X img

官方微信

img

群号:冰川冻土交流群

QQ群:218834310

高级检索
作者投稿 专家审稿 编辑办公 编委办公 主编办公

冰川冻土 ›› 2021, Vol. 43 ›› Issue (4): 1243-1252.doi: 10.7522/j.issn.1000-0240.2021.0079

• 冰冻圈技术 • 上一篇    下一篇

基于KNN机器学习方法对青藏高原唐古拉地区表层土壤水热状况的模拟

刘宏超1(), 马俊杰2,3, 李韧2()   

  1. 1.兰州大学 大气科学学院 半干旱气候变化教育部重点实验室,甘肃 兰州 730000
    2.中国科学院 西北生态环境资源研究院 冰冻圈科学国家重点实验室/青藏高原冰冻圈观测研究站,甘肃 兰州 730000
    3.中国科学院大学,北京 100049
  • 收稿日期:2020-01-08 修回日期:2020-05-15 出版日期:2021-08-31 发布日期:2021-09-09
  • 通讯作者: 李韧 E-mail:liuhch18@lzu.edu.cn;liren@lzb.ac.cn
  • 作者简介:刘宏超,硕士研究生,主要从事边界层与陆面过程研究. E-mail: liuhch18@lzu.edu.cn
  • 基金资助:
    自然科学基金项目(42071093);冰冻圈科学国家重点实验室创新群体项目(41721091)

Simulation of the water-thermal features within the surface soil in Tanggula region, Qinghai-Tibet Plateau, by using KNN model

Hongchao LIU1(), Junjie MA2,3, Ren LI2()   

  1. 1.Key Laboratory for Semi-Arid Climate Change of the Ministry of Education,College of Atmospheric Sciences,Lanzhou University,Lanzhou 730000,China
    2.Cryosphere Research Station on the Qinghai-Tibet Plateau,State Key Laboratory of Cryosphere Sciences,Northwest Institute of Eco-Environment and Resources,Chinese Academy of Sciences,Lanzhou 730000,China
    3.University of Chinese Academy of Sciences,Beijing 100049,China
  • Received:2020-01-08 Revised:2020-05-15 Online:2021-08-31 Published:2021-09-09
  • Contact: Ren LI E-mail:liuhch18@lzu.edu.cn;liren@lzb.ac.cn

摘要:

利用唐古拉站2004—2012年气象观测资料,基于KNN算法,结合机器学习思想,建立了一个气象回归模型,模拟了2005年唐古拉地区表层土壤水热变化趋势,结合实测数据,将模拟值与观测值进行对比,并对模型模拟效果进行了评估。结果表明:KNN模型能够较好地模拟活动层表层土壤水热状况,各层土壤温度的模拟值与观测值的相关系数均在0.99以上,均方根误差在1.25 ℃以内;不同深度土壤水分的模拟值与观测值的相关系数均在0.95以上,均方根误差在0.02 m3?m-3以内。总体上,KNN模型能够对青藏高原多年冻土区唐古拉地区表层土壤水热状况进行较为精确地模拟,该模型对于青藏高原其他地区的适用性有待进一步研究验证。

关键词: KNN, 机器学习, 青藏高原, 多年冻土, 土壤水热过程

Abstract:

Based on the meteorological observation data from 2004 to 2012 at Tangula Station, using the KNN algorithm, combined with machine learning ideas, a meteorological regression model was established to simulate the surface soil water-theamal change trend in Tangula region in 2005, and the measured values were combined with the observed values. The value comparison is made to evaluate the simulation effect of the model. The results show that the KNN model can well simulate the hydrothermal conditions of the soil in the active layer. The results show that the KNN model can well simulate the soil water and heat conditions in the active layer, and the correlation coefficients between the simulated values of the simulated soil temperature and the observed values are above 0.99, and the root mean square error is within 1.25 ℃; The correlation coefficients between the simulated and observed values of soil moisture are above 0.95, and the root mean square error is within 0.02 m3?m-3. In general, the KNN model can simulate the hydrothermal process of Tangula station in the permafrost region of the Qinghai-Tibet Plateau, and its applicability to other parts of the Qinghai-Tibet Plateau needs further research and verification.

Key words: KNN, machine learning, Qinghai-Tibet Plateau, permafrost, hydrothermal process

中图分类号: 

  • P642.14