ArticlePDF Available

Multiclass Classification of Agro-Ecological Zones for Arabica Coffee: An Improved Understanding of the Impacts of Climate Change

Authors:

Abstract and Figures

Cultivation of Coffea arabica is highly sensitive to and has been shown to be negatively impacted by progressive climatic changes. Previous research contributed little to support forward-looking adaptation. Agro-ecological zoning is a common tool to identify homolo-gous environments and prioritize research. We demonstrate here a pragmatic approach to describe spatial changes in agro-climatic zones suitable for coffee under current and future climates. We defined agro-ecological zones suitable to produce arabica coffee by clustering geo-referenced coffee occurrence locations based on bio-climatic variables. We used random forest classification of climate data layers to model the spatial distribution of these agro-ecological zones. We used these zones to identify spatially explicit impact scenarios and to choose locations for the long-term evaluation of adaptation measures as climate changes. We found that in zones currently classified as hot and dry, climate change will impact arabica more than those that are better suited to it. Research in these zones should therefore focus on expanding arabica's environmental limits. Zones that currently have climates better suited for arabica will migrate upwards by about 500m in elevation. In these zones the up-slope migration will be gradual, but will likely have negative ecosystem impacts. Additionally, we identified locations that with high probability will not change their climatic characteristics and are suitable to evaluate C. arabica germplasm in the face of climate change. These locations should be used to investigate long term adaptation strategies to production systems.
Content may be subject to copyright.
RESEARCH ARTICLE
Multiclass Classification of Agro-Ecological
Zones for Arabica Coffee: An Improved
Understanding of the Impacts of Climate
Change
Christian Bunn
1
*, Peter Läderach
2
, Juan Guillermo Pérez Jimenez
1
,
Christophe Montagnon
3
, Timothy Schilling
3
1International Center for Tropical Agriculture (CIAT), Km 17, Recta Cali-Palmira, Apartado Aéreo, 6713,
Cali, Colombia, 2International Center for Tropical Agriculture (CIAT), Hotel Seminole, 2 Cuadras al Sur,
Managua, Nicaragua, 3World Coffee Research, 578 John Kimbrough Blvd, College Station, Texas, 77843
2477, United States of America
*Christian.bunn@gmail.com
Abstract
Cultivation of Coffea arabica is highly sensitive to and has been shown to be negatively
impacted by progressive climatic changes. Previous research contributed little to support
forward-looking adaptation. Agro-ecological zoning is a common tool to identify homolo-
gous environments and prioritize research. We demonstrate here a pragmatic approach to
describe spatial changes in agro-climatic zones suitable for coffee under current and future
climates. We defined agro-ecological zones suitable to produce arabica coffee by clustering
geo-referenced coffee occurrence locations based on bio-climatic variables. We used ran-
dom forest classification of climate data layers to model the spatial distribution of these
agro-ecological zones. We used these zones to identify spatially explicit impact scenarios
and to choose locations for the long-term evaluation of adaptation measures as climate
changes. We found that in zones currently classified as hot and dry, climate change will
impact arabica more than those that are better suited to it. Research in these zones should
therefore focus on expanding arabica's environmental limits. Zones that currently have cli-
mates better suited for arabica will migrate upwards by about 500m in elevation. In these
zones the up-slope migration will be gradual, but will likely have negative ecosystem
impacts. Additionally, we identified locations that with high probability will not change their
climatic characteristics and are suitable to evaluate C.arabica germplasm in the face of cli-
mate change. These locations should be used to investigate long term adaptation strategies
to production systems.
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 1/16
OPEN ACCESS
Citation: Bunn C, Läderach P, Pérez Jimenez JG,
Montagnon C, Schilling T (2015) Multiclass
Classification of Agro-Ecological Zones for Arabica
Coffee: An Improved Understanding of the Impacts of
Climate Change. PLoS ONE 10(10): e0140490.
doi:10.1371/journal.pone.0140490
Editor: Juan A. Añel, Universidade de Vigo, SPAIN
Received: June 8, 2015
Accepted: September 25, 2015
Published: October 27, 2015
Copyright: © 2015 Bunn et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any
medium, provided the original author and source are
credited.
Data Availability Statement: According to CIAT's
institutional policy, all raw data are available on
dataverse: http://thedata.harvard.edu/dvn/dv/CIAT.
Funding: World Coffee Research funded this
research through the project "Identifying Long Term
Variety Trial Locations, Provide Climate Information to
Support World Coffee Research Variety Trials and
Support on Trial Data Analysis." http://
worldcoffeeresearch.org/who-we-are/contact-us.
World Coffee Research was represented by CM and
TS and continuously interacted with CB and PL to
design, implement and evaluate the research. CM
and TS from World Coffee Research supported the
Introduction
Most of the world's coffee comes from the perennial tree Coffea arabica [1], plantations of
which are productive for 2050 years. Climate controls where coffee can be grown [2]. Arabica
coffee requires a climate with annual mean temperatures of about 20°C and over 1200 mm
annual rainfall to be economically viable [2]. Temperatures over 30°C for extended periods
reduce yields[3], while frost for a few days damages or even kills the plant [4]. A short dry
period of less than 40mm precipitation per month increases yield and promotes uniform flow-
ering, but more than 3 dry months reduces yield [5].
The livelihoods of about 25 million small producers globally depend on arabica coffee [6].
Because arabica requires a specific climate within narrow limits, these growers will see both
yield and quality fall as the climate changes. Rising temperatures were predicted to reduce
yields below economic viability in Veracruz, Mexico by 2020 [7]. For Chiapas in Mexico pro-
jections suggest that rising temperatures will eliminate coffee below 1100 masl by 2050 [8]. In
the center of origin of C. arabica, the Ethiopian highlands, the climatic niche of the indigenous
varieties may disappear by 2080 [9]. Brazil, which produces a third of the global crop of arabica
coffee, is projected to lose up to 95% of the suitable area by 2100 [10]. A study of the global
impact of climate change predicted that the area suitable for arabica coffee will be reduced by
50% by 2050, mostly caused by higher temperatures [11]. A more recent study showed that
these impacts vary by region [12].
It is a challenge to develop unambiguous strategies to adapt to the projected changes in cli-
mate. For well researched crops like wheat, rice or maize, mechanistic simulation models are
available to estimate the effects of changed climate on crop performance [13]. They can also be
used to forecast how agronomic management might be adapted to changed conditions. The
model available for coffee (Caf2007), however, could not be applied on larger spatial scales due
to limited spatial data [14], limited knowledge of the crops physiology [15], and its specifica-
tion for plot scale application [16].
Alternatively, the Veracruz [7] and Brazil [10] studies extrapolated known climatic limits of
arabica coffee in their respective regions to forecast the distribution of future climates suitable
for arabica coffee. While useful, the results could not be readily extrapolated globally. Other
studies used larger data sets and methods of machine learning to predict the spatial distribution
of coffee species [11,12,17]. As applied to coffee, these analyses relied on whether a sites future
climate lay within the range of those that determine current distribution. The models estimated
the future distribution of arabica coffee with high confidence, but they only considered a binary
distinction between suitable and unsuitable climate. This allowed the reliable identification of
locations that will likely transition to other crops in the future. However, to guide adaptation
research a better identification of the climatic characteristics of the impacted regions will be
necessary. Furthermore, in regions that remain suitable guidance is needed to distinguish
zones that will require systemic or incremental adaptation measures [18].
To address these issues, we chose an agro-ecological zoning (AEZ) approach. AEZs for arab-
ica coffee were defined in Brazil using overlay maps of limiting climatic factors [19], or by clus-
ter analysis of several variables in Colombia [20]. In viticulture future changes in wine growing
zones have been projected using the Random Forest (RF) algorithm on local [21] and continen-
tal scale [22]. We extended this application to global scale to analyze how the different climates
of each coffee AEZ will be affected by climate change. We used this analysis to suggest options
to adapt and to identify homologous sites to facilitate technology transfer.
A recent review concluded that a globally-coordinated breeding program was needed to
confront the negative impacts of climate change[23]. We therefore show how the AEZ
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 2/16
manuscript preparation with comments. World Coffee
Research is a 501 (c)(5) non-profit, collaborative
research and development program of the global
coffee industry to grow, protect, and enhance
supplies of quality coffee while improving the
livelihoods of the families who produce it. The
program is funded and driven by the global coffee
industry, guided by producers, executed by coffee
scientists around the world and supported by the
Norman Borlaug Institute, part of Texas A&M
University.
