The Impacts of Current Climate Variability on Coffee Production in the Northern and Southern Highlands of Tanzania

Coffee is the most traded commodity in the world. In Tanzania, Coffee is the second largest traditional commodity. However, several climate change studies have predicted that coffee production will be reduced as a result of climate change. Therefore, the study aimed to assess the impact of current climate change on Tanzania’s Arabica coffee production and determine the most significant climatic variables, which influence coffee production in the respective regions. Global interpolated climatic database (Worldclim dataset) and official historical coffee production data from Tanzania Coffee Board for a period of 40 years (1970-2018) were used. Climatic parameters and coffee production were compared through descriptive statistics, correlation analysis, and multiple regressions. The Mann-Kendall method was used to detect significant trends in climatic data. The minimum temperature has been increasing at a higher rate than the maximum temperature in the Northern and Southern Highlands zones. A 1 °C increase in minimum temperature (Tmin) during short rains and annual mean temperature (Tmean) resulted in a significant coffee production decrease (-6,041 and -4,450 tons) in Kilimanjaro and Arusha regions respectively. In the Southern Highlands zone coffee production positively correlated with temperature. A significant reduction in coffee production due to a decline in long rains was also observed in the Kilimanjaro region. The warming and drought trends are likely to continue with significant implications on coffee production and this, calls for the development of suitable adaptation strategies to sustain production. Such strategies may include, re-adapting the coffee agronomic practices to climate change, improving water and nutrient use efficiency in coffee trees, and developing genetically improved coffee cultivars that will tolerate the impact of climate change.

