Tropical Plant Pathology

https://doi.org/10.1007/s40858-018-0271-8

ORIGINAL ARTICLE

Seasonal and altitudinal differences in coffee leaf rust epidemics

on coffee berry disease-resistant varieties in Southwest Ethiopia
Gerba Daba 1,2 & Kenny Helsen 1 & Gezahegn Berecha 2 & Bart Lievens 3 & Adugna Debela 4 & Olivier Honnay 1

Received: 16 August 2018 / Accepted: 17 December 2018

# Sociedade Brasileira de Fitopatologia 2019

Abstract
Coffea arabica is native to the Afromontane forests of southwestern Ethiopia, the leading African country in Arabica coffee production.
The intensity of coffee leaf rust (CLR), a fungal disease of growing concern to coffee farmers, was assessed in eight coffee berry
disease-resistant C. arabica varieties planted at three different altitudes. Disease variables assessed were CLR prevalence (percent of
infected plants), incidence (percent rusted leaves) and severity (percent leaf area affected) at four times within a year (Nov, Jan, Apr and
Jun). The intensity of CLR epidemics was highest at the lowland and lowest at the highland location. We hypothesize that inoculum
may be limited by cooling conditions that negatively affects fungal sporulation. CLR intensity was highest at harvest time (Nov) and
decreased during the following dry season, reaching zero values in Apr and Jun. All coffee varieties were similarly susceptible to CLR.
Our data suggest that more efforts should be employed by breeders to develop CLR-resistant coffee varieties, and that shifting
cultivation from lower to higher altitudes may contribute to achieve to higher yields due to lower impact of CLR epidemics.

Keywords Coffea arabica L. . Hemileia vastatrix . Epidemics . Host plant resistance

Introduction production is estimated to be more than 350,000 tons, of

which 90% comes from home gardens and semi-natural agro-
Among all natural commodities, the monetary value of the forestry systems, and the remaining 10% originates from
coffee trade is only exceeded by that of oil (DaMatta and large-scale intensively managed shade plantations (Alemu
Ramalho 2006; Gichimu 2012). The crop is also important et al. 2016).
export revenue and provides labor opportunities in many trop- Despite its huge economic importance, Ethiopian coffee pro-
ical countries (Kushalappa and Eskes 1989; Ribeyre and duction faces limiting factors, including weather, pests and dis-
Avelino 2012; Talhinhas et al. 2017). Coffea arabica is native eases constrains together with poor agronomic practices (Aerts
to the Afromontane forests of SW Ethiopia and the country et al. 2017; Alemu et al. 2016; Davis et al. 2012; Moat et al.
ranks first in Africa and fifth in the world in terms of Arabica 2017). Plant diseases are a major cause of yield losses in
coffee production. The average annual Ethiopian coffee Ethiopia, including coffee berry disease (caused by
Colletotrichum kahawae), coffee wilt disease (caused by
Section Editor: Nik Cunniffe Gibberella xylarioides), and coffee leaf rust. These diseases may
account for yield losses of 24–30%, 29% and 20–25%, respec-
* Olivier Honnay tively (Alemu et al. 2006; Rutherford 2006; Zeru et al. 2012).
[email protected] Coffee leaf rust (CLR) is a fungal disease caused by
1
Hemileia vastatrix (Basidiomycota, Pucciniaceae) (Arneson
Plant Conservation and Population Biology, Biology Department, 2000; Matovu et al. 2013). H. vastatrix is a hemicyclic fungus
KU Leuven, Leuven, Belgium
2
and the urediniosporic life cycle is the most important source
College of Agriculture and Veterinary Medicine, Department of of inoculum (Talhinhas et al. 2017). Chlorotic spots are the
Horticulture and Plant Sciences, Jimma University, Jimma, Ethiopia
3
first macroscopic symptoms, preceding the differentiation of
Laboratory for Process Microbial Ecology and Bioinspirational suprastomatal uredinia. Ultimately, the disease leads to defo-
Management, Department of Microbial and Molecular Systems, KU
Leuven, Leuven, Belgium liation, and when acute it can lead to the death of branches and
4 heavy crop losses (Avelino et al. 2015). The peak of the CLR
Ethiopian Horticulture and Agriculture Investment Authority, Addis
Ababa, Ethiopia symptom development in Ethiopia occurs during fruit
 Trop. plant pathol.