Competing Interests: The authors have declared
that no competing interests exist.
approach might be used to select sites for multi-location variety trials that such a program
would require.
Data and Methods
We used WorldClims bioclimatic variables [24] to define AEZs suitable for arabica coffee.
Although soil attributes, aspect, and local microclimate determine crop performance at local
scales, they are unimportant in defining the global distribution of AEZs. We assembled a data-
base of geo-references of sites where C.arabica is currently grown throughout the world. On
the climate data for these sites we then used cluster analysis to define the AEZs that are suitable
to grow arabica coffee. Next we trained the RF algorithm on the AEZ definition as response
variable and the climate variables as independent variables. These RF models were extrapolated
on maps of both current and future climates to predict the changes that each AEZs will con-
front with climate change. As a demonstration of how the method might be used, we identified
possible sites for an international multi-location variety trial (IMLVT).
Database of locations of arabica coffee
We used data of the current distribution of arabica coffee to define the climates that are suitable
for cultivation. The data came from four sources:
1. Geo-referenced coffee farms from a database of the location of 100,000 farms developed by
the International Center for Tropical Agriculture (CIAT) and its collaborators [11];
2. Geo-referenced municipalities in Brazil that produce arabica coffee [25]; Similarly, we pro-
duced a set of 5,666 locations for Indonesia by sampling within polygons where we knew
arabica is grown and stratification based on region-specific ranges of altitude [17]
3. For those regions where neither (1) nor (2) were available, we identified coffee plantations
from Google Earth images [11]; and
4. The Global Biodiversity Information Facility [26].
The raw database contained a total of 124,820 geo-referenced locations growing C.arabica.
Because coffee farms are often small, to avoid spatial bias we reduced the database to unique
pixel cells on a 5 arc-minute grid, which we call occurrence pixels. We stratified the database
to contain only locations with elevations above 100 masl. We also removed as outliers sites for
which one or more environmental variable exceeded 3.5 standard deviations from the mean.
Fig 1 shows the distribution of the 3545 occurrence pixels in the final dataset together with the
distribution of arabica area harvested.
Climate data
For the current climate (19502000), we used the WorldClim data set at 5 arc-minute resolu-
tion [24]. WorldClim provides data of monthly precipitation, mean monthly minimum and
maximum temperatures, and 19 bioclimatic variables derived from these data. We comple-
mented the latter with another derived variable of consecutive months with less than 40mm
precipitation (Table 1). As we point out in the introduction, a short dry season increases yields
but dry periods longer than three months reduce yields or require irrigation.
To predict climate in the period 2040 to 2069 (2050s), we used 19 global climate models
(GCMs) from the Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate
Change (IPCC) [28]. We chose the representative concentration pathway (RCP) 6.0, an inter-
mediate scenario in which radiative forcing continues to increase until the end of the century
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 3/16
[29].We downscaled the GCM outputs using the delta method [30], which computes the differ-
ence between model outputs for current conditions and the 2050s. We interpolated these data
Fig 1. Distribution of occurrence pixels (5 arc minutes, red dots). The brown colors indicate producer groups by area of arabica coffee harvested in each
country [27].
doi:10.1371/journal.pone.0140490.g001
Table 1. Bioclimatic variables used and mean values at occurrence pixels under current and 2050s conditions; the values for 2050s were calcu-
lated as mean impact across 19 GCMs.
Type Bioclimatic
variable
Description Current
mean
2050s
mean
Unit Clustering
1
BIO 1 Annual mean temperature 20.6 22.5 °C
BIO 2 Mean diurnal range (mean of monthly (max tempmin
temp))
11.6 11.7 °C
BIO 3 Isothermality (BIO2/BIO7) (*100) 72 71 - X
BIO 4 Temperature seasonality (standard deviation *100) 136.9 143.9 °C
BIO 5 Max temperature of warmest month 28.5 30.5 °C X
Temperature BIO 6 Min temperature of coldest month 12.2 13.8 °C X
BIO 7 Temperature annual range (BIO5-BIO6) 16.3 16.7 °C
BIO 8 Mean temperature of wettest quarter 21.6 23.4 °C
BIO 9 Mean temperature of driest quarter 19.3 21.2 °C
BIO 10 Mean temperature of warmest quarter 22.1 24.0 °C
BIO 11 Mean temperature of coldest quarter 18.7 20.4 °C
BIO 12 Annual precipitation 1637 1645 mm X
BIO 13 Precipitation of wettest month 280 289 mm
BIO 14 Precipitation of driest month 33 31 mm
Precipitation BIO 15 Precipitation seasonality (coefcient of variation) 66 67 -
BIO 16 Precipitation of wettest quarter 739 749 mm
BIO 17 Precipitation of driest quarter 122 121 mm
BIO 18 Precipitation of warmest quarter 492 479 mm X
BIO 19 Precipitation of coldest quarter 211 224 mm
BIO 20 Number of consecutive months <40mm precipitation 2.5 2.6 - X
1
X = variables used for agglomerative clustering.
doi:10.1371/journal.pone.0140490.t001
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 4/16
to 5 arc-minutes resolution and applied them to the WorldClim data for current climate and
recalculated the bioclimatic variables (Table 1).
AEZs for arabica coffee
We transformed all 20 bioclimatic variables to z-scores. For those variables that were highly
correlated (Pearson coefficients |r|>0.7), we kept the one that we judged most informative in
the coffee context and discarded the others.
We then performed an agglomerative cluster analysis on the occurrence pixels using the
Ward algorithm in the statistics software R [31]. We determined the final number of clusters
using the indices of Ratkowsky and Lance [32] and Calinski and Harabasz [33] and by judging
the dendrogram of distances between clusters. From the clusters we described AEZs classified
by the bioclimatic variables that define each of them. We tested statistical significance of differ-
ences of climate data between the AEZs by one-way analysis of variance in R [31]. We based
the AEZ descriptions on the differences of the group means from the grand mean and calcu-
lated their confidence intervals using Rs multcomp package [34].
Current and future spatial distribution of AEZs for arabica coffee
We used the Random Forest package [35] to classify the climate in each pixel into AEZs for
arabica coffee. We then examined the spatial distribution of each of the defined AEZs to assess
the climate of global coffee growing regions, and to evaluate how climate change will affect
them.
The Random Forest package creates an ensemble of decision trees and selects the mode of
the individual trees, which reduces the risk of generating an overconfident classification (over
fitting) [35]. We trained the algorithm with random samples of occurrence pixels within the
AEZs and a random background sample of pixels within coffee-producing countries that did
not have coffee. From the occurrence pixels in each AEZ group, we selected samples the same
size of the smallest AEZ group and used 2.5 times as many background samples. For binary
classification problems a 1:1 sampling ratio is recommended to avoid the preferential predic-
tion of the majority class [36]. The sampling ratio we chose accounted for the trade-off between
the multi-class AEZ classification, as well as the binary classification in to suitable classes and
unsuitable background locations. Additionally, we constrained the background samples to con-
tain only pixels with annual mean temperatures within the range as the occurrence pixels to
exclude unfeasible locations [36]. We used all 20 bioclimatic variables as independent variables
(Table 1). For each RF model we grew 1000 decision trees with seven variables selected at each
node and replicated this process three times.
To make the process more robust, we divided the training sample into five random groups
and trained the package five times, withholding one of the groups each time. In summary, we
drew three random samples from the entire population of occurrence pixels, from each of
which we drew five random subsamples to give 15 individual models trained. We extrapolated
these onto maps of the 20 bioclimatic variables and determined the modal value across all 15
model results. We obtained maps of each AEZ plus class 0background sites where Arabica
coffee is unlikely to be cultivated. We repeated the process for the 2050 data. The most likely
future AEZ was determined for each pixel by the mode across the results for the 19 GCMs.
To evaluate the classification we used the area under receiver operating characteristic curve
(AUC), which has values 01. An AUC of 0.5 indicates that the performance was no better
than random sampling, while 1.0 is perfect classification. We used the standard AUC to evalu-
ate the capacity of individual models to correctly discriminate occurrence pixels from the back-
ground sample. The definition of the AUC measure can be extended to multiclass problems by
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 5/16
averaging all pairwise AUC comparisons to a multiclass AUC [37]. We used this measure to
evaluate the discrimination of AEZs by the models.
