塔里木盆地北缘荒漠土壤有机质含量的高光谱估测
作者:
作者单位:

(1.新疆师范大学地理科学与旅游学院,新疆 乌鲁木齐 830054;2.新疆维吾尔自治区重点实验室“新疆干旱区湖泊环境与资源实验室”,新疆 乌鲁木齐 830054)

作者简介:

玉米提·买明(1992-),男(维吾尔族),新疆吐鲁番人,硕士研究生,研究方向为资源环境遥感。E-mail:812814409@qq.com。

通讯作者:

基金项目:

基金项目:国家自然科学基金(41561051、41261051);新疆师范大学研究生创新基金项目(XSY201902007)。


Hyperspectral estimation of desert soil organic matter content in the northern margin of Tarim basin
Author:
Affiliation:

(1.College of Geographic Science and Tourism,Xinjiang Normal University,Urumqi Xinjiang 830054;2.Xinjiang Uygur Autonomous Region Key Laboratory “Xinjiang Arid Lake Environment and Resources Laboratory”,Urumqi Xinjiang 830054)

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
    摘要:

    土壤有机质含量的多少是衡量土壤肥力的重要指标,了解土壤有机质的状况及动态变化,为指导干旱区绿洲农业生产及生态环境保护提供科学依据。基于在塔里木盆地北缘绿洲-荒漠过渡带采集的80个土壤样品,测定有机质含量和光谱反射率。在原始反射率R的基础上,进行光谱反射率的一阶微分R′、倒数对数lg(1/R)、倒数对数的一阶微分[lg(1/R)]′以及去除包络线C(R)处理,并将处理后的光谱数据与土壤有机质进行相关性分析,从而选取568、578、803、806、845、955 nm等敏感波段构建土壤有机质含量的估测模型。结果表明:(1)土壤有机质与土壤反射率呈负相关,有机质含量越高反射率越低。(2)光谱变换处理可有效提升光谱对土壤有机质含量的敏感性,其相关系数最高可达0.654(P<0.001)。(3)比较多元线性逐步回归、偏最小二乘回归和反向传播神经网络(BPNN)3种建模方法发现,反向传播神经网络模型精度较高,稳定性更好,且以倒数对数的一阶微分[lg(1/R)]′为自变量的模型最优,决定系数为0.864,均方根误差为1.86,这表明[lg(1/R)]′-BPNN模型相较于其它模型可以更为准确地预测荒漠区土壤有机质含量。

    Abstract:

    The content of soil organic matter is an important index to measure soil fertility.Understanding the status and dynamic changes of soil organic matter can provide scientific basis for guiding oasis agricultural production and ecological environmental protection in arid areas.Based on 80 soil samples collected in oasis-desert ecotone on the northern margin of Tarim basin,organic matter contents and spectral reflectance were measured.Based on the original reflectance R,first order differential R′,reciprocal logarithm lg(1/R),reciprocal logarithm’s first order differential[lg(1/R)]′and envelope removal processing C(R)were carried out,and the processed spectral data was correlated with soil organic matter for analysis.568,578,803,806,845,955 nm and other sensitive bands were selected to construct estimation models of soil organic matter content.The results showed that there was a negative correlation between soil organic matter and soil reflectance,and the higher the content of organic matter was,the lower the reflectance was.The spectral transformation improved the sensitivity of the spectrum to the content of soil organic matter,and the correlation coefficient could be as high as 0.654(P<0.001). In the comparison of multiple linear stepwise regression,partial least square regression and back propagation neural network(BPNN),it was found that the back propagation neural network model had higher accuracy and better stability.The model with the first order differential [1g(1/R)]′of the reciprocal logarithm as the independent variable was the best.The coefficient of determination was equal to 0.864,and root mean square error was equal to 1.86.This shows that[lg(1/R)]′-BPNN model can predict the content of soil organic matter in desert area more accurately than other models.

    参考文献
    相似文献
    引证文献
引用本文

玉米提·买明,王雪梅.塔里木盆地北缘荒漠土壤有机质含量的高光谱估测[J].中国土壤与肥料,2021,(4):318-326.

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
历史
  • 收稿日期:2020-04-03
  • 最后修改日期:
  • 录用日期:2020-05-30
  • 在线发布日期: 2021-09-15
《中国土壤与肥料》招聘启事
关闭