Comparison of three diagrammatic keys for the quantification of late blight in tomato leaves

Abstract

Three diagrammatic grading keys were designed for the assessment of the severity of late blight (caused by Phytophthora infestans) in tomato leaves. Simplified and broad keys considered, respectively, six (3, 12, 22, 40, 60 and 77%) and eight (3, 6, 12, 22, 40, 60, 77 and 90%) levels of disease severity, whilst a modified key based on a previous proposal for potato late blight considered six levels (1, 5, 10, 16, 32 and 50%). The keys were validated by 24 evaluators who assessed digital images of tomato leaves exhibiting different areas with lesions. Evaluator errors were compared using a mixed model in which evaluators were considered as random effects and the keys and evaluations as fixed effects. The accuracy and precision of the evaluators were compared by simple linear regression between the estimated and actual values of disease severity. The repeatability of evaluators was assessed using Pearson’s correlation coefficient. There was significant (P ≤ 0·001) variability amongst the errors made by evaluators, although the precision of each of the three keys was high with a coefficient of determination (R2) of 0·96, 0·93 and 0·83 for the simplified, broad and modified key, respectively. Repeatability of estimations amongst the evaluators was adequate (correlation coefficients of 0·91, 0·91 and 0·90 for the three keys, respectively). The simplified and broad keys resulted in higher precision and accuracy for the estimation of severity than did the modified key. Since the simplified key considers a smaller number of disease severity levels, its use is recommended in the assessment of late blight in tomato leaves.