Future Climate Scenarios for Tanzania s Arabica Coffee Growing Areas. Final report Cali, Colombia: October 2012

Transcription

1 Future Climate Scenarios for Tanzania s Arabica Coffee Growing Areas Final report Cali, Colombia: October 2012

2 Table of Contents 1 Authors and contact details Executive summary Project Background and Objectives Key objectives Methodology a Study area b Current climate... 4 Bioclimatic variables c Future climate... 5 Global circulation models... 5 Generation of future climate d Crop prediction... 6 Maximum Entropy e Measure of confidence f Environmental factors driving change in suitability Result I: Climate change summary of coffee production sites a The summary climate characteristics for all coffee factory sites in Tanzania... 8 General climate characteristics... 8 Extreme conditions... 9 Climate Seasonality... 9 Variability between models b Regional changes in the mean annual precipitation (2030) c Regional changes in the mean annual precipitation (2050) d Regional changes in the mean annual temperature (2030) e Regional changes in the mean annual temperature (2050) Result II: Suitability maps of coffee production areas a Current suitability of coffee production areas b Future suitability of coffee production areas c Address uncertainty of MaxEnt output using multiple GCM Result III: Environmental factors which drive the suitability of coffee Conclusions I

3 10 References II

4 Table of figures Figure 1: Study sites of Arabica coffee-growing regions in Tanzania Figure 2: Climate trend summary 2030 and 2050 for sample sites Figure 3: Mean annual precipitation change by 2030 in the coffee-growing regions of Tanzania Figure 4: Mean annual precipitation change by 2050 in the coffee-growing regions of Tanzania Figure 5: Mean annual temperature change by 2030 in the coffee-growing regions of Tanzania Figure 6: Mean annual temperature change by 2050 in the coffee-growing regions of Tanzania Figure 7: Current suitability for Arabica coffee production in the coffee-growing regions of Tanzania Figure 8: Suitability of Arabica coffee production by Figure 9: Suitability of Arabica coffee production by Figure 10: Predicted changes in coffee suitability and breadth of climate models uncertainty by Figure 11: Predicted changes in coffee suitability and breadth of climate models uncertainty by Figure 12: Relation between the suitability of the coffee production areas and the altitude for the current climate and forecasts for 2050 in Tanzania Figure 13: Available area at different altitudes for marginal (20-40%), good (40-60%), and very good (60-80%) suitability III

5 Table of tables Table 1: Contribution of different bioclimatic variables to the predicted shift in suitability decrease for Arabica coffee in Tanzania, between the present and the 2030s Table 2: Contribution of different bioclimatic variables to the predicted shift in suitability decrease for Arabica Coffee in Tanzania, between today and the 2050s IV

6 1 Authors and contact details The Decision and Policy Analyses (DAPA) group at CIAT conducted the analyses presented here under the leadership of Peter Läderach, with the collaboration of Anton Eitzinger, Oriana Ovalle, Stephania Carmona and Eric Rahn. For further information please contact: Dr. Peter Laderach or Anton Eitzinger International Center for Tropical Agriculture (CIAT) A.A. 6713, Cali Colombia and 2 Executive summary This document reports on the methods and results of a consultancy with the title Future Climate Scenarios for Tanzania s Coffee Growing Areas conducted for the Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH (GIZ). The methodology applied was based on the combination of current climate data with future climate change predictions from 19 models for 2030 and The data of the current climate and the climate change was used as input to MAXENT, a crop prediction model. The evidence data used for MAXENT were presence points of Arabica coffee collected by GPS through fieldwork in Tanzania. The analysis focused on the specific regions that were of interest to the client and provide predictions of the future climate and suitability of current coffee-growing areas to continue growing coffee by 2030 and 2050 whereby climate suitability is defined as the level of certain climatic characteristics that determine which areas have potential to grow a crop successfully. The results show that the change in suitability as climate change occurs is site-specific. There will be areas that become unsuitable for coffee (Kagera, Mara and Kigoma). There will be areas with a slight loss in suitability for coffee (Arusha, Iringa), but may remain suitable if farmers adapt their agronomic management to the new conditions the area will experience. In Tanzania only few areas present a slight increase in suitability, these are located in Arusha and Iringa. Finally, there will be areas where currently no coffee is grown but which in the future will become suitable especially in higher altitudes around Mount Kilimanjaro and on the northwest side of the Usambara Mountains, where important natural forest reserves are located. Since many of these areas are protected, it is not recommended to clear forests or invade protected areas in order to produce coffee. Climate change brings not only bad news but also a lot of potential. The winners will be those who are prepared for change and know how to adapt. In Tanzania the mean annual and the minimum and maximum monthly temperature will increase by 2030 and continue to increase by 2050 like that the annual precipitation. The overall climate will 1

