林业科学2016，Vol.52Issue(12)：39-49,11.DOI:10.11707/j.1001-7488.20161205
基于光谱水分指数的核桃叶片含水量估算模型
Estimation Models for Water Content of Walnut Leaves Based on Spectral Moisture Index
摘要
Abstract
[Objective]Foliar water content of walnut was monitored quantitatively by the spectral technique to provide the technical support for water-saving agricultural production and water management in artificial oasis in southern Xinjiang basin. [Method]Based on both spectral reflectance measurement by UniSpec-SC portable spectrometer and foliar water content measured by dry-weight method in laboratory, the foliar water content estimation models for Juglans regia‘Wen185’at the fruit setting period,fruit rapid-growth period,fruit hard-core period,fruit fat-change period and fruit near-mature period were established with monadic linear regression model,stepwise multiple linear regression model, principle component regression model and partial least squares regression model. The six water indexes of spectral indices:water index ( WI) ,water band index ( WBI) ,normalized difference water index ( NDWI) ,the ratio index ( WI/NDWI) , center wavelength water index (Ratio975) were used for constructing the estimation models. The accuracy of these models was tested and evaluated by independent samples. [Result]The results showed that the spectral reflectance of J. regia‘Wen185’increased with leaf water content increasing in visible waveband,while weaken with the foliar water content increasing in the near infrared wave band. In addition,the green light,yellow light,red light and near infrared light were the sensitive spectral bands for the J. regia‘Wen185’. There were significant correlations (P <0. 01) between foliar water content and spectral water index but with different values of correlation coefficients. The results also showed that stepwise multiple linear regression model was suitable for fruit setting period,fruit rapid growth period,fruit hard-core period and fruit fat-change period with R2 all above 0. 82,while principle component regression model had the highest determination coefficient R2 of 0. 877 4 at the fruit near-mature period. Independent samples were used to test the 2 regression models,and the results showed that the root mean square error (0. 613 7 g·kg -1≤RMSE≤5. 774 6 g·kg -1 ) and relative error ( 0. 818 2% ≤RE≤2. 977 5%) were relative smaller. The models were all tested by means of confidence ellipse test and the model estimated values were consistent with the measured values. The accuracy of simple linear regression model and partial least squares regression model was lowest with relatively high root mean square error (1. 021 0 g·kg -1≤RMSE≤11. 205 6 g·kg -1) and relative error (1. 109 7%≤RE≤3. 680 8%),indicating that the model estimated results were poor. [Conclusion]The screened leaf water content SMLR spectral estimation model near mature at the fruit setting period,fruit hard-core period,fruit fat-change period and fruit near-mature period,and the leaf water content PCR spectral estimation model at fruit rapid-growth period have good stability and high estimation accuracy. The foliar water content models constructed by spectral water index can be applied in walnut foliar moisture monitoring quantitative in artificial oasis southern Xinjiang basin. The results also suggest that the spectrum technology possesses great application potential in monitoring foliar water content status of J. regia.关键词
核桃/叶片/含水量/光谱/光谱水分指数/回归分析/估算模型Key words
Juglans regia/foliar/water content/spectrum/water index of spectral/regression analysis/estimation model分类
农业科技引用本文复制引用
胡珍珠,潘存德,潘鑫,朱白雪..基于光谱水分指数的核桃叶片含水量估算模型[J].林业科学,2016,52(12):39-49,11.基金项目
新疆维吾尔自治区“十二五”重大科技专项课题(201130102-2) （201130102-2）
新疆研究生科研创新项目(XJGRI2014082)。 （XJGRI2014082）
