多年冻土区活动层是地表水和地下水相互转化中十分重要的交换通道，活动层土壤含水量是多年冻土区水文循环中重要的组成部分，其动态变化与寒区生态环境密切相关。在气候变化背景下，深入了解活动层土壤含水量的动态变化特征具有重要意义。本文利用ELM（Extreme Learning Machine）模型对青藏高原腹地不同海拔高度多年冻土区土壤含水量进行模拟分析，结果表明：与BP神经网络模型相比，二输入变量ELM模型的模拟精度更高；ELM模型模拟后1天土壤含水量的NSE值在0.69~0.87之间，其中坡下20 cm深度处模拟NSE取得最大值（0.87），并且模拟精度随着推后时间的增加有所提升，模拟后3天和后7天的NSE值分别在0.76~0.92和0.75~0.93之间；坡下各深度含水量的模拟效果优于坡上。在此基础上，通过设置不同的气候变化情景，研究土壤含水量在气候变化背景下的动态变化规律及响应特征。研究发现，升温导致冻结初期以及融化初期不同深度的土壤含水量均出现增大的趋势，在完全冻结期和完全融化期变化不明显。且随着气温增幅的加大，冻结初期以及融化初期的土壤含水量变化也逐渐增大，深层土壤含水量较浅层土壤含水量的增加更加显著。在降水增加的情景下，降水增加越大，土壤含水量的增加趋势越明显，但整体变化幅度较小；坡上各深度土壤含水量的增加主要发生在融化初期和完全融化期，坡下则主要集中在融化初期，相比于深层土壤，浅层土壤对降水增加的响应更加强烈。

The active layer in the permafrost region plays a critical role in surface water and groundwater exchange. Soil water content within the active layer is essential for the hydrological cycle in the permafrost region and has a significant impact on the ecological environment in this cold region. Understanding the dynamic characteristics of soil moisture in the active layer is crucial in the context of climate change. This paper employs the Extreme Learning Machine （ELM） model to analyze the soil moisture within the permafrost regions at various heights in the hinterland of the Qinghai-Tibet Plateau. Results indicate that the ELM model， with two input variables， has higher simulation accuracy than the BP neural network model. The Nash-Sutcliffe Efficiency （NSE） values of soil water content during the first day after the ELM model simulation range between 0.69 and 0.87. The simulated NSE value at the depth of 20 cm below the slope is the maximum （0.87）， and simulation accuracy improves with the increase of delay time. The NSE values for the third and seventh days after the simulation are 0.76~0.92 and 0.75~0.93， respectively. The simulation effect of water content at different depths under the slope is better than that on the slope. Subsequently， by establishing different climate change scenarios， the study explores the dynamic change law and response characteristics of soil moisture in the background of climate change. Results indicate that the soil water content at different depths increases during the initial freezing stage and the initial thawing stage due to temperature rise. However， no notable change occurs during the complete freezing period and the complete thawing period. Temperature increase affects the soil water content in the early freezing and early melting stages， with deeper layers experiencing more significant changes than shallower ones. Moreover， under the scenario of precipitation increase， the greater the precipitation increase， the more apparent the trend of soil water content， but the overall change range is small. The increase in soil water content at each depth on the slope mainly occurs in the early melting stage and the complete melting stage， whereas that at the lower slope mainly happens in the early melting stage. Compared with deep soil， shallow soil responds more strongly to precipitation increase.