7 become more seasonal in terms of variation in temperature through the year with temperature in specific districts increasing by about 1.3 ºC by 2030 and 2.3 ºC by 2050 and more seasonal in precipitation with the maximum number of cumulative dry months decreasing from 6 months to 5 months. The implications are that the distribution of suitability within the current coffee-growing areas in Tanzania for Arabica coffee production in general will decrease quite seriously in some regions by The suitable areas will migrate up the altitudinal gradient. Areas that retain some suitability will experience decreased to between by 20 to 50%, while the current suitability is between 40 and 70%. The optimum coffee-producing zone is currently at an altitude between 900 and 2400masl and will by 2050 increase to an altitude between 1300 and 2800 masl. Increasing altitude compensates for the increase in temperature. Compared with today, by 2050 areas at altitudes between 900 and 1300masl will suffer the highest decrease in suitability and the areas around 1800 masl will not change significantly in suitability. 3 Project Background and Objectives Gustav Paulig Ltd, Joh. Johannson Kaffe AS, Löfbergs Lila AB, Neumann Gruppe GmbH, Tchibo GmbH, Fondazione Giuseppe e Pericle Lavazza Onlus, Ecom Coffee, Franck d.d., Tim Hortons and the Swedish Development Agency (SIDA) together with the Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH are implementing a supra regional Development Partnership with a focus on coffee and climate change. The objective of the project is that producers and service providers along selected green coffee supply chains in key coffee regions have access to adequate knowledge and instruments that enable them to apply and finance effective climate change adaption and mitigation strategies. This objective will be achieved by combining effective adaptation and mitigation approaches in a toolbox to be adapted to local specific conditions, training producers and service providers on the use of the toolbox and running capacity building activities for stakeholders along the green coffee supply chains. A self-financing institutional framework shall be established as a long-term host for the initiative and a platform strengthening the worldwide dissemination of the toolbox. In order to be better prepared for the impacts of climate change, scientific research on a regional basis is necessary. Therefore future climate scenarios for the coffee growing regions in the four pilot countries, Brazil, Vietnam, Tanzania and Guatemala, are necessary. The set up of an international data base on climate change and its impacts in the coffee production accessible for coffee actors at all levels shall be initialized and supported. The objective of this study is to develop future climate scenarios indicating the climate change suitability of Arabica coffee variety under changing climatic conditions for Tanzania s coffee-growing regions. Currently, the Arabica coffee-growing areas in Tanzania are located in three different zones: (i) the northern highlands including the Arusha, Kilimanjaro and Tanga regions; (ii) the southern highlands in 2

8 the Iringa and Ruvuma regions and the highlands between Lake Tanganyika and Lake Nyassa in the Mbeya region; and (iii) on the Burundi border in the Kigoma region (Figure 1). Figure 1: Study sites of Arabica coffee-growing regions in Tanzania. 4 Key objectives Predict the climate change for the areas cultivating Arabica coffee in Tanzania. Predict the impact of progressive climate change on cultivation suitability of Arabica coffee in Tanzania. 5 Methodology 5.a Study area Four hundred twenty-two coordinates of coffee farms across the main eight Arabica coffee-growing regions in Tanzania have been compiled (figure 1). The regions included Mbeya, Kilimanjaro, Ruvuma, Arusha, Kigoma, Iringa, Tanga and Manyara. The coordinates were distributed between an altitudinal range of 945 to 1973 masl, being 1300 to 1700 masl the altitude with the highest frequency of points. In Tanzania much of the crop under cultivation is rain-fed, household food depends on the streams and rivers which are the sources of irrigation water. The coffee is one of the variety of crops is under 3

