Abstract: Soil moisture is the main driver of sand-fixing vegetation,which is affected by the interaction of topography,vegetation and climate. In this study,soil moisture in semi-fixed sand dunes in the Gurbantunggut Desert was used as the research object,and the relationship between soil moisture and topography-vegetation factors at different micro-landform types and depths was studied using generalized linear,generalized additive and random forest models. The results showed that:Soil moisture,except for the surface layer (0~40 cm) on the windward slope,had no significant difference between different micro-landform types and depths of the dunes. Compared with vegetation factors,topographic factors had more significant effect on soil moisture at different depths. The slope and elevation difference of topographic factors showed a negative trend on soil moisture,while the slope aspect showed a positive trend. Shrub coverage and biomass of vegetation factors also showed a negative trend on soil moisture. For the evaluation of soil moisture at different depths,the random forest model was better than the generalized linear and generalized additive models. The results of this study provided theoretical guidance for the future protection of sand-fixing vegetation in this area.

Key words: Soil moisture, Impact factor, Correlation analysis, Generalized linear models, Generalized additive model, Random forest model

摘要： 土壤水分是固沙植被的主要驱动因子，受地形、植被和气候的相互作用影响。本研究以古尔班通古特沙漠半固定沙丘上的土壤水分为研究对象，采用广义线性、广义加性和随机森林等模型，研究了不同微地貌类型和深度的土壤水分与地形-植被因子之间的关系。结果表明：除了迎风坡上表层(0~40 cm)土壤水分之外，沙丘不同微地貌类型和深度的土壤水分之间差异不显著；与植被因子相比，地形因子对不同深度土壤水分的影响更显著；地形因子坡度和高差与土壤水分含量呈负相关关系，而坡向与土壤水分含量呈正相关关系；植被因子的灌木盖度和生物量也对土壤水分呈现负向趋势；针对不同深度的土壤水分与地形-植被因子的关系建模上，随机森林模型优于广义线性和广义加性模型。研究结果为该地区未来固沙植被保护工作提供了理论指导。

关键词: 土壤水分, 影响因子, 相关性分析, 广义线性模型, 广义加性模型, 随机森林模型