Example: Identify potential trial sites
We classified pixels to be appropriate sites for variety evaluation in the long-term if the charac-
teristics of the climate were stable over time and the classification was unambiguous. We there-
fore defined three conditions:
1. All repeats of the RF classification step had to agree on the AEZ classification under current
conditions;
2. All modal classifications across the 19 GCMs had to agree, and
3. The classifications of (1) and (2) had to be identical in both time steps.
We produced additional maps that indicated sites with less stringent classification if there
was 80% agreement with the conditions above.
Results
AEZs for arabica coffee production
Among the temperature factors the annual mean temperature and the annual total precipita-
tion are most frequently mentioned in the literature to describe the spatial distribution of cli-
matic suitability for coffee production (e.g. [5]). Furthermore, maximum and minimum
temperatures were shown to be influential, but also temperature variability [15].
At the 3545 occurrence pixels of C.arabica Bio5 (the mean maximum temperature of the
warmest month) and Bio6 (the mean minimum temperature of the coldest month) were corre-
lated with Pearson coefficients |r| = 0.42. This was below the threshold (|r|>0.7), so we
included them for clustering. In contrast, Bio1 (the annual mean temperature) correlated with
several other temperature variables so we excluded all of them. Of the variables that represent
temperature variability, Bio3 correlated least with Bio5 and Bio6. It is the ratio of the mean
monthly temperature range to the annual range (Bio2/Bio7100) (|r|0.4), and we included it
in the analysis (Table 1).
The literature identifies annual precipitation (Bio12) and the length of the dry season
(Bio20) as the most important precipitation-related variables that influence the yield of coffee.
Their Pearson coefficient was |r| = 0.56 so we included them in the analysis. In general, coeffi-
cients among precipitation variables were high and we excluded them. The exception was vari-
able Bio18, the precipitation of the warmest quarter, which had acceptable coefficients with
both Bio12 (|r
BIO12
| = 0.45) and Bio20 (|r
BIO20
| = 0.21) and low coefficients with most other
precipitation variables; we therefore included it in the analysis (Table 1).
We obtained five distinct agro-ecological zones (AEZs, Fig 2) for arabica coffee within the
3545 occurrence pixels based on six standardized bioclimatic variables. We described the AEZs
in terms of their climatic characteristics. For some better insight into of Bio3 we included the
related variables Bio2 (Mean diurnal range) and Bio7 (Annual temperature range).
One way analysis of variance showed highly significant differences between all AEZs for all
variables (at p <0.001). We summarized the five AEZs as follows (the list numbers correspond
to the groups in Fig 2):
1. Hot-wet(HW), characterized high maximum temperature in the warmest month, high
annual precipitation, a short dry season and a humid warmest quarter of the year (Fig 3).
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 6/16
2. Constant(Ct), characterized by lacking temperature seasonality. It had the highest iso-
thermality, and the lowest mean values for daily and annual temperature range. Cts precipi-
tation is similar to the HW zone with high annual precipitation and a short dry season (Fig
3).
3. Hot-dry(HD), characterized by high maximum temperatures were high and no cold
month. The annual total precipitation is low and has the lowest precipitation in the warmest
quarter of all groups and a long dry season (Fig 3).
The last two groups of occurrences are characterized by low minimum temperatures.
5. Cool-variable(CV) characterized by the highest annual temperature range, and the lowest
mean isothermality. Precipitation is moderate.
6. Cool-dry(CD), characterized by the lowest minimum temperature of the coldest month
but also the lowest annual precipitation with a long dry season.
Current and future spatial distribution of clusters
The RF classification gave maps of current distribution of the AEZs (Fig 4) and the changes
that climate change will bring as forecast by the 19 GCMs (Fig 5).
The Ct AEZ is mostly in highlands close to the equator: Colombia, Ethiopia, Kivu lake
region in Central Africa, Kenya and Indonesia (Fig 4B, 4D and 4E). Regions towards the latitu-
dinal margins were characterized by more variability. Southern Brazil was mostly dominated
by the related AEZs HD, CV and CD. The latter two AEZs show a high seasonal temperature
Fig 2. Dendrogram of agglomerative clustering. On the vertical axis is Euclidean distance. The final definition of clusters is indicated by the color codes:
Green-Hot-Wet, Violet-Constant, Red-Hot-Dry, Yellow-Cool-Variable, Turquoise–“Cool-Dry.
doi:10.1371/journal.pone.0140490.g002
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 7/16
variation and low minimum temperatures. While in southern Brazil the CV AEZ has a long
dry season, CD AEZ has lower maximum temperatures and a shorter dry season. In central
Brazil, the HD AEZ region has high maximum temperatures and a strong dry season (Fig 4A).
In the Central America region all AEZs were represented but HW and HD with high maximum
temperatures are most prevalent. Some regions experience high rainfall, however, while others
are characterized by lower precipitation and longer dry seasons. Southern Mexico was typical
for the HW AEZ with a hot and wet climate. Nicaragua was representative of the HD AEZ with
a long dry season and high temperatures. Costa Rica is an exception in this region with most of
its coffee region being of Ct climate with stable and moist conditions (Fig 4C).
By 2050, the spatial distribution of AEZs changes little, with Ct still mostly around the equa-
tor. The extent of the distribution will be reduced, however, especially in Brazils coffee growing
regions, with most of the HD AEZ pixels becoming unsuitable. The CD AEZ will also be
reduced substantially, especially in western São Paulo state. The areas that are currently in the
CV AEZ zone largely remain of the same climate. There will not be any migration to the south
(Fig 5A). Changes in Central America will be similar to those in Brazil with most of the HD
AEZ mostly becoming unsuitable. The location of the other AEZs persists but reduced extent
in extent (Fig 5C).
Under current conditions, arabica coffee grows on 7.2% of the total pixels in the latitudinal
belt 30°N33°S. The Ct AEZ accounts for 26% of all suitable pixels, which is a little more than
the HD AEZ with 25%. The other three AEZs shared the remainder 49%. The total share of
suitable pixels in the 30°N33°S belt will be halved to 3.6% by the 2050s. The Ct AEZ will be
Fig 3. Confidence intervals of group compared with the grand mean for descriptor variables. Group labels are the cluster numbers (C 15) in the
dendrogram (Fig 2). The dashed line is the grand mean, the value for which is given at the bottom.
doi:10.1371/journal.pone.0140490.g003
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 8/16
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 9/16
less affected by climate change than the others and will make up for 34% of the area that
remains suitable for arabica coffee. Most of the loss will be in the HD AEZ, while the other
three AEZs will be reduced proportionally so that their relative share changes little (Table 2).
The median elevation of all suitable pixels was 1024 masl but the median elevations differed
between AEZs. The Ct AEZ lay at 1575 masl compared with the CD AEZ at 700 masl. By 2050,
the median elevation of pixels that remain suitable for arabica coffee will be more than 300m
higher. The effect will differ by AEZ; the elevation of the CV AEZ will not change, while the
HW AEZ will be nearly 500m higher (Table 2).
We disaggregated the AEZs for now and 2050 to determine in which AEZ each pixel will be
classified in 2050 (Fig 6). The Ct AEZ was least affected with 59% of the pixels unchanged, 4%
becoming each HW and HD, and 34% becoming unsuitable. The HD AEZ will be most
affected, with only 16% of current pixels remaining suitable by 2050s and 78% becoming
unsuitable. Of the other three AEZs about 40% of current grid cells remain suitable and 46
49% will become unsuitable. The remainder shifted to the other AEZs (Fig 6).
Only half of the pixels that will be in the HD AEZ in 2050 currently belong to this AEZ,
with the remainder coming from other zones. In contrast, nearly all future area the Ct AEZ cur-
rently belong to it. Only a small part of the area that will be suitable for arabica coffee in 2050
will be land that was previously unsuitable. Half of this novel area will be in the CV AEZ (Fig
6), such as in the southern margin of Brazilian coffee zones (Fig 4A).
The average ability of the RF algorithm to discriminate AEZs was satisfactorily high across
all individual models, with the multiclass AUC averaging 0.84, which is much better than
chance. The conventional AUC measure averaged 0.91, which demonstrates the robustness of
the algorithm to discriminate between suitable and unsuitable pixel cells.
Example application: Spatial distribution of robust sites
We overlaid the maps in Fig 4 with pixels that met the 80% and 100% stability criteria defined
above in the Data and methods section.