to the above problems, Tanzania Coffee Research Institute (TaCRI) has developed coffee hybrid varieties that are resistant to CLR and CBD (TaCRI, 2011).
The production and productivity of both species largely depend on the climate to attain high yields and quality (Killeen & Harper, 2016). Arabica coffee grows well in an area with an optimum temperature range of 18-21 °C (Magrach & Ghazoul, 2015) and an optimum annual rainfall range of 1200-1800 mm (Alegre, 1959). Due to its narrow climatic requirements coffee crop, is expected to be the most affected by the increasing temperatures (Malyadri, 2016) and reduced rainfall (Wagner et al., 2021). According to IPCC (2014), the global climate has changed over the past century and is projected to continue changing throughout this century. Furthermore, global circulation models (GCMs) all point to higher mean temperatures and changes in precipitation regimes. Africa is one of the continent's most severely affected by climate change due to its geographical characteristics of having the majority of land lying across the warming tropics (Filho & Rao, 2014). In this continent, the temperature is projected to rise faster than the rest of the world, which could exceed 2 °C by mid-21st century and 4 °C by the end of the 21st century (Niang et al., 2014). East Africa will be increasingly affected by climate change in the coming decades, with temperatures already increasing and predicted to rise further (Adhikari et al., 2015;Craparo et al., 2015). Countries within East Africa are also experiencing reduced rainfall due to the shortening of long rains (Wainwright et al., 2019). As a part of the tropical region, Tanzania has experienced sustained warming particularly since 1970 (IPCC, 2007). Based on downscaled climate models, Tanzania is projected to experience a mean temperature increase of 2-4 °C by 2100 (IPCC, 2007;Läderach et al., 2012).
The increase in temperatures and precipitation shortages has negative impacts on coffee flowering and fruiting. However, global studies indicate that precipitation factors such as annual and seasonal precipitation are of less importance compared with temperatures in determining suitability (Ovalle-Rivera et al., 2015;Rao, 2016). In Tanzania, research has shown that coffee yields are especially affected by elevated night temperatures (Craparo et al., 2015) and droughts due to a shift in seasons (Wagner et al., 2021). It is predicted that in Tanzania every 1 °C increase in minimum temperature will result in annual yield losses of nearly 140 kg ha -1 (Craparo et al., 2015). The severity of pests and disease spread is likely to increase with advancing climate change, a significant challenge in coffee production (Jaramillo et al., 2011). According to Ovalle-Rivera et al. (2015), generally, the influence of weather variations on coffee-producing countries is predicted to be negative. This will jeopardize coffee quantity and quality hence endangering coffee producers who occupy 90 percent of the population and their livelihoods mostly rely on coffee.
Climate change projections also suggest that some areas would lose suitability for growing coffee while others would gain from temperature increases and possibly in rainfall (Ovalle-Rivera et al., 2015). According to Killeen and Harper (2016), there would be a change in coffee production areas because suitable areas will become too warm or prone to periodic drought. Furthermore, Ovalle-Rivera et al. (2015) in their study documented future global loss of Arabica coffee area by 2050 as follows: Mesoamerica (30%), South America (16-20%), and Africa (9-25%). They have also reported that Mexico from Mesoamerica and Brazil from South America would lose about 29% of its suitable Arabica coffee growing areas respectively. Pacific countries such as India and Vietnam will also experience a loss of suitability areas and be highly affected. Davis et al. (2012) proposed a substantial reduction in the area suitable for Indigenous Arabica varieties in Eastern Africa. Land suitable for Arabica coffee in East Africa is predicted to shift from 400-2000 m above sea level to 800-2500 m above sea level. Moreover, there would be a modest change in the suitability of the areas in Ethiopia, Kenya, Rwanda, and Burundi that currently grow Arabica coffee. Tanzania and Uganda would lose suitable areas at elevations below 1400 m above sea level. For the case of the Northern Highlands of Tanzania, the optimum coffee-producing zone would need to shift upwards attitudinally by 150-200 m, to sustain coffee quality and quantity (Craparo et al., 2015). However, this pushes coffee into a higher altitudinal zone that currently hosts substantial biodiversity of (mostly protected) forest species (Hemp, 2005), thus limiting upslope coffee expansion in northern Tanzania and indeed much of the coffee regions in tropical countries.
Farmers at lower elevations will no longer be able to grow quality coffee and may have to abandon it (Läderach et al. (2017). This suggests that actors along the coffee supply chain will have to adapt to the changes that climate change will bring. According to Stafford et al. (2011), farmers can adjust by making incremental adaptations and innovations based on their experiences to deal with climate variability. Incremental adaptation occurs in a short timeframe at lower altitudes, whereas the same areas may undergo transformative adaptation in the long term. At higher altitudes incremental adaption may be needed in the long term based on two adaptation strategies to two levels of climate change (incremental adaptation for lower levels of progressive climate change and transformative adaptation for higher levels of progressive climate change). Läderach et al. (2017) has developed a two-dimensional adaptation framework in time and space for coffee production in Nicaragua.
According to the author the same principle and framework are applicable across coffee-growing regions around the world, as the patterns of decreasing exposure with higher altitudes are the same globally; only the magnitude and timeframe changes. In Vietnam, Phuong Le and Nguyen (2018), identified adaptation processes more broadly as long-term strategies and analyzes temporary coping responses to drought of coffee growers.
Generally, different studies indicate that smallholder coffee farmers have been responding to climate change impacts through a range of interventions, including agronomic practices such as planting shade trees, pruning, planting drought-tolerant varieties, and application of organic fertilizers (Kajembe et al., 2016;Wagner et al., 2018;Mbwambo et al., 2021). In the study by Pham et al. (2019), adaptation and management practices were identified by more than 70% of total studies (25 papers), of which agro forestry, either through intercropping or shading, was most common (18 papers), followed by irrigation and efficient use and management of water (12 papers), development of new cultivars that are drought and heat-stress resistant and/or pest and disease tolerant (10 papers) and diversification of cropping patterns or livelihood activities (9 papers). Other measures included the relocation of coffee plantations to more bio-climatically suitable areas (6 papers), crop insurance (3 papers), off-farm livelihoods (2 papers), and shifts from Arabica to Robusta or cocoa (2 papers). According to Mbwambo et al. (2021) the adaptation measures used by smallholders coffee farmers in the Northern and Southern Highlands of Tanzania are in the order of; use of shade tree, use of mulching, use of organic manure, planting disease resistance varieties, use of cut-off drains, use of terraces and irrigation. However, according to Craparo et al. (2015), the increase in T min challenges the common notion that shade trees are always a beneficial aspect of climate change adaptation. Furthermore, responses towards adopting adaptations practices have been influenced by factors such as education level, farming experience, farm size, access to extension services, and time awareness of climate change information (Mbwambo et al., 2021).
Even though, in Tanzania, the general feeling is that the climate has been changing over the years and may be responsible for current low production and productivity (Craparo et al., 2015;Wagner et al., 2021;Mbwambo et al., 2021), this perception has largely remained anecdotal, with limited assessment covering the whole country's coffee-growing areas. Studies by Craparo et al. (2015) and Wagner et al. (2021) dealt with specific coffee-growing zones and pointed to the possible future climatic trajectories in those zones. Nevertheless, climate change studies are necessary for the formulation of climate change adaptation strategies for coffee farming in Tanzania, which does not exist yet. This study used official historical coffee production data to understand whether the general production patterns have any bearing on the historical climatic trends in major Arabica coffee growing areas in Tanzania.

