<p>Permafrost degradation has profound implications for hydrological dynamics. Understanding influences of meteorological factors on soil-water content is essential for capturing active layer dynamics. While extensive research has examined individual meteorological factors, there remains a gap in comprehensively investigating how multiple factors and their interactions collectively influence soil-water content. In this study, we enhance understanding of the influences of meteorological factors on soil-water content by analyzing the combined effects and interactions of various factors in the Tanggula region, Tibetan Plateau using explainable machine learning. We first predict the soil-water content of shallow soils by considering multiple meteorological factors using the Extreme Gradient Boosting (XGBoost) model. We then analyze the individual contributions and interactions of meteorological factors to soil-water content using SHapley Additive exPlanations (SHAP). Results indicate that soil temperature (average SHAP value: 0.0414), water vapor pressure (0.0312), and net radiation (0.0295) are the dominant factors influencing soil-water content. Specifically, soil-water content significantly increases by nearly 35% when soil temperature exceeds 0&#xa0;°C; soil-water content rises by approximately 11% when water vapor pressure exceeds 0.86&#xa0;kPa; and when net radiation surpasses 400&#xa0;W/m², soil-water content increases by 12%. Soil temperature also exhibits strong interactions with other factors, amplifying or diminishing their effects on soil-water content. Conversely, precipitation has the limited impact on soil-water content, primarily due to the low annual rainfall and high evaporation rates in the Tanggula region. These insights from the Tanggula region highlight the importance of key meteorological factors and their interactions in governing soil-water content dynamics in permafrost regions.</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

Influence of meteorological factors on soil-water content in permafrost regions using explainable machine learning: insights from the Tanggula region, Tibetan Plateau

  • Yue Lu,
  • Gang Mei,
  • Zhengjing Ma,
  • Nengxiong Xu,
  • Jianbing Peng

摘要

Permafrost degradation has profound implications for hydrological dynamics. Understanding influences of meteorological factors on soil-water content is essential for capturing active layer dynamics. While extensive research has examined individual meteorological factors, there remains a gap in comprehensively investigating how multiple factors and their interactions collectively influence soil-water content. In this study, we enhance understanding of the influences of meteorological factors on soil-water content by analyzing the combined effects and interactions of various factors in the Tanggula region, Tibetan Plateau using explainable machine learning. We first predict the soil-water content of shallow soils by considering multiple meteorological factors using the Extreme Gradient Boosting (XGBoost) model. We then analyze the individual contributions and interactions of meteorological factors to soil-water content using SHapley Additive exPlanations (SHAP). Results indicate that soil temperature (average SHAP value: 0.0414), water vapor pressure (0.0312), and net radiation (0.0295) are the dominant factors influencing soil-water content. Specifically, soil-water content significantly increases by nearly 35% when soil temperature exceeds 0 °C; soil-water content rises by approximately 11% when water vapor pressure exceeds 0.86 kPa; and when net radiation surpasses 400 W/m², soil-water content increases by 12%. Soil temperature also exhibits strong interactions with other factors, amplifying or diminishing their effects on soil-water content. Conversely, precipitation has the limited impact on soil-water content, primarily due to the low annual rainfall and high evaporation rates in the Tanggula region. These insights from the Tanggula region highlight the importance of key meteorological factors and their interactions in governing soil-water content dynamics in permafrost regions.