Aaron P. Davis; Tadesse Woldemariam Gole; Susana Baena; Justin Moat

“ There could be a buffer influence for wild populations of Arabica, because the forest micro-environment itself is not included in recorded climate data and therefore not modelled. That is, certain variables such as the mean temperature (s) will be lower inside the forest than in exposed areas in the same general bioclimatic space. In addition, the long generation time of Arabica (c. 50 years, and perhaps up to 100 years [4] ) will mean that even if reproduction, dispersal and colonization are reduced, or neutralized, individual trees may persist for some time. ”

“ Ethiopia is the fifth largest global exporter of Arabica and the main producer of coffee in Africa; coffee accounts for c. 33% of Ethiopia's total export earnings [1] . The largest and most diverse populations of indigenous (wild) Arabica occur in the highlands of south-western Ethiopia ( [4] , [15] – [16] , but the native range includes satellite population in south-eastern South Sudan (Boma Plateau) and northern Kenya (Mt Marsabit) , at altitudes between 950 and 1950 m, although 1200 m is the most frequent lower altitude limit [17] . ”

“ Detailed modelling of Arabica cultivation is required, on local and regional scales, in order to inform famers and decision makers as to the requirements for future-proofing the sustainability of their crop. The methodology used here could be adapted for coffee plantations on a regional scale, by substituting the location of plantations for indigenous populations, and by applying a modified threshold approach based on the parameters encountered and employed in cultivation. ”

“ Re-locating coffee plantations is likely to bring economic benefits within a realistic time frame; the assisted migration of natural populations of Arabica coffee is not.
What we have shown here is that under a range of emission scenarios some populations of Arabica (occurring in optimal bioclimatic space) might be able to resist climate change until 2080, at least in the absence of severely negative influences (e.g. deforestation) . ”

“ Fragmentation reduces the progress and success of migration for many forest species, and would hinder the establishment of new areas corresponding to Moist Evergreen Afromontane Forest and Transitional Rain Forest vegetation types. Managed relocation of Arabica individuals or even populations by human effort is conceivable, although as with any other form of dispersal a suitable habitat would have to be available during this process and these may be limited (see above) and localized. ”

“ Data for cultivated Arabica confirms these observations, and shows that this species is sensitive to environmental variables, particularly temperature and precipitation [2] , [6] , [8] . Soil-water balance is a key factor for the growth of Arabica, although until now this data is only available for the species in cultivation [2] , [8] . In addition, observation and reported distribution show that coffee species are rarely adventive, are only scarcely found in secondary (regenerating) vegetation, and quickly become stressed in degraded habitats [17] . ”

“ Unlike cultivated Arabica coffee, the distribution of indigenous populations is controlled almost entirely by natural, biotic parameters, even though these factors are influenced by anthropogenic actions. ”

“ The area analysis predicts a general northward concentration through time, i.e. an increase in suitable bioclimatic space in the northern part of the distribution, although the likelihood of migration and establishment by Arabica is assumed to be extremely limited based on insubstantial dispersal and colonization ability, especially in stressed environments. Arabica has a relatively long generation time: even in cultivation it requires a minimum of three to four years to produce fruit and at least five to eight years to reach maximum reproductive potential [50] . ”

“ For more precise modelling of future distribution trends for Arabica, mapping and modelling work should incorporate actual (remaining) vegetation, as opposed to a potential vegetation [46] , although these data are not presently available. These mapping data would be used to ‘clip’ suitable remaining habitat for Arabica from the models, and provide more accurate estimates for loss of suitable bioclimatic space. More detailed ecological data (subcanopy temperatures, soil/water balance, soil type, humidity, evapotranspiration) is also required. ”

“ The evidence from coffee farmers, from numerous coffee growing regions around the world, is that they are already suffering from the influences increased warming [12] , [11] . Precise modelling of the influence of climate change for either Arabica or robusta is limited. So far work has been largely restricted to climatic envelope forecasting based on optimum and suboptimum (tolerable) temperature and rainfall averages [11] , [13] , although in Sao Paulo (Brazil) future climatic scenarios have been applied across the federal agricultural zoning system for cultivated Arabica [10] . ”

“ We use bioclimatic modelling based on locality data, in combination with future climate change scenarios across predetermined time intervals until 2080. Bioclimatic models typically compare large-scale species distributions (distribution maps) against present-day modelled climate parameters, in order to generate a predictive statistical model [23] . ”

“ These thresholds were used to classify each Arabica locality though time, within each scenario, into 68% (optimal) , 95% (intermediate [suboptimal] ; includes the 68% threshold) , 100% (marginal [extreme environments] ; includes 68% and 95% thresholds) . These thresholds were applied to the models to assess the future predicted distribution in terms of bioclimatic suitability (see locality analysis and area analysis, below) , with the following assumptions and limitations. (1) Individual plants will not move to different localities that are less suitable than the present day. ”

