The relationship between elevation, soil temperatures, soil chemical

characteristics and green coffee bean quality and biochemistry in southwest

Ethiopia

Merkebu Getachewab (merkeb2020@[Link]), Kassaye Tolassac, Pieter De Frennea, Kris Verheyena,

Ayco J.M. Tackd, Kristoffer Hylanderd, Biruk Ayalewd, Pascal Boeckxe
a
Forest and Nature Lab, Department of Environment, Faculty of Bioscience Engineering, Ghent
University, Belgium
b
Department of Horticulture and Plant Sciences, College of Agriculture and Veterinary Medicine,
Jimma University, Ethiopia
c
Ethiopian Institute of Agricultural Research, Addis Ababa, Ethiopia
d
Department of Ecology, Environment and Plant Sciences, Stockholm University, Stockholm, Sweden
e
Isotope Bioscience Laboratory, Department of Green Chemistry and Technology, Faculty of
Bioscience Engineering, Ghent University, Belgium

Abstract

Green coffee bean quality and biochemistry are influenced by environmental variables. The present
study was designed to study the influence of soil temperatures and soil chemistry on bean physical
attributes, bean quality (assessed by three internationally trained, experienced and certified Q-grade
cuppers licensed by the Specialty Coffee Association (SCA) Coffee Quality Institute (CQI) and
biochemistry of green coffee beans). The study was performed in 53 farms in southwest Ethiopia
distributed along an elevational gradients (1500 – 2160 m a.s.l.) and with varying shade canopy cover
(open to dense shade). A total of 159 individual coffee trees were sampled. Shade tree canopy cover,
soil temperature, and soil chemistry as well as coffee management intensity were quantified as
explanatory variables. Green bean quality was negatively correlated to soil temperatures. On the other
hand, hundred bean mass and green bean biochemistry (caffeine, trigonelline and chlorogenic acid
contents) were negatively correlated to soil temperatures but positively to soil chemistry. During the
coffee fruit development period (flowering to fruit maturity), temperature appeared to be a driving factor
influencing coffee bean quality and biochemistry. Total specialty quality was significantly associated
with soil chemistry , in which 84% of the variation could be explained by soil chemical variables. This
study is the first to demonstrate the relationship between soil temperatures and chemistry in coffee bean
quality and green bean biochemical compositions. Although the relative importance of factors such as
air temperatures and humidity and soil moisture are missing from this study, we find that soil
temperatures and soil chemistry have a strong effect on coffee bean quality and biochemistry. Overall,
climate change, which generally involves a substantial increase in mean temperatures of tropical
regions, could be expected to have a negative impact on coffee bean quality and biochemistry.

Keywords: Arabica coffee, Ethiopia, cup quality, soil properties, microclimate

1
1. Introduction

Coffee is one of the most important global commodities providing livelihood opportunities to millions
of people in the global South (Legesse 2020; Ovalle-Rivera et al. 2015; Davis et al. 2012). In addition
to being an important cash crop to farmers in Ethiopia, coffee generates about a quarter of the country's
export earnings (Davis et al. 2012). Southwestern Ethiopia is known as a primary center of origin and
diversity of Arabica coffee (hereafter referred to as coffee), where the species grows naturally as
understory shrub in moist Afromontane forests (Hundera et al. 2013; Davis et al. 2012). Coffee in this
region is mainly grown under natural shade trees or in farming systems that deliberately incorporate
specific shade trees. Common coffee shade tree species in this region include Albizia gummifera, Acacia
abyssinica, Cordia africana, Croton macrostachyus, and Millettia ferruginea.

Coffee bean quality is mainly described by its physical attributes, cup quality (sensory quality attributes)
and chemical bean constituents (Cheng et al. 2016; Dos Santos Scholz et al. 2016). Cup quality,
determines the desirability of a coffee for consumption which impacts the price setting and
competitiveness in the international markets (Tolessa et al. 2018). These quality attributes, such as cup
cleanness, acidity, body and flavor can be distinguished by sensory organs and are assessed by
professional cup tasters based on established procedures. Green bean chemical constituents like
caffeine, trigonelline and chlorogenic acids (CGA) are the dominant precursors for coffee aroma and
flavor development and these chemical compounds determine the flavor, aroma and bitterness, and
ultimately the quality of the final coffee beverage (Chemura et al. 2021; Cheng et al. 2016; Bertrand et
al. 2012; Joet et al. 2010). On the other hand, bean physical attributes are mainly determined by mass,
defects, length and diameter (Teklu et al. 2011).

Regional warming and increasingly erratic rainfall have already increased the frequency of poor
harvests, affecting coffee productivity (DaMatta et al. 2018; Ramalho et al. 2018; Laderach et al. 2017;
Craparo et al. 2015). Rising temperatures, droughts, and erratic weather patterns are predicted to reduce
the overall land suitable for growing Arabica coffee in Ethiopia by 50% between 2040-2070 primarily
in the Harar, Sidamo and Jimma areas (Moat et al. 2017). Nearly half of the current coffee growing
areas of the country would lose 20-40% climate suitability, mainly in areas of low to medium elevations
(Ovalle-Rivera et al. 2015; Davis et al. 2012). The productivity and quality of coffee depends on
complex spatiotemporal interactions of climatic, topographic, edaphic and biological components
(DaMatta et al. 2019; Cerda et al. 2017), which complicates the prediction of climate change impacts.
Hence, in the context of the observed and predicted climate changes, continuous adaptation will remain
a significant challenge for coffee production and bean quality.

The spatial distribution of air temperature is controlled by elevation (Navarro-Serrano et al. 2020).
Likewise, soil temperature is the function of heat flux in the soil as well as heat exchanges between the

2
soil and atmosphere (Elias et al. 2004). Although soil temperatures are affected by the interaction of
multiple factors like soil inputs, mulching, irrigation, air temperature was found to be the main driving
variable for the changes in soil temperatures (Onwuka and Mang 2018; Barman et al. 2017). Coffee
trees growing at high elevations produce beans with greater bean quality because elevation affects
coffee quality via the growing temperatures (Avelino et al. 2007; Muschler 2001). Lower temperature
slows down bean ripening processes and allows more time for bean filling (Muschler 2001). Literature
relating soil temperatures with coffee bean quality and biochemistry are rarely available and hence, our
main focus was given to measure this variable. Meanwhile, soil temperature is an important
environmental variable but it is a rarely reported indicator of climate change. In this regard, the ideal
growing temperature of Arabica coffee was appeared to be between 18 and 21°C (Camargo 2010) and
temperatures outside this range will cause incomplete maturation. Effect of elevation on biochemical
composition of green coffee beans has also been demonstrated in that a higher caffeine content with the
increasing elevation, while chlorogenic acids content decline (Avelino et al. 2007). Bertrand et al.
(2012), on the other hand, reported a higher chlorogenic acid content at higher elevations than at the
low. Meanwhile, soil nutrients perform specific and essential functions in plant metabolism, and their
deficiency may lead to metabolic disturbances that can affect coffee bean quality and biochemistry.
These soil nutrients are influenced by environmental changes particularly, temperature changes might
have a pronounced effect on soil nutrient dynamics which might have effect on coffee bean quality and
biochemistry (Clemente et al. 2015).

There are several alternatives to reduce the negative impacts of high temperatures on coffee production
and to produce high-quality beans. One of the option is moving its cultivation areas to higher elevations
as suitability moves upslope to compensate for the increased temperatures (Moat et al. 2017; Ovalle-
Rivera et al. 2015). The other option is via managing the canopy shade and the associated cooling of
the understorey microclimate (De Frenne et al. 2019, Davis et al. 2012). Shifting coffee-growing areas
upslope might bring conflicts with land uses with crops other than coffee (Magrach and Ghazoul 2015).
Besides, it can induce adverse socio-economic and environmental impacts associated with deforestation
for new coffee cultivation (Meyfroidt et al. 2013). Open land at high elevations might be remote or too
steep for growing coffee (Bunn et al. 2015; Rahn et al. 2014). The latter option (manipulating sub-
canopy microclimate), could be more realistic opportunity to modulate the increasing air temperatures
(Ruslim et al. 2017; Ovalle-Rivera et al. 2015; Davis et al. 2012), considering elevation as a space-for-
time substitute for climate warming (De Frenne et al. 2013). Meanwhile, there are also some trade-offs
and adverse effects of shade where the decrease in crop yield and quality is likely to occur as the trees
canopy cover increases due to the competition for light, water, and nutrients (Sarmiento-Soler et al.
2020; Blaser et al. 2017).