Description of the Study Area
The study area is comprised of two major Arabica coffee growing zones. The Northern Highland zone involved the Kilimanjaro region (Hai, Moshi rural, Siha, and Rombo Districts) and Arusha Region (Arumeru, Longido, Monduli, and Karatu Districts). The Southern Highland zone included the Songwe region (Mbozi and Ileje districts), Mbeya (Mbeya and Rungwe districts), and Ruvuma (Mbinga, Songea, and Nyasa districts) ( Figure 1). In these zones, Arabica coffee production is exclusively rain-fed. The Northern Highlands zone is characterized by a bimodal rainfall pattern, while the Southern Highland zone experiences a unimodal rainfall pattern. jas.ccsenet. Note. T max = Maximum temperature; T min = Minimum temperature.

Data C
Descriptive statistics (frequencies and percentiles), correlation analysis, and multiple regressions were performed in STATA 13.0 (StataCorp LP, College Station, TX, USA) and SPSS 21.0 (IBM-SPSS Inc, Chicago, IL, USA) software. The rank-based nonparametric Mann-Kendall (Mann, 1945;Kendall, 1975) method was applied to the long-term climatic data to detect statistically significant trends. In this test, the null hypothesis (H 0 ) was that there has been no trend in rainfall and temperature over time; the alternative hypothesis (H 1 ) was that there has been a trend (increasing (+) or decreasing (-) over time. In the Northern Highlands zone, correlation analysis was used to examine the relationship of long rains, short rains, and annual rainfall with coffee production. T min (°C), T max (°C), and T mean (°C), both in the long rains, short rain seasons, and annually were also correlated with coffee production. In the Southern Highlands zone the climatic parameters (rainfall, T min (°C), T max (°C), and T mean (°C) were also correlated with coffee production in the growing season and annually. On the other hand, a multiple regression model was used to see the effect of independent variables (amount of rainfall and temperature) on the dependent variable (amount of coffee produced in tons). The model with the best statistical quality and highest adjusted R-squared was chosen. Regressors with higher p-values (smaller t-statistic values) were excluded one by one. If the exclusion of a regressor produced a positive change in the adjusted R squared value, it was left out and subsequently tried with the next regressor that had the highest p-value. Regressors with the highest p-values were excluded until the change in the adjusted R-squared was negative (Gay et al., 2006). Mbeya and Songwe were one region (Mbeya) before they split in 2015. Therefore, to have a 40-year coffee production data, these two regions (Mbeya and Songwe) were combined. The regression equation is represented as; Where, Y = Dependent variable (Coffee (t)); B o = Intercept; B (1-n) = Coefficients of regression line; X (1-n) Predictor variables (rainfall and temperature parameters).

