Monitoring and detection of soybean rust by spectroradiometry and prediction model construction for decision-making of control

Author(s)

Matheus Mereb Negrisoli

• Document type: Doctoral Thesis
• Press: Botucatu. 2022-08-17.
• Institution: Universidade Estadual Paulista (Unesp). Faculdade de Ciências Agronômicas. Botucatu
• Defense date: 2022-05-20
• Advisor: Carlos Gilberto Raetano

Abstract

Monitoring of soybean rust (SBR) by remote sensing can improve disease detection in the field and be used as a decision support system for control, which can influence the quality of application and the effectiveness of control. Thus, the objective of this study was to obtain a correlation between SBR progress and the reflectance values of the crop under different conditions of disease severity levels and soybean cultivars, in order to propose a disease prediction model and use it as a monitoring method to aid in the decision making of chemical control. The research project was conducted between 2018 and 2021 at the School of Agriculture (FCA/UNESP), Botucatu, SP, divided into three phases. Initially, the effects of SBR on susceptible and partially resistant soybean cultivars was evaluated in the field and in the laboratory, in terms of disease severity, crop yield, chlorophyll fluorescence, gas exchange, and the applicability of remote sensing for determination of these effects. After then, we evaluated the detection of soybean rust by remote sensing and the construction of a prediction and classification model based on leaflet reflectance at different levels of SBR severity. Different statistical techniques to reduce data dimensionality and classification algorithms were also evaluated. At the end, the application of the disease prediction model in the field was carried out evaluating the effect of different application timings on the application technology and control effectiveness. The applicability of the prediction model was evaluated, comparing it with conventional monitoring methods as the source for decision making of control (calendarized applications at a pre-defined period and at the first appearance of the symptoms). The effect of different application timings was evaluated in terms of spray deposit, spray coverage, disease control effectiveness and crop yield. Significant differences were observed regarding the effect of the disease on the two soybean cultivars, with a reduction in the negative effects of the pathogen on the soybean cultivar with partial resistance. It was possible to distinguish between healthy and SBR-infected plants based on leaf reflectance at different severity levels with up to 93% accuracy and precision, and these data were successfully used to create a disease classification and prediction model. The different application timings influenced the control effectiveness and application technology, in which treatments with application at times of higher leaf area index obtained less uniform spray distribution. The use of the proposed prediction model and remote sensing techniques are effective and promising to be integrated into disease management programs. (AU)

• FAPESP's process: 18/26486-0 - Monitoring and detection of Asian soybean rust by spectroradiometry and construction of prediction model for control decision making
• Grantee: Matheus Mereb Negrisoli
• Support Opportunities: Scholarships in Brazil - Doctorate