| 基于BP神经网络和FEFLOW模型模拟预测多年冻土活动层温度 |
| |
| 引用本文： | 郭林茂,常娟,徐洪亮,叶仁政.基于BP神经网络和FEFLOW模型模拟预测多年冻土活动层温度[J].冰川冻土,2020,42(2):399-411. |
| |
| 作者姓名： | 郭林茂 常娟 徐洪亮 叶仁政 |
| |
| 作者单位： | 兰州大学 资源环境学院，甘肃 兰州 730000 |
| |
| 基金项目： | 国家自然科学基金项目(41671015) |
| |
| 摘 要： | 土壤温度是陆面过程中地-气系统间能量与物质交换的重要参数， 它的动态变化及其对气候变化的响应也是研究陆面过程的关键问题之一。在全球变暖背景下， 研究青藏高原多年冻土活动层土壤热状况动态变化， 对深入了解高原活动层厚度的变化特征及下垫面的热力作用均有重要意义。利用BP神经网络模型， 对青藏高原风火山地区的地表温度进行了模拟， 并利用输出的地表温度驱动FEFLOW模型对研究区活动层不同深度土壤温度进行了模拟。与各深度土壤温度观测值对比发现， 均方根误差介于0.09 ~ 1.78 ℃， 纳什效率系数介于0.86 ~ 0.98， 模拟效果良好。结合BP神经网络模型和FEFLOW模型预测了研究区未来50年活动层热状况的动态变化过程， 结果表明： 在0.02、 0.048、 0.07 ℃·a-1三种升温情景下， 50年后研究区活动层厚度将分别增加19.4、 51.8、 64.7 cm， 土壤升温幅度随着深度的增加逐渐减小。同时发现， 随着气温不同程度的升高， 土壤开始融化的时间在不断提前， 开始冻结的时间则不断延迟， 这种规律随着土壤深度的增加而减弱， 但不同深度土壤冻融过程对气温升高的响应差异却随着增温速率的增大而逐渐减小。
|
| 关 键 词： | 青藏高原 活动层 土壤热状况 气候变化 BP神经网络 FEFLOW模型 |
| 收稿时间： | 2019-01-17 |
| 修稿时间： | 2019-06-05 |
Simulation and prediction of permafrost active layer temperature based on BP neural network and FEFLOW model： take the Fenghuoshan area on the Tibetan Plateau as an example |
| |
| Linmao GUO,Juan CHANG,Hongliang XU,Renzheng YE.Simulation and prediction of permafrost active layer temperature based on BP neural network and FEFLOW model： take the Fenghuoshan area on the Tibetan Plateau as an example[J].Journal of Glaciology and Geocryology,2020,42(2):399-411. |
| |
| Authors: | Linmao GUO Juan CHANG Hongliang XU Renzheng YE |
| |
| Institution: | College of Earth and Environmental Sciences，Lanzhou University，Lanzhou 730000，China |
| |
| Abstract: | Soil temperature is an essential index for energy and material exchange between the earth-atmosphere system during land surface processes. Its dynamic changes and its response to climate change are also one of the key issues in the study of land surface processes. Under the background of global warming， the dynamic changes of soil thermal conditions in the permafrost active layer on the Tibetan Plateau are of great significance for understanding the variation of active layer thickness and the thermal effects of the underlying surface. In this paper， the BP neural network model is used to simulate the surface temperature of the Fenghuoshan area on the plateau. Using the output surface temperature to drive the FEFLOW model， the soil temperature at different depths in the active layer was simulated. Comparisons between the simulated and observed data in situ showed that： the root mean square error ranged from 0.09 to 1.78 ℃， and the Nash-Sutcliffe efficiency coefficient ranged from 0.86 to 0.98. Combined with the BP neural network model and the FEFLOW model， the dynamic changes of the active layer thermal conditions in the study area in the next 50 years are predicted. The results show that under the three warming scenarios of 0.02 ℃·a-1， 0.048 ℃·a-1， 0.07 ℃·a-1， the thickness of the active layer in the study area will increase by 19.4， 51.8， 64.7 cm 50 years later， respectively. Soil temperature at different depths showed a significant increase with the increase of air temperature， and the influence of air temperature on soil temperature decreases with depth. At the same time， it is found that under different warming scenarios， the time for the soil to start thawing is continuously advanced， and the time to start freezing is continuously delayed， and this law also decreases with the soil depth， but the difference of response of soil freezing and thawing process to temperature rise at different depths decreases with increasing of the warming rate. |
| |
| Keywords: | Tibetan Plateau active layer soil thermal condition climate change BP neural network FEFLOW model |
|
| 点击此处可从《冰川冻土》浏览原始摘要信息 |
| 点击此处可从《冰川冻土》下载免费的PDF全文 |
