Scholarlink

<p>Gully erosion poses a significant environmental challenge, requiring accurate identification of high-risk areas for effective management. This study aimed to map gully erosion susceptibility and compare the performance of five entropy models: Shannon, Tsallis, Rényi, Cross-Validation Combined, and Similarity-Based Entropy. After extracting 35 influential variables using Google Earth Engine (GEE) data and mitigating severe multicollinearity (by removing 13 variables with VIF &gt; 10), objective weights for the 22 final factors were calculated using different entropy methods. The models were evaluated using an independent set of 396 newly field-collected points. Results indicated that the Similarity-Based Entropy model achieved the highest performance metrics among the five models tested, including accuracy (0.6237), precision (0.6541), area under the ROC curve (AUC = 0.684), F1-Score (0.5826), Matthews correlation coefficient (0.251), geometric mean (0.615), and balanced accuracy (0.624), although its advantage over Tsallis entropy (AUC = 0.661) was modest. This model demonstrated balanced performance in distinguishing high-risk and low-risk zones, supported by a stable and well-distributed weighting scheme that emphasized key variables such as flow accumulation (Flow_Accum), time of concentration (TC), height above the nearest drainage (HAND), upstream catchment area (Catcharea), and stream power index (SPI). The Tsallis entropy model showed competitive performance, ranking second. In conclusion, the Similarity-Based Entropy approach, by considering structural relationships among variables, is proposed as a promising method that, after further validation in other regions, could be useful for conservation planning and resource allocation.</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

Comparative assessment of similarity-based, Tsallis, Rényi, Shannon, and cross-validation combined entropy models for gully erosion susceptibility mapping

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月15日

查看全文
  • Mohammad Kazemi,
  • Narges Kariminejad,
  • Adolfo Quesada-Román

摘要

Gully erosion poses a significant environmental challenge, requiring accurate identification of high-risk areas for effective management. This study aimed to map gully erosion susceptibility and compare the performance of five entropy models: Shannon, Tsallis, Rényi, Cross-Validation Combined, and Similarity-Based Entropy. After extracting 35 influential variables using Google Earth Engine (GEE) data and mitigating severe multicollinearity (by removing 13 variables with VIF > 10), objective weights for the 22 final factors were calculated using different entropy methods. The models were evaluated using an independent set of 396 newly field-collected points. Results indicated that the Similarity-Based Entropy model achieved the highest performance metrics among the five models tested, including accuracy (0.6237), precision (0.6541), area under the ROC curve (AUC = 0.684), F1-Score (0.5826), Matthews correlation coefficient (0.251), geometric mean (0.615), and balanced accuracy (0.624), although its advantage over Tsallis entropy (AUC = 0.661) was modest. This model demonstrated balanced performance in distinguishing high-risk and low-risk zones, supported by a stable and well-distributed weighting scheme that emphasized key variables such as flow accumulation (Flow_Accum), time of concentration (TC), height above the nearest drainage (HAND), upstream catchment area (Catcharea), and stream power index (SPI). The Tsallis entropy model showed competitive performance, ranking second. In conclusion, the Similarity-Based Entropy approach, by considering structural relationships among variables, is proposed as a promising method that, after further validation in other regions, could be useful for conservation planning and resource allocation.