harvesting in the dry season but primary crop losses during Belachew et al. 2015). These CBD-resistant varieties were
initial infection are generally low. Secondary crop losses as a identified through long-term observational and experimental
result of decreased plant fitness following previous infections research during the 1970s and 1980s (Van der Graaff 1981).
tend to be more substantial (Talhinhas et al. 2017). Although several CBD resistant C. arabica variaties from East
CLR was first reported in Ethiopia in 1934 (Sylvain 1958). Africa (such as Ruiru 11 or Batian in Kenya) are known to
The current prevalence of CLR in Ethiopia was estimated in also express some level of resistance to CLR (Gichimu 2015),
27% at Mechara (Hararghe) in the eastern part of the country, the susceptibility of CBD resistant C. arabica varieties to
and from 33% to 96% in the montane rainforests of SW CLR is not known for the Ethiopian varieties. Given the high
Ethiopia (Zeru et al. 2012). Furthermore, the prevalence of prevalence of CBD in SW-Ethiopia, almost all newly planted
CLR is currently increasing in Ethiopia; percentage of infect- coffee is CBD resistant (Getu et al. 2017). Hence, it is critical
ed trees increased from 12.9% in 2000 year to 36.0% in 2010 to assesses and compare their susceptibility related to CLR
(Zeru et al. 2012). The growing CLR incidence has been as- before starting new plantations. Also, identifying the optimal
sociated with a changing climate, especially increased temper- time of assessment for CLR infection is important for its man-
atures and annual rainfall, earlier rain in the season and sun- agement strategy. The objective of this study was to monitor
shine duration reduction, which favor the life cycle of and compare CLR epidemics in eight CBD-resistant
H. vastatrix (Avelino et al. 2015). An early onset of the rainy C. arabica varieties planted at three locations (altitudes)
season favors early development of the disease, as reported for throughout one annual seasonal cycle in SW Ethiopia. Both
CLR epidemics in Nicaragua (Avelino et al. 2015). A reduc- single and interaction effects were of interest.
tion in the daily thermal amplitude during the rainy season
onset further shortens the CLR latent period (Avelino et al.
2015; Boudrot et al. 2016). In the Americas, CLR prevalence Material and methods
is strongly linked to rainfall, with severe outbreaks of CLR
occurring during the two annual rainy seasons (Boudrot et al. Study area and experimental setting
2016; Vandermeer et al. 2017). After each rainy season, CLR
strongly declines and the coffee shrubs shed infected leaves Our study was conducted at three sites in the Mana district,
(Vandermeer et al. 2017). Jimma Zone, Oromia Regional State, Southwestern Ethiopia
Detailed information on the seasonality of CLR in (Table 1 and Fig.1). The average annual rainfall, mean daily
Afromontane forests at different altitudes has not yet been minimum temperature, and maximum temperature are
documented for SW-Ethiopia. In Rwanda, CLR was reported 1523 mm, 13 °C and 24.8 °C, respectively (Bote and Jan
to decrease with altitude; CLR severity (percent leaf area dam- 2017). Mana is one of the major coffee producing districts in
aged) decreased by 1.5% per 100 m increase in altitude, when the Jimma zone with a total cultivated area of 48,000 ha of
limited between 1400 and 1800 m of altitude (Bigirimana which 23% is in lowland areas, 65% in midland, and the
et al. 2012). The increasing altitude would be a surrogate for remaining 12% in highlands (Bote and Jan 2017). There are
lowering temperature, negatively affecting sporulation, thus two seasons in Ethiopia (the dry and the rainy season). The
reducing inoculum levels (Bigirimana et al. 2012). Whether onset of the dry season is in September and the onset of the
the same pattern occurs in SW Ethiopia is not entirely known, rainy season is in June. In June 2012, eight different CBD
but data from Bonga, Berhane-Kontir,Yayu and Harenna areas resistant C. arabica varieties which are the most commonly
in Ethiopia corroborates this hypothesis (Jefuka et al. 2010; used in the region (74–1, 75–227, 74–54, 74–112, 74–140,
Zeru et al. 2012). 74–158, 74–165 and 74–148) (Table 2) were planted at three
The most commonly used C. arabica varieties in Ethiopia different sites with different altitudes: Degelu (Low: 1450 m),
exhibit resistance to CBD, as other means to control the dis- Gembe (Middle: 1610 m) and Buture (High: 2063 m)
ease are scarce (Belachew et al. 2015; Van der Graaff 1981). (Table 1, Fig.1). The study was conducted in semi-plantation
Out of the 40 C. arabica varieties used in Ethiopia, 13 are coffee systems where the coffee management practices were
CBD-resistant (G. Bekele Personal communication; uniform. Agronomic practices such as weeding, slashing,