3
Although the effect of elevation and shade canopy cover on coffee bean yield has been well studied,
information on the effects of soil temperatures and soil chemistry on cup quality and green bean
biochemistry is scarce. A few available studies suggest that elevation and shade influence green coffee
bean quality and biochemistry when they were treated independently or in combination (Worku et al.
2018; Leonel and Philippe 2007; Avelino et al. 2007). Certain studies reported that more shaded
conditions reduce temperature stress in the canopy and prolongs the maturation period of coffee berries.
It also reduces periodic over-bearing and a subsequent die back of coffee plants (Leonel and Philippe
2007). Likewise, higher shade promotes slower and more balanced fruit maturation by the mother plant,
thus yielding a better-quality product than unshaded coffee plants at lower elevations (Barbosa et al.
2012; Geromel et al. 2008; Leonel and Philippe 2007; Muschler 2001). On the other hand, larger bean
size and mass were obtained when canopy cover increased from 0 to 80%. A recent study from
southwest Ethiopia indicated that dense shade at high elevations produced beans with lower cup quality,
whereas higher shading at lower elevations increased cup quality (Tolessa et al. 2017). Another recent
study from Ethiopia revealed that coffee plants grown under shade produced greater acidity than those
grown under the full sun (Worku et al. 2018). Besides, recent studies from southwest Ethiopia revealed
that soil fertility parameters affected cup quality of wild Arabica coffee in its natural habitat and that
available P, K and the ratio between Mg and K were the most important soil chemical variables that
influenced bean size, cupping scores, and green bean biochemistry (Yadessa et al. 2020; Clemente et
al. 2015).

Here we examined how soil temperatures and soil chemistry affect green coffee bean biochemical
composition, hundred bean mass and sensory quality attributes. The following research question was
specifically addressed: how does shade tree canopy cover influence coffee bean physical attributes,
brew quality, and green bean biochemistry along elevational gradients via its impacts on the soil
temperature and chemistry? We hypothesized that:
1. Cooler temperatures lead to better bean quality and higher caffeine, trigonelline, and chlorogenic
acids content.

2. Soil fertility has a positive effect on bean quality and has a higher potential to improve caffeine,
trigonelline, and chlorogenic acids content.

3. Materials and Methods

Study area
The study was conducted in the Goma and Gera districts of Jimma Zone in southwestern Ethiopia
(7°37’48’’– 7°56’37’’N latitude and 36°13’41’’– 36°39’17’’E longitudes). In total 53 coffee farms along an
elevational gradient ranging from ca. 1500 to 2160 m asl were selected. The region is characterized by
a humid and warm subtropical climate with a yearly rainfall between 1500 and 2000 mm. The main
rainy season is from May to September (monomodal rainfall), accounting for about 85% of the annual
rainfall. Coffee cultivation in the region is rain-fed. Soils are classified as Eutric Nitisols, which are

4
deep and well-drained soils with a clay content of more than 30% and a pH (measured in H2O) between
4.2 and 6.2 (Kebede et al. 2018; Kufa 2011).

Coffee plot selection and characterization

The study included agroforestry sites distributed across a landscape, comprising approximately 2500
km2 area. In order to encompass an elevational and shade tree canopy cover gradient, 53 coffee farms
(Fig 1) along an elevational gradient ranging between 1500– 2160 m asl. were selected. The sites vary
considerably in their coffee management intensity ranging from little human interference to intensive
management, including herbicides and mineral fertilizers, for the purpose of increasing coffee yield
(Zewdie et al. 2020). The farms are also characterized by dense and diverse shade tree canopy cover.
In the semi-plantation coffee production system, there is high anthropogenic disturbance resulting in a
relatively species poor canopy consisting of tree species such as Albizia schimperiana, Albizia
gummifera and Croton macrostachyus. Mulching and organic fertilizers are commonly used soil
fertility management strategies. On the other hand, a plantation coffee system represents shade
plantations with only few and large canopies. Shade trees are mainly indigenous species (most
dominantly Acacia spp. and Croton macrostachyus), though recently, fast-growing exotic species have
been introduced such as Grevillea robusta. High coffee stem density is also assumed to be a component
of this production system. Herbicides and commercial fertilizers are applied regularly and mulching is
also a common practice (personal observation). Accordingly, management intensities were classified
into a categorical variable as plantation (n=8) and smallholder farms (n=45). To avoid spatial
autocorrelation, the selected farms were at least 3-4 km apart. Sampling was conducted in 30 x 30 m
coffee plots per farm, in which three coffee trees were selected from each coffee farm giving a total of
159 coffee trees. The selected three coffee trees were spread within the 30 x 30 m coffee plots at each
site, and all of them were consistently positioned inside the plantation to avoid edge effects. The study
was conducted from February to December 2019.

Environmental drivers
The following environmental variables were measured to describe elevation, shade tree canopy cover,
soil temperature, and soil properties per coffee tree. The elevation of each coffee plot was measured
with a Global Positioning System (GPS) (Garmin-60). Hemispherical photography was used to quantify
shade tree canopy cover. Pictures were taken looking upwards from the center above each coffee tree,
using a hemispherical fish-eye lens with the camera (Canon DS126091, ICES-003, Sigma 4.5 mm lens,
Canon, Japan). Three photos were taken oriented towards the North. All pictures were taken under
solidly overcast sky between 09:00 and 16:00. The analysis was performed in the blue color channel to
improve the contrast between tree branches, foliage and sky (Schleppi et al. 2007; Juncker et al. 2004).
The pictures were processed and analyzed using Gap Light Analyzer (GLA) (Jarcuska 2008;
Jonckheere. et al. 2004; Frazer et al. 1999).

5
Fig. 1. Study area showing an aerial view of the study landscape representing the distribution of coffee farms
established within the landscape; the dots represent coffee farms (both smallholders and commercial plantation).

As a proxy for microclimate, soil temperatures were recorded in each coffee farm at a three hour
intervals between February and December 2019 (period in which coffee been development from
flowering onwards takes place) using miniature temperature sensors iButton (DS192H, Maxim/Dallas
Semi-conductor Corp., USA) buried in the soil at 10 cm depth and at 40 cm distance from the coffee
tree trunks. We could not measure air temperatures because of theft of visible devices. The daily
minimum, mean and maximum temperature were computed to ensure the best representation of
temperature experienced by the coffee plant. The daily mean temperatures were calculated as the
average of the daily maximum and minimum temperatures.

For each soil sample, an oven-dried sub-sample was used for the measurements of soil carbon, total N,
Olsen-P, Ca, Mg and K. All the soil samples were dried to a constant weight at 65°C for 48 h, ground
and sieved over a 2 mm mesh. For soil total C and N, the soil samples were combusted at 1200°C and
the gases were measured by a thermal conductivity detector in a CNS elemental analyzer (Variomax
Cube, Elementar, Germany). Olsen-P, which is a measure for plant-available P (Gilbert et al. 2009) was
extracted in NaHCO3 (ISO 11263:1994) followed by calorimetric measurement according to the

6
malachite green procedure (Lajtha et al. 1999; Robertson et al. 1999). Soil total Ca, K and Mg was
measured by atomic absorption spectroscopy after the destruction of the soil samples with HClO4
(65%), HNO3 (70%) and H2SO4 (98%) inteflon bombs for four hour at 150°C. Exchangeable K+, Ca2+,
Mg2+, Na+ and Al3+ concentrations were measured by atomic absorption spectroscopy (AA240FS, Fast
Sequential AAS) after extraction in 0.1 M BaCl2 (NEN 5738:1996).

Coffee berry sampling and measurements

All fully ripe, red-colored coffee berries were hand-picked once at peak harvest between
October and November 2019 from each selected coffee tree using local coffee bags. Berries
that ripe earlier were harvested first from lower elevation sites and subsequently the high
elevation sites followed. The berries were dry-processed, i.e. sun-dried (on raised mesh wire)
immediately after harvest (harvesting was in the morning and subjected to drying started in the
afternoon). The berries were returned to traditional coffee bags before sunset and stored in
clean rooms (to prevent spoilage) and were exposed to the sun in the morning until green beans
attained 11.5% moisture content measured using coffee moisture tester (mini GAC, Dickey -
John, USA). The berries were regularly turned to maintain uniform drying and the dried coffee
berries were separately labeled and packed for analysis. The dried coffee berries were dehusked
using a hulling coffee machine (coffee huller, McKinnon, Scotland) at Jimma University,
cleaned and stored at room temperature.

Bean physical attributes: Bean length (mm) and diameter (mm) were measured using a bean
measuring caliper (Mitutoyo, IP 67, CD-20-PPX, Kawasaki, Japan) using 10 beans per sample.
Additionally, the mass of the beans was recorded by taking 100 beans from each sample. Finally, the
green bean samples were submitted to the Ethiopian Commodity Exchange (ECX) for raw and sensory
quality analyses.

Raw quality (40% of the total preliminary quality): A green coffee bean sample of 100 g was used
for physical quality evaluation before roasting. Primary and secondary defects and odor, were assessed
according to the procedures developed by the ECX (2011). The rating was based on a scale from 0 to
15 for the defects and 0 to 10 for odor.

Cup quality (60% of the total preliminary quality): Coffee bean samples were evaluated for cup
quality attributes by a panel of three internationally trained, experienced and certified Q-grade cuppers
in Jimma ECX center. Acidity, body, cup cleanness and flavor were assessed following a standard
method (ECX 2011). This Q-grade standard method involves Q-certified cuppers, i.e., cuppers licensed
by the Specialty Coffee Association (SCA) Coffee Quality Institute (CQI).

