and long-term patterns.

## Contribution

The study introduces a data-driven framework using Random Forest and SHAP analysis to quantify multi-scale SWC impacts and identify benefit thresholds.

## Key findings

- Runoff reduction peaked at 53.8% in July, while sediment reduction remained above 84% year-round.
- Antecedent rainfall had a stronger influence on runoff and sediment than current-day rainfall.
- Significant runoff reduction occurred when terrace coverage reached 4.74 × 10⁴ hm², and peak sediment reduction at 18.94 × 10⁴ hm² forest area.

## Abstract

Quantifying the multi-scale effects of Soil and Water Conservation (SWC) measures is essential for managing soil erosion and water resources in the Yellow River Basin. This study evaluates the multi-scale (daily, monthly, annual) impacts of SWC on runoff and sediment reduction in the Sanchuan River Basin from 1960 to 2019 by integrating a Random Forest (RF) model with SHapley Additive exPlanations (SHAP) analysis. Key findings include: (1) Runoff reduction showed distinct seasonal variation, peaking at 53.8% in July, whereas sediment reduction remained consistently high (> 84%) year-round. (2) SHAP analysis quantitatively demonstrated that antecedent rainfall (R1d) exerted a stronger influence than current-day rainfall (R0d) on both runoff and sediment responses, underscoring the importance of cumulative hydrological conditions—a finding robust despite moderate model R² values (0.552 for runoff, 0.452 for sediment). (3) The analysis revealed two distinct benefit thresholds: significant runoff reduction emerged during 2001–2003 when terrace coverage reached 4.74 × 10⁴ hm², while peak sediment reduction occurred in 2013–2015 when forest area attained 18.94 × 10⁴ hm². These thresholds, derived from polynomial trend analysis and validated by Pettitt change-point detection, mark periods when cumulative SWC implementation is associated with significant hydrological benefits. The study provides a mechanistic, data-driven framework for understanding SWC effects and offers scale-specific, quantitative targets for adaptive watershed management in erosion-prone regions.