9 irrigation during both dry and rainy season (Kagubila, 1994), therefore lack of adequate rainfall affects coffee production. Mild Arabica coffee produced in Kilimanjaro and Arusha regions of the Northern Highlands by smallscale coffee farmers and large estates has decreased, whereas production has increased in the Southern Highlands among small-scale coffee farmers in the Ruvuma and Mbeya regions (Ikeno, 2007). 5.b Current climate As current climate (baseline) we used historical climate data from the database (Hijmanset al., 2005). The WorldClim data are generated through interpolation of average monthly climate data from weather stations on a 30 arc-second resolution grid (often referred to as "1 km" resolution). Variables included are monthly total precipitation, and monthly mean, minimum, and maximum temperature, and 19 bioclimatic variables (Hijmanset al., 2005). In the WorldClim database, climate layers were interpolated using: Major climate databases compiled by the Global Historical Climatology Network (GHCN), the FAO, the WMO, the International Center for Tropical Agriculture (CIAT), R-HYdronet, and a number of additional minor databases for Australia, New Zealand, the Nordic European Countries, Ecuador, Peru, Bolivia, amongst others. The SRTM elevation database (aggregated to 30 arc-seconds, "1 km") The ANUSPLIN software. ANUSPLIN is a program for interpolating noisy multivariate data using thin plate smoothing splines. We used latitude, longitude, and elevation as independent variables. For stations where records for multiple years were available, the averages were calculated for the period. Only records with at least 10 years of data were used. In some cases the time period was extended to the period to include records from areas for which there were few recent records available (e.g., DR Congo) or predominantly recent records (e.g., Amazonia). After removing stations with errors, the database consisted of precipitation records from 47,554 locations, mean temperature from 24,542 locations, and minimum and maximum temperature for 14,835 locations. The data on which WorldClim is based on in Tanzania are from 606 stations with precipitation data, 566 stations with mean temperature, and 58 stations with minimum and maximum temperatures. Bioclimatic variables Within the WorldClim database, there are bioclimatic variables that were derived from the monthly temperature and rainfall values to generate more biologically meaningful variables, which are often used in ecological niche modeling (e.g., BIOCLIM, GARP). The bioclimatic variables represent annual trends (e.g., mean annual temperature, annual precipitation), seasonality (e.g., annual range in temperature and precipitation) and extreme or limiting environmental factors (e.g., temperature of the coldest and warmest month, and precipitation of the wettest and driest quarters). A quarter is a period of three months (1/4 of the year). 4

10 The derived bioclimatic variables are: Bio1 = Annual mean temperature Bio2 = Mean diurnal range (mean of monthly (max temp - min temp)) Bio3 = Isothermality (Bio2/Bio7) (* 100) Bio4 = Temperature seasonality (standard deviation *100) Bio5 = Maximum temperature of warmest month Bio6 = Minimum temperature of coldest month Bio7 = Temperature annual range (Bio5 Bi06) Bio8 = Mean temperature of wettest quarter Bio9 = Mean temperature of driest quarter Bio10 = Mean temperature of warmest quarter Bio11 = Mean temperature of coldest quarter Bio12 = Annual precipitation Bio13 = Precipitation of wettest month Bio14 = Precipitation of driest month Bio15 = Precipitation seasonality (coefficient of variation) Bio16 = Precipitation of wettest quarter Bio17 = Precipitation of driest quarter Bio18 = Precipitation of warmest quarter Bio19 = Precipitation of coldest quarter 5.c Future climate Global circulation models A global circulation model (GCM) is a computer-based model that calculates and predicts what climate patterns will be like in a number of years in the future. GCMs use equations of motion as a numerical weather prediction model, with the purpose of numerically simulating changes in the climate as a result of slow changes in some boundary conditions (such as the solar constant) or physical parameters (such as the concentration of greenhouse gases). The model focuses on each grid cell and the transfer of energy between grid cells. Once the simulation is calculated a number of climate patterns can be determined; from ocean and wind currents to patterns in precipitation and rates of evaporation rates that affect, for example, lake-levels and growth of agricultural plants. The GCMs are run in a number of specialized computer laboratories around the world. We used data in our analyses from these laboratories. Generation of future climate The Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report was based on the results of 21GCMs, data that are available through an IPCC interface, or directly from the institutions that developed each individual model. The spatial resolution of the GCM results is inappropriate for analyzing the impacts on agriculture as in almost all cases the grid cells measure more than 100 km a side. This is especially a problem in heterogeneous landscapes such as those of the Andes, where, in some places, one cell can cover the entire width of the range. 5