7
Roasting, grinding, and brew preparation: This was performed by the ECX laboratory in Jimma,
Ethiopia. A roaster equipped with a cooling system, in which air was forced through a perforated plate,
capable of roasting up to 500 g of coffee beans, was used for roasting the coffee beans. An amount of
100 g green beans was used for each sample and the beans were put into the roasting machine with six
cylinders (Probat, 4 Barrel Roaster, Germany). They were carefully roasted for 7-8 minutes to medium
roast at temperatures of 200°C. Subsequently, the roasted bean samples were ground to a medium level
using a Guatemala SB electrical grinder, which were cleaned well after each sample. The medium
roasted coffee was tipped out into a cooling tray and allowed to cool down for 4 minutes rapidly by
blowing cold air through it. Then, eight grams of coffee powder was put into a 250 mL cup and 5 cups
per coffee sample were used. Next 125 ml boiled water (93°C) was poured onto the ground coffee,
followed by stirring the content to ensure homogeneity of the mixture. Then, the cups were filled with
an additional 125 mL and left to settle. After three minutes, floating coffee was skimmed, and the brew
was ready for cup tasting. Finally, the five prepared cups were cup tasted by three professional Q-grade
cuppers operating in ECX. Each panelist gave their independent judgment using a cupping form and
the average score of the three cuppers was used.

Finally, the total preliminary quality was calculated using raw and cup quality scores. Coffee samples
of grades 1-3 (specialty 1, 2 and 3) were assessed for total specialty quality. Accordingly, aroma, flavor,
acidity, body, uniformity, cup cleanness, overall preference, aftertaste, balance and sweetness were
rated on a scale from 0 to 10. The sum of all these cup quality attributes gave a total specialty quality
ranging from 0 to 100 ([Link]

Bean biochemistry: Caffeine, trigonelline and chlorogenic acid content were analyzed at the Ethiopian
Institute of Agricultural Research as described by Vignoli et al. (2011). Ground green coffee beans (0.5
g) were submitted to direct hot water (95°C, 50 mL) extraction and stirred for 20 minutes on a hot plate.
The extract was filtered through No. 4 Whatman filter followed by a 0.45 μm PTFE filter prior to
injection into High Performance Liquid Chromatography (HPLC) (Surveyor, Thermo Finnigan, USA)
having a prevail C18 (250 × 4.6 mm, i.d. 5 μm, 25°C) column. The mobile phase is composed of 5%
acetic acid in water (v/v) (solvent A) and acetonitrile (solvent B). Flow rate was 0.5 mL min-1 and the
injection volume was 10 μL. Both caffeine and trigonelline acid were detected by UV detector at 272
nm wavelength, whereas chlorogenic acids were detected at 320 nm. The reported chlorogenic acid was
the total chlorogenic acids content. In all the cases, the standards were used for quantification and
retention time determinations.

Data analyses

8
As response variables, we retained six key variables representing the physical bean attribute (hundred
bean mass), brew quality (total preliminary and specialty quality), and bean biochemical compositions
(caffeine, trigonelline and chlorogenic acids content). First, the effects of elevation and shade tree
canopy cover was assessed. Second, relationships with soil chemical variables and soil temperature
were tested. Third, the link between green bean biochemistry and quality was examined.

Linear mixed-effect models (LMMs) were fitted for all analyses and response variables. First, elevation
and shade tree canopy cover were used as fixed effect predictors and coffee farm as a random effect
term. These models allowed us to explicitly model our structured and nested data, containing clusters
of non-independent observational units that are hierarchical in nature (farms in this case). As coffee
trees were clustered in different farms, the variation across these sampling farms is assumed to be
random and uncorrelated with the predictor variables. Therefore, coffee farm was included as a random
factor. Both full and reduced linear mixed models (LMM) were performed in which the full model
consisted of elevation, canopy cover, and elevation-by-canopy cover interactions as fixed effects in
linear mixed-effect models. We also included management intensity as a categorical fixed effect. All
the data were presented to the tree level (n = 159). The models were then fitted using maximum-
likelihood methods in the ‘lme4’ packages using the ‘lmer’ function (Harrison et al. 2018). The p-values
of the fixed effects (elevation, canopy cover, and their interaction) and the overall model significance
were estimated based on the denominator degrees of freedom calculated with the Satterthwaite
approximation, in the ‘lmerTest’ package (Bates et al. 2018). Moreover, model assumptions were
checked after fitting the models. For all models, the distributions of error terms (residuals) were
approximately normal having a constant variance, with zero means, indicating adequate model fit
(Nakagawa and Schielzeth 2013). To test the explanatory power of several different predictor variables
for the variation in response variables, the coefficient of determination (R2) was quantified using the
‘[Link]’ function in the package ‘MuMIn’ (Barton and Barton 2015). Accordingly, both
marginal and conditional R2 values were determined to describe the proportion of variance explained
by the fixed effects alone as well as the fixed and random factors together, respectively (Nakagawa and
Schielzeth 2013).

Finally, we tested the effects of the environmental predictors (soil chemical characteristics and soil
temperature). First Principal Component Analyses (PCA) were conducted to reveal groupings and
relationships between the main environmental predictor variables for soil temperatures and soil
chemical variables separately. To select prominent variables for subsequent regression analyses, the
first two principal components from soil temperature variables were taken (Supporting Information Fig.
S1). Likewise, the first two principal components for soil chemical variables were taken (Supporting
Information Fig. S2). Then, the score values obtained from the above considered principal components
were utilized as independent variables in LMM to identify the main driving variables determining bean

9
quality attributes and biochemistry using backward variable selection procedures. For the analysis of
principal components, the function ‘prcomp’ was used from the packages ‘factoextra’ and ‘stats’.

The R function ‘fviz’ was used to create a ggplot2-based visualization in the biplot analysis. The R fu
nction ‘ggplot’ was used from the packages ‘ggplot2’ and ‘broom’ to produce plotting. To combine m
ultiple graphics, the package gridExtra was utilized. The R version 3.6.1 (R Development Core Team
2019) was used for all analyses.

3. Results and Discussion

The total preliminary quality was used to classify the coffee samples into different quality grades.
According to ECX (2011), dry-processed coffees can be categorized as follows based on total
preliminary quality: 91-100 (grade 1), 81-90 (grade 2) and 71-80 (grade 3) whereas the specialty coffee
achieving scores between 85-100 are classified as specialty 1 (Q1) and 80-84 is specialty 2 (Q2). In our
present study, the minimum total preliminary quality was 71, and the maximum was 91, whereas the
minimum specialty quality was 80.5 and the maximum score was 90.5. Hence, all our samples fell under
grade 2 and 3, and Q1 and Q2 (Table 1).

Table 1. Fixed effects, coffee bean physical attributes, green bean biochemistry and total preliminary and specialty
quality.

Environmental Soil chemistry Response variables Descriptive statistics for 159

variable coffee trees
1. Bean physical attributes Min Mean Max
Elevation (m) pH (H2O) - Hundred bean mass (g) 9.4 15.2 19.9
Canopy cover Soil C (%) - Bean length (mm) 8.0 9.1 10.4
(%)
Soil N (%) - Bean diameter (mm) 5.8 6.3 6.7
Soil Soil K (mg kg-1) 2. Bean biochemistry
temperature
(mean,
min and max) Soil Mg (mg kg-1) - Caffeine (g kg-1 dry mass) 0.5 1.2 2.3
in 0C
Soil Ca (mg kg-1) - Trigonelline (g kg-1 dry mass) 0.9 1.5 2.7
Olsen-P (mg kg-1) - Chlorogenic acids (g kg-1 dry 3.8 7.4 14.1
mass)
3. Cup quality attributes
3.1. Preliminary attributes
- Primary defects (score 0-15) 12.0 15.0 15.0
- Secondary defects (score 0-15) 1.0 6.0 15.0
- Odour (score 0-10) 10.0 10.0 10.0
Raw total 23.0 30.9 40.0
- Acidity (score 0-15) 9.0 12.2 15.0
- Body (score 0-15) 9.0 12.0 13.0
- Flavour (score 0-15) 9.0 10.2 15.0
- Cup cleanness (score 0-15) 15.0 15.0 15.0
Cup total 41.0 49.3 58.0

10
 Total preliminary quality 71.0 80.3 91.0
3.2. Specialty attributes (score 0-10)
- Aroma 7.3 7.8 8.5
- Flavour 7.0 7.8 8.8
- Aftertaste 7.0 7.8 8.8
- Acidity 7.0 7.9 8.8
- Body 7.3 7.8 8.5
- Cup cleanness 10.0 10.0 10.0
- Overall 7.0 7.8 8.8
- Sweetness 10.0 10.0 10.0
- Balance 7.0 7.8 8.8
- Uniformity 10.0 10.0 10.0
Total specialty quality 80.5 84.6 90.5

Relationship between elevation, shade canopy cover and coffee bean quality and biochemistry

In order to link coffee bean quality and biochemistry to elevation and shade canopy cover, the associated
soil temperature and chemical variables were considered separately as variables explaining changes in
hundred bean mass, total preliminary and specialty quality. A clear relationship of hundred bean mass
with elevation and shade canopy cover was observed, confirming that a greater hundred bean mass was
produced in response to the increasing elevation at different shade levels. Hundred bean mass was
increased under light shade (10-35% canopy cover) and intermediate shade levels (35-65% canopy
cover) with elevation. The interaction explain 28% of the variation (Table 2). However, under dense
shade (65-100% level), the hundred bean mass did not change with elevation (Table 2 and Fig 2a).
Similarly, the interaction effect of elevation and canopy cover significantly influenced the total
preliminary quality. An increased total preliminary quality was obtained at higher elevations under
dense shade conditions while the total preliminary quality did not change with elevation in more open
coffee forests (Table 2 and Fig 2b). Yet, only 3.6% of the variations in the total preliminary quality was
explained by the model. The conditional R2 (40.2%) indeed indicates a considerable spatial random
variability for total preliminary quality.

