Bock, C. H.  UNIV. OF FLORIDA  
Gottwald, Timothy  
Parker, P.E.  USDAAPHISPPQ  
Ferrandino, F.  CONN. AGRI. EXP. STATION  
Parnell, S.  ROTHAMSTED RESEARCH  
Van Den Bosch, F.  ROTHAMSTED RESEARCH 
Submitted to: European Journal of Plant Pathology
Publication Type: Peer Reviewed Journal Publication Acceptance Date: January 1, 2009 Publication Date: December 20, 2009 Citation: Bock, C., Gottwald, T.R., Parker, P., Ferrandino, F., Parnell, S., Van Den Bosch, F. 2009. HorsfallBarratt recalibration and replicated severity estimates of citrus canker . European Journal of Plant Pathology. 125:2338. Interpretive Summary: Citrus canker causes severe damage in citrus growing regions. Accurate and precise assessment of citrus canker and other plant pathogens is needed to obtain good quality data. Citrus canker assessment data were used to ascertain some of the mechanics and effects of the Horsfall and Barratt (HB) assessment scale. There was a loss in precision when data were recalibrated through the HB scale, but there was little difference in accuracy. Estimates of disease based on the cumulative mean for each leaf were generally similar for both direct estimation and HB recalibration. At low replication (<10) direct estimation was more precise than HB recalibration, but precision of both methods was similar at high replication (28 replicates). Variance was related to actual disease magnitude, and HB recalibration resulted in greater variance, particularly at low replicates (<5). Cumulative mean citrus canker severity for each rater showed that HB recalibration did not improve the estimate of the mean or its variance at any level of replication. Disease scales may be expedient to use in some situations, individual estimates of disease using the HB recalibration tend to be less precise in estimates of citrus canker severity. Replication improves the quality of direct estimates and HB recalibrated data. The HB recalibration does not appear to offer any advantages in accuracy or precision to the assessment of citrus canker symptoms over direct visual estimation, and can detract from the precision of the estimate. Technical Abstract: Citrus canker is a serious disease of citrus in tropical and subtropical citrus growing regions. Accurate and precise assessment of citrus canker and other plant pathogens is needed to obtain good quality data. Citrus canker assessment data were used to ascertain some of the mechanics of the Horsfall and Barratt (HB) assessment scale, and to detect effects of the recalibration on the mean value and variance of the estimate of citrus canker severity. Twentyeight raters directly assessed each of twohundred cankerdiseased leaves of varying severity once. These data were used as replicated direct visual estimates, and as the corresponding category of the HB scale. Correlation coefficients showed individuals lost precision when data were recalibrated through the HB scale, but there was little difference in accuracy. Estimates of disease based on the cumulative mean for each leaf were generally similar for both direct estimation and HB recalibration, regardless of disease severities. The mean values show that at low replication (<10) direct estimation was more precise than HB recalibration (at 3 replicates, r = 0.94 and 0.91, r2 = 0.91 and 0.88, respectively), but precision of both methods was similar at 28 replicates (r = 0.96 for both, respectively). The variance was closely related to actual disease magnitude although HB recalibration tended to cause greater variance at low replicates (<5). Correlation analysis showed that with 3 replicates direct estimations (0.80) was superior compared to HB recalibration (0.64), supported by the coefficient of determination from regression analysis (r2 = 0.63 and 0.41, respectively). The correlation coefficient and coefficient of determination improved with increasing replication for both methods, and at >/=10 replicates they were similar. The cumulative mean citrus canker severity for each rater showed that HB recalibration did not improve the estimate of the cumulative mean at any level of replication for any rater and the variance of the cumulative mean estimates on a leafbyleaf basis show that HB recalibration had neither a consistent nor beneficial effect. The cumulative mean by direct estimation was closer to the actual mean disease 57.0% of the time for the first onehundred leaves assessed, and 78.7% of the time for the second onehundred leaf set. The variance was most often lower using direct estimation compared to the HB recalibration (54.7 and 58.9% of the time, respectively). The actual proportion of estimates that were further from the actual using HB recalibration was greater (59.5% of estimates) suggesting greater discrepancy was more frequent using HB recalibration. Although disease scales may be expedient to use in some situations, individual estimates of disease using the HB recalibration tend to be less precise in estimates of citrus canker severity. Replication and averaging improves the quality of direct estimates and HB recalibrated data. The HB recalibration does not appear to offer any advantages in accuracy or precision to the assessment of citrus canker symptoms over direct visual estimation, and can detract from the precision of the estimate.
