Response of arabica coffee populations on coffee leaf rust in two weather conditions in North Sumatra, Indonesia

Coffee leaf rust (CLR) disease is seriously threatening the sustainability of coffee production in many countries. The incidence and severity as CLR parameters generally depend on the coffee plant, pathogen, and environment. Therefore, this study aimed to determine the effect of arabica coffee populations and weather conditions on CLR parameters. The study was carried out using a field experiment with seven groups of arabica coffee populations in different weathers in North Sumatra Province, Indonesia. The results showed that the arabica coffee population P7 from District Toba had high resistance to wet weather. Furthermore, all the population groups showed higher degree of CLR parameters in dry than wet weather, while a significant interaction between the population and the weather was observed on branch rust incidence, leaf rust incidence, and severity. Moreover, H. vastatrix races were probably varied and had different adaptability to weather changes, which was the most decisive factor for the dispersal and severity of CLR. The CLR parameters in the dry weather did not correlate with the parameters in the wet weather and vice versa. The total rainfall and the number of the rainy day reduced the CLR dispersal and severity.


Introduction
Fungus Hemileia vastatrix Berkeley and Broome causes coffee leaf rust and is an epidemic in many coffee-producing countries [1] [2].Due to the epidemic, around two million coffee workers lost their jobs and caused an income loss of approximately around three billion USD [3].Coffee plants in some farms in North Sumatra were infected by these fungus [4] [5].Coffee is very important for North Sumatra Province, Indonesia, meanwhile, this Province had a large number of coffee farmers (households) [6].They produced a huge number of quantity of green beans.North Sumatra Province is the main production area of arabica coffee in Indonesia [6].
The H. vastatrix fungus causes rust on the abaxial face of the leaf (Figure 1) which can cover up to 100%.A previous study in Indonesia has shown that there are ten physiological races of H.vastatrix [7].This fungus absorbs nutrients from the leaf resulting nutritional deficiency, damage, premature falls off and a decrease in photosynthesis.In a prolonged drought condition, the production of coffee fruit can be reduced by more than 50% [1].
The degree of CLR parameters of plants depend on the genetic, pathogen, environment, and their interaction.Globally, the use of resistant cultivars has been generally accepted to overcome this fungus.Since fully resistant cultivars are not easily produced, creating incompletely resistant cultivars Coefficient of variation (CV) (%) (RTK5 0.5 /GM) x 100

Combined analysis of variance
The results showed that host, pathogen, and environment influenced CLR incidence and severity.
Similarly, the analysis of variance also showed that the weather influenced IRB, IRL, and SRL significantly (Table 2).The populations were also highly significantly different in CLR parameters and there was a significant interaction of population and weather (PxW) on IRB, IRL, and SRL.  3) with approximately onetenth of the population had high resistance, while half showed moderate resistance based on the SRL.This result showed that it is possible to discover resistant genotypes among the unknown genotypes in North Sumatra Province, which has a fairly rich genotypic variability in CLR parameters [5] and adaptability to climate change [15].There are differences in the CLR severity of the arabica coffee cultivars population [16], and in Liberica coffee in Meranti Islands of Indonesia [17].Similarly, resistant coffee plants with high beans production is also possibly discovered among existing populations [18].Based on the Table 3 results, all the population had higher CLR parameters in dry weather rather than in wet weather.Moreover, due to the interaction between population and weather, the rank of CLR parameters of the populations changed depend on the weather.

Pathogen
A study showed that there are 10 races of H. vastatrix in Indonesia [7].However, there is no information on the number of races available in North Sumatra and in the districts where this study was conducted.Based on the results, the races at different locations were supposedly different, which was indicated by the different adaptability of H. vastatrix to weather changes.Therefore, to measure the adaptability of the pathogen to weather changes, a correlation test between the CLR parameters in both dry and wet weather conditions was conducted (Table 4).A significant correlation coefficient is a measure of adaptability such as in Environment-1, the IRB of population P1 in the wet weather had a significant positive correlation with its IRB in the dry weather, which showed that the pathogen races had good adaptability to weather changes in Environment-1.The number of significant correlation coefficients is a relative measure of the adaptability of a pathogen compared to other pathogens.Meanwhile, the pathogen race discovered in Environment-1 had higher adaptability (number of significant r = 7), followed by the pathogen race in Environment-2 (significant r = 5), and Environment-4 (significant r = 5), while the lowest was the pathogen race in Environment-3 (significant r = 4).This showed that pathogen races are different between environments (locations).