On the other hand, neither the interactive effects of elevation and canopy cover nor the canopy cover
alone had relationship with total specialty quality, caffeine, trigonelline and total chlorogenic acids.
Positive relationships were observed between elevation and total specialty quality (explaining 3.8% of
the variance), caffeine (explaining 6.1% of the variance), trigonelline (explaining 9.5% of the variance)
and total chlorogenic acids (explaining 7.5% of the variance) and this variability is probably due to the
climatic variables associated with elevation (Table 2 and Fig 2c and Fig 3). The low marginal R² values
and relatively high conditional R² values again imply a strong random “farm” effect in our data.

A study conducted in southwest Ethiopia demonstrated that coffee trees grown under open conditions
at 1780 m elevation produced lower quality beans, particularly with respect to acidity, body, and flavor
compared to coffee from plants grown under dense shade (Bote 2016). However, this study was

11
conducted under an artificially developed shade net. A reduced shade level at lower elevations was
found to have a negative effect on total preliminary quality. Avelino et al. (2007) found a positive
relationship between increasing elevations (lower temperature) and quality for Arabica coffee cultivated
under intermediate shade levels in Central America. A study in Uganda (Sarmiento-Soler et al. 2020)
revealed that higher temperatures and vapor pressure deficit (increased transpiration) at lower elevations
led to sub-optimum growing conditions for coffee. Studies conducted in southwest Ethiopia indicated
that coffee beans grown at mid-elevation with less shaded conditions accumulated higher content of
total chlorogenic acid. Based on their reports, the highest caffeine content was obtained from mid-
elevation at dense shade (Tolessa et al. 2017).

12
Table 2: Overview of linear mixed effect model testing of elevation, canopy cover and management intensity on
hundred bean mass, total preliminary and specialty quality and biochemical composition. The p-values are two-
tailed at 0.05 from the linear mixed-effect model, Marg. R2 is the proportion of variance explained by the fixed
effects only (i.e. elevation, canopy cover and management) whereas Cond. R2 is the proportion of the variance
explained by both fixed and random effects (the whole model), both Marg and Cond R2 values were given only
for the full models. Management intensity refers to the extent of farm management among the coffee farms.

Response Effect Effect F (p-value) Explained variance (%)

estimate
Marg. R2 Cond. R2
Hundred bean mass Elevation 0.01 31.4 (<0.001)
Canopy cover 0.03 14.3 (<0.001)
Elevation x Canopy cover -0.006 15.9 (<0.001) 27.9 30.2
Management Intensity -0.068 0.04 (0.847)
Intercept -11.4

Total preliminary quality Elevation 0.006 1.4 (0.185)

Canopy cover -0.03 5.4 (0.014)
Elevation x Canopy cover 0.002 4.9 (0.019) 3.6 40.2
Management Intensity 1.262 1.18 (0.282)
Intercept 93.58

Total specialty quality Elevation 0.004 3.6 (0.014)

Canopy cover 0.05 0.6 (0.409)
Elevation x Canopy cover -0.005 0.8 (0.342) 3.8 25.7
Management Intensity -0.16 0.21 (0.646)
Intercept 78.21

Caffeine Elevation 0.001 4.5 (0.044)

Canopy cover 0.02 1.6 (0.216)
Elevation x Canopy cover -0.003 1.8 (0.177) 6.1 67.6
Management Intensity 0.039 0.138 (0.71)
Intercept -0.566

Trigonelline Elevation 0.001 3.62 (0.023)

Canopy cover 0.02 0.34 (0.238)
Elevation x Canopy cover -0.006 0.51 (0.479) 9.5 68.5
Management Intensity -0.24 0.04 (0.857)
Intercept -0.51

Chlorogenic acids Elevation 0.003 1.3 (0.024)

Canopy cover -0.002 0.001 (0.323)
Elevation x Canopy cover -0.002 0.005 (0.944) 7.5 61.6
Management Intensity -0.096 0.032 (0.860)
Intercept 1.944
On the other hand, the lower bean mass at higher elevation for dense shade cover might be due to cooler
temperatures (14.8°C – 16.7°C) and low incoming radiation, resulting in reduced coffee berry growth
(although these are influenced by many other climatic variables like light intensity, light quality,
sunshine hours, cloud cover, mist, etc.). Hence, a balanced shade canopy cover provides a means to
keep coffee plants closer to their optimum temperatures ranges (18°C – 21°C) and prevent damage from
extreme minimum and maximum temperatures (Lin 2006).

13
Fig 2. The relationship between elevation and shade canopy cover with (a) hundred bean mass, (b) total
preliminary quality, and (c) total specialty quality. Data points represent a particular response variable at a single
coffee tree (n=159) in which the fitted regression lines and 95% confidence intervals are from linear mixed-effect
models at p<0.05.

Fig 3. The relationship between elevation and (a) caffeine (b) trigonelline, and (c) chlorogenic acids content of
green coffee beans. Data points represent a particular response variable at a single coffee tree (n=159) in which
the fitted regression lines and 95% confidence intervals are from linear mixed-effect models at p<0.05.

14
Relationship between soil temperatures, chemical variables and bean quality and biochemistry

In the biplot where temperatures and soil chemical variables were condensed into PCA axes, the length
of the vector indicates the variance explained by the variable. The angle between the vectors lines
corresponds to the degree of association between the variables. The chemical soil variables that
clustered to the right (soil Ca, Mg and C, K-CEC and CEC-K) have large positive loadings, i.e, the
relationship among these variables is strong and positive. A strong positive relation was observed
between soil N and soil C/N. Likewise, a positive relationship was observed between soil pH and Olsen-
P (Appendix Fig S2). On the other hand, the soil temperature variables cluster to the right, minimum
and maximum temperatures, have large positive loadings and are strongly correlated with each other
(Appendix Fig. S1).

Soil chemical variables were found to have a significantly positive relationship with hundred bean mass
whereas a significantly negative relationship was observed with soil temperatures, in which 14.7% of
the variation in hundred bean mass could be explained by soil chemical variables and soil temperatures
together (Table 3). Similarly, the total preliminary quality was found to be correlated significantly with
soil temperatures both positively (PC1) and negatively (PC2) while it had a significant positive
relationship with soil chemical variables. 71% of the variation could be explained by soil chemical
variables and soil temperatures (Table 3 and fig 4). Total specialty quality showed a significantly
positive relationship with soil properties, in which 84% of the variation could be explained by soil
chemical variables (Table 3). Soil temperatures were found to have a strong and significant negative
relationship with caffeine, trigonelline and total CGA (Table 3 and fig 4). On the other hand, soil
chemical variables were found to have a significantly positive relationship with caffeine, trigonelline
and total CGA (Table 3 and fig 4).

As part of the tropical climatic belt, southwestern Ethiopia is characterized by a reduced seasonal
temperature variation, with elevation being the major driving factor of temperature at larger spatial
scales. At the local scale, shade tree canopy cover is the main determinant of local microclimate
temperature in this system. During the course of this study, an attempt was made to measure air
temperature but due to theft it was impossible to get the loggers in the field. However, literature relating
soil temperatures with coffee bean quality and biochemistry are rarely available and hence, our main
focus was given to measure this variable. Meanwhile, soil temperature is an important environmental
variable but it is a rarely reported indicator of climate change. Based on previous reports, soil
temperature was found to be equally important with the local air temperature under shaded environment
in explaining tree performance. The findings showed that there is a strong correlation between local air
temperature and soil temperature at different depths. The annual and seasonal mean local air and soil
temperatures mainly displayed significant trends of increase over years, although the rise of the mean
air and soil temperatures was a bit asymmetric (Leeper et al. 2021; Zhan et al. 2019). Shade tree canopy

15
cover modulates macroclimatic trends through effects on local microclimates. In this regard, dense tree
canopies not only lower ground-layer temperatures but also increase relative humidity and shade in the
understory. Higher relative humidity in dense shaded environments can also protect trees underneath
from drought. All together it can buffer the impacts of regional climate change effects (De Frenne et al.
2013; Von Arx et al. 2013). Our present study confirmed that high-quality coffee beans were mostly
produced at high elevations (cool temperatures). Temperature during fruit development period
(flowering to fruit maturity) is a crucial driving variable influencing brew quality and green bean
biochemistry. In the present study, cool climates (i.e., more elevated conditions and dense canopy
cover) had a high potential to produce coffee beans having superior total preliminary quality, higher
caffeine, total CGA contents, and trigonelline concentrations as compared to coffee growing in warmer
microclimates (low elevations with light shade).