We used these to make a list of recommended sites for trials for each country that is an
important arabica producer. For each site within an AEZ we listed geographic location, admin-
istrative description, altitude and values of bioclimatic variables were given (Table 3).
We could not identify robust sites for all countries that are important arabica producers.
Nor could we identify potential trial sites for all the AEZs that occur in each country, for exam-
ple in Nicaragua the only robust sites were in the HD AEZ. Although a part of Nicaraguas cof-
fee is produced in the HW AEZ, none of these pixels were classified as robust.
Discussion
The agro-ecological zoning model developed here adds to our understanding of the climate
change impacts on the production of arabica coffee globally. The global AEZ approach we used
goes beyond previous regional AEZ research by demonstrating how different AEZs will be
affected differently, which can be used to guide research into adaptation to climate change. The
global impacts we project here agree with previous studies on the magnitude of impacts, which
is a reduction of area suitable for coffee production by about 50% until the 2050s [9,11,12].
Most areas that will become unsuitable to grow arabica coffee in the future now have cli-
mates with high maximum temperatures and long dry seasons (AEZ HD). These include some
Fig 4. Distribution of AEZs in regions important for arabica coffee; A Brazil; B East Africa; C Central America; D Indonesia; E Colombia; F India;
Colored grid cells represent the agro-ecological zone; dots indicate sites recommended for trial sites, hatching alternative locations with less
model agreement.
doi:10.1371/journal.pone.0140490.g004
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 10 / 16
Fig 5. Change of agro-ecological zone in important arabica coffee production regions until the 2050s; A Brazil; B East Africa; C Central America; D
Indonesia; E Colombia; F India
doi:10.1371/journal.pone.0140490.g005
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 11 / 16
areas that currently give high yields of arabica coffee (northern Minas Gerais state in Brazil,
parts of India, and Nicaragua). In contrast, substantial areas that currently lie in other AEZs
will become HD in the future. These scenarios will offer both challenges to and opportunities
for the coffee sector. On the one hand, important coffee-producing areas will struggle to remain
productive while currently less-favored areas may become more productive. Research to adapt
coffee production to climate change will thus have to make arabica coffee better adapted to
heat and drought stress. Other regions may have to change their agronomic practices to remain
competitive, for example by learning from farmers who are currently productive in the HD
AEZ.
The constant (Ct) AEZ, which has neither high nor cold temperatures, will be least affected
by climate change. It occurs close to the equator in Colombia, Ethiopia, Kenya and Indonesia
and produces high quality coffee. Despite the comparatively small effects of climate change on
Table 2. Distribution of grid cells in the agro-ecological zones under current and 2050s conditions.
Climate Unit Hot-wet Constant Hot-dry Cool-variable Cool-dry Total
Current Pixel count - 8943 14869 14337 11637 6479 56265
Pixel share % 16 26 25 21 12 7.2
1
Median elevation masl 946 1578 807 825 704 1024
2050 Pixel count - 4992 9710 4248 6944 2859 28753
Pixel share % 17 34 15 24 10 3.6
1
Median elevation masl 1429 1954 1185 812 835 1362
1
Percent share of all grid cells within a latitudinal belt 30°N33°S
doi:10.1371/journal.pone.0140490.t002
Fig 6. Transition plot of the fate of suitable pixels in coffee AEZs from current conditions to 2050s; size of boxes and width of transition arrows is
representative of the number of pixels; the size of the box of the unsuitablecategory includes only pixels that become suitable by 2050 or were
suitable with current climate.
doi:10.1371/journal.pone.0140490.g006
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 12 / 16
this AEZ, challenges arise due to decreased coffee quality [38]. Climate change will bring few
novel pixels in the Ct AEZ so that its increase in median elevation was caused by it losing low-
elevation pixels. Due to the conical form of land with elevation, there is more agricultural area
at lower elevations than there is upslope. The actual area lost to production may therefore be
larger than the number of pixels lost suggests. Moreover, land at high elevations is often diffi-
cult to access, is too steep for cultivation, or has shallow soil so that upward migration may not
always be an option. Furthermore, land at higher elevations often has high ecosystem value or
is used for forestry, which could be further barriers to upward migration of coffee.
The Hot-wet (HW) AEZ has high precipitation similar to the Ct AEZ. Temperatures are
higher, but seasonal variation is average. Typical locations are southern Mexico and central
Ethiopia. The HW AEZ occurs associated with both Ct and HD AEZs, sharing high maximum
temperatures with the latter. The strong effect of climate change on elevation in the HW and
Ct AEZs suggests that higher temperatures will be the most limiting factor. High temperatures
induce dehiscence of flowers and fruits [4] and make attacks from pests like the coffee berry
borer [39] more likely.
About half of the novel area for arabica coffee will be dispersed among existing coffee
regions at higher elevations. Substantial areas will become suitable south of the southern mar-
gin of the Brazilian coffee region. These areas have short dry seasons but temperature variabil-
ity and especially low minimum temperatures currently make them unsuitable for coffee. The
higher temperatures that climate change will bring will reduce frost risk [19]. Nevertheless,
there will be little expansion beyond the present latitudinal limits because temperature variabil-
ity will remain a limitation.
We aimed to identify homologous climatic zones in current and future conditions, which is
the basis of the AEZs. Locations that become another AEZ in the future may adapt to it guided
by how growers that are currently in that AEZ manage their crops. The AEZs that we defined,
however, are based on current climates, and we use the same criteria to classify future climates.
But climates that we consider unusual or marginal for arabica coffee today may become more
common by 2050. For example, the AEZ that dominates southern Brazil, CV, has low mini-
mum temperatures and high temperature variability. With climate change bringing reduced
frost risk, we speculate that in the future a novel AEZ with high temperature variability and
high maximum temperatures could become important.
We defined the AEZs using a database of occurrence pixels where C.arabica is grown that
included data from all the worlds coffee-producing regions. We excluded as outliers locations
with unusual climates. It might be useful to include marginal locations, however, as the envi-
ronmental limits for arabica coffee may provide insights into possible adaptation strategies.
The addition of occurrence records from marginal locations such as Zambia or Yemen may
Table 3. List of recommended trial sites for Nicaragua.
ID AEZ Lon Lat Country State District Elev bio_2 bio_3 bio_5 bio_6 bio_7 bio_12 bio_18 bio_20
° ° - - - masl 0.1°C - 0.1°C 0.1°C 0.1°C mm mm months
12 Hot-dry -85.958 13.125 Nicaragua Jinotega Jinotega 1330 98 70 252 113 139 1797 398 3
13 Hot-dry -86.458 13.208 Nicaragua Estelí Estelí 1164 105 73 274 132 142 1669 638 5
15 Hot-dry -86.125 13.292 Nicaragua Jinotega San Sebastián de Yalí 1369 101 70 262 119 143 1738 393 3
16 Hot-dry -86.042 13.292 Nicaragua Jinotega San Rafael del Norte 1032 105 73 274 131 143 1486 372 3
17 Hot-dry -86.625 13.375 Nicaragua Madriz San Lucas 1131 106 74 278 135 143 1929 777 3
19 Hot-dry -86.542 13.708 Nicaragua Nueva Segovia Macuelizo 1290 103 70 276 130 146 1864 558 2
20 Hot-dry -86.542 13.792 Nicaragua Nueva Segovia Dipilto 1370 101 69 271 126 145 1864 546 2
doi:10.1371/journal.pone.0140490.t003
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 13 / 16
improve our ability to differentiate between marginal and unsuitable climates. This could con-
tribute to our ability to adapt to more extreme climates in the future.
Machine learning approaches, like the RF algorithm used here, have been criticized to be
prone to overfit to specific variable states. We applied the algorithm carefully, choosing vari-
ables that with low levels of correlation and achieved a high classification accuracy as shown by
the AUC metric. Moreover, the overall projected impact of climate change is similar to that
projected in other studies [11]. On the other hand, climate change impacts on coffee will poten-
tially be more severe as was demonstrated by models with more pessimistic emission scenario
choices or when considering a longer time horizon [9,10]).
We specified that robust sites for the long-term variety trials must unambiguously represent
an AEZ and that the fundamental climate characteristics will be unaffected by climatic change.
We specified these conditions because variety improvement in coffee can take several decades.
