The Characteristics Of Near-infrared Spectral Of Citrus Huanglongbing And Application Of Reagent Efficacy Evaluation

Citrus Huanglongbing(HLB) is a devastating disease of citrus, has seriously restricted the healthy development of the citrus industry in Guangdong and South China, is a problem urgently to be solved in our country and even the world agricultural economy and technology development. Rapid diagnosis of Citrus Huanglongbing(HLB) is the precondition for HLB control. Therefor, rapidly and effective diagnosis and detection technology of HLB, has great practical significance to timely find and take corresponding measures to prevent and control HLB. At present, there is need a simple and suitable for large-scale diagnostic techniques. This study collected near infrared spectrum of sugar orange leaves, establishing datasets of standard samples, and combined with chemometrics methods, different qualitative discrimination model was established, and verified the validation set, finally came to the conclusion that discriminant analysis results for the field of HLB rapid diagnosis and provided a scientific basis.The main results were as follows:(1) Establishing the PCA model of Micro NIR. Comparing the classification results of PCA models which have different scanning methods and different spectral prepossessing methods. Determining the Back blade midrib of citrus leave was the best collection method for the micro near infrared spectrum. It was mean using the light source of Micro NIR 1700 Spectrometer along back blade midrib collecting spectrum. The best parameters were as follows: Savitzky-Golay and first derivative, point smoothing number was 3, the principal component number 15. The PCA model, after prepossessing the raw data by the optimal parameters, has the highest discriminant analysis results and the best stability. It’s correct recognition rate of the validation set was as high as 100%, and the correct rejection rate can reach 90%.(2) Establishing the PLS-DA model of Micro NIR. The near infrared spectrum scanning method was back blade midrib and the principal component number was 15, as same as the PCA model. Comparing the classification results of PLS-DA models under different spectral prepossessing methods. It was found that the bast PLS-DA model was obtained after prepossessing raw data, and the best parameters were as follows: Savitzky-Golay and first derivative, point smoothing number was 3. The number of R2, RMSEP and SEP was 0.970317,0.080198 and 0.074656, respectively, of the best PLS-DA model, the correct recognition rate and correct rejection rate of the validation set were as high as 100%.(3) Results of clustering analysis and identify test of FT-NIR.180 samples, that have finished scanning Micro NIR, their Fourier transform near infrared spectra were collected by FTIR. The result of cluster analysis identify test show that, the correct classification rate reached 100%, at the same time, HLB-infected samples HLB-uninfected samples fell in different regions, obviously. This suggested that two different types of samples can be clustering analysis to classify accurately. By comparison, the discriminant classification results of PCA model and PLS-DA model on samples were as same as cluster analysis, which good supported of the reliability of the classification results of PCA model and PLS-DA model.(4) Applicability validation of Micro NIR models in the field. In the experiment, the best PCA model and PLS-DA model all can distinguish between HLB-infected and HLB-uninfected samples in each case. The validation results of a same citrus orchard but different azimuths(different repeat groups) were as follows: the correct recognition rate were more than 93%, the highest reached 100%, and the correct rejection rate were more than 80%(including 80%), the highest also up to 100%. It showed that the two models have no special requirements for the orientation of the samples, and the stability was better, which can meet the requirements of field testing. The validation results of different citrus orchards were as follows: the correct recognition rate of the PCA model and PLS model were more than 95%, the highest reached 100%, and the correct rejection rate are more than 80%, the highest also up to 86.67%. Both comparison results showed that the the classification accuracies for the field samples were as good as the validation sets of two models, which proved that the two models have high classification power, and at the same time, the discriminant analysis ability of PLS-DA model was better than that of PCA model.(5) The test of chemical treatment efficacy evaluation. In this study, tetracycline hydrochloride, gentamycin sulfate and oxytetracycline hydrochloride were selected for chemical treatment, which Candidatus Liberibacter asiaticus(CLas) inhibition effect is relatively higher then others, the drug concentration was 1000 mg/L. Low content of CLas and the citrus leaves was asymptomatic were predicted and judged with near infrared spectroscopy models in early. Then the chemical control of the disease leaves in a timely, and spectral models for real-time monitoring. The real-time prediction model monitoring results show that with increase of the chemical treatment days, except the CK group, the three kinds of chemical treatment groups were reducing the number of samples continuously, indicating that the medicament of CLas in continuous play inhibition, and three chemical treatment effect were: tetracycline hydrochloride > gentamycin sulfate > oxytetracycline hydrochloride. Statistical results showed that, without the PCA model of oxytetracycline hydrochloride, the correct rate was 96%, PCA and PLS-DA models were 100% correct judgment of the changes in the other treatment groups of samples. This showed that combinated with the advantages of model early detection and high efficient chemical control at the same time, using real-time monitoring model efficacy to discovered Citrus Huanglongbing in time, chemicals reagent control and stop in time, which can solve the problem that due to the discovery of late and reduce the effect, and excessive use of chemicals reagent into the environment caused pollution. For the early diagnosis, prevention and cure of Citrus Huanglongbing, this study provided a new idea and method.