Table 1 Description of the experimental sites

Location Altitude Altitude (m.a.s.l.) Longitude Latitude T max (°C) T min (°C) T mean (°C) Humidity (%)

Degelu Lowland 1450 37°02′43″ E 08°67′96” N 26.51 17.61 21.5 69.57

Gembe Midland 1610 37°07′44″ E 08°67′10” N 22.17 13.01 18.8 72.80
Buture Highland 2063 37°02′50″ E 08°56′96” N 20.86 10.37 17.1 75.64
Trop. plant pathol.

Fig. 1 Map of the study sites in the Mana district, Southwestern Ethiopia

pruning and planting distance were identical. Addition of fer- Disease assessments
tilizer in the coffee farms is not practiced and they were not
applied to the experimental sites in all the three locations. The Three disease variables: CLR prevalence, incidence and se-
dominant shade tree species in all the three locations were verity (Jefuka et al. 2010) were assessed four times during the
Cordia Africana, Acacia abyssinica, Albizza grandibtateata seasons: Nov 2015, Jan 2016, Apr 2016 and Jul 2016 on five
and Vernonia amygdalina. The shade level in all the three coffee plants selected at random in each plot. CLR prevalence
locations was uniform. At each altitude, a randomized com- was operationally defined as the percentage of plants affected
plete block design was established, with three blocks per lo- by CLR per plot. For assessing incidence, six branches, one
cation. Each block consisted of 8 plots (10 × 10 m each), to- pair in the upper, middle and lower canopy layer, were then
taling 24 plots per location (Fig. 2). Each variety was random- selected and marked and the percentage of diseased leaves per
ly assigned to the plots, with each plot composed of 25 plants branch was recorded. CLR severity refers to the percent leaf
of the same variety spaced at 2-m from each other (Fig. 2). area affected, which was estimated with the aid of a diagram-
matic scale with five severity levels: 1, 3, 5, 7 or 10%
(Kushalappa and Chaves 1980). Both incidence and severity
estimates were averaged across branches to obtain one value
Table 2 Coffee berry disease resistant Coffea arabica varieties used in
this study, including typical grain yield and growth habit per shrub. All measures were then expressed (averaged) at the
plot level.
Variety Grain yield (kg ha−1) Habit

On research station On farm

Data analyses

74–1 1200 600–700 Open The effects of C. arabica varieties, altitude and assessment time
75–227 1700 800–900 Open on the disease were assessed by fitting generalized linear mixed
74–54 1800 800–900 Intermediate effects models (GLMM) with maximum likelihood estimation
74–112 1800 900–1000 Compact to data from each experimental site. For ratio variables such as
74–140 1900 900–1000 Compact proportions/percentages that are constrained between 0 and 1, a
74–148 1800 600–700 Compact beta rather than a normal distribution was used, by setting a
74–158 1900 900–1000 Compact logit link function. Because proportional boundaries 0 and 1
74–165 1700 800–900 Compact are not allowed in a beta distribution, the original responses
(Y) were transformed using Y′ = (Y x (N-1) + 0.5) / N, where
 Trop. plant pathol.

Fig. 2 Experimental set-up. Eight C. arabica varieties are replicated three times at each location (lowland, midland and highland). Twenty-four plots at
each location, each plot having a 10x10m size. Numbers indicate the C.arabica variety ID