As already demonstrated by the PCA, the total preliminary quality is strongly and significantly
correlated with temperature (in terms of effect size and beta-values), where 72.3% of the variations are
explained by temperature and soil chemical variables. This implies that the share of farm variability
other than soil temperature and soil chemical variables tend to be relatively small as demonstrated from
the marginal and conditional R2 values (Table 3 and Fig 4). The results further support that coffee cup
quality attributes are sensitive to soil temperature changes and soil chemical properties. However, there
are many climatic variables that were not measured which might have effects on cup quality. The study
of Bertrand et al. (2012) demonstrated that, among the climatic factors, mean temperature during coffee
seed development greatly influenced the sensory profile and cup quality attributes such as acidity, fruity
character and flavor were obtained from the cool climates.

Coffee does not tolerate a wide range of temperatures in that mean temperatures below 16°C and above
23°C are suboptimal and the optimum temperature for Arabica coffee growth ranges between 18°C to
21°C (Camargo 2010; Kirkpatrick 1936). In our present study, the minimum, mean and maximum soil
temperatures during the coffee fruit development period (flowering to fruit maturity) ranges between
9.5°C – 17.5°C, 14.5°C – 21.0°C, and 17.5°C – 36.5°C, respectively, about 1.5°C cooler than the lower
limit (16°C) for coffee Arabica. In our network of coffee plots, we did not detect an apparent effect of
management intensity on the studied coffee quality attributes.

16
Fig 4. Scatterplots of (a) caffeine, (b) trigonelline, (c) total preliminary quality and (d) hundred bean mass as a
function of soil nutrients (PCnutrient1) and soil temperatures (PCtemp1) principal components. Data points
represent a particular response variable at a single coffee tree (n=159) in which the fitted regression lines and 95%
confidence intervals are from linear mixed-effect models at p<0.05. The principal component biplots are included
in supplementary information.

On the other hand, a hundred bean mass showed both positive and negative relationships with
temperature and only 14.3% of the variations are explained by both temperature and soil chemical
variables (Table 3). This implies that temperature could have both positive and negative associations
with hundred bean mass and the magnitude of this relationships on coffee quality attributes depend
more on other farm variables other than what temperature and soil chemical variables explained. Our
findings tend to show that dense shading conditions reduced hundred bean mass at very high elevations.
In contrast, dense shading at warmer (lower elevations) increased bean mass (Fig 2).

As indicated in Vaast et al. (2006) and Laderach et al. (2017), the faster fruit development due to
elevated temperature results in malformed beans with poorer cup quality, which is the result of the
consequence of excessive demand for resources from the seed endosperm in a compressed timeframe.
According to a report of Ramalho et al. (2018), prevalence of high temperatures in the field conditions
often coincides with decreased water availability and high atmospheric vapor pressure deficit, resulting
in lower coffee bean quality because high VPD causes water limited conditions which in turn can
worsen the stress effects on coffee plants by increasing the transpiration rate or by reducing carbon
uptake (Avila et al. 2020, Xu et al. 2016; Belko et al. 2012). The findings of Silva et al. (2005), showed

17
that temperature was likely the most important factor to bring variations in cup quality and bean
biochemistry of coffee from the southwest region of Ethiopia.

Green bean biochemical compositions (caffeine, trigonelline and total CGA) were negatively correlated
with temperature, and around 45% of the variation in the data was explained by temperature and soil
chemical variables. However, the share of random farm variability other than temperature and soil
chemical variables was still large (roughly 20% of the variation in the data) (Table 3). The results further
support that green bean biochemical compositions are very sensitive to temperature changes and soil
chemical variables. Caffeine, trigonelline and total CGA were found to be favored by shading (Farah et
al. 2006; Somporn et al. 2012), whereas other studies obtained contradictory findings, for instance, total
CGA was favored by open sun (Tolessa et al. 2017; Somporn et al. 2012; Vaast et al. 2006). Some other
studies reported a negligible correlation between total CGA content and shading (Avelino et al. 2007).
The effect of shade canopy cover and elevation on soil temperatures and chemistry was reported in an
earlier study (Getachew et al. 2022). We then found that the interactive effect of elevation and shade
tree canopy cover had a significant and strong effect on mean soil temperatures. On the other hand,
Olsen-P decreased with increasing elevation. Contrarily, soil C/N increased with increasing shade
canopy cover (Getachew et al. 2022). Although we here do not find a direct link between shade canopy
cover and bean biochemistry, shading had a significant indirect influence on bean biochemistry through
modification of the local microclimatic conditions and soil chemical variables (Fig 3, Table 3). This
confirms our expectations that increased light and warming below a more open canopy can negatively
influence the synthesis of various bean quality precursors (such as caffeine, trigonelline and total CGA)
by accelerating berry maturation and disfavoring bean-filling and brew quality (Bertrand et al. 2012,
Joet et al. 2010, Geromel et al. 2008, Vaast et al. 2006).

Caffeine in green beans co-determines flavor and the quality of the final coffee beverage. Caffeine is
associated with the bitterness and flavor of the coffee. The caffeine content in the green coffee beans
showed a great variability with values ranging from 0.5 to 2.3 g per 100 g. Trigonelline contents
obtained in this study showed a large variability too with values ranging from 0.9 to 2.7 g per 100 g.
Derivatives of trigonelline are known to be an important precursor of the volatile compounds that
contribute to the aroma and taste of roasted coffee (Cheng et al. 2016; Joet et al. 2010; Malta and Chagas
2009; De Castro and Marraccini 2006). Indeed, most studies reported a decreasing quality with
decreasing trigonelline levels. Green coffee beans of arabica genotypes possesses chlorogenic acid
contents that range between 3.5 – 7.5 g per 100 g (Gloess et al. 2013) while the total CGA content
obtained in our present study ranged from 3.8 to 14 g 100 g. Total CGA contents is related with the
astringency and bitterness in coffee brews (Gloess et al. 2013; Farah et al. 2006) and the content showed
an increasing trend with increasing brew quality. Although most literatures have often associated high
levels of CGA with lower brew quality (Farah et al. 2006; De Castro and Marraccini 2006), suggesting

18
that it is possible to obtain coffee brews with various levels of total CGA. Barbosa et al. (2019); Corso
et al. (2016); Dos Santos Scholz et al. (2016); Zanin et al. (2016) and Kitzberger et al. (2013) also
reported that good cup quality roasted coffee beans could also show wide variability in their CGA
contents. Possible explanations for this difference could include the genetic differences of the Arabica
genotypes, coffee tree age variations; time and temperature of bean roasting, and brew preparation
processes.

19
1 Table 3. Effects of soil temperature and chemical variables on hundred bean mass, total preliminary and specialty quality and biochemistry of green coffee beans. Marg. R2 is
2 the proportion of variance explained by the four PC’s used as predictors; whereas Cond. R2 is the proportion of the variance explained by both the PC’s used as predictors and
3 the random effect (full model), both Marg. and Cond. R2 values were given only for the full models. PCnutrient1 = the 1st principal component of soil chemical variables;
4 PCnutrient2 = the 2nd principal component of soil chemical variables; PCtemp1= the 1 st principal component of soil temperatures; PCtemp2= the 2 nd principal component of
5 soil temperatures.
6
Response Predictors Effect Std. Standardized regression P-value Explained variance (%)
estimate Err coefficient
Marg. R2 Cond. R2
Hundred bean mass (g) PCtemp1 4.02 1.19 3.35 0.0012 14.7 28.7
PCtemp2 -0.41 0.14 -2.89 0.005
PCnutrient1 4.29 1.36 3.95 < 0.001
Intercept 14.22

Total preliminary quality PCnutrient2 0.321 0.112 3.323 0.020 71.1 85.2
PCtemp1 -1.048 0.068 -15.322 < 0.001
PCtemp2 1.709 0.129 13.175 < 0.001
Intercept 84.26

Total specialty quality PCnutrient1 0.52 0.008 70.52 < 0.001 83.5 88.5
PCnutrient2 0.29 0.02 13.25 < 0.001
Intercept 88.65

Caffeine (g kg-1) PCtemp1 -1.30 0.19 -6.88 < 0.001 45.8 68.5
PCtemp2 -0.19 0.02 -8.47 < 0.001
PCnutrient1 2.13 0.26 7.74 < 0.001
PCnutrient2 0.09 0.01 7.05 < 0.001
Intercept 1.29

