基于光谱的库尔勒香梨叶片氮素含量估算模型

Prediction models for total nitrogen content in Korla Fragrant Pear leaves based on spectra

为了给遥感技术应用于库尔勒香梨叶片氮素的营养诊断提供技术支撑，利用便携式光谱仪（SVCHR-768）测定了树龄为20a的库尔勒香梨叶片的光谱反射率，结合室内分析的叶片氮素含量，采取逐步回归法对叶片氮素含量与叶片光谱参数之间的相关性进行了分析，确定了敏感波段，并构建了基于光谱参数的氮素含量模型。结果表明：库尔勒香梨叶片氮素含量与敏感波长720nm处的原始光谱和敏感波长703nm处的一阶微分光谱构建了线性模型（调整决定系数R。值均〉0．9）；通过比较分析和检验基于720nm的原始光谱反射率、703nm-阶微分的光谱反射率、比值植被指数（Rg／Ro）（Rg是波长在510～560n／n范围内的最大相对反射率，Ro是波长在640～680nm范围内的最小相对反射率）和归一化植被指数[（Rg—Ro）／（Rg＋Ro）]等4个高光谱参数对于叶片氮素含量估算模型的有效性，最终确定了在703nm处的一阶微分光谱建立的模型为香梨叶片氮素含量的最佳预测模型。

In order to provide a technical support for applying remote sensing techniques to N nutrient diagnosis in Korla Fragrant Pear leaves, spectral reflectance of leaves in 20-year-old Korla Fragrant Pear tree was determined by SVC HR- 768, and total N content in leaves was analyzed in the lab. Correlation between spectral parameters and N content in leaves was analyzed, some sensitive bands were determined, and a model for N content was established based spectral parameters. The results show that two linear models of N content in leaves with the raw spectral reflectance at 720 nm sensitive bands and 703 nm sensitive bands are established, and adjustment decisive coefficient （R2） is over 0.9. The prediction models were further compared using four spectral parameters, original spectral reflectance of 720 nm, the first derivative spectral reflectance of 703 nm, （Rg/Ro） （Rg： the largest relative reflectance at 510-560 nm, Ro： the minimal relative reflectance at 640-680 nm） and （Rg -- Ro）/（Rg d- Ro）. The linear model with first derivative reflectance at 703 nm was determined as the best model for estimating N content of Korla Fragrant Pear leaves.