N is the sample size (Cribari-Neto and Zeileis 2009). Measures location x time interaction for all disease measures showed
assessed in the same individuals at different times lack indepen- significant reduction in disease only for lowland and midland
dence. A plot ID variable, since the same plot was assessed locations, especially CLR prevalence (lowland z = 12.38,
through time, was included as a random intercept in each P < 0.001; midland z = 7.80, P < 0.001) and incidence (low-
GLMM. Block ID was furthermore included as a second ran- land z = 11.76, P < 0.001; midland z = 4.56, P < 0.001), and
dom intercept. Altitude (lowland, midland or highland), only for lowland locations for CLR severity (z = 9.21, P <
C. arabica variety and time of evaluation and their first order 0.001) (Table 4, Fig. 3a, b, c).
interactions were included as fixed factors. Final models were The differences in disease intensity among the three loca-
obtained after stepwise model reduction, based on Akaike’s tions in November were reduced in January, with only signif-
information criteria (AIC). All models were fitted using the icant differences between highland and midland (z = −3.70,
glmmTMB packag (Magnusson et al. 2017) under R 3.3.3 sta- p < 0.001); between highland and lowland (z = −2.37, P =
tistical environment (R Development Core Team, 2017). 0.046) for CLR prevalence (Table 4 and Fig. 3c); between
Pairwise comparisons between factor levels were based on highland and midland (z = −2.58, P = 0.027) for CLR inci-
Tukey test at 5% significance. Note that due to the nested block dence (Table 4, Fig. 3b); and between highland and midland
design of the experiment, collinearity problems across the fixed (z = −2.63, P = 0.024) for CLR severity (Table 4, Fig. 3b).
factors did not occur. Neither the main effect of coffee variety, nor the interaction
between coffee variety and location significantly affected any
of the disease measures (Fig. 4a, b, c). Note that because of
Results this non-significance (P < 0.05), these factors were removed
from the final GLMM models presented in Table 3, following
The mixed models results showed that coffee varieties did not the stepwise model reduction.
affect CRL epidemics, while time, location and their interac-
tion effects did affect (P < 0.0001) all CLR variables (Table 3).
On average, CLR tended to decrease during the dry season Discussion
(January) following harvest (Table 3, Fig. 3), reaching zero
values in April and July (data not shown). Disease intensity Our findings contribute new knowledge to CLR epidemics in
was highest at the lowland locations and lowest for highland SW Africa. We corroborate previous reports of the effect of
locations during November (Fig. 3). However, the significant altitude shaping CLR development on Arabica coffee and

Table 3 Parameter estimates of the generalized linear mixed effect models (GLMM) on the coffee leaf rust (CLR) infection measures. Beta
coefficients, marginal means (MM) and test statistics of contrast predictor = 0 (z) given for each retained predictor after model reduction (N = 144)

Coefficient CLR Incidence CLR Severity CLR Prevalence

beta MM z beta MM z beta MM z

Location Highland −3.59 0.028 7.29*** −3.84 0.022 5.73*** −2.10 0.11 16.51***
Midland −1.89 0.086 23.28*** −2.79 0.052 14.39*** −0.87 0.47 36.48***
Lowland −0.57 0.13 27.96*** −0.98 0.092 16.95*** 2.85 0.66 36.12***
Time November −3.59 0.12 28.9*** −3.84 0.074 20.30*** −2.10 0.63 45.34***
January −3.51 0.041 12.96*** −3.77 0.030 8.79*** −2.15 0.17 25.94***
Location*Time Highland:Nov −3.59 0.027 5.88*** −3.84 0.021 4.66*** −2.10 0.11 13.09***
Highland:Jan −3.51 0.029 6.33*** −3.77 0.023 5.03*** −2.15 0.10 12.48***
Midland:Nov −1.89 0.13 23.66*** −2.79 0.058 12.23*** −0.87 0.70 32.23***
Midland:Jan −4.52 0.056 12.70*** −4.09 0.046 10.50*** −4.10 0.25 20.64***
Lowland:Nov −0.57 0.36 46.88*** −0.98 0.27 36.01*** 2.85 0.95 29.37***
Lowland:Jan −6.17 0.041 9.17*** −6.47 0.027 4.97*** −6.43 0.18 17.72***

*
0.05 ≥ P > 0.01; ** 0.01 ≥ P > 0.001; *** 0.001 ≥ P
Trop. plant pathol.

