Scholarlink

<p>Endorheic lakes are critical environmental indicators, yet their responses to regional pressures vary significantly. The neighboring Lake Urmia (LUB) and Lake Van (LVB) basins present a natural experiment: LUB has experienced catastrophic desiccation, while LVB remains hydrologically stable. This study leverages this divergence to disentangle anthropogenic drivers from climatic pressures. We conducted a 22-year (2000–2022) comparative analysis using a harmonized Landsat composites. Three water indices (NDWI, AWEI, MBWI) and Land Use/Land Cover (LULC) transitions into a Random Forest regression model, interpreted via SHapley Additive exPlanations (SHAP) Results reveal vastly different transformations. LUB underwent a massive landscape conversion, losing 1.89 million water pixels while gaining 1.30 million 'Agriculture' and 1.44 million 'Bare &amp; Saline Land' pixels. Conversely, LVB remained remarkably stable. SHAP analysis pinpointed dissimilar causal mechanisms: LUB's hydrological variance was dominated (Normalized SHAP &gt; 0.38) by the 'Water to 'Bare Land' transition. In contrast, LVB's dynamics were defined by persistence in stable classes. Model performance served as a diagnostic: LUB models achieved high predictive power (R2 = 0.53–0.59), by learning the acute desiccation signal, whereas LVB models yielded negligible R2 values (0.03–0.06), confirming the absence of a strong LULC-driven hydrological signal. We conclude that the divergent fates of these basins, given their climatic similarity, are not solely a result of differing climatic trends but are strongly attributable to intense, unsustainable anthropogenic pressures, primarily agricultural expansion and water management within the Lake Urmia basin.</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

Surface water desiccation vs. hydrological equilibrium: assessing the role of LULC transitions in the Lake Urmia and Lake Van basins through machine learning

  • 期刊论文

  • 发表日期：

    2026年7月11日

查看全文
  • Mohammad Mahdi Pourhanifeh,
  • Farid Shahidinejad,
  • Hassan Khosravi

摘要

Endorheic lakes are critical environmental indicators, yet their responses to regional pressures vary significantly. The neighboring Lake Urmia (LUB) and Lake Van (LVB) basins present a natural experiment: LUB has experienced catastrophic desiccation, while LVB remains hydrologically stable. This study leverages this divergence to disentangle anthropogenic drivers from climatic pressures. We conducted a 22-year (2000–2022) comparative analysis using a harmonized Landsat composites. Three water indices (NDWI, AWEI, MBWI) and Land Use/Land Cover (LULC) transitions into a Random Forest regression model, interpreted via SHapley Additive exPlanations (SHAP) Results reveal vastly different transformations. LUB underwent a massive landscape conversion, losing 1.89 million water pixels while gaining 1.30 million 'Agriculture' and 1.44 million 'Bare & Saline Land' pixels. Conversely, LVB remained remarkably stable. SHAP analysis pinpointed dissimilar causal mechanisms: LUB's hydrological variance was dominated (Normalized SHAP > 0.38) by the 'Water to 'Bare Land' transition. In contrast, LVB's dynamics were defined by persistence in stable classes. Model performance served as a diagnostic: LUB models achieved high predictive power (R2 = 0.53–0.59), by learning the acute desiccation signal, whereas LVB models yielded negligible R2 values (0.03–0.06), confirming the absence of a strong LULC-driven hydrological signal. We conclude that the divergent fates of these basins, given their climatic similarity, are not solely a result of differing climatic trends but are strongly attributable to intense, unsustainable anthropogenic pressures, primarily agricultural expansion and water management within the Lake Urmia basin.