11 Downscaling is therefore needed to provide higher-resolution surfaces of expected future climates if the likely impacts of climate change on agriculture are to be more accurately forecasted. We used a simple downscaling method (named delta method), based on the sum of interpolated anomalies to high-resolution monthly climate surfaces from WorldClim (Hijmanset al., 2005). The method, basically, produces a smoothed (interpolated) surface of changes in climates (deltas or anomalies) and then applies this interpolated surface to the baseline climate (from WorldClim), taking into account the possible bias due to the difference in baselines. The method assumes that changes in climates are only relevant at coarse scales, and that relationships between variables are maintained towards the future (Ramirez and Jarvis, 2010). CIAT downloaded the data from the Earth System Grid (ESG) data portal and applied the downscaling method on over 19 GCMs from the IPCC Fourth Assessment Report (2007) for the emission scenario SRES-A2 and for 2 different 30 year running mean periods (i.e [2030s], [2050s]). Each dataset (SRES scenario GCM timeslice) comprises 4 variables at a monthly time-step (mean, maximum, minimum temperature, and total precipitation), on a spatial resolution of 30 arc-seconds (Ramirez and Jarvis, 2010). 5.d Crop prediction Maximum Entropy Maximum entropy (MAXENT) is a general-purpose method for making predictions or inferences from incomplete information. The idea is to estimate a target probability distribution by finding the probability distribution of maximum entropy, subject to a set of constraints that represent (one s) incomplete information about the target distribution. The information available about the target distribution often presents itself as a set of real-valued variables, called features, and the constraints are that the expected value of each feature should match its empirical average - average value for a set of sample points taken from the target distribution (Phillips et al., 2006). Similar to logistic regression, MAXENT weights each environmental variable by a constant. The probability distribution is the sum of each weighed variable divided by a scaling constant to ensure that the probability value ranges from 0 to 1. The program starts with a uniform probability distribution and iteratively alters one weight at a time to maximize the likelihood of reaching the optimum probability distribution. MAXENT is generally considered to be the most accurate model (Elith et al., 2006). MAXENT makes use of presence points of the crop, inferring distribution probabilities of the species in order to obtain a probability for the predicting zone. However, the program does not take into account irrigation. In using the presence points, MAXENT assumes that the current climatic conditions are appropriate for the development of the crop. 5.e Measure of confidence Availability of high-quality and less uncertain climate predictions is less likely at the current state of science. GCMs do not provide realistic representations of climate conditions in a particular site, but 6

12 rather provide estimated conditions for a large scale. Ramirez-Villegas and Challinor (2012) state that climate model outputs cannot be inputted directly into plot-scale agriculture models, but support the idea that higher resolution climate modelling largely improves results and can be adequately used if: (1) scales between models are matched, (2) skill of models is assed and ways to create robust model ensembles are defined, (3) uncertainty and models spread are quantified in a robust way, and (iv) decision making in the context of uncertainty is fully understood (Ramirez-Villegas and Challinor, 2012). Therefore it is very important to address uncertainty of climate prediction models used. Jarvis et al. (2012) state that impact assessment methods are sensitive to uncertainties and assessing the climateinherent uncertainty in climate change impact assessment projects explicitly entails the usage of different GCMs. To consider climate-inherent uncertainty we used 19 different GCMs in our study. Future crop suitability is predicted using each of the GCM models via MAXENT algorithms described above, we calculated the change of suitability (compared to current suitability results using climate baseline worldclim) for each GCM. As final maps to show uncertainty of MaxEnt modelling using future climate predictions we produced on pixel basis: (i) the change of the ensemble mean, (ii) the percentile rank using first quartile (25 th percentile) and third quartile (75 th percentile), and (III) the agreement among (19) climate models (GCM) calculated as percentage of models predicting changes in the same direction as the average of all models at a given location. 5.f Environmental factors driving change in suitability In order to understand the relative importance of different climatic drivers, we then carried out a forward, step-wise regression analysis with the suitability shift per data point as the dependent variable and the model-average changes in the bioclimatic variables between the present and future as the independent variables, and calculating the relative contribution of each variable to the total predicted suitability shift in terms of the proportion of R-square explained when adding each variable to the linear regression model. This analysis was carried out separately for the data points showing positive and negative shifts in suitability. 7