Northern Highlands Zone
The average annual total rainfall for the past 40 years  in Kilimanjaro and Arusha regions was found to be 1435.77 mm and 733.4 mm respectively (Appendix A). In the Kilimanjaro region, the highest amount of rainfall was 2131.5 mm observed in 2006, which is higher than the long-term average of 1979-2018 by 695.73 mm. Nevertheless, out of 40 years, 22 years were below the average annual rainfall. In the Arusha region, the highest amount of rainfall was 1033 mm in 1988 which is higher than the long average of 1979-2018 by 299.6 mm. In addition, in the Arusha region, 23 years were below the average annual rainfall. T mean in Kilimanjaro and Arusha regions were observed to be 19.67 °C and 20.3 °C respectively (Appendix A).
Over the 40 years, T max in the Northern Highlands zone has been increasing significantly (P < 0.001) at the rate of 0.018 °C year -1 and 0.017 °C year -1 in Kilimanjaro and Arusha regions respectively (Figures 2a and 2b). Arusha and Kilimanjaro regions have also experienced a significant (P < 0.01) increase of T max during the long rain season at the rate of 0.01 °C season -1 . The short rains season, on the other hand, has been characterized by significant (P < 0.01) increases of T max (0.02 °C season -1 ) in Kilimanjaro and Arusha regions. The study has also revealed a significant (P < 0.01) increase in T min at the rate of 0.023 °C year -1 in the Kilimanjaro and Arusha regions (Figures 2c and 2d). Furthermore, a significant (P < 0.01) increase in T min during the short rain season was observed at Kilimanjaro and Arusha regions at the rate of 0.03 °C season -1 . Significant (P < 0.01) increase in T min during the long rain season was also observed in the Arusha region only at the rate of 0.02 °C season -1 . The results from the Mann-Kendall trend analysis for the Kilimanjaro region indicate statistically significant positive trends (P < 0.05) for T max in all months while for T min only two months (March and April) showed insignificant trends (P > 0.05). In the Arusha region, significant positive trends for T max were observed in most of the months except for February. For the case of T min , only two months (April and November) did not show positive significant trends. Mean warming for the Arabica growing regions of Northern Tanzania over the 40 years  has been + 0.819 °C (Kilimanjaro) and + 0.702 °C (Arusha). The box plots below (Figures 3 and 4) indicate that the highest T max was observed in the Arusha region (30.67 °C) followed by the Kilimanjaro region (30 °C) in February. Furthermore, the months of June, July, and August had the lowest T max (20-22 °C). The highest T min was observed in April in Kilimanjaro (17.4 °C) and Arusha (17.7 °C) regions while the lowest T min was in July and August (12 °C  However, only the months of August and October exhibited significant positive trends in T min (P < 0.05). In the Ruvuma region, seven months exhibited a significant upward trend in T max and T min . In this zone, October and November have been the hottest months in terms of T max (Figures 9a and 9b). The months of June and July on the other hand had the lowest T max up to 22 °C. The highest T min (17 °C) was also observed in January and February in Mbeya and Songwe regions while in the Ruvuma region the T min of above 20 °C was also observed in the same months. On the other hand, Mbeya and Songwe regions had the lowest T min (8 °C) in July followed by the Ruvuma region (11 °C) in the same month (Figures 10a and 10b).

Coffee
Based on t trend has b zone has s region was 14b). On 11,974 t re     Note. Total change is the difference between the trend line value of the first and last year.

Correlation Analysis
The relationship between the amounts of coffee (t) produced and the amount of rainfall (mm) was positively significant (P < 0.05) during the long rain season at Kilimanjaro and Arusha region. Results also showed a significant negative relationship (P < 0.05) between the amount of coffee produced and average T min long and short rain season and average T min in Kilimanjaro and Arusha regions. Furthermore, the analysis indicates a significant negative relationship (P < 0.05) between the annual average T max and the amount of coffee produced in both regions. Average T max had also a significant relationship (P < 0.05) with coffee production during short rain in the Kilimanjaro region and during long rain in the Arusha region. Pearson correlation analysis also resulted in the negative correlation between Annual mean temperature and coffee production in Arusha and Kilimanjaro regions (Table 3).

Table 3. Pearson correlation values for the amount of coffee (t) and climatic data in the Northern Highlands zone
Coffee production in Kilimanjaro Coffee production in Arusha T min short rain Pearson correlation -0.694** -0.558** T min long rain -0.472 ** -0.427** Average T min -0.700** -0.628** T max short rain -0.321* * -0.440** Average T max -0.518** ** -0.553** T mean -0.654** -0.708** Long rains 0.347* 0.320* Note. *: Significant at 0.05 level; **: Significant at 0.01 level; T min = Minimum temperature, T max = Maximum temperature, T mean = Average temperature. Table 4 indicated a significant positive relationship (P < 0.05) between average T min rain season, annual average T min , T mean , and the amount of coffee produced at Songwe-Mbeya and Ruvuma regions. The analysis also shows the positive relationship between the annual averages T max and the amount of coffee produced in the Ruvuma region. No significant correlation between coffee production and rainfall (P > 0.05).

