Evolution of apple scab prediction models from empirical systems to Artificial Intelligence based predictive frameworks

Abstract

Apple scab poses a significant threat to apple production worldwide, necessitating effective management strategies. Over the decades, the evolution of apple scab prediction models has transitioned from empirical and rule-based approaches to sophisticated simulation and machine learning models. Early models, such as Mills’ infection criteria, established foundational guidelines for predicting infection risk based on environmental conditions. Subsequent advancements introduced simulation-based models that integrated real-time data, enhancing decision-making capabilities for growers. The emergence of machine learning techniques has further revolutionized disease detection, enabling automated and accurate identification of apple scab through image analysis. Recent studies highlight the integration of Artificial Intelligence (AI), the Internet of Things (IoT), and sensor technologies as transformative approaches in apple scab management, providing real-time insights and actionable recommendations. This review identifies that simulation and machine learning models currently dominate apple scab prediction research due to their improved accuracy and adaptability. However, key research gaps persist, including limited generalizability across regions, high implementation costs, data dependency, and insufficient field-level validation. This paper reviews the historical development of apple scab prediction models, discusses the integration of modern technologies, and explores potential future directions for enhancing disease management strategies.