Prediction of Severe Epidemics of Chickpea Ascochyta Blight Using Weather Variables

Chickpea production is threatened by severe epidemics of Ascochyta blight occurring in main chickpea growing lands under appropriate weather conditions worldwide. In this 4‐year research, occurrence of Ascochyta blight was monitored across nine main chickpea growing areas of Kermanshah province, western part of Iran. Each year, commercial chickpea fields were studied on a weekly basis from March to June. Disease data were collected as disease incidence (percentage of infected plants) and severity (percentage of infected tissues) and occurrence of epidemics. Weather data were collected as air temperature, rainfall, and relative humidity (RH) on a daily basis. According to a factor analysis, which explained 83% of data variance, 13 weather predictors were selected to estimate disease epidemics developed across different areas. Before modeling, a principal component analysis determined predictive values for these selected weather variables. Then, eight predictors of rainy days in March and April, mean RH in February, mean minimum temperature in January–March–April, and rainfalls in May and June were involved in model based on their predictive values. Current findings advanced our knowledge on the best weather predictors of severe epidemics of Ascochyta blight in chickpea crops at large scale.


| Introduction
Chickpea (Cicer arietinum L.) is cultivated in 141,575 lands producing 66,158 tons in Kermanshah Province, Iran (Anonymous 2022).Chickpea cultivation is threatened by epidemics of Ascochyta blight (Ascochyta rabiei; teleomorph: Didymella rabiei) occurring in this main chickpea producer in Iran.This disease can be highly destructive under appropriate weather conditions in early planted chickpea fields in March (Younesi et al. 2004).Ascocarps are matured in March, and then, ascospores are released from the middle of March to late May under prevailing conditions in Kermanshah.Because chickpea is commonly planted from late February to late April, ascospore releases are coincided with the emergence of chickpea seedlings in early planted fields (Younesi, Safaee, and Sheikholeslami 2011).This observation demonstrates that the two mating types of D. rabiei are present in Kermanshah, suggesting a high potential of the development of severe disease epidemics following conducive conditions.Furthermore, 48% of seed samples collected from diverse regions of Kermanshah was infected by A. rabiei, with 7% of seed samples showing infections greater than 15% (Younesi 1997).In addition, Ascochyta blight reduced the number of seeds per pod by 45%, and seed infections averaged within 8%-100% across commercial chickpea fields studied.
Ascochyta blight has been reported as one of the most destructive diseases of chickpea crops worldwide (Pande et al. 2013).In Syria, using tolerant chickpea genotypes followed by twice applications of chlorothalonil at seedling and podding stages showed the best disease management outcomes.Considering these Syrian findings, chickpea growers were recommended to early planting of tolerant genotypes, late planting of susceptible genotypes, seed disinfestations by fungicides, and application of chlorothalonil at seedling stage (Reddy and Singh 1990).In Iran, seed disinfestations by carbendazim and chlorothalonil applications have been advised to chickpea growers when the disease epidemics develop on tolerant genotypes (Younesi and Sheikholeslami 2009).However, predicting models are still needed to minimize the number of fungicide applications in Ascochyta blight epidemics occurred in susceptible genotypes.
The predicting model developed by Diekmann (1992) involved weather variables described in 23 countries from Asia, America, and Europe according to a discriminant analysis of Ascochyta blight pressure.In Australia, Salam et al. (2011) predicted epidemics of this chickpea disease based on the pathogen inoculum.In Spain, no ascocarp was produced by D. rabiei at temperatures above 10°C and RH lower than 100% (Navas-Cortés, Trapero-Casas, and Jiménez-Díaz 1998).In Israel, a 4-year research predicted the timing of ascocarps' maturity according to accumulated degree-days within 0-15°C with more than 10 mm rainfalls (Shtienberg et al. 2005).This finding reduced applications of chlorothalonil to 1-2 times per season.In India, the maximum temperature and RH in the evening for a period of time from the 9th to 14th week of planting were used to predict Ascochyta blight epidemics (Jhorar et al. 1997).However, there is no predicting model specifically developed for Ascochyta blight epidemics occurring in Iranian chickpeas.Therefore, this research tried to predict the occurrence of severe epidemics of this destructive disease according to wisely selected weather predictors described in Kermanshah.

