Smallholder coffee-based farmers’ perception and their adaptation strategies of climate change and variability in South-Eastern Ethiopia

ABSTRACT

Recent studies suggest that smallholder farmers’ perceptions rather than exact meteorological data strongly influence how they adapt to the changing climate. Therefore, we explored the climate change (CC) perceptions and adaptation strategies of coffee farmers in dependence on the meteorological data (1983–2020) along an elevation gradient (1,600–2,000 masl) in the Sidama region, Ethiopia. In total, 351 coffee farmers were randomly selected for household interviews, complemented with key informants (KIs), focus group discussions (FGDs), and field observations. Severity Index (SI) was computed to measure farmers’ perception of CC, followed by a Mann–Kendall test to ascertain climate trends. Weighted Average Index (WAI) was also used to rank adaptation strategies. We detected an increasing temperature and annual rainfall trend. Nevertheless, while farmers agreed on rising temperatures, they perceived rainfall reduction, contradicting the meteorological data. The highest SI was recorded for the rising temperature, followed by the uncertainty of rainfall distribution, increasing number of hot days, late-onset, and reduced amount of rainfall. The SI results with KIs and FGDs confirmed that weather events seemed more variable than in the past two to three decades and affected coffee production. As the most important CC adaptation strategies, the respondents practise agroforestry, application of compost, terrace construction, modification of farming calendar, and crop diversification. Our results also revealed that gender, education, farming experience, family size, access to agricultural and farmer-to-farmer extensions, and credit services affected adopting adaptation strategies. This study confirms that farmers’ perception is more important in shaping the applied adaptation strategies.

KEYWORDS:

• Adaptation strategies
• agroforestry
• elevation gradient
• farmers’ perception
• severity index
• Sidama

1. Introduction

East African countries including Ethiopia, Kenya, Uganda, and Rwanda are among the leading producer-exporters of high-quality highland arabica coffee (Coffee arabica L.) (Nzeyimana et al. 2013; Wang et al. 2015), which is a strategic commodity for these countries with significant contributions to foreign currency earnings. East African countries account for over 80% of Africa’s total coffee production (2018) and share 26% of the world’s coffee market (Hoebink and Ruben 2015). The livelihoods of an estimated 30 million people in smallholder households in East Africa depend directly on coffee production (Hoebink and Ruben 2015). Coffee smallholders usually produce a wide variety of annual and perennial food crops and fruit species for household consumption or income in diverse farming systems called coffee-based farming systems. However, many studies predict a future drastic reduction of areas suitable for coffee growing (Zullo et al. 2011; Schroth et al. 2015; Grüter et al. 2022; Mulinde et al. 2022), mainly caused by an increase in the mean temperature or prolonged drought period, particularly at low latitudes and altitudes (Bunn et al. 2015; Ovalle-Rivera et al. 2015; Mulinde et al. 2022). Studies indicate that the average temperature will rise between 1.8°C and 4°C by the end of the century globally (2007) and 2.7°C to 3.4°C by 2080 in Ethiopia (Tadege 2007). Such temperature changes will pose an enormous threat to coffee production and smallholder coffee-based farmers’ livelihood. Thus, CC adaptation is of the utmost importance for most major coffee producing regions (Grüter et al. 2022), notably in Ethiopia, which is the origin of the worldwide arabica coffee gene pool (Stellmacher and Grote 2011).

Coffee is the main cash crop in Ethiopia, and about 95% is produced by smallholding farmers (Tefera 2020). Ethiopia is Africa’s largest coffee producer and the world’s fifth largest exporter of arabica coffee (ICO 2015). In 2014, the country produced 398,000 tons of coffee (ICO 2016; Hirons et al. 2018) with an export value of approximately 1 billion US$ (2014; Hirons et al. 2018) and contributed about 7% to 10% of total world coffee production (Tefera and Tefera 2013). Arabica coffee dominates the total export earnings contributing 25–30% (Worku 2019). Coffee production creates 25% of the employment opportunity and 4–5% of GDP in Ethiopia (Worku 2019), supporting the livelihood of 15 million people (Hirons et al. 2018). Consequently, concerns about the impact of CC on coffee production are growing exponentially as CC will likely reduce coffee yields and quality and increase the occurrence of pests and diseases (Baca et al. 2014; Bunn et al. 2015). Grüter et al. (2022) and Mulinde et al. (2022) study revealed that CC will impact and shift growing regions of arabica coffee more than those of other plantation crops (such as banana, avocado, and cashew) because of the narrow ecological niche of arabica coffee. It becomes clear that the increasing climate variability and more frequent extreme weather events in the near future require immediate action. Understanding smallholder coffee-based farmers’ perception of such changes and their adaptive capacity is a prerequisite for successfully implementing sustainable agricultural strategies.