Weather
The role of weather is greater than the function of population and their interaction on CLR parameters.This is because the weather had the highest variance component (32.39, 49.19, and 58.90% in IRB, IRL, and SRL, respectively) and greater variance than population and GxW interaction in CLR parameters (Table 5), therefore, the weather was the main determinant of the CLR dispersal and severity.
Table 4. Coefficient correlation and regression between coffee leaf rust (CLR) parameters in dry and wet weathers in each environment (location), (n = 4, CLR = coffee leaf rust, r value at the 5% level of significance = 0.950, r value at the 1% level of significance = 0.990, ns = not significant, * = significant at the 5% level of significance, ** = significant at the 1% level of significance, y = CLR parameter in wet weather, x = .CLR parameter in dry weather,  Furthermore, the CLR parameters were higher in the dry weather than in the wet weather for all populations (Table 3).This is supported by the results of a study by [19] which stated that CLR severity was higher in the dry season than the wet season in Chanchamayo (Junin-Peru) from April 2017 to March 2018.
Generally, the CLR parameters in the dry weather did not correlate with the wet weather and vice versa (Table 6).
Therefore, CLR parameters in dry weather can not be used to predict CLR parameters in wet weather and vice versa (Table 6).Table 6.Correlation coefficient and determination between dry weather and wet weather in coffee leaf rust (CLR) parameter (n = 7, r value at the 5% level of significance = 0.754, r value at the 1% level of significance = 0.874, ns = not significant, * = significant at the 5% level of significance, ** = significant at the 1% level of significance).A previous study has shown that weather influenced the biology and epidemiology of CLR [8].In 2015-2018, the weather pattern in the Environment-1, 2, 3, and 4 was similarly low rainfall due to fewer rainy days from June to August compared with October to December (Figure 2), which indicated the presence of dry and wet weather.According to [20], the peak of the dry and wet weather period in these districts was from July to September and from October to December, respectively.
In terms of the biological life cycle, CLR infection on coffee leaves involves the adhesion of urediniospores to the leaf surface of the host plant, its germination, formation of appressorium over stomata, penetration, and intercellular colonization, however, weather can affect each phase of this CLR life cycle [7] [8].To determine the effect of rainfall on dispersal and severity of CLR, a correlation test was conducted between CLR parameters and rainfall at one month (Rf-1) and two months (Rf-2) before observation (Table 7a, Table 7b).Moreover, a negative correlation coefficient between IRB and IRL with rainfall showed that rain reduced the dispersal of urediniospores, while in SRL and rainfall, it indicated that rain reduced severity.All populations had higher CLR parameters in the dry weather than in wet weather (Table 3) because the rain eroded the urediniospores from the plant in the wet weather.Therefore, the more the rainfall was, the less the CLR parameters were.These results showed that the wet weather conditions are not conducive to the biological life cycle of CLR because rain water usually wipes out the urediniospores from the plants which lead to a decrease in the dispersal of urediniospores and reduction of severity.However, during dry weather, urediniospores can stay on the leaf surface, germinate, and penetrate leaves through stomata, and sporulate.These repeated cycles in dry weather increase both leaf rust severity and urediniospores dispersal within plants.
The number of significant negative correlation coefficients at Rf-1 was more than Rf-2 (Table 7a, Table 7b), which showed that Rf-1 was more decisive than Rf-2 on reducing the CLR parameters.Furthermore, the environments were different in the relationship between Rf-1 and Rf-2 with CLR parameters.Based on the results, Rf-1 rainfall in Environment-3 had a negative to highly significant negative correlation with IRB, IRL,and SRL (except for SRL in P4 and P5).This can be used as a regression model using the highest coefficient of determination (r 2 ).For IRB (r 2 = 0.75, P5), the regression of IRB (y) with rainfall (x) was y = 90.58-0.20x (84<x<381 mm/month), which showed that an increase in rainfall of 1 mm/month reduced the IRB by 0.20%.Subsequently, for IRL (r 2 0.82, P4), the regression of IRL (y) with rainfall (x) was y = 62.45 -0.12x (84<x<381 mm/month), while for SRL (r 2 = 0.83, P3), the regression of SRL (y) with rainfall (x) was y = 32.71-0.06x (84<x< 381 mm/month).
Moreover, the effect of the number of rainy days on dispersal and severity of CLR was determined by conducting a correlation test between CLR parameters and the number of rainy days in one month (Rd-1) and two months (Rd-2) before the observation of CLR parameters.The results showed that the number of significant negative correlation coefficients at Rd-1 was more than Rd-2 (Table 8a, Table 8b), which indicated that Rd-1 was more capable than Rd-2 to reduce the dispersal and severity of CLR.Therefore, the more rainy days are, the less the CLR parameters are.In agronomic practices, site selection for coffee cultivation concerning CLR needs to be based on weather considerations rather than population because weather contributed more variance than population (Table 5).Hence, the site for coffee cultivation needs to be dominated by wet weather with high rainy days throughout the year.