| Materials and Methods
In this 4-year research (2017-2021), at least 16 commercial chickpea fields per year were studied across the nine geographical regions involving Sarpolzohab and Gilangharb (subtropical), Islamabad and Javanrood (cool temperate), Mahidasht (cool semi-arid), Sahneh and Ravansar (temperate), and Sararood and Songhor (cold).The local chickpea cultivar, Bivanij, is widely cultivated in rainfed lands of Kermanshah.In each region, four commercial chickpea fields were selected in random to examine the occurrence of Ascochyta blight on a weekly basis from late February to early July.The disease incidence was defined as the percentage of infected chickpea plants showing symptoms of brown to black lesions on aerial parts, dried, and hanged leaves of 20 plants per observation.Five observations were made per field that resulted in the examination of 100 plants per field.The disease severity was determined as the percentage of infected tissues per plant, assessing five plants per observation.Weather data involving daily minimum, mean and maximum air temperature, mean daily RH, wind speed, leaf wetness, and daily rainfalls during winter and spring were obtained from adjacent weather stations.
The other disease datasets were recorded as follows: the disease onset, occurrence of disease epidemics (0, 1, and 2), and maximum disease development.Simple correlations between the disease and weather variables were examined for time periods of 1-2 weeks and 1, 2, and 3 months.Among the disease variables tested, the discontinuous data for the occurrence of Ascochyta blight were considered for the disease predicting model, 0 = no evidence of disease, 1 = sparse and low levels of disease occurrence, 2 = severe disease epidemics greater than 40% severity ratings.This disease predictor was used as dependent variable to fit an ordinal logit regression model for predicting severe epidemics of Ascochyta blight developed in chickpea fields.A factor analysis (FA) was performed on the disease and weather variables using correlation matrix (Naseri and Sharifi 2019).This assisted with removing less predictive weather variables to estimate the occurrence of Ascochyta blight epidemics.In the next step, a principal component analysis (PCA) was performed on the disease and weather variables (Naseri and Sharifi 2019).This simplified the selection of most predictive variables to model the development of Ascochyta blight epidemics.Those principal factors or components with eigenvalues greater than 1.00 were considered for interpretations.Variables contributed significantly in these considered principal components if their loading values were greater than 0.35 (Kranz 2003).These FA and PCA provided predictive values for 19 and 13 disease and weather variables, respectively, defined for the current Ascochyta-chickpea pathosystem.Finally, the ordinal logit regression model was developed according to those weather predictors with high contributions in the principal components.To evaluate the fitness of regression model, the percentage of correct predictions of Ascochyta blight epidemics based on the fitted model was determined (Landschoot et al. 2013).

