欢迎访问《中国农学通报》，
中国农学通报

中国农学通报 ›› 2013, Vol. 29 ›› Issue (9): 105-111.doi: 10.11924/j.issn.1000-6850.2012-2164

• 农学 农业基础科学 • 上一篇    下一篇

不同类型土壤全氮含量的高光谱预测研究

彭杰 向红英 周清 王家强 柳维扬 迟春明 庞新安   

  • 收稿日期:2012-06-10 修回日期:2012-07-05 出版日期:2013-03-25 发布日期:2013-03-25
  • 基金资助:
    国家自然基金“新疆南疆主要类型耕作土壤肥力主导因子光谱定量反演研究”(41061031)；湖南省教育厅重点项目“湖南省主要土壤类型理化性质高光谱特性及其预测模型研究”(03A015)。

Prediction on Total Nitrogen Content in Different Type Soils Based on Hyperspectrum

  • Received:2012-06-10 Revised:2012-07-05 Online:2013-03-25 Published:2013-03-25

PDF (PC)

807

摘要： 为了探明土壤全氮的敏感波段，对比不同统计方法建立的预测模型的反演精度与稳定性，以红壤、石灰土、潮土和水稻土4个土类的土壤为研究对象，利用ASD Pro FR地物光谱仪，在室内条件下测定350~2500 nm波段范围的土壤高光谱数据，经分析不同光谱指标与全氮含量数据的相关性，确定全氮的敏感波段，并建立相应的反演模型。结果表明，反射系数、反射系数对倒的一阶微分、反射系数倒数的一阶微分、反射系数的一阶微分、反射系数对数的倒数、反射系数对数的一阶微分与全氮的最高相关系数分别出现在2153、1079、1853、528、1392、438 nm；所有预测模型中，以895、1079、1138、1149、2163、2183、2336、2337 nm波段反射率对倒的一阶微分建立的多元逐步回归模型为最佳模型；逐步回归与一元线性回归相比较而言，逐步回归建立的预测模型的精度和稳定性更佳。

关键词: 产量, 产量

Abstract: In order to find out the sensitive bands for total soil nitrogen content (TN), and to observe the inversion accuracy and stability of prediction models established by various statistical methods. Red soil, lime soil, fluvo-aquic soil and paddy soil were selected. Hyper spectral data of these soils were obtained through a ASD Pro FR field spectrometer within the range of 350-2500 nm in the laboratory. Relativity between spectra indices and TN were analyzed. Inverse models used to estimate TN based on sensitive bands were established. Results showed that: the TN was highly related with reflection coefficient (R), reciprocal of logR (1/lgR), the first order differential of the reciprocal of R, the first order differential of the R, the first order differential of 1/lgR and the first order differential of lgR at 2153, 1392, 1853, 528, 1079 and 438 nm, respectively. The stepwise multiple regression equation including reflectance at 895, 1079, 1138, 1149, 2163, 2183, 2336 and 2337 nm was the best model for fitting to the data of TN. The inversion accuracy and stability of prediction model established by stepwise regression method was better than that of prediction model established by the simple linear regression analysis.