Regression Analysis
(1) Northern Highlands Zone In the Kilimanjaro region, three independent variables were significantly predictive of coffee production according to ANOVA statistics [F (3, 37) = 21.04, P < 0.01]. The model's percent of explaining the variance in coffee production in the Kilimanjaro region was found to be 63% (R 2 = 0.63). In the regression analysis results, the absolute value of Beta indicates the order of importance of the independent variables. The variable with the highest beta value is the relatively most important independent variable. Therefore, analyzing the contributions made by the independent variables in the model, it was found that T min short rain season made the biggest contribution with the value of (Beta = 0.473). It was followed by the average Tmin and long rains respectively (Table 5). Note. B = Un standardized Beta; t = Statistical T; T min : Minimum temperature.
Based on the regression analysis results, the regression equation is represented as, Coffee (t) in Kilimanjaro region = 162367.5 + 6.79 (long rains) − 6040.78 (T min short rain) − 4761.4 (Average T min ) The regression relationship between climatic data and coffee production in the Arusha region was also highly significant and the model a good fit for the data [F (5, 32) = 9.454, P < 0.01]. Five independent variables were found to predict coffee production significantly in the region. Analyzing the relationship it was found that the model's degree of explaining the variance in the dependent variable was 58% (R 2 = 0.57.5). The contribution of the independent variables to the model was in the order of T mean > average T min > T min long rains > T max short rain > long rains. Although the contribution made by T mean was the only one significant, the contribution made by other independent variables entered the model due to the property of regression analysis, and they were found to make the smallest contributions to the model (Table 6).
(2) Southern Highlands Zone The regression relationship for Mbeya-Songwe regions was highly significant and a good fit of the data [F (2, 38) = 6.05, P < 0.01]. The regression model showed that 24% (R 2 = 0.241) of the changes in the coffee production in the Mbeya-Songwe regions are explained by the combined effect of T min rain season and average T min , Among the two independent variables, average T min made the biggest contribution to the model as compared to T min during the rainy season (Table 7).
The regression model for Ruvuma region was also highly significant and a good fit for the data [F (2, 38) = 7.52, P < 0.01]. The model showed that 28% (R 2 = 0.284) of the changes in the coffee production are explained by the combined effect of T min rain season and average T min . As observed in the Mbeya-Songwe regions, the average T min resulted in the biggest contribution as compared to T min long rains in Ruvuma region (Table 8).

