X img

官方微信

img

群号:冰川冻土交流群

QQ群:218834310

高级检索

冰川冻土 ›› 2019, Vol. 41 ›› Issue (1): 215-226.doi: 10.7522/j.issn.1000-0240.2019.0103

• 寒旱区水文水资源 • 上一篇    

基于BP神经网络的青藏高原土壤养分评价

杨文静1, 王一博1,2, 刘鑫1, 孙哲1,2   

  1. 1. 兰州大学 资源环境学院, 甘肃 兰州 730000;
    2. 中国科学院 西北生态环境资源研究院 冻土工程国家重点实验室, 甘肃 兰州 730000
  • 收稿日期:2018-07-19 修回日期:2019-01-29 发布日期:2019-03-16
  • 通讯作者: 王一博。E-mail:wangyib@lzu.edu.cn E-mail:wangyib@lzu.edu.cn
  • 作者简介:杨文静(1994-),女,黑龙江齐齐哈尔人,2016年在东北农业大学获学士学位,现为兰州大学在读硕士研究生,从事寒旱区水文过程研究.E-mail:wjyang2017@lzu.edu.cn
  • 基金资助:
    国家自然科学基金项目(41877149;41530752;91547203;41563005);冻土工程国家重点实验室开放基金项目(SKLFSE201501);兰州大学中央高校基本科研业务费专项基金(LZUJBKY-2018-KB41)资助

Nutrient evaluation of the soil in the Qinghai-Tibet Plateau based on BP neural network

YANG Wenjing1, WANG Yibo1,2, LIU Xin1, SUN Zhe1,2   

  1. 1. College of Earth and Environment Sciences, Lanzhou University, Lanzhou 730000, China;
    2. State Key Laboratory of Frozen Soil Engineering, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China
  • Received:2018-07-19 Revised:2019-01-29 Published:2019-03-16

摘要: 土壤养分在养分循环和土壤-植物关系中起着重要作用,在高海拔生态系统中,由于缺乏系统的实地观测,土壤养分在高山草原中仍然知之甚少。为了了解青藏高原多年冻土区高寒草地土壤养分的基本情况以及土壤养分的等级划分,利用青藏高原腹地西大滩至安多地区采集的154个土壤样品数据,基于BP神经网络模型建立具有3层网络,10个中间层节点的土壤养分评价模型。在MATLAB软件中进行BP神经网络的训练和验证后,对青藏高原多年冻土区高寒草地土壤养分进行综合评价。结果表明:2009年青藏高原高寒草地的土壤养分综合评价等级为4级,属于较低水平。综合评价结果与基于主成分分析方法的土壤质量指数(SQI)基本一致,说明BP神经网络模型对青藏高原土壤养分的评价结果是合理的。对评价结果与海拔、植被盖度和植被类型的关系分析表明,海拔越高或植被盖度越高,土壤养分的评价等级越高;不同植被类型的评价等级表现出高寒沼泽草甸(2级)>高寒草甸(4级)>高寒草原(5级)的趋势。BP网络作为一种简单又准确的识别方法,不仅可以评估土壤养分等级,还可以比较不同地区的土壤养分高低状况,希望为青藏高原的土地资源管理与保护提供基本的科学依据。

关键词: 青藏高原, BP神经网络, 土壤质量, 海拔, 植被

Abstract: Soil nutrients play an important role in nutrient cycling and soil-plant relationships. In high-altitude ecosystems, soil nutrients are still poorly understood in alpine grasslands due to lack of systematic field observations. In order to understand the basic conditions of soil nutrients and the classification of soil nutrients in alpine grassland in the permafrost regions of the plateau, 154 soil samples collected from the Xidatan to Amdo in the hinterland of the Qinghai-Tibet Plateau were used to establish a 3-layer model based on BP neural network. After training and verifying BP neural network in MATLAB software, the soil nutrients of alpine grassland in permafrost regions of the Qinghai-Tibet Plateau were comprehensively evaluated. The results showed that in 2009, the comprehensive evaluation of soil nutrients in the alpine grassland of the plateau was Grade 4, a lower level. The comprehensive evaluation results are basically consistent with the soil quality index (SQI) based on the principal component analysis, indicating that the BP neural network model is reasonable for the evaluation of soil nutrients in the Qinghai-Tibet Plateau. The relationship between the evaluation results and altitude, vegetation coverage and vegetation types showed that the higher the altitude or the higher the vegetation coverage, the higher the evaluation level of soil nutrients; the evaluation grades of different vegetation types showed an order as alpine marsh meadow (level 2) > alpine meadow (level 4) > alpine grassland (level 5). As a simple and accurate identification method, BP network can not only assess the soil nutrient grade, but also comprehensively compare the soil nutrient level, providing a scientific basis for the management and protection of the Qinghai-Tibet Plateau.

Key words: Qinghai-Tibet Plateau, BP neural network, soil quality, elevation, vegetation

中图分类号: 

  • S158.3