Empirical studies in Africa (e.g. Bryan et al. 2009; Deressa et al. 2011; Shiferaw et al. 2014; Asare-Nuamah and Botchway 2019; Mulinde et al. 2019) confirmed that smallholding farmers have already perceived the impacts of changing climate and employed adaptation strategies to cope with harsher and more unpredictable weather events. Shade trees in the coffee-based agroforestry systems (AFS) ameliorate microclimatic fluctuations and protect coffee plants from extreme weather conditions (Lin 2007). Thus, agroforestry has been recognized as a promising way to sustain coffee production under CC scenarios (Lin 2007; 2014).

While some site-specific studies attempted to analyse how Ethiopian smallholder farmers integrating annual crop and livestock adapt to CC (e.g. Bryan et al. 2009; Tesfahunegn et al. 2016; Alemayehu and Bewket 2017; Belay et al. 2017; Teklewold et al. 2019) and both perception of and adaptation strategies (e.g. Deressa et al. 2011; Ayal and Leal 2017; Berhe et al. 2017) but studies are very limited especially on perceptions of smallholding coffee producing farmers to CC and their adaptation strategies in Ethiopia. As coffee is the main cash crop in the region, the changing climate can significantly affect the income of those smallholder farmers; thus, it is crucial to know current CC perceptions and adaptation strategies. For instance, Eshetu et al. (2021) studied the determinants of smallholder coffee producers’ adaptation options to CC in southwest Ethiopia but failed to explain explicitly farmers’ perceived impact of CC and variability on coffee production. Moreover, in CC adaptation discourse, the concept of ‘one size fits all’ does not usually work, therefore, there is a need for conducting micro-level assessments(Asfaw et al. 2018). The study at the micro-level plays an immense role by providing empirical evidence of how smallholder farmers perceive and adapt to CC and variability. It also helps in designing appropriate adaptation strategies and effective policy interventions to lessen the adverse impact of changing climate and enhance smallholder farmers’ adaptive capacities. Therefore, the main objective of this study was to assess the climate change perception and its relationship with applied adaptation strategies of smallholder coffee-based farmers along elevation gradients in Sidama National Regional State, Ethiopia, which is one of the main coffee-producing regions. More specifically, our objectives were as follows: (i) to assess coffee farmers’ perceptions of climate change and the impacts on coffee production, (ii) its comparison with long-term meteorological data, and (iii) identification of suitable adaptation strategies, their biophysical and socioeconomic determinates, and barriers hindering their adoption.

2. Materials and methods

2.1. Description of study area

This study was conducted in Sidama National Regional State, Ethiopia. Sidama region has a total population of 3.4 million (2012). The region’s total area under coffee cultivation is 73,030 ha, and the total production obtained per annum is 50,433 tons, with an average yield of 0.64 tons ha⁻¹ (Tadesse et al. 2020). The study districts, Dale and Wensho, are one of the districts from Sidama National Regional State, Ethiopia (Figure 1). Dale and Wensho are found between 6°50′30′′N and 38°32′0′′E and 06°45′11′′N and 38°30′16′′E, respectively. Wensho elevation ranges from 1,750 to 2,149 masl, whereas Dale elevation ranges from 1,500 and 1,850 masl. Wensho topography and agroecology are characterized as cooler and to some extent milder compared to other districts in Sidama region (Doda 2019). Wensho district has a mean annual rainfall and temperature ranging from 1200 mm to 1600 mm (Molla and Asfaw 2014) and 18°C to 21°C, respectively (Moges et al. 2013). Dale district receives a mean annual precipitation of 858–1,600 mm and a mean annual temperature ranges from 11°C to 28.4°C (Atinafu et al. 2017).

Figure 1. Map of the study area.

In both districts, coffee and enset (Ensete ventricosum L.) are mainly integrated into traditional AFS. Enset, also called ‘false banana,’ is a herbaceous perennial crop primarily grown in southern Ethiopia, which supports the livelihoods of around 20 million people. Albizia gummifera, Cordia africana, and Millettia ferruginea are the most commonly used shade trees. Coffee production is accompanied by subsistence production of maize (Zea mays L.), wheat (Triticum aestivum L.), haricot bean (Phaseolus vulgaris L.), and soybean (Glycine max L.) together with animal husbandry. Avocado (Persea americana M.), mango (Mangifera indica L.), and banana (Musa spp.) are the main fruit species cultivated in the area both for household consumption and income generation.