Fundamental changes in the climate during the course of a long-term trial would invalidate
comparison of data gathered over many years. For the identification of robust sites for each
AEZ we took account of variation between GCMs and selected only those sites that could be
classified unambiguously. Only a small number of such pixels could be identified.
In conclusion, we therefore urge coffee research to consider climatic change carefully when
taking decisions with a long time horizon such as selecting sites for variety trials. When com-
paring data from previous experiments, analysis often considers only the environmental
parameters of interest. This approach may result in erroneous conclusions if fundamental char-
acteristics of the climate change over the time interval under consideration.
We urge that tests of strategies to improve varieties and other agronomic measures consider
the locations we identify here. Nevertheless, future research will also have to expand the envi-
ronmental limits of arabica coffee to novel or marginal climates to minimize the worst impacts
of climate change on the coffee sector.
Acknowledgments
This research was conducted under the CGIAR Research Program on Climate Change, Agri-
culture and Food Security (CCAFS). It was funded and initiated by World Coffee Research
through the project Identifying Long Term Variety Trial Locations, Provide Climate Informa-
tion to Support World Coffee Research Variety Trials and Support on Trial Data Analysis.
We thank our colleagues at the Data and Policy analysis area at CIAT for their support and
helpful comments, and Myles Fisher for his suggestions that greatly improved the original
manuscript.
Author Contributions
Conceived and designed the experiments: CB PL CM TS. Performed the experiments: CB. Ana-
lyzed the data: CB JGPJ. Contributed reagents/materials/analysis tools: PL. Wrote the paper:
CB PL CM.
References
1. USDA. PSD OnlineHome. In: Production, Supply and Distribution Online [Internet]. 2012 [cited 21
Dec 2012]. Available: http://www.fas.usda.gov/psdonline/
2. DaMatta FM, Ramalho JDC. Impacts of drought and temperature stress on coffee physiology and pro-
duction: a review. Braz J Plant Physiol. 2006; 18: 5581.
3. DaMatta FM. Exploring drought tolerance in coffee: a physiological approach with some insights for
plant breeding. Braz J Plant Physiol. 2004; 16: 16.
4. DaMatta FM. Ecophysiological constraints on the production of shaded and unshaded coffee: a review.
Field Crops Res. 2004; 86: 99114. doi: 10.1016/j.fcr.2003.09.001
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 14 / 16
5. Wintgens JN, editor. Coffee: Growing, Processing, Sustainable ProductionA guidebook for Growers,
Processors, Traders, and Researchers. 2nd ed. Weinheim: Wiley-VCH; 2009.
6. Pendergrast M. Uncommon grounds: the history of coffee and how it transformed our world. New York:
Basic Books; 2010.
7. Gay Garcia C, Estrada F, Conde C, Eakin H, Villers L. Potential Impacts of Climate Change on Agricul-
ture: A Case of Study of Coffee Production in Veracruz, Mexico. Clim Change. 2006; 79: 259288. doi:
10.1007/s10584-006-9066-x
8. Schroth G, Läderach P, Dempewolf J, Philpott S, Haggar J, Eakin H, et al. Towards a climate change
adaptation strategy for coffee communities and ecosystems in the Sierra Madre de Chiapas, Mexico.
Mitig Adapt Strateg Glob Change. 2009; 14: 605625.
9. Davis AP, Gole TW, Baena S, Moat J. The Impact of Climate Change on Indigenous Arabica Coffee
(Coffea arabica): Predicting Future Trends and Identifying Priorities. Fenton B, editor. PLoS ONE.
2012; 7: e47981. doi: 10.1371/journal.pone.0047981 PMID: 23144840
10. Assad ED, Pinto HS, Zullo J Junior, Ávila AMH. Impacto das mudanças climáticas no zoneamento
agroclimático do café no Brasil. Pesqui Agropecuária Bras. 2004; 39: 10571064. doi: 10.1590/S0100-
204X2004001100001
11. Bunn C, Läderach P, Ovalle Rivera O, Kirschke D. A bitter cup: climate change profile of global produc-
tion of Arabica and Robusta coffee. Clim Change. 2015; 129: 89101. doi: 10.1007/s10584-014-1306-x
12. Ovalle-Rivera O, Läderach P, Bunn C, Obersteiner M, Schroth G. Projected Shifts in Coffea arabica
Suitability among Major Global Producing Regions Due to Climate Change. Loyola RD, editor. PLOS
ONE. 2015; 10: e0124155. doi: 10.1371/journal.pone.0124155 PMID: 25875230
13. White JW, Hoogenboom G, Kimball BA, Wall GW. Methodologies for simulating impacts of climate
change on crop production. Field Crops Res. 2011; 124: 357368. doi: 10.1016/j.fcr.2011.07.001
14. Eriyagama N, Chemin Y, Alankara R. A methodology for quantifying global consumptive water use of
coffee for sustainable production under conditions of climate change. J Water Clim Change. 2014; 5:
128. doi: 10.2166/wcc.2013.035
15. Van Oijen M, Dauzat J, Harmand J- M, Lawson G, Vaast P. Coffee agroforestry systems in Central
America: I. A review of quantitative information on physiological and ecological processes. Agrofor
Syst. 2010; 80: 341359.
16. Van Oijen M, Dauzat J, Harmand JM, Lawson G, Vaast P. Coffee agroforestry systems in Central
America: II. Development of a simple process-based model and preliminary results. Agrofor Syst. 2010;
118.
17. Schroth G, Läderach P, Blackburn Cuero DS, Neilson J, Bunn C. Winner or loser of climate change? A
modeling study of current and future climatic suitability of Arabica coffee in Indonesia. Reg Environ
Change. 2014;
18. Vermeulen SJ, Challinor AJ, Thornton PK, Campbell BM, Eriyagama N, Vervoort JM, et al. Addressing
uncertainty in adaptation planning for agriculture. Proc Natl Acad Sci. 2013; 110: 83578362. doi: 10.
1073/pnas.1219441110 PMID: 23674681
19. Zullo J, Pinto HS, Assad ED, Ávila AMH. Potential for growing Arabica coffee in the extreme south of
Brazil in a warmer world. Clim Change. 2011;
20. García L. JC, Posada-Suárez H, Läderach P. Recommendations for the Regionalizing of Coffee Culti-
vation in Colombia: A Methodological Proposal Based on Agro-Climatic Indices. Hui D, editor. PLoS
ONE. 2014; 9: e113510. doi: 10.1371/journal.pone.0113510 PMID: 25436456
21. Gaál M, Moriondo M, Bindi M. Modelling the impact of climate change on the Hungarian wine regions
using random forest. Appl Ecol Env Res. 2012; 10: 121140.
22. Moriondo M, Jones GV, Bois B, Dibari C, Ferrise R, Trombi G, et al. Projected shifts of wine regions in
response to climate change. Clim Change. 2013; 119: 825839. doi: 10.1007/s10584-013-0739-y
23. Van der Vossen H, Bertrand B, Charrier A. Next generation variety development for sustainable pro-
duction of arabica coffee (Coffea arabica L.): a review. Euphytica. 2015; 114. doi: 10.1007/s10681-
015-1398-z
24. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A. Very high resolution interpolated climate sur-
faces for global land areas. Int J Climatol. 2005; 25: 19651978.
25. Bunn C, Läderach P. Climate Change Impacts on Arabica Coffee in Brazil. Proceedings of the 25th
International Conference on Coffee Science. Armenia, Colombia; 2014.
26. GBIF PortalHome [Internet]. [cited 6 May 2014]. Available: http://www.gbif.org/
27. Bunn C, Läderach P, De Zegher J, Kirschke D. Where on earth is coffee grown? Spatial disaggregation
of harvested area statistics using suitability data. Proceedings of the 25th International Conference on
Coffee Science. Armenia, Colombia; 2014.
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 15 / 16
28. Stocker TF, Qin D, Plattner G-K, Tignor M, Allen SK, Boschung J, et al. Climate Change 2013. The
Physical Science Basis. Working Group I Contribution to the Fifth Assessment Report of the Intergov-
ernmental Panel on Climate Change-Abstract for decision-makers. Groupe dexperts intergouverne-
mental sur levolution du climat/Intergovernmental Panel on Climate Change-IPCC, C/O World
Meteorological Organization, 7bis Avenue de la Paix, CP 2300 CH-1211 Geneva 2 (Switzerland);
2013.
