|Cook, A. Z.|
Submitted to: Phytopathology
Publication Type: Abstract Only
Publication Acceptance Date: 3/3/2008
Publication Date: 5/1/2008
Citation: Bock, C.H., Parker, P.E., Cook, A., Riley, T., Gottwald, T.R. 2008. Characteristics of Multi-rater Estimates of Citrus Canker Severity. Phytopathology. 98:S23. Interpretive Summary:
Technical Abstract: Citrus canker (CC, caused by <i>Xanthomonas citri</i>) was under eradication for 10 y in Florida. A total of 28 CC surveyors and plant pathologists rated severity of CC symptoms on 200 images to investigate the range of abilities and some factors that influence canker severity estimation. Actual disease was measured on each leaf image by image analysis. Estimates of lesion number (LN) were most accurate and precise, (<i>r2</i>=0.88-0.98), followed by area necrotic+chlorotic (NC, <i>r2</i>=0.63-0.93) and area necrotic (N, <i>r2</i>=0.44-0.94). Raters over- or under-estimated disease, and displayed heterogeneity of variance. Lin’s concordance analysis showed characteristics of individual error in precision and accuracy due to location and scale shifts, and tests for heteroscedasticity (White’s and Breusch-Pagan) confirmed it was ubiquitous with actual disease severity (<i>P</i><0.0001). Averaging raters estimates provided an accurate and precise mean estimate (<i>r2</i>=0.97, 0.87 and 0.92 for LN, N and NC, respectively). LN influenced most raters causing over-estimation of disease area, especially apparent at low severity. A small number of raters also showed a tendency to overestimate due to LN across the spectrum of severity. Although averaging severity scores gave an accurate and precise mean estimate, the rater, and various aspects of the symptoms affect accuracy and precision of individual estimates. Training assessors should improve estimation of CC severity estimation.