2.2. Analytic framework of the study

The present study focuses on how different factors influence smallholder coffee-based farmers’ adoption of adaptation strategies in the phase of CC (Figure 2). Climate change, which manifests as rising temperature, harsh weather events, uneven rainfall distribution, and increased pests and diseases, affects coffee production and agroecosystems. Hence, adaptation strategies implemented by the farmers reduce the adverse impact of CC on coffee production. Farmer adoptions of climate adaptation strategies are not only determined by farmers’ perceptions of CC, but also demographic (gender, age, education, farming experience, family size), socioeconomic (annual family income, income from coffee production, the area under coffee production), and institutional factors (access to agricultural extension, access to credit services) (Figure 2). Finally, the framework illustrates barriers to adaptation (dot line) limit smallholder coffee-based farmers’ adaptive capacity to CC and variability and challenge coffee production (Figure 2).

Figure 2. Analytical framework of the study.

2.3. Study design and sample size

This study employed a multistage sampling technique. First, Dale and Wensho districts were selected due to their well-defined elevational gradients covering low (Dale district), mid (Dale district), and high elevations (Wensho district) while having high coffee production in Sidama region (Table 1). Consequently, a purposive sampling method was employed to select representative from the lowest administrative units in Ethiopia, which is known as peasant association (PA) from each elevation. Altogether, nine representative PAs were randomly selected, six from Dale district and three from Wensho district, corresponding to two PAs from the low (1,600 to 1,750 masl), four PAs from the mid (1,750 to 1,850 masl), and three PAs from the high elevation (1,850 to 2,000 masl). In each elevation range, we randomly selected a representative number of households engaged in coffee production for farmers’ survey. Geographical Information System (GPS) was used to ensure the interviewed households were located in each elevation range.

Table 1. Districts and elevation gradients selected for the study in Sidama National Regional State, south-eastern Ethiopia.

District	Elevation zone	Elevation (masl)	n
Dale	Low	1,600 to 1,750	86
	Mid	1,750 to 1,850	138
Wensho	High	1,850 to 2,000	127
Total			351

n: number of respondents.

The number of respondents included in the study was determined using the methodology of Yamane (1967) as follows:

(1) $\mathrm{n}=\frac{\mathrm{N}}{1+\mathrm{N}(\mathrm{e}{)}^2}$(1)

Where n represents the sample size (number of respondents), N represents the total number of households and e is the level of precision (allowable error, 8%). In total, 351 respondents were selected. The sample sizes from each of the three elevations were determined proportionally based on the total number of households. Further on, 10 KIs, who were knowledgeable farmers (n = 6) and/or coffee experts (n = 4), were also selected for the interview based on the length of their function in PAs, years of coffee farming experience, and basic knowledge of climate variability. Moreover, three focus group discussions (FGDs) (one organized in each elevation region), each encompassing 10 persons, were conducted. KIs and farmers involved in FGDs were selected by PAs administrative officials, development agents, and district coffee experts.

2.4. Data collection

2.4.1. Meteorological data

We used rainfall and temperature data (1983–2020) to evaluate their trends. Rainfall and temperature data from the nearby stations’ observations of our study sites were inadequate due to incomplete temporal coverage and characterized by very too many missing data. Hence, owing to the limited availability of long-term field-based meteorological data in the study sites, we used Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) version 2.0 (Funk et al. 2015) and the maximum (Tmax) and minimum (Tmin) temperature from the observational reanalysis hybrid (Chaney et al. 2014). These are the most accurate data products for east Africa. The CHIRPS data is a 30+ year quasi-global rainfall dataset incorporating 0.05° resolution satellite imagery with in-situ station data. For the temperature, we used time-series temperature data from ERA5 monthly aggregates. Of these, the 2 m air temperature was used in this study. Both the rainfall and temperature data for the study area were extracted in the Google Earth Engine environment (Gorelick et al. 2017).

2.4.2. Farmers’ survey

We used a mixed-method approach, comprising both quantitative and qualitative methods, allowing us to ask a wider range of research questions and collect the essential information (Creswell 2014). Primary data were collected using Participatory Rural Appraisal (PRA) tools, including KIs, FGDs, semi-structured questionnaires (Table S1), and field observations. Checklist questions were prepared to ensure each KI had equal opportunities to provide consistent and accurate information. The major discussion topics for KIs were perception of CC, the impact of CC on coffee production and sources of weather information. FGDs were conducted to get in-depth information and insight about trends of CC and variability, the impact of CC on coffee production, major adaptation strategies, and factors hindering CC adaptation. FGDs were also used to complement the responses acquired using the questionnaire.

