Optimizing visual estimation of peanut late leaf spot severity with online training sessions and standard area diagrams

Abstract

Quantification of plant disease severity is key for plant pathology research, particularly in the evaluation of disease management strategies. Visual estimation of severity remains widely used, especially in field experiments. Training sessions and the use of standard area diagram sets (SADs) are known to enhance rater accuracy. In this study, we aimed to quantify and compare the benefits of these tools, either used alone or in combination, when visually assessing peanut late leaf spot severity. We designed and validated SADs to aid in disease severity estimation and also evaluated the training tool TraineR2, a web-based app that contains actual images of the disease with known severity. Our results show that both tools led to a significant improvement in rater accuracy after their use. For TraineR2, the gains in overall accuracy (ρc from 0.82 to 0.91) and precision (Pearson's r from 0.73 to 0.88) were slightly lower than those obtained with the SADs (ρc from 0.89 to 0.96 and Pearson's r from 0.85 to 0.95). When training and SADs were combined, the overall accuracy was 0.97, and Pearson's r was 0.96, values statistically similar to those achieved using SADs alone. Regarding inter-rater reliability, evaluated based on the intraclass correlation coefficient (ICC), using SADs and training together resulted in an ICC of 0.95, which was higher than using SADs alone (0.93) or training alone (0.84). Our study confirms the utility of combining training sessions and SADs for improving the accuracy of plant disease assessments.