Evaluation of disease severity and/or disease incidence is crucial to many phytopathological studies. Disease severity is defined as the ‘area of a sampling unit affected by disease, expressed as a percentage or proportion of the total area’, while disease incidence is defined as the ‘number of plant units sampled that are diseased expressed as a percentage or proportion of the total number of units assessed’ (10). The amount of disease, either expressed as severity or incidence, is referred to as disease intensity (10) . Incidence is used in pathosystems in which a single lesion per plant is critical, as well as for many wilt and systemic virus diseases (4). On the other hand, severity is more difficult to estimate, but it is used in many pathosystems, such as potato late blight (LB), to conduct research related to disease management (4).

Severity can be estimated with special equipment or with the naked eye. The use of equipment (e.g. 11) is normally too expensive or too labor-intensive for large-scale disease evaluations (5). For this reason, severity of foliar diseases is generally estimated visually.

Scales have been developed to improve visual estimation of severity for many diseases (see Campbell and Madden [4]). Different scales can be compared based on the accuracy and precision of their estimates. Accuracy is defined as the ‘measure of the closeness of an estimate (disease assessment) to the true value’, and precision is defined as a ‘measure of reliability and/or repeatability of disease assessments’ (10). Hau (cited in Campbell and Madden [4]) proposed the following method to determine the accuracy and precision of disease severity estimates: plant units with known severity are assessed by an evaluator with no previous knowledge of the actual severity. The estimates are then regressed on the actual values with a no-intercept model (regression through origin). The slope of the regression represents the accuracy: the closer to 1.0, the more accurate the estimation made by the evaluator. Precision is directly related to the coefficient of determination (r2).

Severity of LB is routinely evaluated using two types of scales: Horsfall-Barratt scales (Table 1) and direct percentage. Horsfall and Barratt (7), through application of the Weber-Fechner law to visual assessment of disease severity, proposed that the human eye would estimate high and low disease severities with greater precision than mid range severities (5). To correct this problem, they suggested that scale increments should be logarithmic rather than linear (5). On the other hand, the percentage scale estimates disease severity directly. Forbes and Korva (5) compared a Horsfall-Barratt scale formerly used in the International Potato Center (CIP) and direct percentage on LB under field conditions. They found that evaluators tended to linearize the Horsfall-Barratt scale and, therefore, direct percentage estimation was more accurate. The Horsfall-Barratt scale used at CIP has intervals that double in size until 50% infection and then reduce in size symmetrically (Figure 1) (5). Intervals on the scale are roughly similar to unit intervals of the linearizing transformation for the logistic model (or logit = ln [x / {100 ─ x}], where x = percentage severity [5]).

Table 1. Horsfall-Barratt scale formerly used at the International Potato Center to estimate severity of potato late blight (5).

An example of a percentage scale used to evaluate LB severity is the ‘modified blight rating system’ developed by W. E. Fry and co-workers (unpublished data) (Table 2). This scale was based on data reported by Fry (6), James (8), and the British Mycological Society (3). The methodology to estimate LB severity consists in dividing the plot (experimental unit) in small quadrants, each containing 20 to 25 plants. The percentage of infected tissue is then estimated in each quadrant according to Table 2, and averaged to obtain the estimated severity in the plot. The number of quadrants to be evaluated varies according to plot size: in 4.5 x 4.5 m plots containing 100 plants, the number of quadrants is 4 (6), i.e., 80 to 100 plants are evaluated. In bigger plots, the number of quadrants to be evaluated depends on the variability in disease distribution: more variability, more quadrants. This scale was developed for epidemics in which the first symptoms appeared when plants had about 200 leaves (6) (approximately two-month old plants), but it has been used successfully with younger plants (1,2). The modified blight rating system may be used just as a reference, because there is evidence suggesting that it is better to simply estimate the percentage of foliage (everything green: stems, leaves, etc.) which is affected by disease rather than using some kind of scale (5).

Table 2. Modified blight rating system used to estimate severity of potato late blight (W. E. Fry and co-workers, unpublished data). See text for references.

Visual estimation of disease severity is subject to an important source of error. The portion of the disease that is really evaluated is the one that has produced visible symptoms and that is still on the plant. Symptomless tissue and infected leaves that had fallen off are not evaluated. This is an important consideration when severity is used as a criterion to decide an intervention (e.g., fungicide application), because the estimated severity may be much lower than the actual severity.

Practical considerations for estimating LB severity in the field are explained elsewhere.