Trigonelline (g kg-1) PCtemp1 -1.67 0.26 -6.39 < 0.001 42.7 68.23
PCtemp2 -0.21 0.03 -7.01 < 0.001
PCnutrient1 1.69 0.26 6.31 < 0.001
PCnutrient2 0.15 0.02 6.67 < 0.001
Intercept 1.54

18
 Chlorogenic acids (g kg-1) PCtemp1 -7.12 1.07 -6.66 < 0.001 47.2 65.7
PCtemp2 -0.91 0.12 -7.33 < 0.001
PCnutrient1 7.01 1.07 6.53 < 0.001
PCnutrient2 0.65 0.08 8.49 < 0.001
Intercept 4.52
7
8 The present study showed that soil chemical properties are positively associated with hundred bean mass, green bean quality, and biochemical composition.
9 This implies that fertile soils produce heavier, high quality beans with higher levels of caffeine, trigonelline and chlorogenic acids. Interestingly, selected soil
10 chemical properties are strongly correlated with the total specialty quality (Table 3 and Fig 4). The measured soil chemical variables explained 84% of the
11 variations in total specialty quality. This suggests that soil chemical fertility is decisive for improving total specialty quality. Hence, we can maximize specialty
12 1 (Q1) by growing coffee at higher elevations, while sustaining soil fertility. According to Yadessa et al. (2020), soil fertility parameters considerably affected
13 the cup quality of wild Arabica coffee in its natural habitat. The study indicated that coffee with improved cup quality was collected from coffee farms having
14 greater available P, K, Mg, and Zn levels. The compounds are considered important for the brew quality and also N and K certainly played a significant role in
15 the final bean quality (Clemente et al. 2015).
16
17 Low temperature at high elevated areas retarded root and microbial activity and thus less phosphatase production compared with lower elevated sites. Besides,
18 water-soluble phosphorus decreased with the decreased soil temperature due to the reduced movement of phosphorus in the soil controlled by diffusion (Onwuka
19 and Mang 2018). As both the roots and microorganisms synthesize and release phosphatases to soil, soil available P might be reduced at higher elevations.
20 However, higher levels of available P content in lower elevations had no contribution in improving the brew quality and green bean biochemistry, suggesting
21 that the indirect relationships due to the natural soil P variation and the possible effects of P-fixation in the soil might render soil P less available for optimal
22 coffee plant growth and fruit quality and composition (Rekik et al. 2019).
23
24
25
26
27

19
28

29 The novelty of this study lies in the investigation of soil temperatures on coffee quality and bean
30 biochemical compositions in that intermediate to deep shade moderates responses to the increasing
31 temperatures at medium and higher elevations and improved coffee bean quality. Hence, we can conclude
32 that coffee beans from cool climates (higher elevations) would have a greater bean quality and better bean
33 biochemical compositions which in turn will have a better commercial value. The main weakness of this
34 particular study is the missing data particularly on essential climatic variables such as ambient air
35 temperature, relative humidity, vapor pressure deficit, soil moisture, solar radiation, and sunshine hours.
36
37 Conclusion and implication for management
38
39 We found that the highest cup scores (better quality coffee) are produced at higher elevations with the
40 highest concentrations of caffeine, chlorogenic acids and trigonelline. It seems that coffee quality from
41 farms at lower elevations (mainly below 1600 m) will suffer most from changing climatic conditions for
42 the areas under study (southwest Ethiopia). Besides, fertile soils produce heavier beans, beans possessing
43 good brew quality and green bean biochemistry. Hence, we can conclude that coffee beans from cool
44 climates and good fertile soils would have a better commercial value. Spatial random variability at the farm
45 level is an important consideration in our studied coffee bean attributes. Hence, future studies need to
46 consider coffee tree genetic differences, coffee tree age and essential climatic variables such as ambient air
47 temperature, relative humidity, vapor pressure deficit, soil moisture, solar radiation, and sunshine hours for
48 a better understanding of brew quality and green bean biochemistry. In addition, shade tree species and the
49 genetic variation of the coffee trees should not be forgotten.
50
51 Acknowledgements
52
53 This study has been supported by the Belgian Development Cooperation (VLIRUOS-NASCERE program)
54 and Ethiopian Ministry of Science and Higher Education (MoSHE). The authors are profoundly grateful
55 for the Ethiopian government for this support. P.D.F. received funding from the European Research Council
56 (ERC) under the European Union’s Horizon 2020 research and innovation programme (ERC Starting Grant
57 FORMICA 757833). We would like to thank coffee owners for allowing us to work in their coffee plots
58 and the local and regional administration for providing the permits to work in the coffee farms. We also
59 thank Beyene Zewdie for establishing the coffee plots. Lastly, we are very grateful to the ECX for their
60 support in evaluating coffee cup quality and Ethiopian Institute of Agricultural research for analyzing our
61 green bean biochemical compositions.
62
63

20
 64 Code availability: Not applicable.

65 Authors’ contribution: Conceptualization, M.G., P.D.F., K.V. and P.B.; methodology, M.G., K.T., K.V.,
66 K.H., A.T., B.A. and P.B.; formal analysis, M.G.; investigation, M.G. and K.T., writing—original draft,
67 M.G. and P.D.F.; writing—review and editing, P.B., K.H., A.T., and K.V.; visualization, M.G., and B.A.;
68 Funding Acquisition, P.B.; Supervision, K.V., P.B. and P.D.F.
69
70 Funding: The study was funded by the joint collaboration of European Research Council (ERC) under the
71 European Union’s Horizon 2020 research and innovation programme and the Ethiopian Ministry of Science
72 and Higher Education.
73

74 Data availability: The datasets analyzed during the current study are available from the corresponding
75 author on reasonable request. Moreover, the data will be made available after publication via an online
76 repository such as figshare.

77 Declarations

78 Conflict of interest: The authors declare no potential conflict of interest.

79 Ethics approval: Not applicable.

80 Consent to participate: Not applicable.

81 Consent for publication: Not applicable.

82
83
84 References
85
86 Avelino J, Barboza B, Davrieux F and Guyot B (2007) Shade effects on sensory and chemical characteristics
87 of coffee from very high altitude plantations in Costa Rica. In Second International Symposium on Multi-
88 strata Agroforestry Systems with Perennial Crops. September 17-21. Turrialba, Costa Rica. Oral and poster
89 presentations. Turrialba: CATIE
90
91 Avila RT, Cardoso AA, De Almeida WL, Costa LC, Machado KL, Barbosa ML, De Souza RP, Oliveira LA,
92 Batista DS, Martins SC and Ramalho JD (2020) Coffee plants respond to drought and elevated [CO2]
93 through changes in stomatal function, plant hydraulic conductance, and aquaporin expression. Environ Exp
94 Bot:177:104148. [Link] envexpbot.2020.104148
95
96 Barbosa JN, Borem FM, Cirillo MA, Malta MR, Alvarenga AA and Alves HMR (2012) Coffee quality and its
97 interactions with environmental factors in Minas Gerais, Brazil. J Agr Sci. 4(5):181.
98 [Link]
99
100 Barbosa MSG, Dos Santos Scholz MB, Kitzberger CSG and De Toledo Benassi M (2019) Correlation between
101 the composition of green Arabica coffee beans and the sensory quality of coffee brews. Food Chem.
102 292:275-280. [Link]
103

21
104 Barman D, Kundu DK, Pal S, Chakraborty AK, Jha AK, Mazumdar SP, Saha R and Bhattacharyya P (2017)
105 Soil temperature prediction from air temperature for alluvial soils in lower Indo-Gangetic plain. Int
106 Agrophys 31(1):9-22. [Link]
107
108 Barton K and Barton MK (2015) Package ‘MuMIn’. Version. 1.18.
109 [Link]
110
111 Bates D, Maechler M, Bolker B, Walker S, Christensen RHB, Singmann H, Dai B Scheipl F, Grothendieck G
112 and Green P (2018) Package ‘lme4’. Version. 1.17. [Link]
113 13/web/packages/lme4/[Link]
114
115 Belko N, Zaman AM, Cisse N, Diop NN, Zombre G, Ehlers JD and Vadez V (2012) Lower soil moisture
116 threshold for transpiration decline under water deficit correlates with lower canopy conductance and higher
117 transpiration efficiency in drought-tolerant cowpea. Funct Plant Biology, 39(4):306-322.
118 [Link]
119
120 Bertrand B, Boulanger R, Dussert S, Ribeyre F, Berthiot L, Descroix F and Joet T (2012) Climatic factors
121 directly impact the volatile organic compound fingerprint in green Arabica coffee bean as well as coffee
122 beverage quality. Food Chem. 135(4):2575-2583. [Link]
123
124 Blaser WJ, Oppong J, Yeboah E and Six J (2017) Shade trees have limited benefits for soil fertility in cocoa
125 agroforests. Agr Ecosys Environ. 243:83-91. [Link] .04.007
126
127 Bote A (2016) Examining growth, yield and bean quality of Ethiopian coffee trees: towards optimizing
128 resources and tree management. Doctoral dissertation, Wageningen University
129
130 Bunn C, Laderach P, Rivera OO and Kirschke D (2015) A bitter cup: climate change profile of global
131 production of Arabica and Robusta coffee. Climatic Change. 129:89-101. [Link]
132 014-1306-x
133
134 Camargo MBD (2010) The impact of climatic variability and climate change on Arabic coffee crop in Brazil.
135 Bragantia. 69(1):239-247. [Link]
136
137 Cerda R, Allinne C, Gary C, Tixier P, Harvey CA, Krolczyk L, Mathiot C, Clement E, Aubertot JN and Avelino
138 J (2017) Effects of shade, altitude and management on multiple ecosystem services in coffee
139 agroecosystems. Eur J Agron. 82:308-319. [Link] 09.019
140
141 Chemura A, Mudereri BT, Yalew AW and Gornott C (2021) Climate change and specialty coffee potential in
142 Ethiopia. Sci Rep-UK, 11(1): 1-13. [Link]
143
144 Cheng B, Furtado A, Smyth HE and Henry RJ (2016) Influence of genotype and environment on coffee quality.
145 Trends Food Sci Tech. 57:20-30. [Link]
146
147 Clemente JM, Martinez HEP, Alves LC, Finger FL and Cecon PR (2015) Effects of nitrogen and potassium
148 on the chemical composition of coffee beans and on beverage quality. Acta Sci-Agron. 37(3):297-305.
149 [Link]
150