13 6 Result I: Climate change summary of coffee production sites 6.a The summary climate characteristics for all coffee factory sites in Tanzania Figure 2: Climate trend summary 2030 and 2050 for sample sites. In figure 2 monthly precipitation is represented by bars and monthly mean temperatures by lines for the three time points (present, 2030 and 2050). The figure illustrates a constant increase of temperature through time. The plus (+) and minus (-) signs on the bars symbolize the increase and decrease of monthly precipitation for the year 2050 with reference to the present climate according to WorldClim (climate data for the period ). Thereafter, the months between December and April are rainier while June to October is less rainy, affecting various annual crops. General climate characteristics The rainfall increases from 1277 millimeters to 1325 millimeters in 2050 passing through 1309 in 2030 Temperatures increase and the average increase is 2.3 ºC passing through an increment of 1.3 ºC in 2030 The mean daily temperature range decreases from 11 ºC to 10.4 ºC in

14 The maximum number of cumulative dry months decreases from 6 months to 5 months Extreme conditions The maximum temperature of the year increases from 27.8 ºC to 30.4 ºC while the warmest quarter gets hotter by 2.4 ºC in 2050 The minimum temperature of the year increases from 11.1 ºC to 13.6 ºC while the coldest quarter gets hotter by 2.4 ºC in 2050 The wettest month gets wetter with 308 millimeters instead of 291 millimeters, while the wettest quarter gets wetter by 36 mm in 2050 The driest month gets drier with 10 millimeters instead of 10 millimeters while the driest quarter gets drier by 2 mm in 2050 Climate Seasonality Overall this climate becomes more seasonal in terms of variability through the year in temperature and more seasonal in precipitation Variability between models The coefficient of variation of temperature predictions between models is 3.6% Temperature predictions were uniform between models and thus no outliers were detected The coefficient of variation of precipitation predictions between models is 3.3% Precipitation predictions were uniform between models and thus no outliers were detected 6.b Regional changes in the mean annual precipitation (2030) 9

15 Figure 3: Mean annual precipitation change by 2030 in the coffee-growing regions of Tanzania. The edges of the boxes indicate the mean maximum and mean minimum values and the ends of the line the maximum and minimum values. The mean maximum and mean minimum values are defined by the mean + or the standard deviation. Annual precipitation increases by 2030 on average by 32 mm and in the year 2050 it increases on average by 48 mm. By 2030 the regions that represent the highest increase in precipitation are Arusha and Manyara of approximately 45 to 50 mm on average (figure 3). In addition, an even greater difference between the regions can be observed by 2050, since annual precipitation continues to increase by approximately 20 mm. Kilimanjaro, in addition to Arusha and Manyara, are the regions with greatest precipitation increase by 2050 with an increment of almost 69 mm (Figure 4). Although the regional analysis suggest a mean annual increase in precipitation by 2030 and 2050, this does not mean that water will be available during the dry season, but rather that the seasons will be more pronounced. 6.c Regional changes in the mean annual precipitation (2050) Figure 4: Mean annual precipitation change by 2050 in the coffee-growing regions of Tanzania. The edges of the boxes indicate the mean maximum and mean minimum values and the ends of the line the maximum and minimum values. The mean maximum and mean minimum values are defined by the mean + or the standard deviation. 10

16 6.d Regional changes in the mean annual temperature (2030) Figure 5: Mean annual temperature change by 2030 in the coffee-growing regions of Tanzania. The edges of the boxes indicate the mean maximum and mean minimum values and the ends of the line the maximum and minimum values. The mean maximum and mean minimum values are defined by the mean + or the standard deviation. The mean annual temperature will increase progressively. The increase by 2050 is between 2.1 and 2.4ºC (Figures 6) and for 2030 between 1.1 and 1.4ºC (Figures 5). 6.e Regional changes in the mean annual temperature (2050) Figure 6: Mean annual temperature change by 2050 in the coffee-growing regions of Tanzania. The edges of the boxes indicate the mean maximum and mean minimum values and the ends of the line the maximum and minimum values. The mean maximum and mean minimum values are defined by the mean + or the standard deviation. 11