“ In this study we model the indigenous distribution of Arabica for the present day, and for the future under the influence of climate change until 2080, in order to identify priorities (for in situ and, and ex situ conservation, monitoring, and future research) and facilitate appropriate decision making. ”

“ For interpretation of the results generated by the locality analysis we need to consider the assumptions used in the downscaling methods when applied to the HadCM3 model [48] , which are: (1) changes in climates vary only over large distances, i.e. as large as Global Circulation Model cell size; (2) relationships between variables in the baseline (‘current climates’) are likely to be maintained towards the future. That is, any generalisations or problems in the original BIOCLIM data will be maintained. ”

“ These changes coincided with the appearance and chronic spread of a parasitic infestation, apparently correlated to the change in local climate, which has seriously undermined the growth and reproductive potential of this coffee species, with severe consequences for the long-term survival of the species [56] . For Arabica, the coffee berry borer (Hypothenemus hampei (Ferrari) [Coleoptera] ) poses a significant compounding threat to indigenous populations and plantations. ”

“ This type of modelling has been judged as ecologically naive by some [25] , [26] , as species occurrence is dependent not only on climate, but also on ecological processes such as dispersal, colonization, and complex interactions with other organisms. In this respect, process based modelling, including ecological data, are often favoured on theoretical grounds, by injecting ecological realism into the modelling framework [23] . ”

“ Global temperatures have increased by an average of 0.74°C (+0.56°C to 0.92°C) in the last 100 years (1906–2005) , and this increase appears to have accelerated since the 1970s [9] . On this basis it has been forecast that the sustainability of the coffee industry faces serious challenges in the coming decades [2] , [10] – [11] . ”

“ Generally, the locality analysis also requires fewer assumptions. In the area analysis, the 68% (optimal) threshold is likely to be very exclusive. For example, if in 2050 a pixel falls into the 95% (intermediate) threshold from the 68% threshold it would stay within the former threshold in subsequent dates. ”

“ At temperatures above 23°C, development and ripening of fruits are accelerated, often leading to the loss of beverage quality [5] , although in some locations higher temperatures (24–25°C) can still produce satisfactory yields of beans, such as in northeast Brazil [6] . ”

“ Genetic diversity in the coffee wilt pathogen ( Gibberella xylarioides ) populations: differentiation by host specialization and RAPD analysis. J Plant Dis Protect 112: 134–145 1 2 Taye K (2006) Ecophysiological diversity of wild Arabica populations in Ethiopia: growth, water relations and hydraulic characteristics along a climatic gradient. Ecology and Development Series No. 46. Cuvillier Verlag: Göttingen. 305 p.1 2 Montagnon C, Bouharmont P (1996) Multivariate analysis of phenotypic diversity of Coffea arabica . ”

“ Locality analysis overview I.Predicted climate change outcomes for indigenous Arabica localities for the year interval 2000, 2020, 2050 and 2080. Stacked bar-charts based on Table 1. Green = optimal [bioclimatic] localities (68%) ; yellow = intermediate (suboptimal) [bioclimatic] localities (95%) ; red = marginal (extreme) [bioclimatic] localities (100%) ; grey = unsuitable bioclimatic localities.Locality analysis overview II.Predicted climate change outcomes for indigenous Arabica localities (349 in total) for the year intervals 2000, 2020, 2050 and 2080. ”

“ Predicting the biodiversity response to climate change: challenges and advances. Syst Biodivers 9: 307–317 1 2 3 4 Morin X, Thuiller W (2009) Comparing niche- and processbased models to reduce prediction uncertainty in species range shifts under climate change. ”

“ Forest herbs in the face of global change: a single-species-multiple-threats approach for Anemone nemorosa . Plant Ecol Ev 143: 19–30 ↑ Walck JL, Hidayati SN, Dixon KW, Thompson K, Poschlod P (2011) Climate change and plant regeneration from seed. Glob Change Biol 17: 2145–2161 ↑ Jetz W, Wilcove DS, Dobson AP (2007) Projected impacts of climate and land-use change on the global diversity of birds. PLoS Biol 5: e157. pmid:17550306↑ Davis AP, Mvungi E (2004) Two new and endangered species of Coffea (Rubiaceae) from the Eastern Arc Mountains (Tanzania) and notes on associated conservation issues. ”

“ Ecological niche modeling in MAXENT: the importance of model complexity and the performance of model selection criteria. Ecol Appl 21: 335–342 pmid:215635661 2 Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas. Int J Climatol 25: 1965–1978 1 2 3 4 Hijmans RJ (2012) Cross-validation of species distribution models: removing spatial sorting bias and calibration with a null model. ”