DO STANDARD AREA DIAGRAMS IMPROVE THE ACCURACY AND PRECISION OF DISEASE ASSESSMENT?
FW NUTTER, DL MILLER and SN WEGULO
Department of Plant Pathology, 351 Bessey Hall, Iowa State University, Ames, Iowa 50011, USA
Background and objectives
Quantitying the amount of damage (crop loss) caused by plant pathogens is a prerequisite to properly estimating the benefits and costs associated with specific disease management practices and programmes. One factor limiting our knowledge of crop losses is the absence of quantitative information concerning the accuracy and precision with which disease severity is measured. James  published a series of black and white standard area diagrams that can be used as assessment aid by raters to assess disease severity. Although it is assumed that the use of standard area diagrams improves the accuracy and precision of disease severity assessments, no definitive studies have been conducted to show that data quality (accuracy and precision) is actually improved. The objectives of this project were: (1) to develop a computerized disease assessment training program (SeverityPro) that allows users to generate diseased leaf images that vary in leaf shape, lesion type, and lesion size and (2) to use this computer program to determine the impact of using standard area diagrams on rater accuracy and precision.
Materials and methods
Ten raters used the computer program SeverityPro to assess 30 leaf images of diseased alfalfa leaves Phoma spring black stem, random lesion size option) without receiving feedback as to the actual (true) disease severity level (number of non-green pixels divided by the total number of pixels comprising the leaf image X100). A graph depicting each rater's error (estimated minus actual severity) on the Y-axis versus the actual severity on the X-axis and a graph relating the actual severities (x) to the rater's estimates) were printed and labelled 'without standard area diagram results'. These results were saved to a file and printed by a laboratory supervisor without raters seeing their results. Each rater was then asked to assess another set of 3E diseased leaf images; however, this time raters used standard area diagrams for Phoma spring blackstem as a disease assessment aid. The data were saved to a file and graphs and equations from these data were printed and compared to the data sets obtained without using standard area diagrams to quantify improvement in accuracy (intercepts close to zero and slopes closer to 1 for actual versus estimated severities) and precision (coefficients of determination R2) and standard errors of the estimate for Y (SEEY).
Results and conclusions
Raters' estimates of disease severity were closer to 'actual' severity levels when raters used standard area diagrams as an assessment aid. Statistical and graphed results indicated that eight out of the 10 raters achieved intercepts closer to zero and seven of the 10 raters produced slopes closer to 1 when standard area diagrams were used as an aid to assess disease severity. Seven of the ten raters achieved higher R2 values and eight of the 10 had lower SEEY values, indicating that these raters' disease severity estimates had greater precision where standard area diagrams were used as an assessment aid.
1. James WC, 1971. Canadian Plant Disease Survey 51, 39-5.