基于日光诱导叶绿素荧光的柑橘黄龙病原位诊断

柑橘黄龙病（Huanglongbing, HLB）是柑橘生产中的毁灭性病害，柑橘植株遭到黄龙病菌侵染后光合能力发生变化而后表现出相应的黄化症状。及早实现HLB的原位快速诊断是防控HLB的重要手段。为探究黄龙病菌侵染柑橘叶片的光合响应机制并实现HLB的原位诊断，该研究分析了健康（Healthy）、未显症HLB（asymptomatic HLB, aHLB）、显症HLB（symptomatic HLB, sHLB）以及黄斑病（Macular，症状与黄龙病相似）柑橘叶片的光合参数与光合色素含量差异。利用光谱技术与日光诱导叶绿素荧光（Sun-induced Chlorophyll Fluorescence, SIF）技术分析了4种类型柑橘叶片的反射率光谱与SIF光谱差异。采用竞争自适应重加权采样（Competitive Adaptive Reweighted Sampling, CARS）算法结合反射率光谱筛选出特征波段，采用SIF光谱的峰值位置（687和741 nm）构建了上行（Upward, Up）和下行（Downward, Dw）SIF产量指数（Up687, Up741, Dw687, Dw741, Up687/741, Dw687/741）。进一步分别利用特征波段的反射率和SIF产量指数，结合K最邻近（K-nearest Neighbor, KNN）分类算法构建了柑橘黄龙病的诊断模型。结果表明，黄龙病菌的侵染使柑橘叶片的光合作用明显减弱，在未显症时期已经表现出来，证明了SIF技术在诊断早期HLB的优势。基于特征波段反射率的KNN模型对未显症HLB和显症HLB的诊断精度为72.7%和75.6%，健康叶片和黄斑病叶片分别为82.2%和64.1%，而基于687和741 nm波长处的上行比值SIF产量指数Up687/741构建的KNN模型对未显症HLB和显症HLB的诊断精度为84.8%和91.1%，健康和黄斑病叶片分别为88.9%和82.1%，均优于反射率光谱模型。结果证明了SIF技术用于诊断柑橘HLB的潜力，为实现柑橘HLB的田间原位、快速、早期诊断提供了可能。

In-situ diagnosis of citrus Huanglongbing using sun-induced chlorophyll fluorescence

Huanglongbing (HLB, or citrus greening) has been one of the devastating diseases in citrus production. The infection of HLB can seriously affect the photosynthetic capacity and growth of citrus plants. Thus, early, accurate, and rapid in-situ diagnosis of HLB can be highly critical to reducing the economic losses for the citrus industry. This study aims to develop an in-situ diagnosis of citrus HLB disease for the photosynthetic response of citrus leaves using sun-induced chlorophyll fluorescence. Four groups were labelled on the citrus trees with the asymptomatic HLB (aHLB), symptomatic HLB (sHLB), and macular (with symptoms similar to HLB) infection, as well as the healthy trees in the orchard. A Li-6800 portable photosynthetic system was utilized to measure the photosynthetic parameters (net photosynthetic rate (Pn), intercellular CO2 concentration (Ci), transpiration rate (Tr), stomatal conductance (Gs)) of citrus leaf samples. The photosynthetic CO2 response curve (A-Ci) was obtained under the different carbon dioxide concentrations, further to calculate the maximum leaf carboxylation rate (Vcmax) and the maximum electron transfer rate (Jmax). Then, the upward and downward sun-induced chlorophyll fluorescence (SIF) and reflectance spectra of leaf samples were collected using an analytical spectral devices (ASD) spectrometer combined with a FluoWat clip. Finally, the photosynthetic pigments content (chlorophyll a, chlorophyll b, and carotenoids) of the leave samples were measured using the spectrophotometry in the laboratory, and the true infection status of the leaves was confirmed using the real-time quantitative polymerase chain reaction (qPCR). A systematic investigation was made on the photosynthetic parameters and the pigments content of the leave samples. An optimal combination of wavebands was selected using the competitive adaptive reweighted sampling (CARS) algorithm and reflectance spectra. The upward (Up) and downward (Dw) SIF yield indices (Up687, Up741, Dw687, Dw741, Up687/741, Dw687/741) were constructed using the peak position (687 nm and 741 nm) of the SIF spectra. Furthermore, the classification models of HLB were established to combine with the K-nearest neighbor (KNN) algorithm, according to the optimal wavebands of reflectance spectra and SIF yield indices. The results showed that the infection of HLB pathogen led to a significant decrease in the photosynthesis of citrus leaves even at the asymptomatic stage, indicating an excellent performance of SIF signals in the early diagnosis of HLB. The diagnostic accuracies of KNN models with the leaf reflectance using the optimal wavebands were 72.7% and 75.6% for aHLB and sHLB leaves, and 82.2% and 64.1% for healthy and macular leaves, respectively. By contrast, the diagnostic accuracies of KNN models using the Up687/741 (ratio of upward SIF yield at 687 nm to 741 nm) SIF yield index for aHLB and sHLB were 84.8% and 91.1%, and that of healthy leaves and macular leaves were 88.9% and 82.1%, respectively. Consequently, the KNN models with the leaf SIF spectra presented a higher potential in the early diagnosis of HLB than those with leaf reflectance spectra. These findings can provide a strong reference for the early, rapid, and in-situ diagnosis of citrus HLB in the incubation period.