The broad bean (Vicia faba) is the seventh most important grain legume in the world. It is consumed green in the pod and as dry grain (^{SIAP, 2020}). In 2020, the State of Mexico was the main producer of green broad beans with 39111 t (47.4%), followed by Puebla with 28529 t (34.6%) and Michoacán with 6388 t (7.7%). These three states accounted for 29.9% of total national production (^{SIAP, 2020}).

According to a study by ^{López (2013}) in the Valley of Toluca, in recent years, diseases caused by phytopathogenic fungi have caused increasing damage to crops. Some particularly damaging diseases are chocolate spot (Botrytis fabae), brown spot (Ascochyta fabae), root rot (Rhizoctonia solani) and rust (Uromyces viciae-fabae). The latter can attack plants from the seedling stage through maturity. It affects any part of the plant, preferably the leaves in the middle and basal parts (Gautam et al., 2020), but it has also been observed attacking the upper leaves, although less frequently (Lopez, 2013). The pustules are brownish-brown in color, located in the center of chlorotic halos that appear on the upper and lower sides of the leaves (^{Gautam et al., 2020}). In other studies, within the halo there are several pustules arranged in circles or irregularly (^{Gautam et al., 2020}). The reduction in the yield of the crops affected by this disease ranges from 40% (^{López, 2013}) to 80%, but its effect goes beyond crop losses since production costs can exceed 30 thousand pesos per hectare depending on the geographical area (^{SIAP, 2020}).

To date, no scale for quantifying the severity of rust in broad bean crops has been published. Such a scale would allow standardizing the evaluation of crop damage. Previously used scales were developed for beans and soybeans (^{Godoy et al., 2006}; ^{Stavely, 1984}). Several methods allow estimating the severity of a phytopathogenic disease as the proportion of affected tissue to the total area of the plant host, but most have precision, accuracy and reproducibility issues (^{Bock et al., 2022}). ^{Rico et al. (2019}) point out that logarithmic diagrammatic scales are among the most widely used tools for measuring damage. These scales consist of an illustrated representation of a series of plants, or parts of plants, showing symptoms of a disease in different degrees of severity.

This type of measurement system has the advantage of being practical, easy to use and manage (^{Lavilla et al., 2022}). Not all measurement systems based on diagrammatic scales are logarithmic, some are linear or arbitrary. These systems are important tools for evaluators working in the field, which are crucial participants in epidemiological surveillance systems (^{Ortega-Acosta et al., 2016}; ^{Lavilla et al., 2022}).

The construction of these scales must take into account aspects such as the upper and lower limits, which should correspond to the maximum and minimum amount of the disease found in the field, respectively. The representation of the symptoms must be as close as possible to what is observed in the plants, and the intermediate levels of disease severity should consider the limitations of human vision, as defined by Weber-Fechner laws (^{Nutter and Schultz, 1995}). These diagrammatic scales must be tested (validated) before being proposed as a standard method of disease quantification. If they produce unsatisfactory results, they should be corrected (^{Bock et al., 2022}). The present study aimed to generate and validate a diagrammatic scale to evaluate the severity of U. viciae-fabae in broad bean crops.

During the spring-summer agricultural cycle of 2020-202, broad bean leaves naturally infected by U. viciae-fabae were collected in one-hectare commercial plots located in the municipalities of Zinacantepec (2,123,460 UTM N, 417,460 UTM W), Toluca (2,123,580 UTM N, 426,120 UTM W), and Calimaya (2,115,960 UTM N, 427,990 UTM W). A directed sample of 20 leaves was taken from each plot, each carefully selected to have a wide diversity of damage degrees and healthy leaves. For their transfer, the leaves were covered with newspaper and placed in a botanical press to avoid damage or contamination. They were transferred to the Faculty of Agricultural Sciences of the Autonomous University of the State of Mexico. From this collection, a sample of 110 leaves with different degrees of damage was selected. This sample was subjected to a visual classification process using variable ranges. Subsequently, the leaves were digitized as images using a multifunctional printer (HP^® LASERJET PRO MFP M127fn) (^{Scapin et al., 2014}).

