Rapid detection of citrus Huanglongbing based on chlorophyll fluorescence imaging technology

Citrus Huanglongbing (HLB) was considered as the ‘cancer’ of citrus trees with highly contagious that had caused a significant economic loss to the citrus industry. The research was aimed to investigate the changes of the photosynthetic response of the absorption, partition, and utilization of excited energy to the HLB infection as well as to develop a method to rapidly detect HLB disease. Chlorophyll fluorescence images of citrus leaves with different infected stages collected from a commercial orchard were measured using a Pulse-Amplitude-Modulation (PAM) chlorophyll fluorescence imaging system. The starch, sucrose, glucose, and fructose within leaves in different infected stages were also determined for carbohydrate metabolic analysis. Results showed an abnormal carbohydrate metabolism with an accumulation of starch, sucrose, glucose, and fructose in HLB infected leaves. It showed a high correlation between the abnormal carbohydrate metabolism and the infection of HLB disease based on Pearson correlation analysis. Among 98 different chlorophyll fluorescence parameters related to the functional and structural information of Photosystem II (PSII), the parameters of the minimum fluorescence, the ratio of variable fluorescence to minimum fluorescence, the maximum quantum yield of PSII and the quantum yield of non-regulated energy dissipation presented relatively high sensitivity to HLB pathogen infection by analyzing the loading coefficient of principle component1 based on the Principal Component Analysis (PCA). Compared with healthy leaves, the minimum fluorescence increased in HLB infected ones, while the ratio of variable fluorescence to minimum fluorescence and the maximum quantum yield of PSII decreased. The increasing value of minimum fluorescence in HLB infected leaves indicted a structural alternation at the PSII pigment level, while the decreasing values of the ratio of variable fluorescence to minimum fluorescence and the maximum quantum yield of PSII in HLB infected leaves implied a decreasing number of active photosynthetic centers in the chloroplasts and lower photosynthesis efficiency of PSII. The pathogen of Huanglongbing changed the photosynthetic energy partitioning in citrus leaves due to decreasing the quantum yield of PSII but increasing the quantum yield of non-regulated energy dissipation. The increasing value of the quantum yield of non-regulated energy dissipation in HLB infected leaves demonstrated irreversible damage to PSII after suffering from the HLB pathogen attack. The modification of chlorophyll fluorescence kinetics indicated a disfunction of PSII in HLB infected leaves. Additionally, the chlorophyll parameters provided a good ability to predict starch, sucrose, glucose, and fructose content in citrus leaves based on the Random Forest regression model. The highest values of coefficient of determination of prediction set were 0.90, 0.84, 0.85 and 0.82 with the Residual Prediction Deviation (RPD) of 3.43, 2.50, 2.57 and 2.39 for starch, sucrose, glucose, and fructose, respectively, after optimizing the parameters (the number of trees and the number of trees) of the random forest regression model. The chlorophyll parameters were then used to build a random forest discriminant model to identify HLB disease, which achieved a good detecting performance with the overall accuracies of 100% and 97.50% for calibration set and prediction set, respectively. The detecting performance based on the machine learning method using chlorophyll fluorescence parameters was equivalent to that using carbohydrate metabolic fingerprints (the contents of starch, sucrose, glucose, and fructose) in citrus leaves with the overall accuracies of 100% and 98.75% for calibration set and prediction set, respectively. The results demonstrated that the chlorophyll fluorescence imaging could be used for nondestructive citrus Huanglongbing disease detection, and could provide a guideline for an early warning of citrus Huanglongbing in the orchard.