Scholarlink

<p>Future projections of water erosion are fundamental for land management in semi-arid areas. However, relying solely on rainfall erosivity (R) fluctuations in the RUSLE model can lead to misleading conclusions. In this study, conducted in the Ksob watershed, part of the Moroccan Western High Atlas, we used an integrated approach based on the RUSLE model and machine learning to predict water erosion risk in the 2030s and 2050s in relation to 2001–2020 as a reference period. Linear Regression (LR), Support Vector Machine (SVM), and Random Forest (RF) were compared to predict R and vegetation cover (C) factors under two climate change scenarios: SSP2-4.5 and SSP5-8.5. The RF model outperformed the LR and SVM models in predicting R (R² = 0.989, MAE = 0.930&#xa0;MJ.mm/ha.h.year, RMSE = 1.328&#xa0;MJ.mm/ha.h.year) and C factors (R² = 0.515, MAE = 0.085, RMSE = 0.109). Considering both periods, 70 to 80% of the watershed areas are projected to experience a decrease in R, with more pronounced reductions in the SSP5-8.5 scenario. Vegetation cover is expected to improve in 59.5 to 67.5% of areas, with negligible difference between scenarios, due to reduced rainfall erosivity in rugged terrains as well as a decrease in annual temperature range, with localized divergent responses largely governed by slope. In the 2030s (2050s), 15 to 15.5% (17.5 to 19%) of the areas will experience a change in water erosion risk, with shifts mainly to lower levels (e.g., from high to moderate-high or from moderate-low to low), while 84.5 to 85% (81 to 82.5%) of the watershed areas will remain stable when accounting for both emission scenarios. The findings of the investigation emphasize the value of integrating machine learning and C-factor predictions into projections of water erosion risk for more accurate estimates.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrated RUSLE-machine learning modeling for water erosion risk assessment under climate change in a Mediterranean semi-arid region: a comparison of LR, SVM, and RF models

  • 期刊论文

  • 发表日期：

    2026年3月24日

查看全文
  • Youssef Boussalim,
  • Youssef Dallahi

摘要

Future projections of water erosion are fundamental for land management in semi-arid areas. However, relying solely on rainfall erosivity (R) fluctuations in the RUSLE model can lead to misleading conclusions. In this study, conducted in the Ksob watershed, part of the Moroccan Western High Atlas, we used an integrated approach based on the RUSLE model and machine learning to predict water erosion risk in the 2030s and 2050s in relation to 2001–2020 as a reference period. Linear Regression (LR), Support Vector Machine (SVM), and Random Forest (RF) were compared to predict R and vegetation cover (C) factors under two climate change scenarios: SSP2-4.5 and SSP5-8.5. The RF model outperformed the LR and SVM models in predicting R (R² = 0.989, MAE = 0.930 MJ.mm/ha.h.year, RMSE = 1.328 MJ.mm/ha.h.year) and C factors (R² = 0.515, MAE = 0.085, RMSE = 0.109). Considering both periods, 70 to 80% of the watershed areas are projected to experience a decrease in R, with more pronounced reductions in the SSP5-8.5 scenario. Vegetation cover is expected to improve in 59.5 to 67.5% of areas, with negligible difference between scenarios, due to reduced rainfall erosivity in rugged terrains as well as a decrease in annual temperature range, with localized divergent responses largely governed by slope. In the 2030s (2050s), 15 to 15.5% (17.5 to 19%) of the areas will experience a change in water erosion risk, with shifts mainly to lower levels (e.g., from high to moderate-high or from moderate-low to low), while 84.5 to 85% (81 to 82.5%) of the watershed areas will remain stable when accounting for both emission scenarios. The findings of the investigation emphasize the value of integrating machine learning and C-factor predictions into projections of water erosion risk for more accurate estimates.