The digitized leaves were subjected to a first evaluation by 20 evaluators without prior knowledge who assigned them a disease percentage. This evaluation was made based on the percentage of actual severity (S) of the disease, based on the formula: severity = (disease area/total area of the image) * 100 (^{Nutter et al., 2006}). The minimum and maximum severity values were used to define the minimum and maximum values of the proposed scale. These results were compared with the values assigned by the ^©ASSESS 2.0 software (Image Analysis Software for Plant Disease Quantification).

The scale was validated using 58 digitized leaves that represented the different degrees of severity. The images were randomized and projected as individual slides using the Power Point program (Microsoft 365^®) to 10 of the 20 evaluators involved in the first evaluation, selected according to the accuracy of their evaluation and their commitment to the project. Using a simple linear regression, the accuracy and precision of the severity evaluation of each evaluator were estimated and compared with the actual severity value obtained with the ©ASSESS 2.0 software.

The accuracy of the evaluation was determined using a T test (severity estimated by each evaluator compared with actual values of severity estimated by the software) applied to the intercept of the linear regression (b₀) to verify the intercept hypothesis H₀: b₀=0, and by checking that the coefficient of the slope (b₁) was different from 1 (H1: b₁=1), with P≤0.01. The precision of the estimates was determined based on the coefficient of determination (r²) of the regression and the absolute error (1-r²) (^{Hernández and Sandoval, 2015}). The linear regression analysis was carried out with the statistical software SAS (Statistical Analysis System) version 9.0.

The diagrammatic severity scale for U. viciae-fabae obtained was made up of six classes, represented by the following ranges of values: 0 (0.0), 1 (>0.1-6.0), 2 (>6.1-12.0), 3 (>12.1-24.0), 4 (>24.1-56.0), and 5 (>56.1-<100). The maximum severity value was 56%, associated with discoloration to a pale green hue and coalesced pustules, which meets the characteristics for this degree of damage. The lower limit observed was 0% damage (Figure 1).

The values of r² (coefficient of determination) of the first evaluation ranged from 0.575 to 0.892, with a mean of 0.738 (Table 1). In the evaluation using the proposed scale, the values ​​ranged from 0.90 to 0.97, with a mean of 0.93 (Table 1). Therefore, the evaluations can be considered accurate, according to ^{Belan et al. (2014}). ^{Ortega-Acosta et al. (2016}) reported average results >0.80 in 100% of the combinations of the two evaluations. This indicates that the diagrammatic scales are reproducible and their implementation in the field is possible since the results of the evaluations improved with the use of the proposed scale.

Figure 1 Diagrammatic scale to evaluate the severity of broad bean rust caused by Uromyces viciae-fabae. 

The reproducibility of the evaluation of the degree of damage using the proposed scale is considered high because the evaluators estimated similar percentages of severity (^{Hernández and Sandoval, 2015}). The determination coefficients between the different evaluators ranged from 0.90 to 0.97. Thus, it can be said that the diagrammatic scale proposed to evaluate the severity of broad bean rust fits well with the actual values of severity, which are similar to those produced by the software ^©ASSESS 2.0 (R²=0.98). This indicates that the evaluators were highly accurate, which reduced the absolute errors (Figure 2). This is consistent with what is reported about the regional coffee rust (Hemileia vastatrix) severity assessment system, which concentrates accuracy and precision data under an epidemiological surveillance system (^{DGSV-CNRF, 2018}).

Figure 2 Actual severity vs. absolute error of the disease quantification evaluations a) without proposed diagrammatic scale and b) with proposed diagrammatic scale. 

Regarding the null hypothesis (b₀=0 or b₁=1), the coefficient b1 was statistically close to 1 (P<0.01) (Table 1) for most of the evaluators, indicating that the visual evaluation yielded results close to the actual values, even when there was a tendency to overestimate them. Acceptable levels of precision and accuracy (b0 and b1) were achieved when evaluating the severity of broad bean rust using the six-class diagrammatic severity scale. The largest absolute errors were obtained when no scale was used (Figure 2). Therefore, the diagrammatic logarithmic scale generated in the present study can be considered a standardized method of quantifying the severity of rust in broad bean plants. This scale could be used as reference material, for support in the evaluation of control methods, comparison of assays in different locations, reproducible epidemiological studies, and breeding programs trying to produce resistant broad bean varieties.