冻融作用通常发生在高寒区域，由于温度的变化导致土壤的水分状态和体积发生变化，土体发生冻胀融沉现象，从而造成土壤结构破坏和性质改变。冻融侵蚀是指因冻融作用而导致土壤侵蚀过程改变和强度增加的过程。在我国，冻融侵蚀总面积已达172.48万km²，占陆地面积的17.97%。已有的研究多通过遥感影像对冻融侵蚀面积进行界定，或结合室内实验和野外调查，探究冻融作用对土壤可蚀性的影响，但是对冻融作用强度的空间分布缺乏定量化研究，且对于小流域尺度冻融作用的关注较少，这导致难以定量化评价冻融作用对流域内土壤侵蚀的贡献。所以本论文关注流域尺度冻融作用强度，构建模糊逻辑综合评价体系对小流域尺度冻融作用强度空间分布进行评价，并探究冻融作用强度与地形、土地利用和土壤的关系。研究结果提高冻融作用强度评价精度，并为土地利用精细化管理与规划提供参考。

基于文献调研和野外调查，本研究选取海拔、坡向、地形湿度因子、土壤机械组成和土壤有机碳含量这几个因子评价流域尺度冻融作用强度分级。通过实测数据和Random Forest(RF)模型预测获得土壤因子空间分布数据，构建双层模糊逻辑评价体系预测冻融作用强度，探究其空间分布规律与土壤侵蚀的关系。本研究结论如下：

（1）   研究区内土壤砂粒含量与有机碳含量的空间分布与流域内土壤侵蚀-沉积格局类似。土壤有机碳含量的高值主要集中在流域地势低洼处，土壤沉积区和林地草地等土地利用类型区域；土壤砂粒含量的空间分布规律与土壤有机碳含量相反。同时，通过统计和空间分析对比Ordinary Kriging(OK)和RF模型预测的土壤属性空间分布，发现RF模型能够灵敏捕捉环境协变量局部突变导致的土壤属性变化，适用于地形或土地利用复杂区域；而OK模型仅考虑土壤属性空间自相关性而更适合土壤属性均匀变化区域。

（2）   研究区内冻融作用强度主要为轻度和中度，冻融作用强度空间分布存在聚集性，与流域内土壤侵蚀-沉积格局类似。其中地形通过影响水分和热量的空间再分配影响冻融作用强度。海拔与冻融作用强度呈显著负相关，海拔越高，地形湿度指数越低，冻融作用强度越小。坡向也通过影响太阳辐射的分配影响冻融作用，阴坡接受太阳辐射越少，土壤水分蒸腾作用损失较少，冻融作用强度增加。

（3）   不同土地利用类型影响冻融作用强度。林地草地的植被覆盖虽然减少了地表径流和水土流失，但是在流域内会造成林地和草地附近积雪遇阻堆积，融雪融化补充土壤水分，林带后耕地土壤含水量增加，受冻融作用影响加强，土壤可蚀性增加，春季融雪径流造成林带附近耕地土壤侵蚀加剧，所以需考虑利用低矮密集的灌木构建防护带，在防治水土流失的同时能有效减少融雪侵蚀对植被带后耕地的影响。土壤冻融作用与土地利用类型和土壤理化性质密切相关，可以通过施加有机肥、免耕等保护性耕作措施保持土壤结构，减弱冻融作用强度，防治水土流失。

Freeze-thaw cycle refers to the process that occurs in cold regions, where temperature variations cause changes in the phase and volume of soil moisture, leading to soil expansion and settlement during freezing and thawing. This results in the destruction of soil structure and alteration of properties. Freeze-thaw erosion describes the changes and intensification in soil erosion processes caused by Freeze-thaw cycles. In China, the total area affected by freeze-thaw erosion has reached 1.7248 million km², accounting for 17.97% of China's land area. Existing studies often combine laboratory experiments with field investigations to explore the impact of freeze-thaw erosion on soil erodibility or use remote sensing imagery to delineate areas of freeze-thaw erosion, but there is a lack of evaluation of the spatial distribution of freeze-thaw cycle intensity. Moreover, there is less focus on freeze-thaw cycle at the small watershed scale, making it difficult to quantitatively assess its impact on watershed runoff and sediment yield. Therefore, this thesis focuses on the evaluation of freeze-thaw cycle at the watershed scale, combining measured soil data and remote sensing data, and uses fuzzy comprehensive logic to assess the intensity of freeze-thaw cycle at the small watershed scale. The results are significant for exploring the relationship between freeze-thaw cycle and environmental factors such as micro-topography, improving the accuracy of freeze-thaw cycle intensity evaluation, and fine-tuning land use management and planning. Based on literature research and field investigations, this study selects altitude, slope direction, topographic wetness factor, soil mechanical composition, and soil organic carbon content to evaluate the classification of freeze-thaw cycle intensity at the watershed scale. Using measured data and Random Forest (RF) model predictions to obtain the spatial distribution of soil factors, a dual-layer fuzzy logic evaluation system is constructed to predict freeze-thaw cycle intensity and explore its spatial distribution patterns and prevention measures. The conclusions of this study are as follows:

(1) The spatial distribution of soil sand content and organic carbon content in the study area resembles the soil erosion-deposition pattern within the watershed. High values of soil organic carbon are mainly concentrated in low-lying areas of the watershed, soil deposition areas, and land use types such as forests and grasslands. The spatial distribution pattern of soil sand content is opposite to that of soil organic carbon content. At the same time, by comparing the Ordinary Kriging (OK) model and RF model in predicting soil factors through statistical and spatial analysis, it is found that the RF model can sensitively capture soil property changes caused by local sudden changes in environmental covariates, suitable for areas with complex terrain or land use; whereas the OK model only considers the spatial autocorrelation of soil properties and is more suitable for areas with uniform changes in soil properties.

(2) The freeze-thaw cycle intensity in the study area is mainly light and moderate, and the spatial distribution of freeze-thaw cycle intensity shows clustering, similar to the soil erosion-deposition pattern of the watershed. Topography affects freeze-thaw cycle intensity by influencing the spatial redistribution of moisture and heat. Altitude is significantly negatively correlated with freeze-thaw cycle intensity; the higher the altitude, the lower the topographic wetness index, and the lower the freeze-thaw cycle intensity. Slope direction also affects freeze-thaw cycle by influencing the distribution of solar radiation; the less solar radiation received on shady slopes, the less loss of soil moisture through transpiration, and the increase in freeze-thaw cycle intensity.

(3) Different land use types affect the intensity of freeze-thaw cycle. Although vegetation cover in forested and grassland areas reduces surface runoff and soil loss, it often causes snow to accumulate near forests and grasslands in the watershed, deepening snow thickness, continuously replenishing soil moisture during snowmelt, and increasing soil erodibility near forests. In spring, rising temperatures and melting snow form snowmelt runoff, causing severe soil erosion near forested bands, necessitating consideration of constructing protective bands with low, dense shrubs to prevent soil erosion while effectively reducing snowmelt erosion on cultivated lands behind vegetation bands. Soil freeze-thaw cycle is closely related to land use types and soil physicochemical properties; applying organic fertilizers and no-till conservation tillage measures can maintain soil structure, reduce freeze-thaw erosion, and prevent soil erosion.