with increasing altitude has been linked to decreasing temper-

atures that affect rust spore germination (Kushalappa and
Eskes 1989). The optimal temperature for H. vastatrix spore
germination is 22 °C, or in the range of 20 to 25 °C
(Bigirimana et al. 2012; Ribeyre and Avelino 2012;
Talhinhas et al. 2017; Toniutti et al. 2017). This matches the
average temperatures at our lowland site (21.5 °C). Bote and
Jan (2017) reported average temperatures at our highland and
midland locations of 17.1 °C and 18.8 °C, respectively. These
lower temperatures at both locations likely resulted in extend-
ed latent period and slower disease development. Also, larger
seasonal or daily temperature fluctuations and lower night-
time temperatures, commonly observed at higher latitudes,
have been associated with reduced CLR incidence
(Kushalappa and Eskes 1989). The observed gradual decrease
in CLR with increasing altitude has also been observed in the
Americas (Avelino et al. 2015). Although out study did not
allow to disentangle the effects of altitude and inoculum pres-
sure, and although we have no measures of ambient spore
amounts, the higher altitude is likely a surrogate variable for
lower inoculum pressure.
As to the seasonal variation of CLR, we found higher in-
tensity during coffee harvest time (November), decreasing
during the following dry season (January). The disease was
not observed during the subsequent rainy seasons (April and
July assessments). The peak of CLR epidemics in November
may be related to spore release due to the physical disturbance
introduced by the operations of coffee harvesting, which oc-
curs from late September to October. Similar findings were
reported in coffee farms in Costa Rica and Mexico, where
secondary infection cycles occur right after harvesting time
and dramatically expands from November to January
Fig. 3 Incidence (a), Severity (b) and Prevalence (c) of CLR as
influenced by the interaction effect of altitude and time (all interactions (Boudrot et al. 2016; Vandermeer et al. 2017). During our
were significant, p < 0.0001). Error bars represent standard error of the study, weather was generally dry in January and probably
mean (N = 144) there was no free water to support rust disease development,
agreeing with previous data from Ethiopia (Jefuka et al. 2010).
further demonstrated that CBD-resistant C. arabica behaved We found no significant differences in CLR intensity
similarly in relation to CLR. The decrease in CLR intensity among the C. arabica varieties across the altitudinal and

Table 4 Results of the pairwise

contrasts (Tukey tests) for the CLR Prevalence CLR Incidence CLR Severity
generalized linear mixed effect
models (GLMM) on the coffee z p z p z p
leaf rust (CLR) infection
measures. Test statistics (z) and p- Location highland-midland −9.67 <0.0001 −6.01 <0.0001 −4.23 <0.0001
highland-lowland −11.19 <0.0001 −8.69 <0.0001 −7.73 0.0001
values given for each pairwise
lowland-midland 3.93 0.0003 3.41 0.0019 3.41 0.0019
contrast (N = 144) Time November–January 11.77 <0.0001 8.57 <0.0001 6.36 <0.0001
Location*Time highland:Nov-Jan 0.18 0.86 −0.32 0.75 −0.28 0.78
midland:Nov-Jan 7.80 <0.0001 4.56 <0.0001 1.12 0.26
lowland:Nov-Jan 12.38 <0.0001 11.76 <0.0001 9.21 <0.0001
Time*Location Nov: highland-midland −10.47 <0.0001 −6.73 <0.0001 −3.94 <0.0001
Nov: highland-lowland −13.03 <0.0001 −11.69 <0.0001 −10.65 0.0002
Nov: Lowland-Midland 6.21 <0.0001 8.16 <0.0001 8.56 <0.0001
Jan: highland-midland −3.70 0.0006 −2.58 0.027 −2.63 0.024
Jan: highland-lowland −2.37 0.046 −1.34 0.37 −0.58 0.83
Jan: lowland-midland −1.38 0.35 −1.38 0.35 −1.87 0.15
 Trop. plant pathol.

furthermore also provide producers with better coffee bean

qualities and better prices (Tolessa et al. 2017; Worku et al.
2018). Yet, CLR disease was recorded at the highland site but
at reduced levels. The eight C. arabica varieties evaluated
here were equally susceptibley to CLR, suggesting that further
efforts need to be directed towards developing CBD resistant
varieties that are also resistant to CLR.
Coffee in Ethiopia is grown across a huge area, under a wide
variety of production systems and cultivation practices. Since
this research was limited to one production system and particu-
lar sites in SW Ethiopia, further assessment of CLR in relation to
altitude and inoculum pressure is encouraged. Although our
study was limited to one production cycle, we do not expect
that interannual variability in weather would drastically influ-
ence the results, but it will be instructive to keep collecting and
analyzing data in the following seasons in order to confirm our
findings. Finally, if CLR epidemics are further confirmed to not
occur during April and June, evaluations in November and
January may suffice for collecting epidemiological data.