| Results
Severe disease epidemics mostly with 100% severity ratings occurred during 2017-2018 growing season of chickpea.There was no disease epidemics developed in chickpea fields during 2018-2019 and 2019-2020 growing seasons.The symptoms of Ascochyta blight (lower than 40% severity rating) were evident in chickpea fields across Islamabad, Mahidasht, Sararood, and Sahneh during these two growing seasons.There was no disease symptoms observed in commercial chickpea fields across all the nine regions studied during 2020-2021 growing seasons.
Although the disease incidence and severity datasets were collected for the development of Ascochyta blight in commercial chickpea fields across the nine regions studied, the discontinuous data on the occurrence of Ascochyta blight epidemics (0, 1, and 2 levels) were considered for the remainder of statistical analyses, because this disease variable provided a greater contribution in the principal components, which were regarded as linear combinations of weather variables (data not shown).Leaf wetness was ignored in the current research due to the lack of equipment required to record this weather variable.
With the help of FA results, the best predictors of Ascochyta blight epidemics among 19 weather variables were determined (Table 1).The four principal factors of this FA explained 83% of total variance in Ascochyta blight and weather datasets collected during the four growing seasons of chickpea growth.The first principal factor accounting for 32% of diseaseweather data variance provided moderate loading values for the number of rainy days and mean RH in April.Mean minimum temperatures for January, February, March, and April significantly contributed in the second factor accounting for 24% of variance.The third principal factor explaining 20% of variance provided significant contributions of rainfalls in May and June and the number of rainy days in June.Mean RH in February and March and the number of rainy days in March significantly contributed in the fourth principal factor explaining 7% of data variance.
Therefore, the first principal factor defined rainy days and RH in April, the second factor for mean monthly minimum temperatures in January-February-March-April, the third factor for rainfalls in May and June, and the fourth factor for rainfalls and RH in February and March (Table 1).These weather variables significantly contributed in the four principal factors were considered for the next statistical analysis.Hence, the 13 weather variables were subjected to the PCA to determine their predictive values for developing the ordinal logit regression model.
The PCA provided predictive values for 13 weather variables selected for the ordinal logit regression model according to the FA results.The four principal components explained 89% of total variance in the occurrence of Ascochyta blight epidemics during four growing seasons across nine chickpea growing regions (Table 2).The first principal component accounting for 32% of data variance provided moderate loading values for the number of rainy days in April and mean RH in February and April.Mean minimum temperatures in January, March, and April significantly contributed in the second principal component accounting for 27% of variance.Rainfalls in May and June and the number of rainy days in June corresponded with the third principal component explaining 22% of total variance.Mean RH in February and March and the number of rainy days and rainfalls in March were linked to the fourth principal component accounting for 9% of data variance.According to the accumulated contributions of variables (Table 2) and the strength of contributions, the most predictive variables describing appropriate weather conditions for the occurrence of Ascochyta blight epidemics were selected for the regression model.This method of selecting weather predictors based on the FA and PCA results minimized collinearity among variables involved in the regression model.Therefore, the ordinal logit regression model was developed to predict Ascochyta blight epidemics occurred in commercial chickpea fields across different geographical areas and weather conditions (Table 3).The current model explained 92% of variations in the disease and weather variables defined during the four growing seasons in chickpea fields studied across the nine regions.
The data fitted by the regression model corresponded with the response data collected from commercial chickpea fields during the four growing seasons across the nine regions (Figure 1).In the 2017-2018 seasons with the occurrence of severe Ascochyta blight epidemics, the mean minimum temperatures in January, March, and April were greater than those in the 2020-2021 without the disease occurrence (Figure 2).The highest levels of rainfalls in spring months, April-May-June, were recorded during the 2017-2018 season (severe disease epidemics occurred) compared to the lowest rainfalls in the 2020-2021 season (no epidemic observed; Figure 3).
Due to heavy yield losses to Ascochyta blight epidemics in main chickpea growing regions of Kermanshah, a concise development of predicting model is highly needed to optimize timing of fungicide applications.Thus, it is crucial to determine the best weather predictors of this destructive disease before the development of severe epidemics under agro-ecological conditions in Kermanshah.It is previously known that severe Ascochyta blight epidemics develop in early planted chickpea (susceptible cultivar Bivanij) fields in March when ascocarps are matured and then released from mid-March to late May under prevailing conditions in the current study area (Younesi et al. 2004).Moreover, such an availability of the pathogen inocula threatens young seedlings of chickpea commonly planted early from late February to late March (Younesi, Safaee, and Sheikholeslami 2011).An earlier model predicted Ascochyta blight epidemics according to the mean temperatures, rainy days, and rainfalls in the first and second months of planting chickpea (Diekmann 1992).The weather variables selected in the present findings are in agreement with this old predicting model; however, this research added to the duration of these