CLR Parameters
The parameters of CLR have a significant to highly significant positive correlation (Table 9).An increase in the dispersal (IRB and IRL) increases the severity (SRL), which is estimated by IRB and IRL.This result is in line with a study carried out by [21] which stated that there is a significant correlation between SRL and IRL.Since this incidence is not always phenotypically and genotypically correlated with severity [5], selection needs to be based on SRL.This is because SRL showed the level of damage on the leaf, hence, an increase in SRL causes a significant loss in fruit production.Table 8a.Coefficient correlation and regression between coffee leaf rust (CLR) parameter (y) and rainy day in one (x) and two months (x) before data recording of arabica coffee seven populations in Environment-1 and Environment-2 (n = 8, r value at the 5% level of significance = 0.707, r value at the 1% level of significance = 0.834, ns = not significant, * = significant at the 5% level of significance, ** = significant at the 1% level of significance).(n = 14, r value at the 5% level of significance = 0.754, r value at the 5% level of significance = 0.874, ns = not significant, * = significant at the 5% level of significance, ** = significant at the 1% level of significance).

Conclusions
In conclusion, the arabica coffee population P7 from toba District with SRL of 4.53% has a high resistance in the wet weather, meanwhile, all the population has a higher level of CLR parameters in dry weather rather than in wet weather.Furthermore, the interaction of population and weather in IRB and IRL was significant and was highly significant in SRL.H. vastatrix races had different adaptability to weather changes and varied between the environments (locations).The results also showed that the weather had greater influence than population and their interaction on CLR parameters, therefore, it was the main determinant of the dispersal and severity of CLR.Generally, there was no correlation between CLR parameters in the dry weather with CLR parameters in the wet weather and vice versa.The CLR parameters also had a significantly negative correlation coefficient with rainfall and rainy day, which indicated that rain reduced the dispersal and severity of CLR.Moreover, rainfall and rainy day in one month before observation of CLR parameter was the most decisive factor to reduce dispersal and severity of CLR, while the parameters of CLR had a significant to a highly significant positive correlation.These results can contribute to the efforts of creating sustainable and productive coffee cultivation considering appropriate rainfall and rainy days.The sustainable supply of coffee is very important because billions of cups of coffee are consumed yearly [22] so that coffee is the second rank after petroleum on the list of the world's economic value [23].

Figure 1 .
Figure 1.The fungus H. vastatrix live in leaf tissue and produce reproductive structures on underside surface.

Figure 2 .
Figure 2. Rainfall (A) and rainy day (B) of four environment (locations) from January 2015 to December 2018

Table 1 .
Source of variation, degree of freedom, expected mean squares, F-ratios and variance component for combined analysis of variance of randomized complete blocks experiments (a = number of weather, r = number of locations, t = number of population, GM = grand mean).

Table 2 .
Combined analysis of variance of CLR parameters of seven arabica coffee populations over two weathers (ns = not significant, * = significant at the 5% level of significance, ** = significant at the 1% level of significance).

Table 3 .
Coffee leaf rust (CLR) parameters of seven arabica coffee populations in two weathers LSD 0.05 = Fisher's least significant difference for the 5% level of significance.LSD 0.05 = 8.10 for Incidence of rust in branch (IRB), LSD 0.05 = 10.45 for Incidence of rust on leaf (IRL), LSD 0.05 = 7.56 for Severity of rust on leaf (SRL).The means that followed common letter in Severity of rust on leaf (SRL), Severity of rust on leaf (SRL), and Severity of rust on leaf (SRL) were not significantly different for the 5% level of significance based on Fisher's least significant difference test).

Table 5 .
Variance component of coffee leaf rust (CLR) parameters of seven arabica coffee populations over two weathers.

Table 8b .
Coefficient correlation and regression between coffee leaf rust (CLR) parameter (y) and rainy day in one (x) and two months (x) before data recording of arabica coffee seven populations in Environment-3 and Environment-4 (n = 8, r value at the 5% level of significance = 0.707, r value at the 1% level of significance = 0.834, ns = not significant, * = significant at the 5% level of significance, ** = significant at the 1% level of significance).