22
151 Corso MP, Vignoli JA and De Toledo Benassi M (2016) Development of an instant coffee enriched with
152 chlorogenic acids. J food Sci Tech. 53:1380-1388. [Link]
153
154 Craparo ACW, Van Asten PJ, Laderach P, Jassogne LT and Grab SW (2015) Arabica coffee yields decline in
155 Tanzania due to climate change: Global implications. Agr Forest Meteorol. 207:1-10.
156 [Link]
157
158 DaMatta FM, Avila RT, Cardoso AA, Martins SC and Ramalho JC (2018) Physiological and agronomic
159 performance of the coffee crop in the context of climate change and global warming: A review. J Agr food
160 Chem. 65(2):364-372. [Link] 7b04537
161
162 DaMatta FM, Rahn E, Laderach P, Ghini R and Ramalho JC (2019) Why could the coffee crop endure climate
163 change and global warming to a greater extent than previously estimated? Climatic Change. 152:167-178.
164 [Link]
165
166 Davis AP, Gole TW, Baena S and Moat J (2012) The impact of climate change on indigenous Arabica coffee
167 (Coffea arabica L.): predicting future trends and identifying priorities. Plos One. 93(5):992-1001.
168 [Link]
169
170 De Castro RD and Marraccini P (2006) Cytology, biochemistry and molecular changes during coffee fruit
171 development. J Plant Physiol. 18(1):175-199. [Link]
172
173 De Frenne P, Rodriguez-Sanchez F, Coomes DA, Baeten L, Verstraeten G, Vellend M, Bernhardt-Römermann
174 M, Brown CD, Brunet J, Cornelis J and Decocq GM (2013) Microclimate moderates plant responses to
175 macroclimate warming. Proceedings of the National Academy of Sciences, 110(46):18561-18565.
176 [Link]
177
178 De Frenne P, Zellweger F, Rodriguez-Sanchez F, Scheffers BR, Hylander K, Luoto M and Lenoir J (2019)
179 Global buffering of temperatures under forest canopies. Nature Eco Evol. 3:744–749.
180 [Link]
181
182 Dos Santos Scholz MB, Kitzberger CSG, Pagiatto NF, Pereira LFP, Davrieux F, Pot D, Charmetant P. and
183 Leroy T (2016) Chemical composition in wild Ethiopian Arabica coffee accessions. Euphytica. 209:429-
184 438. DOI 10.1007/s10681-016-1653-y
185
186 ECX (Ethiopian Commodity Exchange) (2011) ECX quality operation manual, Addis
187 Ababa, Ethiopia
188
189 Elias EA, Cichota R, Torriani HH and De Jong Van Lier Q (2004) Analytical soil–temperature model:
190 Correction for temporal variation of daily amplitude. Soil Sci Soc Am J, 68(3):784-788.
191 [Link]
192
193 Farah A, Monteiro MC, Calado V, Franca AS and Trugo LC (2006) Correlation between cup quality and
194 chemical attributes of Brazilian coffee. Food Chem, 98(2):373-380.
195 [Link]
196

23
197 Frazer GW, Canham CD and Lertzman KP (1999) Gap Light Analyzer (GLA), Version 2.0: Imaging software
198 to extract canopy structure and gap light transmission indices from true-color fisheye photographs, user’s
199 manual and program documentation. Simon Fraser University, Burnaby, British Columbia, and the Institute
200 of Ecosystem Studies, Millbrook, New York, [Link]://[Link]/viewdoc/download?
201 doi=[Link].1223&rep=rep1&type=pdf
202
203 Getachew M, Verheyen K, Tolessa K, Ayalew B, Hylander K, Tack A, Garedew W, Bauters M, Boeckx P and
204 De Frenne P (2022) Shade tree canopy cover affects coffee plant traits across elevations in coffee farms in
205 southwest Ethiopia. Nord J Bot, 2022(2):e03383. [Link]
206
207 Geromel C, Ferreira LP, Davrieux F, Guyot B, Ribeyre F, Dos Santos Scholz MB, Pereira LFP, Vaast P, Pot
208 D, Leroy T and Androcioli Filho A (2008) Effects of shade on the development and sugar metabolism of
209 coffee (Coffea arabica L.) fruits. Plant Physiol Bioch. 46(5):569-579.
210 [Link]
211
212 Gloess AN, Schönbächler B, Klopprogge B, Lucio D, Chatelain K, Bongartz A, Strittmatter A, Rast M and
213 Yeretzian C (2013) Comparison of nine common coffee extraction methods: instrumental and sensory
214 analysis. Eur Food Res Technol. 236:607-627. [Link] s00217-013-1917-x
215
216 Harrison XA, Donaldson L, Correa-Cano ME, Evans J, Fisher DN, Goodwin CE, Robinson BS, Hodgson DJ
217 and Inger R (2018) A brief introduction to mixed effects modeling and multi-model inference in ecology.
218 Peer J. 479. [Link]
219
220 Hundera K, Aerts R, Fontaine A, Van Mechelen M, Gijbels P, Honnay O and Muys B (2013) Effects of coffee
221 management intensity on composition, structure, and regeneration status of Ethiopian moist evergreen
222 afromontane forests. Environ Manage. 51:801-809. [Link] 10.1007/s00267-012-9976-5
223
224 Jarcuska, B (2008) Methodological overview to hemispherical photography, demonstrated on an example of
225 the software GLA. Folia Oecologica, 35:66-69. [Link]
226
227 Joet T, Laffargue A, Descroix F, Doulbeau S, Bertrand B and Dussert S (2010) Influence of environmental
228 factors, wet processing and their interactions on the biochemical composition of green Arabica coffee
229 beans. Food Chem. 118(3):693-701. [Link]
230
231 Jonckheere I, Fleck S, Nackaerts K, Muys B, Coppin P, Weiss M and Baret F (2004) Review of methods for
232 in situ leaf area index determination: Theories, sensors and hemispherical photography. Agr Forest
233 Meteorol. 121(1):19-35. [Link]
234
235 Lajtha K, Jarrell WM, Johnson DW and Sollins P (1999) Standard soil methods for long-term ecological
236 research, Oxford University, England, pp.166-182
237
238 Kirkpatrick TW (1936) The ecology of coffee plantations: climate conditions in East African countries. East
239 Afr Agric J. 1(6):476–486. [Link]
240
241 Kitzberger CSG, Dos Santos Scholz MB, Pereira LFP, Vieira LGE, Sera T, Silva JGD and De Toledo Benassi
242 M (2013) Diterpenes in green and roasted coffee of Coffea arabica cultivars growing in the same edapho-
243 climatic conditions. J Food Compos Anal. 30(1):52-57. [Link] 01.007