Predictors
Parameter estimate t prob.weather data recorded for the 2 months in Diekmann's model.Hence, our predicting model developed based on the FA and PCA results fitted rainfall, rainy days, and temperature predictors for a longer period of time, from 1-2 months before planting chickpea to 2-3 months after planting crops.Such remarkable associations of Ascochyta blight development with weather data collected 1-2 months before planting might be attributed to the noticeable impact of weather conditions on inocula survival and formation (Navas-Cortés, Trapero-Casas, and Jiménez-Díaz 1998;Salam et al. 2011;Younesi, Safaee, and Sheikholeslami 2011).This may explain why the mean minimum temperature in January (2-3 months before planting) and mean RH in February (1-2 months before planting) were fitted in the present model as effective as the temperature and rainy days in the first and second month of planting chickpea crops.
Furthermore, the current findings revealed reasonable predictive values of mean monthly minimum temperature and RH in order to model the development of Ascochyta blight epidemics in chickpeas.Moreover, the variable of rainfall was associated with the occurrence of disease epidemics when recorded in 2-3 months after planting chickpeas in the study area, in May and June.This may suggest the important role of this weather predictor in the distribution of infections, plant-byplant and field-by-field.Such noticeable associations of air temperature and rainfall with the occurrence of severe epidemics of Ascochyta blight 1-2 months after planting chickpea may support Indian findings.In India, the maximum temperature and RH in the evening for a period of time from the third to the middle of the fourth month after planting predicted severe Ascochyta blight epidemics (Jhorar et al. 1997).Therefore, it appears that predicting the disease epidemics based on the temperature, RH, rainfall, and rainy days for a longer period of time covering 6 months of chickpea growing season justified 92% of variability in disease-weather datasets.
In Australia, Salam et al. (2011) predicted epidemics of chickpea Ascochyta blight based on the pathogen inoculum and daily temperature and rainfalls.In Spain, no ascocarp was produced by D. rabiei at temperatures above 10°C and RH lower than 100% (Navas-Cortés, Trapero-Casas, and Jiménez-Díaz 1998).
In Israel, a 4-year research predicted the timing of ascocarps maturity according to accumulated degree-days within 0-15°C with more than 10 mm rainfalls (Shtienberg et al. 2005).However, these studies did not consider any time period for these weather predictors.Whereas in India (Jhorar et al. 1997) and in the current research, weather variables recorded during a definite period of time were used to predict severe Ascochyta blight epidemics occurring in chickpea crops.This suggested that the development of a predicting model for Ascochyta blight must be conducted according to specific weather predictors determined for each geographical region.Although the old model developed by Diekmann (1992) was used as a basis of modeling disease epidemics in India (Bal et al. 2008), the current observations might also explain why this old model could not predict severe epidemics of chickpea Ascochyta blight in Iran (not shown data).Furthermore, considering the importance of primary inocula in the forms of seed-borne or stubble-borne in the severity of Ascochyta blight epidemics (Salam et al. 2011;Shtienberg et al. 2005;Younesi 1997), monitoring weather conditions for at least 2 months before planting is required depending on the growing region.

| Conclusion
Therefore, it could be concluded that predicting Ascochyta blight epidemics based on a further weather descriptors involving air temperature, RH, rainfall, and rainy days for a longer period of time covering 6 months of chickpea growing season justified nearly the entire of variability in disease-weather datasets.This improved the predictability of modeling Ascochyta blight epidemics occurring in chickpeas when compared to the earlier models developed in the world as discussed.
Abbreviations: MMT = mean monthly minimum temperature; MRH = mean monthly relative humidity; NRD = number of rainy days.

FIGURE 1 |
FIGURE 1 | Response data correlated to fitted data estimated by ordinal logit regression model predicting Ascochyta blight epidemics in chickpea fields.

FIGURE 2 |
FIGURE 2 | Mean monthly minimum temperatures recorded in nine chickpea growing regions studied during four growing seasons.FIGURE 3 | Monthly rainfalls recorded in nine chickpea growing regions studied during four growing seasons.

TABLE 1 |
Factor analysis of 19 disease and weather variables linked to Ascochyta blight epidemics occurred during four growing seasons across nine main chickpea growing regions, Kermanshah.
Note: Bold values refer to significant loadings.

TABLE 3 |
Ordinal logit regression model to predict occurrence of Ascochyta blight epidemics in chickpea field according to best weather predictors.

TABLE 2 |
Principal component analysis of 13 Ascochyta blight and weather variables to determine their predictive values.