Acknowledgements We want to express our gratitude to VLIR-UOS, the

Institutional University Cooperation programme between the Flemish
University and Jimma University for funding this research work. We
are also indebted to Shafi Abba Fixa, Biya Abba Macca and Awol, the
owners of the coffee farms at three different locations in the Mana district
to let us work in the coffee farm and provide all the necessary information
during the research work.

Author’s contribution GD, AD and GB designed the study. GD per-

formed the field work. OH and GD drafted the manuscript. All authors
contributed to data interpretation. KH and GD performed the statistical
Fig. 4 Incidence (a), Severity (b) and Prevalence (c) of CLR among a analyses. All authors gave final approval for publication.
number of C. arabica varieties with resistance against coffee berry
disease. There were no significant differences among varieties. Error Publisher’s note Springer Nature remains neutral with regard to jurisdic-
bars represent standard error of the mean (N = 144) tional claims in published maps and institutional affiliations.

seasonal gradients. However, since only a limited number of

potentially rather genetically similar, C. arabica varieties test-
References
ed, the possibility remains that other coffee varieties exhibit Aerts R, Geerart L, Berecha G, Hundera K, Muys B, De Kort H, Honnay
some level of resistance to CLR. The Ethiopian montane O (2017) Conserving wild Arabica coffee: emerging threats and
rainforests indeed contain the wild gene pool of Arabica cof- opportunities. Agricultura, Ecosystems & Environment 237:75–79
fee, although this gene pool is currently highly underexploited Alemu K, Adugna G, Lemessa F, Muleta D (2016) Current status of
coffee berry diseases (Colletotrichum kahawae Waller and bridge)
(Aerts et al. 2017). Future research should focus on systemat-
in Ethiopia. Archives of Phytopathology and Plant Protection 49:
ically screening larger numbers of cultivated and wild coffee 421–433
varieties to identify potential CLR disease resistance. Apart Arneson P (2000) Coffee rust. The plant health instructor. The American
from genetic make-up, the susceptibility of coffee varieties to Phytopathological Society. Retrieved August 13, 2011 from http://
CLR is influenced by crop management, physiological status www.apsnetorg/edcenter/in/coffee.rust.Aspx
Avelino J, Cristancho M, Georgiou S, Imbach P, Aguilar L, Bornemann
of the plant, and the production situation (Ribeyre and Avelino
G, Morales C (2015) The coffee rust crises in Colombia and Central
2012; Kushalappa and Eskes 1989; Toniutti et al. 2017). High America (2008–2013): impacts, plausible causes and proposed so-
berry yield and high fruit load, for example, have been report- lutions. Food Security 7:303–321
ed to enhance H. vastatrix infection, as well as a high leaf-to- Belachew K, Teferi D, Livelihood E (2015) Climatic variables and impact of
fruit ratio (Ribeyre and Avelino 2012; Kushalappa and Eskes coffee berry diseases (Colletotrichum kahawae) in Ethiopian coffee
production. Journal of Biology, Agriculture and Healthcare 5:55–64
1989; Toniutti et al. 2017). Our results suggest that moving Bigirimana J, Njoroge K, Gahakwa D, Phiri NA (2012) Incidence and
coffee plantations to higher altitudes may help to protect severity of coffee leaf rust and other coffee pests and diseases in
against CLR. Producing coffee at higher altitude will Rwanda. African Journal of Agricultural Research 7:3847–3852
Trop. plant pathol.

