Hybrid Modeling of Cercospora Leaf Spot Epidemiology: Integrating Mechanistic and Machine Learning Approaches Using...

Area of Science:

• Plant Pathology
• Agricultural Meteorology
• Remote Sensing

Background:

• Cercospora leaf spot poses a significant threat to sugar beet production, impacting yield and quality.
• Existing predictive models often lack integration of diverse data sources like mechanistic, meteorological, and UAV data.

Purpose of the Study:

• To develop and validate a unified predictive framework for Cercospora leaf spot by integrating mechanistic, meteorological, and UAV data.
• To assess the impact of disease severity and onset on sugar beet yield and sugar content.

Main Methods:

• Field trials were conducted from 2020-2022 with susceptible sugar beet varieties under varying fungicide regimes and artificial inoculation.
• Disease severity, airborne inoculum, and yield were monitored throughout the growing seasons.
• Hybrid models were developed, integrating multisource data including climatic variables, UAV spectral-structural indices, and mechanistic covariates.

Main Results:

• Significant treatment differences were observed, with disease incubation lasting 7-12 days and spore peaks preceding rapid severity increases.
• Yield loss was directly correlated with disease severity, impacting root fresh weight and sugar content.
• The integrated hybrid models achieved high prediction accuracy, with notable reductions in Root Mean Square Error (RMSE) across all monitored parameters, showing up to a 39% improvement over simpler models.

Conclusions:

• The developed hybrid model offers a significant advancement in predicting Cercospora leaf spot, outperforming previous methods.
• Understanding disease epidemiology through integrated data analysis can lead to more effective disease management strategies for sugar beets.
• This framework supports enhanced crop protection and improved agricultural decision-making.