17 7 Result II: Suitability maps of coffee production areas 7.a Current suitability of coffee production areas Figure 7: Current suitability for Arabica coffee production in the coffee-growing regions of Tanzania. According to the MAXENT results, there are three main differing zones in Tanzania where currently a suitability range between 50 and 90% is found (figure 7). Areas with this suitability range are mainly distributed on an altitude between 900 to 2400 masl and are located in the (i) southwestern highlands of the Iringa, Mbeya, Ruvuma, and Rukwa regions; (ii) the northeast highlands close to the border with Kenya in the Arusha, Mara, Manyara regions and the lower part of Mount Kilimanjaro, where coffee is usually intercropped with banana; and (iii) towards the northwest of Tanzania in the Kigoma and Kagera regions on the border with Rwanda and Burundi, where at some places around Lake Victoria mainly Robusta is cultivated. Some zones are located in the Morogoro, Tanga, and Mwanza regions, but are generally less significant. The remaining regions are in general less suitable. 12

18 7.b Future suitability of coffee production areas Figure 8: Suitability of Arabica coffee production by By 2030 the main coffee producing regions are predicted to loose suitability by 10 to 30% (figure 8). According to the model, the majority of the areas continue to be moderately suitable for coffee production with similar climatic conditions as today. It is estimated, that in 2050 the suitability of coffee production will be mainly concentrated in the north eastern highlands of the Arusha region, in the surroundings of Mount Kilimanjaro and towards the southwest in the highlands of the Iringa and Mbeya regions (figure 9). In this zone, the most suitable areas will even expand in the highest areas, which are located approximately between 1300 and 2800 masl (figure 9). The north-western part of the coffeeproducing regions close to Lake Victoria on the other hand, is anticipated to loose suitability and in some cases even become entirely unsuitable. 13

19 Figure 9: Suitability of Arabica coffee production by

20 7.c Address uncertainty of MaxEnt output using multiple GCM Figure 10: Predicted changes in coffee suitability and breadth of climate models uncertainty by Uncertainty was calculated for the predicted suitability change, First the agreement among the mean value for each grid cell of all 19 GCMs on emission scenario A2 (business as usual) was calculated, then the same was done with the average of the first quartile of the models which can also be called the pessimistic scenario, and the average of the third quartile also stated the very optimistic scenario. This was done separately for the different prediction direction (negative, no- or positive change) (Figure 10, 11). In the majority of the coffee producing regions of Tanzania, the suitability of coffee is characterized by a decrease in It is to be noted, that in some areas, especially in some zones of the regions Arusha and Iringa some areas gain suitability (high located areas) and some do not present significant changes 15

21 (figure 10.a). In 2050, the suitability of coffee cultivation diminishes dramatically at the border with Rwanda and Burundi and in areas close by the Lake Victoria, mainly in the Kagera and Kigoma regions (figure 11.a). The mean suitability of these areas is estimated to reduce by 20 to 50%, while the current suitability is between 40 and 70%. The areas around Mount Kilimanjaro and in general areas located in the surroundings of mountains continue to present good suitability with respect to the climate of 2050, although there is a general tendency towards higher altitudes. The degree of similarity between the GCM s models prediction in the coffee zones of Tanzania for 2030 and 2050 is of 75% - 100%, with an exception of a small area in the Iringa region with a similarity degree of no more than 25% on the Sao and Mbalwe hills. Figure 11: Predicted changes in coffee suitability and breadth of climate models uncertainty by

22 Figure 12: Relation between the suitability of the coffee production areas and the altitude for the current climate (blue line) and forecasts for 2050 (red line) in Tanzania. The gray lines indicate the projection of the different GCMs. Marginal Good Very Good 17

23 Figure 13: Available area at different altitudes for marginal (20-40%), good (40-60%), and very good (60-80%) suitability. Due to progressive climate change the higher located areas become more suitable for coffee production (figure 14). The variable altitude was not used in the suitability modelling, since it is independent of all other variables. However, altitude is very much correlated with the variables related to temperature. Currently, the optimum coffee-producing zone is at an altitude between 900 and 1800 masl but until 2050 it is estimated to increase to an altitude between 1400 and 2500 masl. In comparison with today, in the year 2050 the area between 900 and 1300 masl will be most affected, while the areas located above 1800 masl will not experience significant changes (figure 14). A further reason that explains why there remain so few adequate areas is that the highest located areas which constitute the most suitable ones are limited (figure 15). The altitudinal shifts of suitable coffee growing regions in Tanzania can have a significant socioeconomic impact with farmers being forced to change their livelihoods. Furthermore, coffee production would have to compete with forestry and natural ecosystems on higher altitudes (e.g. on Kilimanjaro). (Haggar & Schepp, 2012) 18