29. Fujino J, Nair R, Kainuma M, Masui T, Matsuoka Y. Multi-gas mitigation analysis on stabilization sce-
narios using AIM global model. Energy J. 2006; 343354.
30. Ramirez J, Jarvis A. Disaggregation of Global Circulation Model Outputs. Int Cent Trop Agric CIAT Cali
Colomb. 2010;
31. R Core Team. R: A Language and Environment for Statistical Computing [Internet]. Vienna, Austria: R
Foundation for Statistical Computing; 2014. Available: http://www.R-project.org
32. Ratkowsky DA, Lance GN. A criterion for determining the number of groups in a classification. Aust
Comput J. 1978; 10: 115117.
33. Caliński T, Harabasz J. A dendrite method for cluster analysis. Commun Stat-Theory Methods. 1974;3:
127.
34. Bretz F, Hothorn T, Westfall P. Multiple comparisons using R. CRC Press; 2010.
35. Breiman L. Random Forests. Mach Learn. 2001; 45: 532. doi: 10.1023/A:1010933404324
36. Barbet-Massin M, Jiguet F, Albert CH, Thuiller W. Selecting pseudo-absences for species distribution
models: how, where and how many? Methods Ecol Evol. 2012; 3: 327338. doi: 10.1111/j.2041-210X.
2011.00172.x
37. Hand D, Till R. A Simple Generalisation of the Area Under the ROC Curve for Multiple Class Classifica-
tion Problems. Mach Learn. 2001; 45: 171186. doi: 10.1023/A:1010920819831
38. Läderach P, Lundy M, Jarvis A, Ramirez J, Portilla EP, Schepp K, et al. Predicted impact of climate
change on coffee supply chains. The Economic, Social and Political Elements of Climate Change.
Springer; 2011. pp. 703723.
39. Jaramillo J, Chabi-Olaye A, Kamonjo C, Jaramillo A, Vega FE, Poehling H-M, et al. Thermal Tolerance
of the Coffee Berry Borer Hypothenemus hampei: Predictions of Climate Change Impact on a Tropical
Insect Pest. Rands S, editor. PLoS ONE. 2009; 4: e6487. doi: 10.1371/journal.pone.0006487 PMID:
19649255
Arabica Agro Ecological Zones and Climatic Change
PLOS ONE | DOI:10.1371/journal.pone.0140490 October 27, 2015 16 / 16
... Increasing knowledge of coffee cultivated in Yemen is of great significance for the coffee community around the world, including the 12.5 million households that rely on coffee growing for their incomes [15]. Due to climate change, the world is projected to lose 50% of suitable coffee-growing land by 2050 [16]. The climate in the coffee-growing regions of Yemen is one of the driest in the world, with an annual rainfall below 400 mm per year, while the minimum annual rainfall from more than 62,000 points representing arabica coffee-growing area around the world was almost double that, at 754 mm per year [17]. ...
... A central and statistically rigorous description and phenotypical evaluation of the different coffee genetic types in Yemen will pave the way for a sound genetic improvement program in the country and prevent genetic erosion. Furthermore, the resultant knowledge may benefit the wider coffee community facing challenges related to climate change [16]. ...
Article
Full-text available
While Ethiopia and South Sudan are the native habitats for Coffea arabica, Yemen is considered an important domestication center for this coffee species as most Arabica coffee grown around the world can be traced back to Yemen. Furthermore, climatic conditions in Yemen are hot and extremely dry. As such, Yemeni coffee trees likely have genetic merits with respect to climate resilience. However, until recently, very little was known about the genetic landscape of Yemeni coffee. The Yemeni coffee sector identifies coffee trees according to numerous vernacular names such as Udaini, Tufahi or Dawairi. However, the geographical landscape of these names and their correlation with the genetic background of the coffee trees have never been explored. In this study, we investigated the geographic occurrence of vernacular names in 148 coffee farms across the main coffee areas of Yemen. Then, we used microsatellite markers to genotype 88 coffee trees whose vernacular name was ascertained by farmers. We find a clear geographical pattern for the use of vernacular coffee names. However, the vernacular names showed no significant association with genetics. Our results support the need for a robust description of different coffee types in Yemen based on their genetic background for the benefit of Yemeni farmers.
... Changes in the suitability of coffee species accompany the relocation of producing regions. Other regions will become less climatically suitable for growing Arabica coffee (in South and Central America, Africa, and Asia) but more suitable for growing robusta coffee [66]. At least 83% of the total future coffee-growing area meets the requirements for robusta cultivation, but only 17% (±6%) meets requirements for Arabica [63]. ...
Article
Full-text available
Coffee production is fragile, and the Intergovernmental Panel on Climate Change (IPCC) reports indicate that climate change (CC) will reduce worldwide yields on average and decrease coffee-suitable land by 2050. This article adopted the systematic review approach to provide an update of the literature available on the impacts of climate change on coffee production and other ecosystem services following the framework proposed by the Millenium Ecosystem Assessment. The review identified 148 records from literature considering the effects of climate change and climate variability on coffee production, covering countries mostly from three continents (America, Africa, and Asia). The current literature evaluates and analyses various climate change impacts on single services using qualitative and quantitative methodologies. Impacts have been classified and described according to different impact groups. However, available research products lacked important analytical functions on the precise relationships between the potential risks of CC on coffee farming systems and associated ecosystem services. Consequently, the manuscript recommends further work on ecosystem services and their interrelation to assess the impacts of climate change on coffee following the ecosystem services framework.
... Intergroup crosses between "New-Yemen" or "Harrar" with "Typica/Bourbon" might be worth evaluating. Exploring new F1 hybrids in C. arabica with genetic clusters such as "Harrar" or "New-Yemen", both thriving in marginal dry and hot coffee areas (Hararghe and Yemen) is a promising opportunity in the context of climate change in coffee growing areas around the world [63] for which resilient and climate smart varieties are needed [64]. This highlights the importance of the exchange of genetic materials in the framework of collaborative research with equitable benefit sharing. ...
Article
Full-text available
The coffee species Coffea arabica is facing numerous challenges regarding climate change, pests and disease pressure. Improved varieties will be part of the solution. Making optimal use of the scarce genetic diversity of the species is hence essential. In this paper, we present the first study of C. arabica genetic diversity covering its complete native habitat in Ethiopia together with its main domestication centers: Yemen and Hararghe region in Ethiopia. All in all, 555 samples were analyzed with a set of Single Sequence Repeat markers. Through admixture genetic analysis, six clusters were identified. A total of two “Core Ethiopian” clusters did not participate in the domestication of the species. There were four clusters that were part of the “Domestication Pathway” of C. arabica. The first one was named “Ethiopian Legacy” as it represents the genetic link between “Core Ethiopia” and the “Domestication Pathway” in Yemen and Hararghe. The geographic origin of this cluster in Ethiopia was the south of Ethiopia, namely Gedio, Guji and Sidama, which hence appears as the source of coffee seeds that led to the domestication of C. arabica. In Yemen, in addition to the “Ethiopian Legacy” cluster, we confirmed the “Typica/Bourbon” and “New-Yemen” clusters. In Hararghe, the “Harrar” cluster, never described before, likely originates from a re-introduction of domesticated coffee from Yemen into this region of Ethiopia. Cultivated varieties around the world today originate from the “Ethiopian Legacy” and “Typica/Bourbon” clusters and but none are related to the “new-Yemen” and “Harrar” clusters. Implications for breeding strategies are discussed.
... This is relevant as coffee is highly susceptible to changes in climatic conditions. Increases in temperature and changes in precipitation patterns will affect coffee yields and quality and be particularly severe in regions less suitable for coffee production (Bunn et al., 2015a). ...
Thesis
Full-text available
Coffee production is an important source of export revenue for producing countries, especially for small, agriculture-dependent economies like Rwanda. Coffee production is a key driver in the development and improvement of rural livelihoods, serving as a source of cash income for the many coffee-producing households. The coffee value chain in Rwanda changed visibly since the early 2000s. Since then, the number of Coffee Washing Stations increased considerably, as did the share of fully washed coffee production. Rwanda’s coffee sector is a now well-established player in the international speciality coffee market. Despite considerable improvements, productivity remains low, as farmers struggle with pests and diseases, poor soil fertility and insufficient access to fertilisers. These challenges faced by coffee producers call for suitable and sustainable solutions. With coffee farmers also facing the repercussions of progressing climate change, the present dissertation aims to identify ways to support smallholder coffee producers in their efforts to respond to the challenges they face. Therefore, the thesis taps into two important fields of research on coffee producers – climate change adaptation and sustainability certification. First, the thesis evaluates the role of certification in improving farmers’ economic and environmental performance. Secondly, it addresses the question of how farmers respond to climate change, and how they can be supported in their efforts. The dissertation consists of two case studies from Rwanda and one chapter reviewing the literature on climate change adaptation. The data for the empirical research was collected from September to December 2019 in three climatic regions in Rwanda.