Climate Variability
The decline in long rain in the Northern Highlands zone of Tanzania is characterized by the shortening of the rainy season which is caused by earlier cessation as the result of a decline of rainfall in April and May. Other studies also show a decline in the March-to-May seasonal rainfall over eastern Africa (Rowell and Booth 2015;Maidment et al., 2015;Wainwright et al., 2019), however, with drying in March, April, and May (Niang et al., 2014). Contrary to our findings, the study by Wagner et al. (2021) indicated a significant increase in the amount of rainfall during May and a reduction of rainfall in April around Mt. Kilimanjaro over the last 19 years (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019). On the other hand, according to Wainwright et al. (2019), the observed decline in Eastern African long rains is characterized by shortening of the rainy season (with late-onset and earlier cessation) rather than by a decrease in the peak daily rainfall. Different observations observed in different studies can be explained by large regional and local variability in precipitation (Dai, 2018;Macleod & Caminade, 2019), and therefore changes observed in other parts of Tanzania or East Africa do not necessarily contest with what is experienced in Arusha and Kilimanjaro regions.
The suggested link to the decrease in rainfall is the rapid warming of the Indian Ocean which causes an increase in convection and precipitation over the tropical Indian Ocean, contributing to the decrease in rainfall over the continental land surface (Lemma & Megersa, 2021). Over the last three decades rainfall has decreased by around 15% over eastern Africa, in the main growing season (March and May/June) (Williams et al., 2012). On the other hand, the increase in short rain observed in the Arusha region (Northern Highlands zone) could be the result of extreme Indian Ocean Dipole (IOD) events which affect the short rainy season from October to December (Shelleph Limbu et al., 2019). With increasing global mean temperature, the frequency of extreme positive IOD is expected to significantly increase (Cai et al., 2018). The increase in short rains in the Arusha region (Northern Highlands zone) conform with projections from different General Circulation Models (GCMs) which are broadly indicating increases in annual rainfall in Ethiopia (Niang et al., 2014), but these increases are largely due to increased rainfall in the October-December period in southern Ethiopia (McSweeney et al., 2010).
Moreover, despite the prediction that annual precipitation will increase in East Africa (Wainwright et al., 2019;Adhikari et al., 2015), our findings indicate a decrease (not significant) in annual rainfall in the coffee-growing areas of the Northern Highlands zone for the four past decades . The decrease of long rains in the Northern Highlands zone has a direct link with the decline in annual rainfall over the area. According to Liebmann et al. (2014) "long rains" season [March-May (MAM)] is a manifestation of a long-term decline in rainfall totals. In addition, there has been no such downward trend in the "short rains" [October-December (OND)], but this season has continued to exhibit large year-to-year variability, which at times has exacerbated the impact of the long rains decline. This explains why coffee farmers in the Northern Highlands zone perceived drought increase (Mbwambo et al., 2021) despite the short rain increase. This is where successful adaptation measures are critical (Wagner et al., 2019). Furthermore, coffee-growing farmers, in the Southern Highlands zone are also confronted with reduced rainfall which occurs during the growing season and annually. As in Craparo et al. (2015), the decrease in rainfall in the Southern Highlands zone was not statistically significant. Moreover, the study revealed that half of the study period has been experiencing low rainfall (rainfall below long-term annual average). However, even though there has been low rainfall in the coffee-growing area, irrigation has been used at a low scale. The study conducted by Mbwambo et al., 2021 found that over 40 years , only about 17% of coffee farmers in the Northern and 5% in the Southern Highlands zone used irrigation practices in their coffee fields. Other adaptation practices used by coffee farmers in the Northern Highlands zone were in the order of shade trees (96%) > Mulching (94%) > cut-off drains (28%) and Terraces (14%). In the southern Highlands zone, adaptation practices were in the order of shade trees (70%) > Mulching 70%) > Terraces (46%) and cut-off drains (37%) (Mbwambo et al., 2021).
The study revealed further that, there have been increasing trends in monthly, seasonal, and annual temperature in the two zones with T min increasing at a higher rate than T max . These findings are in agreement with those from the Northern part of Tanzania (Craparo et al., 2015), Ethiopia (Mekasha et al., 2014), Kenya (Omondi et al., 2014), and Uganda (Nsubuga et al., 2014) which reported that mean warming is primarily driven by substantial increases in the daily minima composition compared with daily maxima. According to Niang et al. (2014), this is an indicator of continued warming. Additionally, according to the observed results, the T mean in the Northern and Southern Highlands zone of Tanzania seems to have reached the upper limit of the mean temperature bracket (18-21 °C) suitable for coffee cultivation (Alègre, 1959).

Effect of Climate Change on Coffee Production
The decline of coffee production in the Northern Highlands zone is linked to the decrease in long rains and the increase of T min . Generally, both, long and short rains are very important in the reproductive phase of the coffee plant. On one hand, the short rains in October trigger flowering in coffee plants after the dry spell period (Jassogne et al., 2013), and on the other hand, long rainy season (March to May) if delayed and inadequate will negatively affect the expansion stage, during which rainfall is required to sustain berry development. Normally, in the Northern Highlands zone coffee crop enters the reproductive phase during the short rain season (October-December) and so the crop becomes more sensitive to temperature during this period. Generally, high night temperatures increase the rate of respiration so the assimilates which could be used for growth and yield are reduced (Nagarajan et al., 2010;Bapuji Rao et al., 2014). Drought and high temperatures during this period in the Kilimanjaro region will cause fruit abortions, increased bean defects, reduced berry growth, and acceleration of ripening, leading to a reduction in coffee yield and quality (Craparo et al., 2020;Wagner et al., 2021). The study of Craparo et al. (2015), reported similar findings that yield in the Northern Highlands zone is decreasing as the results of the increase in T min , however, in their study, there was no relationship between yield decrease and decrease in long rains as observed in this study.
Additionally, although T mean affected coffee production in the Arusha region, it was T min that accelerated the increase of T mean Therefore, the decrease in coffee production observed in the Northern Highlands zone which is experiencing reduced rainfall is aggravated by higher T min in the area. The inclusion of more shade trees might help to reduce heat stress (Kajembe et al., 2016), however, conservation of heat during the night challenges the common notion that shade trees are always a beneficial aspect of climate change adaptation (Craparo et al., 2015). Other strategies may include, re-adapting the coffee agronomic practices to climate change, use of technologies that will improve water and nutrient use efficiency in coffee trees, and developing genetically improved coffee varieties that will tolerate the impact of climate change.
On the other hand, coffee production in the Southern Highlands zone positively correlated with temperature even though T mean is already out of the optimum range. The observed positive correlation could be explained by the fact that there was no significant decrease in rainfall during the growing season (November to May) and annually. Moreover, the Southern Highlands zone is also characterized by very low T min and T max in June, July, and August. The low temperature in these months can reduce the negative impact of high temperature during the growing seasons. However, from all climatic parameters, it was only the average T min that resulted in a significant increase in coffee production in the Southern Highlands zone. Nevertheless, the findings from this study revealed further that, if T min during the growing season will continue to increase in the Southern Highlands zone, coffee production will also be affected, as the inclusion of this parameter improved the model in both regions of the Southern Highlands zone. Generally, the findings from this study conform to the perceptions of farmers in the Northern and Southern Highlands zone reported by Mbwambo et al. (2021), that reduced rainfall and/or increase in temperature have resulted in coffee production decline. Other non-climatic factors may have contributed to the increase in coffee production in the Southern Highlands zone as discussed below.