24 Arusha Iringa Kigoma Kilimanjaro Manyara Mbeya Ruvuma Tanga 8 Result III: Environmental factors which drive the suitability of coffee The regression analysis identified primarily the bioclimatic variables related to precipitation increase and the general increasing temperature as drivers of the predicted suitability shifts. This analysis was done taking into account presence points of the crops under study. Because the presence points are distributed in different climatic areas, the regression analysis does not show good results, which is why the factors that influence the suitability change have been determined for each region. The results of the regression analysis show that the change in only one variable can result in distinct local impacts. For example, by 2030 in Arusha and Kilimanjaro the increase of the minimum temperature of the coldest month results in a low negative suitability change (on average -20%) in combination with precipitation variables. This is due to the fact that the increase in temperature, although positive for physiological plant growth, is not sufficient to affect a positive suitability change. Table 1: Contribution of different bioclimatic variables to the predicted shift in suitability decrease for Arabica coffee in Tanzania, between the present and the 2030s. The value of R 2 and the standardized coefficients β are inform on the relative contribution and importance of each independent variable with respect to a specific region. % TOTAL R² (Beta) BIOCLIMATIC VARIABLES Minimum temperature of coldest month Maximum temperature of warmest month Precipitation of wettest quarter Precipitation of driest quarter Precipitation of coldest quarter Precipitation of warmest quarter Mean temperature of driest quarter Annual precipitation Temperature annual range Precipitation of wettest month Mean temperature of warmest quarter 90 (0.71) 76 (-0.79) 24 (0.40) 68 (1.00) 32 (-0.56) 27 (1.10) 11 (0.37) 53 (0.82) 9 (0.28) 87 (-0.93) 38 (-0.16) 26 (0.39) 18 (-0.38) 56 (1.39) 20 (0.61) 18 (0.49) 87 (0.60) 8 (-0.43) 19

25 Arusha Iringa Kigoma Kilimanjaro Manyara Mbeya Ruvuma Tanga The dry period corresponds to the quiescent growth phase of the crop and is important to stimulate flowering (Da Matta et al., 2007). The level of flowering depends largely on the volume and number of heavy watering s applied during the dry season (Marsh, 2007). This is why in Iringa, characterized by one of the most distinguished dry seasons, the precipitation decrease affects the suitability of the crop negatively on average by -20%. While the precipitation increase of the coldest trimester in Mbeya and Kigoma, affects suitability negatively, this although precipitation increase could be positive for plant growth, but only if the distribution of precipitation is good. Bad rain distribution could negatively affect flowering. In Ruvuma a small precipitation increase of the coldest trimester does not affect suitability change severely. In Manyara and Tanga, on the other hand, suitability is affected by an increase in maximum and mean temperature of 1 C and 2 C, respectively. With increasing temperature, evapotranspiration also increases and may cause a water deficit for the coffee plant. Areas that are not yet irrigated may need irrigation and areas being irrigated may need an increased supply of water. Increases in temperature are also likely to reduce the length of the growing season, which could reduce crop yields (SABMiller et al., 2010). Table 2: Contribution of different bioclimatic variables to the predicted shift in suitability decrease for Arabica Coffee in Tanzania, between today and the 2050s. The value of R 2 and the standardized coefficients β inform on the relative contribution and importance of each independent variable with respect to a specific region. % TOTAL R² (Beta) BIOCLIMATIC VARIABLES Maximum temperature of warmest month Annual precipitation Minimum temperature of coldest month Precipitation of warmest quarter Precipitation of coldest quarter Precipitation of driest quarter Precipitation seasonality (coefficient of variation) Mean temperature of driest quarter Precipitation of wettest quarter Precipitation of driest month Precipitation of wettest month 83 (0.66) 30 (0.23) 18 (0.53) 21 (-0.80) 20 (-0.92) 41 (-0.55) 34 (0.99) 25 (0.66) 72 (0.743) 8 (0.21) 93 (-0.93) 45 (0.30) 13 (-0.13) 30 (-0.70) 12 (0.81) 21 (1.09) 22 (0.55) 15 (-0.14) 20 (0.60) 18 (0.67) 92 (1.01) 20

