1Dept Fitopatologia, UFV, 36571-000 Vicosa, MG, Brazil; 2Dept Agronomia - Fitossanidade, UFRPE, 52171-900 Recife, PE, Brazil

Background and objectives
Leaf blight, caused by Curvularia eragrostidis, is the most important disease of yam (Dioscorea cayennensis) in the Brazilian north-east. Typical symptoms are leaf spots, which are dark-brown, necrotic, and round (2-3 cm diameter), usually with a yellow halo. Large number of lesions may cause severe leaf fall, and up to 40% reduction in tuber weight. Although blight may induce severe losses, there are no epidemiological studies of the disease in Brazil. As these studies require the use of reliable methods to assess disease in the field, extensive work was conducted to develop a diagrammatic key and to test its precision and accuracy.

Materials and methods
Maximal disease severity, shape, distribution and frequency of lesions of 500 diseased leaves were analyzed, as well as leaf area and diseased area of 150 leaves (MOP image analyzer, Kontron, Germany). Based on these characteristics and on the Weber-Fechner Law [1], a diagrammatic key was prepared. This key is made of diagrams of six diseased leaves, representing the severity values of 1, 2, 4, 8, 16 and 32%. Disease severity on 50 colour models of leaves was assessed twice by three groups of assessors: 'specialists' (work with leaf blight), 'experienced' (work with other diseases), and 'non-experienced' (no experience with plant disease). First they assessed severity without the key, and 7 days later using it. Estimated values of severity were regressed on values generated by the image analyzer. The equation slopes measured accuracy and the coefficients of determination (R2) measured precision. To determine the value of training, 25 colour models of leaves not used yet were assessed by two groups of five people each, different from those before. Members of group 1 were 'trained' with known values of severity, while members of group 2 were not. Accuracy and precision were calculated as before.

Results and discussion
The accuracy of the individual assessors in both assessment times was determined to validate the diagrammatic key. Slopes of the regression equations ranged from 0.42 to 1.08, significantly different (P<0.01) from 1.0. While using the key, most assessors, disregarding experience, underestimating severity. This trend is different from the usually found overestimation of severity. Regarding the precision, visual assessments in both times explained 36.8 to 83.3% of the variation on the value generated by the image analyzer. At the first and second evaluations, the estimates by the specialists explained the average of 77.3 and 72.7%, respectively, of the variation on disease assessment, experienced explained 63.3 and 73.2%, and non-experienced 54.9 and 62%. The repeatability of the assessments by specialists and by the experienced groups was larger than by the non-experienced. Accuracy of leaf blight assessments without the diagrammatic key was consistently lower than the accuracy with the key. All assessors who did not use the key significantly overestimated (P<0.01) disease severity. In general, accuracy by the experienced assessors was larger than by the non-experienced. Precision of estimates with the key was also larger than precision of estimates without it. Both accuracy and precision associated with the trained assessors were larger than accuracy with the untrained. The diagrammatic key was considered simple and easy to use by most people who handled it. It was also observed that low levels of severity (<16%) were estimated with larger accuracy and precision than levels between 16 and 32%. When compared to not using it, the key use improved estimates of severity, disregarding experience of assessors. A general trend was that specialists were more accurate and precise than experienced and non-experienced assessors. As training of assessors increased accuracy and precision of the assessment of the disease, we can anticipate the diagrammatic key as an important tool to produce reliable assessments of yam leaf blight in the field.

1. Horsfall JG, Barratt RW, 1945. Phytopathology 35: 655.