Other Factors Affecting Production
Despite the positive relationship between low coffee productions with weather-related problems, the sharp decrease in coffee production noted in the Northern Highlands zone is likely to have been magnified by factors other than climate change per se. Bureau for Agricultural Consultancy and Advisory Service (BACAS) (2005) noted that the nationalization of estate farms in the Northern Highlands zone contributed to the decline in coffee production in the zone, which used to produce 50% of total coffee in 1972/73 due to the dismal performance under primary cooperatives. Another possible factor is the farmers' disincentive to invest in coffee due to the historic price slump of 1980-2002. The slump, from an average of 5 USD lb -1 in 1980 to 0.77 USD lb -1 in 2002 (Drip Beans, 2020) caused a lot of problems to those who depend on coffee for their livelihoods, including farmers. They could barely meet the cost of production and as a result, production fell steadily, with the area under coffee also declining.
Another area for consideration is land holding per small holder family. BACAS (2005) reported that households owning less than 2 ha in the Northern zone were almost 70% of the sampled households while in the Southern zone they were only 32%, implying that land is scarcer in the northern zone, particularly so in Kilimanjaro. Mbwambo et al. (2021) also reported that the majority of the smallholder coffee farmers from the Northern Highlands zone possess farm sizes between 0.5 and 1ha, while those from the Southern Highlands zone had farm sizes between 1 and 2 ha. Due to land scarcity, coffee farmers may opt to intercrop coffee with other crops and this can reduce the number of coffee trees per area, hence low production. Also, the rate of planting new trees in the Northern Highlands zone is reported to be less than that of the Southern Highlands zone, because the aging coffee trees are owned by the elderly who are naturally risk-averse. These indicate that may be the increase in coffee production observed in the Southern Highlands zone, apart from being favored by climatic factors, has been further boosted by the replanting of new coffee trees and adoption of new, high-yielding varieties.

Conclusion and Recommendations
There has been a decline in long rains and a rise in T min which ultimately affected coffee production in the Northern Highlands zone. The Southern Highlands zone, on the other hand, has not yet suffered from the impact of climate change. Nevertheless, T min is increasing at a higher rate in the area and it may affect the production of coffee shortly. Therefore, without sufficient adaptation measures, coffee production in the Northern Highlands zone will be reduced and the famous brand of Kilimanjaro Coffee will disappear from the Market. This calls for public and private sectors to invest in climate change adaptation strategies that will better sustain this important industry and the livelihoods of millions of smallholder farmers who depend on it. The emphasize should also be given by the Arabica coffee growing region which may have already suffered yield losses due to climate change. Such strategies may include, re-adapting the coffee agronomic practices to climate change, improving water and nutrient use efficiency in coffee trees, and developing genetically improved coffee cultivars that will tolerate the impact of climate change. Note. STD = Standard deviation, T min = Minimum temperature, T max = Maximum temperature.

Copyrights
Copyright for this article is retained by the author(s), with first publication rights granted to the journal.
This is an open-access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/4.0/).