... Se increased the content of catalase Introduction Coffee is one of the most important commodities worldwide with a significant economic impact on over 25 million mostly smallholder farmers in more than 60 countries throughout the tropics (Jayakumar et al., 2017). Coffee plants are highly sensitive to the growing environment and are generally restricted to the "Coffee Belt"-between 25 degrees North and 30 degrees South with an average temperature between 18 and 22 • C for Coffea arabica and 22 and 28 • C for Coffea canephora (DaMatta and Ramalho, 2006;Descroix and Snoeck, 2004;Bunn et al., 2015;Bliss, 2017). Among the 104 Coffea species described (Davis and Rakotonasolo, 2008), the two most economically important species are C. arabica (Arabica) and C. canephora (Robusta). ...
Article
Full-text available
The effects of selenium (Se) on plant metabolism have been reported in several studies triggering plant tolerance to abiotic stresses, yet, the effects of Se on coffee plants under chilling stress are unclear. This study aimed to evaluate the effects of foliar Se application on coffee seedlings submitted to chilling stress and subsequent plant recovery. Two Coffea species, Coffea arabica cv. Arara, and Coffea canephora clone 31, were submitted to foliar application of sodium selenate solution (0.4 mg plant–1) or a control foliar solution, then on day 2 plants were submitted to low temperature (10°C day/4°C night) for 2 days. After that, the temperature was restored to optimal (25°C day/20°C night) for 2 days. Leaf samples were collected three times (before, during, and after the chilling stress) to perform analyses. After the chilling stress, visual leaf injury was observed in both species; however, the damage was twofold higher in C. canephora. The lower effect of cold on C. arabica was correlated to the increase in ascorbate peroxidase and higher content of starch, sucrose, and total soluble sugars compared with C. canephora, as well as a reduction in reducing sugars and proline content during the stress and rewarming. Se increased the nitrogen and sulfur content before stress but reduced their content during low temperature. The reduced content of nitrogen and sulfur during stress indicates that they were remobilized to stem and roots. Se supply reduced the damage in C. canephora leaves by 24% compared with the control. However, there was no evidence of the Se effects on antioxidant enzymatic pathways or ROS activity during stress as previously reported in the literature. Se increased the content of catalase during the rewarming. Se foliar supply also increased starch, amino acids, and proline, which may have reduced symptom expression in C. canephora in response to low temperature. In conclusion, Se foliar application can be used as a strategy to improve coffee tolerance under low-temperature changing nutrient remobilization, carbohydrate metabolism, and catalase activity in response to rewarming stress, but C. arabica and C. canephora respond differently to chilling stress and Se supply.
... Due to changing climate conditions, there have been increasing concerns about the future quantity and quality of the coffee yield in the decades to come. Some publications showed the declining suitable area (in excess of 50%) and extinction of coffee plantations [11], [12]. This study aims to predict the future coffee's land suitability as an agricultural geographical indication product in Indonesia using climate historical big data. ...
... These activities and their consequences in terms of global warming result in increasing the vulnerability of agricultural and food systems [2]. In the particular case of Arabica coffee (Coffea arabica), the forecasted consequences of climate change (CC), and more specifically changes in rainfall patterns, elevated temperatures, more frequent drought periods, and a shift in geographical coffee-growing regions can boost environmental and socioeconomic threats in the near future [3]. C. arabica currently only grows in world regions where the differences in photoperiods do not exceed 2 h, suggesting that this species might not be adapted to large changes in seasonal photoperiods. ...
Article
Full-text available
Climate change (CC) is already impacting Arabica coffee cultivation in the intertropical zone. To deal with this situation, it is no longer possible to manage this crop using industrial agriculture techniques, which has been the main strategy implemented since the Green Revolution. Developing a more sustainable agriculture system that respects people and the environment is essential to guarantee future generations’ access to natural resources. In the case of Arabica coffee, the solution has been found. Agroforestry is proposed as an ecosystem-based strategy to mitigate and adapt to CC. At least 60% of Arabica coffee is produced in agroforestry systems (AFSs), which are the most sustainable way to produce coffee. Nevertheless, AFS coffee cultivation is currently uncompetitive partly because all modern varieties, selected for full-sun intensive cropping systems, have low yields in shaded environments. Here we review the reasons why agroforestry is part of the solution to CC, and why no breeding work has been undertaken for this cropping system. Based on the literature data, for breeding purposes we also define for the first time one possible coffee ideotype required for AFS coffee cultivation. The four main traits are: (1) productivity based on F1 hybrid vigor, tree volume and flowering intensity under shade; (2) beverage quality by using wild Ethiopian accessions as female progenitors and selecting for this criterion using specific biochemical and molecular predictors; (3) plant health to ensure good tolerance to stress, especially biotic; and (4) low fertilization to promote sustainable production. For each of these traits, numerous criteria with threshold values to be achieved per trait were identified. Through this research, an ecosystem-based breeding strategy was defined to help create new F1 hybrid varieties within the next 10 years.
... This is relevant as coffee is highly susceptible to changes in climatic conditions. Increases in temperature and changes in precipitation patterns will affect coffee yields and quality and be particularly severe in regions less suitable for coffee production (Bunn, Läderach, Jimenez, Montagnon, & Schilling, 2015). ...
Article
Full-text available
Sustainability certification has become an important tool for promoting sustainable agricultural value chains. Nevertheless, its economic and environmental effects on the producer level remain unclear. We investigate the relationship of Rainforest Alliance Certification with socio‐economic and environmental outcomes in Rwanda and consider potential tradeoffs between dimensions. To reduce potential selection bias in the econometric estimation, we use inverse probability weighted regression adjustment. We find no significant association between certification and socio‐economic indicators but a significant correlation between certification and good agricultural practices and biodiversity‐related practices. Effects on economic outcomes and biodiversity‐related practices are linked; their relationship differs across climatic regions.
Article
Like many perennial crops, coffee exhibits alternate bearing, a pattern of reproduction in which high-yielding years are followed by low-yielding ones. Alternate bearing threatens farmer livelihoods, yet little is known about the underlying mechanisms in coffee or the potential for farm management to mitigate it. The resource budget model, an ecological theory positing endogenous resource tradeoffs as the driver of reproductive variability, could help fill this gap. On three coffee farms in Santa María de Dota, Costa Rica, we manipulated relative fruit load, fertilizer levels, and shade cover to test whether the model’s core assumptions (i.e., that fruiting depletes resources and limits investment in subsequent reproduction) can elucidate patterns of alternate bearing in coffee, and to assess whether these patterns are impacted by farm management practices. Coffee plants exhibited within-year and between-year tradeoffs of a high fruit load that scaled from decreased bean size during the same season to fewer fruited nodes and fruits per node in both old and new cohorts of branches during the subsequent season. Stem nitrogen concentration was also depleted in response to high fruit loads and recovered during the subsequent season of low fruiting. The findings provide novel evidence that tradeoffs of a high fruit load are manifested in several reproductive traits at both the branch- and plant-level and offer initial support for the resource budget model in the system. While both moderate shade and increased fertilizer levels tended to improve reproductive traits, a lack of interactive effects between either management treatment and the relative fruit load treatment suggests that they do little to mitigate the reproductive tradeoffs underlying alternate bearing.