26 According to the regression analysis, the main factors that determine the change in suitability by 2050 in Manyara is the maximum temperature of the warmest month; in Kigoma and Mbeya it is the precipitation of the coldest trimester; in Iringa it is the precipitation of the driest trimester; and in Arusha and Kilimanjaro it is the minimum temperature of coldest month. They are the same variables that affect the regions by 2030, but cause a more drastic impact on suitability by increasing mean temperature by 1 C and increasing or decreasing of different precipitation variables by In Tanga the variable that most contributes to the suitability decrease of -40% by 2050 is the annual precipitation increase of 50 mm on average. In Ruvuma, on the other hand, the precipitation of the driest month affects a loss in suitability of -45% by Conclusions In Tanzania the yearly and monthly rainfall will increase by 2030 and progressively increase by In Tanzania the yearly and monthly minimum and maximum temperatures will increase by 2030 and progressively increase by The implications are that the distribution of suitability within the current coffee-growing areas in Tanzania for Arabica coffee production in general will decrease quite seriously in some regions by The optimum coffee-producing zone is currently at an altitude between 900 and 2400 masl and will by 2050 increase to an altitude between 1300 and 2800 masl. Compared with today, by 2050 areas at altitudes between 900 and 1300 masl will suffer the highest decrease in suitability and the areas around 1800 masl will not change significantly in suitability. The seasons will be more pronounced; the dry season will be drier and hotter and the rainy season will be wetter and hotter. The precipitation change of the coldest month (normally dry months) affects suitability considerably in some regions of Tanzania by 2030 and The same is true for the temperature increase of the hottest months. 10 References 21

27 Andersen, L Denmark-Vietnam sector program support. In Wastewater Reuse in Agriculture in Vietnam: Water Management, Environment and Human Health Aspects. Workshop IWMI. Da Matta, F., Ronchi, C., Claudio, P., Maestri, M. And Barros, R Ecophidiology of coffee growth and production. Braz J. Plant Physiol., vol19, n4 (cited ) pp Elith, J., Graham, C.H., Anderson, R.P., Dudík, M., Ferrier, S., Guisan, A., Hijmans, R.J., Huettmann, F., Leathwick, R., Lehmann, A., Li, J., Lohmann, L.G., Loiselle, B.A., Manion, G., Moritz, C., Nakamura, M., Nakazawa,Y., Overton, J.McC., Peterson, A.T., Phillips, J., Richardson, K., Scachetti-Pereira, R., Schapire, E., Soberon, J., Williams, S., Wisz, M., Zimmermann, E Novel methods improve prediction of species' distributions from occurrence data. Ecography 29: Haggar, J and Schepp, K Coffee and Climate Change Impacts and options for adaption in Brazil, Guatemala, Tanzania and Vietnam. Hijmans, R.J., S.E. Cameron, J.L. Parra, P.G. Jones, and A. Jarvis Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25:1965. Marsh, A Diversification by smallholder farmers: Tanzania Arabica Coffee. Agricultural Management, marketing and finance service. FAO. pp Phillips, S.J., R.P. Anderson, and R.E. Schapire Maximum entropy modeling of species geographic distributions. Ecological Modelling 190: Ramirez & Jarvis Disaggregation of Global Circulation Model Outputs. Disaggregation of Global Circulation Model Outputs. Julian Ramirez-Villegas, Andy Challinor, Assessing relevant climate data for agricultural applications, Agricultural and Forest Meteorology, Volume 161, 15 August 2012, Pages SABMiller, G.T.Z., WWF (2010) Water futures: Working together for a secure water future. Titus, A. and Pereira, G The fine art of irrigation in Arabica coffee platations. IneedCoffee. Tukey, J.W Exploratory data analysis. Addison-Wesley Publishing Co., Reading, MA. Kagubila M Irrigation Management Transfer: Experiences from Tanzania. Ikeno, J. The declining coffee economy and low population growth in Mwanga district, Tanzania. African Study Monographs, Suppl.5: -9, March