Bote AD, Jan V (2017) Tree management and environmental conditions Matovu R, Kangire A, Phiri N, Hakiza G, Kagezi G, Musoli P (2013)
affect coffee (Coffea arabica L.) bean quality. NJAS-Wageningen Ecological factors influencing incidence and severity of coffee leaf
Journal of Life Sciences 83:39–46 rust and coffee berry disease in major Arabica coffee growing dis-
Boudrot A, Pico J, Merle I, Granados E, Vílchez S, Tixier P, Allinne C tricts of Uganda. Uganda Journal of Agricultural Sciences 14:87–
(2016) Shade effects on the dispersal of airborne Hemileia vastatrix 100
uredospores. Phytopathology 106:572–580 Moat J, Williams J, Baena S, Wilkinson T, Gole TW, Challa ZK, Davis
Cribari-Neto F, Zeileis A (2009) Beta regression in R. CRAN.R-project AP (2017) Resilience potential of the Ethiopian coffee sector under
DaMatta FM, Ramalho JDC (2006) Impacts of drought and temperature climate change. Nature Plants 3:17081
stress on coffee physiology and production: a review. Brazilian Ribeyre F, Avelino J (2012) Impact of field pests and diseases on coffee
Journal of Plant Physiology 18:55–81 quality Specialty Coffee: Managing Coffee: International Plant
Davis AP, Gole TW, Baena S, Moat J (2012) The impact of climate Nutrition Institute Penang, Malaysia, pp 151–176
change on indigenous arabica coffee (Coffea arabica): predicting Rutherford MA (2006) Current knowledge of coffee wilt disease, a major
future trends and identifying priorities. PLoS One 7(11):e47981 constraint to coffee production in Africa. Phytopathology 96(6):
van der Graaff NA (1981) Selection of arabica coffee types resistant to 663–666
coffee berry disease in Ethiopia. In: Dissertation Sylvain PG (1958) Ethiopian coffee its significance to world coffee prob-
Landbouwhogeschool Wageningen. Wageningen, The Netherlands lems. Economic Botany 12:111–139
Talhinhas P, Batista D, Diniz I, Vieira A, Silva DN, Loureiro A, Guerra-
Getu B, Belachew B, Ayano A (2017) Strengthening Arabica coffee
Guimarães L (2017) The coffee leaf rust pathogen Hemileia
(Coffea arabica L.,) quality breeding research strategy in Ethiopia
vastatrix: one and a half centuries around the tropics. Molecular
through optimum use of coffee terroir: farm management. 25th in-
Plant Pathology 18:1039–1051
ternational conference on coffee science, Armenia., pp 42–55
Tolessa K, D’heer J, Duchateau L, Boeckx P (2017) Influence of growing
Gichimu BM (2012) Field screening of selected Coffea arabica L. geno- altitude, shade and harvest period on quality and biochemical com-
types against coffee leaf rust. African journal of. Horticultural position of Ethiopian specialty coffee. Journal of the Science of
Science 6:82–91 Food and Agriculture 97:2849–2857
Gichimu BM (2015) Coffee Breeding in Kenya: Achievements, Toniutti L, Breitler JC, Etienne H, Campa C, Doulbeau S, Urban L,
Challenges and Current Focus Bertrand B (2017) Influence of environmental conditions and genet-
Jefuka C, Fininsa C, Adugna G, Hindorf H (2010) Coffee leaf rust epi- ic background of Arabica coffee (C. arabica L) on leaf rust
demics (Hemileia vastatrix) in montane coffee (Coffea arabica L.) (Hemileia vastatrix) pathogenesis. Frontiers in Plant Science 8:2025
forests in southwestern Ethiopia. East African Journal of Science 4: Vandermeer J, Hajian-Forooshani Z, Perfecto I (2017) The dynamics of
86–95 the coffee rust disease: an epidemiological approach using network
Kushalappa A, Chaves G (1980) An analysis of the development of theory. European Journal of Plant Pathology 5:1–10
coffee rust in the field. Fitopatologia Brasileira 5:95–103 Worku M, de Meulenaer B, Duchateau L, Boeckx P (2018) Effect of
Kushalappa AC, Eskes AB (1989) Advances in coffee rust research. altitude on biochemical composition and quality of green arabica
Annual Review of Phytopathology 27:503–531 coffee beans can be affected by shade and postharvest processing
Magnusson A, Skaug H, Nielsen A, Berg C, Kristensen K, Maechler M, method. Food Research International 105:278–285
van Bentham K, Bolker B, Brooks M (2017) glmmTMB: Zeru A, Assefa F, Adugna G, Hindorf H (2012) Occurrence of fungal
Generalized linear mixed models using template model builder. R diseases of Coffea arabica L. in montane rainforests of Ethiopia.
package version 0.1.1 Journal of Applied Botany and Food Quality 82:148–151