Article
Full-text available
Estimating crop biomass is critical for countries whose primary source of income is agriculture. It is a valuable indicator for evaluating crop yields and provides information to growers and managers for developing climate change adaptation strategies. The objective of the study was to model the impacts of agroclimatic indicators on the performance of aboveground biomass (AGB) in Arabica coffee trees, a critical income source for millions of Ethiopians. One hundred thirty-five coffee tree stump diameters were measured at 40 cm above ground level. The historical (1998–2010) and future (2041–2070) agroclimatic data were downloaded from the European Copernicus climate change services website. All datasets were tested for missing data, outliers, and multicollinearity and were grouped into three clusters using the K-mean clustering method. The parameter estimates (coefficients of regression) were analyzed using a generalized regression model. The performance of coffee AGB in each cluster was estimated using an artificial neural network model. The future expected change in AGB of coffee trees was compared using a paired t-test. The regression model's results reveal that the sensitivity of C. arabica trees to agroclimatic variables significantly differs based on the kind of indicator, RCP scenario, and microclimate. Under the current climatic conditions, the rise of the coldest minimum (TNn) and warmest (TXx) temperatures raises the AGB of the coffee tree, but the rise of the warmest minimum (TNx) and coldest maximum (TXn) temperatures decreases it (P <0.05). Under the RCP4.5, the rise of consecutively dry days (CDD) and TNx increased the AGB of the coffee tree, while TNx and TXx decreased it (P<0.05. Except for TXx, all indicators would significantly reduce the AGB of coffee trees under RCP8.5 (P <0.05). The average values of AGB under the current, RCP4.5, and RCP85 climate change scenarios, respectively, were 26.66, 28.79, and 24.41 kg/tree. Compared to the current climatic conditions, the predicted values of AGB under RCP4.5 and RCP8.5 will increase in the first and third clusters and decrease in the second. As a result, early warning systems and adaptive strategies will be necessary to reduce the detrimental consequences of climate change. More research into the effects of other climatic conditions on crops, such as physiologically effective degree days, cold, hot, and rainy periods, is also required.
Article
Full-text available
Arabica coffees (60 % of current world coffee production) are generally sold at considerably better prices than robustas on account of superior beverage quality. However, costs of production are much higher, mainly due to more stringent demands for soil and climatic conditions, crop management, primary processing and control of several pests and diseases including the potentially very destructive coffee leaf rust (CLR) and berry disease (CBD). Breeding for disease resistance in combination with vigour, productivity and quality started in the early 1920s in India, but especially in the second half of the 20th century comprehensive breeding programmes have been implemented in several other coffee producing countries. Many of the resulting CLR- and CBD + CLR-resistant cultivars (true-breeding lines and F1 hybrids) meet the required standards of profitable and sustainable crop production. Challenges of more recent date include limited access to additional genetic resources of Coffea arabica, breakdown of host resistance to CLR, aggravating insect pest problems and the increasingly negative impact of climate change on arabica coffee production worldwide. This review discusses prospects of breeding and disseminating next generation (hybrid) cultivars of arabica coffee for sustainable coffee production under changing conditions of diseases, pests and climate. International networking on coffee breeding will facilitate sharing of resources (financial, genetic) and scientific information, application of genomics-assisted selection technologies, and pre-breeding for specific characters. Breeding and multiplication of new cultivars well adapted to the local environment will continue to be carried out at national or regional levels. A tree crop like arabica coffee does not lend itself to centralized variety development and dissemination on a global scale.
Conference Paper
Full-text available
Brazil is the world’s largest producer of Arabica coffee. Adverse climatic events in its major production regions have global repercussions through market effects. Climate change impacts on the Brazilian coffee production are thus of high interest to understand long term trends on global coffee markets. Nevertheless, this aspect has not been looked at using a rigid modelling framework. To investigate the climate change impacts on the Brazilian Arabica production, we employed three different modeling approaches to generate an ensemble climate envelope model of current and future Arabica suitability distributions. First, based on high-resolution census data an unbiased database of Arabica production locations was assembled. Three different classification methods from ecological niche theory, Hypervolume, Maxent and Bioclim, were trained and evaluated against this true distribution of current production using various feasible parameter settings. We thus extrapolated these models on downscaled climate data from three global climate models of the 4th IPCC Assessment report for 2030, 2050 and 2080 in the SRES A2 emission scenario and constructed consensus maps for each time slice. The results of the consensus across all models show a marginal migration of areas suitable for coffee production towards the Southern states of Santa Catarina and Rio Grande do Sul, while more Northern locations in Bahia, Rondonia and Goias are projected to experience drastic losses of area. By 2050 50% (70% by 2080) of models in the ensemble project a loss of suitable area in Minas Gerais as compared to current climate if cultivation practices are adapted to progressive climate change. When we applied a more restrictive analysis, demanding a high agreement of models comparable to the agreement for current conditions, we found that area with such ideal conditions could be reduced entirely by 2080 for all of Brazil. Our results confirm previously hypothesized trends caused by climate change, but dispute the extent of opportunities from novel areas in the extreme South of Brazil. Once full climate change effects are experienced, Brazil may face challenges to remain a major coffee producing country.
Conference Paper
Full-text available
To date, knowledge on the physical distribution of global coffee production is limited. Publicly available datasets of coffee production are incomplete or of limited value to coffee research. Spatially explicit data on the physical distribution of coffee production would help to understand and design policies that address challenges to the industry that result from global change processes and resource limitations. Here, we demonstrate a method to spatially disaggregate global harvested area production statistics from FAO using point location based suitability maps. Machine learning methods are applied to distinguish climate in coffee production countries from the climate at geo-referenced locations of coffee production to derive spatially explicit probabilities of presence of coffee production sub-nationally. Using a cross entropy approach harvested area statistics for C. arabica and C. canephora are allocated to areas with suitable climate. Finally, we compare our dataset with other available datasets and demonstrate that our approach best represents the existing knowledge of the global distribution of area dedicated to coffee production. The resulting dataset may serve to more adequately represent coffee production in research that is interested in the spatially explicit analysis of land use, land use change and resource management than before possible.
Article
Full-text available
Coffee has proven to be highly sensitive to climate change. Because coffee plantations have a lifespan of about thirty years, the likely effects of future climates are already a concern. Forward-looking research on adaptation is therefore in high demand across the entire supply chain. In this paper we seek to project current and future climate suitability for coffee production (Coffea arabica and Coffea canephora) on a global scale. We used machine learning algorithms to derive functions of climatic suitability from a database of geo-referenced production locations. Use of several parameter combinations enhances the robustness of our analysis. The resulting multi-model ensemble suggests that higher temperatures may reduce yields of C. arabica, while C. canephora could suffer from increasing variability of intra-seasonal temperatures. Climate change will reduce the global area suitable for coffee by about 50 % across emission scenarios. Impacts are highest at low latitudes and low altitudes. Impacts at higher altitudes and higher latitudes are still negative but less pronounced. The world’s dominant production regions in Brazil and Vietnam may experience substantial reductions in area available for coffee. Some regions in East Africa and Asia may become more suitable, but these are partially in forested areas, which could pose a challenge to mitigation efforts.
Chapter
A quick pick-me-up or a subtle beverage with an aroma that conjures up images of special moments shared with special people? There's more to coffee than that. Apart from being a beautiful tree with fragrant flowers, coffee is also a culture, practically a religion to a certain elite and certainly a source of income to millions of people, rich and poor alike. Coffee professionals around the world will find the specific information they need in this lavishly illustrated and practical work designed to answer all their questions about the coffee plant and how it is grown, harvested, processed and refined. Specialists and experienced professionals were consulted and some 40 renowned international experts have contributed their specific knowledge and expertise to this comprehensive handbook, covering such topics as: Growing Pests, diseases, and their control Harvesting and processing Storage, shipment, quality The latest economical and technological aspects.In addition, special indexes demystify such confusing data as information sources, conversion tables and other technicalities. With its 40 chapters, over 1000 pages and 900 superb illustrations, this is a universally reliable manual, providing basic guidelines and recommendations applicable everywhere, and not geared to any specific country. © WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 2004. All rights reserved.
Article
Regional studies have shown that climate change will affect climatic suitability for Arabica coffee (Coffea arabica) within current regions of production. Increases in temperature and changes in precipitation patterns will decrease yield, reduce quality and increase pest and disease pressure. This is the first global study on the impact of climate change on suitability to grow Arabica coffee. We modeled the global distribution of Arabica coffee under changes in climatic suitability by 2050s as projected by 21 global circulation models. The results suggest decreased areas suitable for Arabica coffee in Mesoamerica at lower altitudes. In South America close to the equator higher elevations could benefit, but higher latitudes lose suitability. Coffee regions in Ethiopia and Kenya are projected to become more suitable but those in India and Vietnam to become less suitable. Globally, we predict decreases in climatic suitability at lower altitudes and high latitudes, which may shift production among the major regions that produce Arabica coffee.