24
244
245 Kufa T, Ayano A, Yilma A, Kumela T and Tefera W (2011) The contribution of coffee research for coffee
246 seed development in Ethiopia. J Agr Res Dev. 1(1):009-016.
247
248 Laderach P, Ramirez–Villegas J, Navarro-Racines C, Zelaya C, Martinez–Valle A and Jarvis A (2017) Climate
249 change adaptation of coffee production in space and time. Climatic Change. 141:47-62.
250 [Link]
251
252 Lajtha K, Jarrell WM, Johnson DW and Sollins P (1999) Standard soil methods for long-term ecological
253 research, Oxford University, England, pp.166-182
254
255 Leeper RD, Matthews JL, Cesarini MS and Bell JE (2021) Evaluation of air and soil temperatures for
256 determining the onset of growing season. J Geophys Res-Biogeo. 126(8):e2020JG006171.
257 [Link]
258
259 Legesse A (2020) Assessment of coffee (Coffea arabica L.) genetic erosion and genetic resources management
260 in Ethiopia. Int J Agr Ext. 7(3):223-229. [Link] 10.33687/ijae.007.03.2972
261
262 Leonel LE and Philippe V (2007) Effects of altitude, shade, yield and fertilization on coffee quality (Coffea
263 arabica L.) produced in agroforestry systems of the Northern Central Zones of Nicaragua. J Food Sci. 2356-
264 2361.
265
266 Lin G, Zheng M, Bai Z and Zhao X (2006) Effect of temperature and pressure on the morphology of carbon
267 dioxide corrosion scales. Corrosion. 62(06): 06060501
268
269 Magrach A and Ghazoul J (2015) Climate and pest-driven geographic shifts in global coffee production:
270 implications for forest cover, biodiversity and carbon storage. Plos One. 10(7):e0133071.
271 [Link] [Link].0133071
272
273 Malta MR and Chagas SDR (2009) Evaluation of non-volatile compounds in different cultivars of coffee
274 cultivated in southern Minas Gerais. Acta Scientiarum. Agron. 31(1):57-61. [Link]
275 4025/actasciagron.v31i1.6629
276
277 Meyfroidt P, Vu TP and Hoang VA (2013) Trajectories of deforestation, coffee expansion and displacement
278 of shifting cultivation in the Central Highlands of Vietnam. Global Environ Chang. 23(5):1187-1198.
279 [Link]
280
281 Moat J, Williams J, Baena S, Wilkinson T, Gole TW, Challa, ZK, Demissew S and Davis AP (2017) Resilience
282 potential of the Ethiopian coffee sector under climate change. Nat plants. 3(17081):1-14.
283 [Link]
284
285 Muschler RG (2001) Shade improves coffee quality in a sub-optimal coffee-zone of Costa Rica. Agroforest
286 Syst. 51:131–139. [Link]
287
288 Nakagawa S and Schielzeth H (2013) A general and simple method for obtaining R2 from generalized linear
289 mixed‐effects models. Methods Ecol Evol. 4(2):133-142. [Link] 2041-
290 210x.2012.00261.x

25
291
292 Navarro-Serrano F, López-Moreno JI, Azorin-Molina C, Alonso-González E, Aznarez-Balta M, Buisán ST
293 and Revuelto J (2020) Elevation Effects on Air Temperature in a Topographically Complex Mountain
294 Valley in the Spanish Pyrenees. Atmosphere-Basel, 11(6).:656. [Link] 11060656
295
296 Onwuka B and Mang B (2018) Effects of soil temperature on some soil properties and plant growth. Adv
297 Plants Agric Res. 8(1):34. [Link]
298
299 Ovalle-Rivera O, Laderach P, Bunn C, Obersteiner M and Schroth G (2015) Projected shifts in Coffea arabica
300 suitability among major global producing regions due to climate change. Plos One. 10(4):e0124155.
301 [Link]
302
303 R Core Team (2019) R: A language and environment for statistical computing. R Foundation for Statistical
304 Computing, Vienna, Austria. [Link]
305
306 Rahn E, Laderach P, Baca M, Cressy C, Schroth G, Malin D, Van Rikxoort H and Shriver J (2014) Climate
307 change adaptation, mitigation and livelihood benefits in coffee production: where are the synergies?. Mitig
308 Adapt Strat GL. 19:1119-1137. [Link] 1007/s11027-013-9467-x
309
310 Ramalho JC, Pais IP, Leitao AE, Guerra M, Reboredo FH, Máguas CM, Carvalho ML, Scotti-Campos P,
311 Ribeiro-Barros AI, Lidon FJ and DaMatta FM (2018) Can elevated air [CO2] conditions mitigate the
312 predicted warming impact on the quality of coffee bean?. Front Plant Sci, 9:287.
313 [Link]
314
315 Rekik F, Van Es H, Hernandez-Aguilera JN and Gomez MI (2019) Linking Coffee to Soil: Can Soil Health
316 Increase Coffee Cup Quality in Colombia?. Soil Sci. 184(1):25-33. [Link]
317 00000248
318
319 Robertson GP, Coleman DC, Sollins P and Bledsoe CS eds (1999) Standard soil methods for long-term
320 ecological research. Oxford University, England
321
322 Ruslim Y, Yusuf S, Lahjie AM and Lisnawati A (2017) Agroforestry system biodiversity of Arabica coffee
323 cultivation in North Toraja district, South Sulawesi, Indonesia. [Link] handle/
324 123456789/1881
325
326 Sarmiento-Soler A, Vaast P, Hoffmann MP, Jassogne L, Van Asten P, Graefe S and Rotter RP (2020) Effect
327 of cropping system, shade cover and altitudinal gradient on coffee yield components at Mt. Elgon, Uganda.
328 Agr Ecosyst Environ. 295:106887. [Link] 106887
329
330 Schleppi P, Conedera M, Sedivy I and Thimonier A (2007) Correcting non-linearity and slope effects in the
331 estimation of the leaf area index of forests from hemispherical photographs. Agr Forest Meteorol.
332 144(3):236-242. [Link]
333
334 Silva EAD, Mazzafera P, Brunini O, Sakai E, Arruda FB, Mattoso LHC, Carvalho CR and Pires RCM (2005)
335 The influence of water management and environmental conditions on the chemical composition and
336 beverage quality of coffee beans. Plant physiol. 17(2):229-238. [Link]
337 04202005000200006

26
338
339 Somporn C, Kamtuo A, Theerakulpisut P and Siriamornpun S (2012) Effect of shading on yield, sugar content,
340 phenolic acids and antioxidant property of coffee beans (Coffea arabica L.) harvested from north‐eastern
341 Thailand. J Sci Food Agr. 92(9):1956-1963. [Link] 1002/jsfa.5568
342
343 Teklu B, Mohammed A and Kufa T (2011) Effect of processing methods and drying materials on bean physical
344 and sensorial quality attributes of coffee (Coffea arabica L.) Doctoral dissertation, Jimma University
345
346 Tolessa K, Dheer J, Duchateau L and Boeckx P (2017) Influence of growing altitude, shade and harvest period
347 on quality and biochemical composition of Ethiopian specialty coffee. J Sci Food Agr. 97(9):2849-2857.
348 [Link]
349
350 Tolessa K, Duchateau L and Boeckx P (2018) Analysis of coffee quality along the coffee value chain in Jimma
351 zone, Ethiopia. Afr J Agr Res. 13(29):1468-1475. [Link] AJAR2018.13118
352
353 Vaast P, Bertrand B, Perriot JJ, Guyot B and Genard, M (2006) Fruit thinning and shade improve bean
354 characteristics and beverage quality of coffee (Coffea arabica L.) under optimal conditions. J Sci Food Agr.
355 86(2):197-204. [Link]
356
357 Vignoli JA, Bassoli DG and Benassi MDT, 2011. Antioxidant activity, polyphenols, caffeine and melanoidins
358 in soluble coffee: The influence of processing conditions and raw material. Food Chem. 124(3):863-868.
359 [Link]
360
361 Von A, G Graf Pannatier E, Thimonier A and Rebetez M (2013) Microclimate in forests with varying leaf area
362 index and soil moisture: potential implications for seedling establishment in a changing climate. J Ecol,
363 101(5):1201-1213. [Link]
364
365 Worku M, De Meulenaer B, Duchateau L and Boeckx P (2018) Effect of altitude on biochemical composition
366 and quality of green arabica coffee beans can be affected by shade and post-harvest processing method.
367 Food Res Int. 105:278-285. [Link] 2017.11.016
368
369 Xu Z, Jiang Y, Jia B and Zhou G (2016) Elevated CO2 response of stomata and its dependence on
370 environmental factors. Front Plant Sci. 7:657. [Link]
371
372 Yadessa A, Burkhardt J, Bekele E, Hundera K and Goldbach H (2020) Influence of soil properties on bean
373 quality of wild coffee (Coffea arabica L.) in the natural coffee forests of southwest and southeast Ethiopia.
374 Ethiopian J App Sci Tech. 11(2):23-38.
375
376 Zanin RC, Corso MP, Kitzberger CSG, Dos Santos Scholz MB and De Toledo Benassi M (2016) Good cup
377 quality roasted coffees show wide variation in chlorogenic acids content. LWT-Food Sci Technol. 74:480-
378 483. [Link]
379
380 Zewdie B, Tack AJ, Adugna G, Nemomissa S and Hylander K (2020) Patterns and drivers of fungal disease
381 communities on Arabica coffee along a management gradient. Basic Appl Ecol. 47:95-106.
382 [Link]
383

27
384 Zhan MJ, Xia L, Zhan L and Wang Y (2019) Recognition of changes in air and soil temperatures at a station
385 typical of China’s subtropical monsoon region (1961–2018). Adv Meteorol. 2019:1-9.
386 [Link]

28