A detailed household survey was administered between September and December 2020 in the selected households (n = 351). Socio-demographic and biophysical characteristics (Table S2), farm-related information, perception of CC, perceived impacts of CC and variability on coffee production, adaptation strategies, determinants of adaptation and adaptation barrier data were collected using a semi-structured questionnaire. Data from the selected respondent households were collected using 5-point Likert scale typology questions, applied similarly to Masud et al. (2017) and Hasan and Kumar (2019). To ensure the validity of the obtained information, field observations were conducted throughout the whole course of the research works.

2.5. Data analysis

2.5.1. Meteorological data analysis

Meteorological data was analysed using R-software. Mann–Kendall test was used to determine whether climate trend exists in time series data, using rainfall and temperature as proxies (Chepkoech et al. 2018). Mann–Kendall trend test is a non-parametric test used to identify trends in a series (Alemu and Dioha 2020) and is less affected by outliers (Salmi et al. 2002). It is also commonly employed to detect monotonic trends in a series of environmental or climate data (Alemu and Dioha 2020).

2.5.2. Severity index (SI) calculation

Severity Index was calculated following Al-Hammad and Assaf (1996), Longe et al. (2009) and Masud et al. (2017) to measure farmers’ perceptions of climate change as follows:

(2) $\mathrm{S}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x},\left(SI\right)=\left(\frac{{\sum }_{i=0}^4{a}_i{x}_i}{{\sum }_{i=0}^4{x}_i}\right)\left(100\mathrm{\%}\right)$(2)

Where:

ai = the index of a class and a constant expressing the-weight given to the class

x_i = frequency of responses

i = 0, 1, 2, 3, 4 and described as - x₀, x₁, x₂, x₃, x₄, are the frequencies of response corresponding to a₀ = 0, a₁ = 1, a₂ = 2, a₃ = 3, a₄ = 4, respectively.

The rating classifications are described as:

a₀ = Strongly Disagree 0.0 < SI < 12.5

a₁ = Disagree 12.5 < SI < 37.5

a₂ = Neutral 37.5 < SI < 62.5

a₃ = Agree 62.5 < SI < 87.5

a₄ = Strongly Agree 87.5 < SI < 100

Based on a 5-point Likert scale, the scores administered to the responses of surveyed households are as follows: strongly disagree (0), disagree (1), neutral (2), agree (3), and strongly agree (4). To simplify the interpretation, each rating is given the following connotation: Strongly Disagree (SDA), Disagree (DSA), Neutral (N), Agree (A), and Strongly Agree (SA).

2.5.3. Weighted average index (WAI) calculation

Weighted Average Index (WAI) was used to rank farmers adaptation strategies and factors hindering CC adaptation. WAI was estimated using Eq. (3) as employed by other studies (Fagariba et al. 2018; Williams et al. 2019).

(3) $\mathrm{W}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{A}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{g}\mathrm{e}\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x},(WAI)=\frac{\sum FiWi}{\sum Fi}$(3)

where F is the frequency of each assessed adaptation response/barriers variables, W is the weight of each score and i is the score.

2.6. Statistical analysis

One-way ANOVA was applied to evaluate the association and differences between the three elevations over different attributes. The 5-point Likert scale used to measure perception and adaptation strategies was aggregated into a continuous variable for the purpose of inferential analysis (ANOVA) (Asare-Nuamah and Botchway 2019). The perception of farmers’ climate variability and its impact on coffee production were evaluated using aggregated mean scores of their response to multidimensional indicators of climate variability (Ayal and Leal 2017) (Table 3). A multiple regression analysis was also conducted to identify determinants of smallholding farmers’ perception on CC and variability and their adaptation strategies. Finally, the qualitative data collected through KIs, FGDs and personal observation were analyzed through qualitative descriptions, narrations, and thematic analysis.

Table 2. Mann–Kendall test for temperature and rainfall (1983–2020) in the study sites.

Elevation zone	Variables	Year	Mann–Kendall test
			Kendall’s Tau	p-value
Low elevation	Annual rainfall	1983–2020	0.226	0.05
Mid elevation	Annual rainfall	1983–2020	0.314	0.006
High elevation	Annual rainfall	1983–2020	0.275	0.02
Low elevation	Tmax	1983–2020	0.447	<0.001
Mid elevation	Tmax	1983–2020	0.458	<0.001
High elevation	Tmax	1983–2020	0.376	0.001
Low elevation	Tmin	1983–2020	0.417	<0.001
Mid elevation	Tmin	1983–2020	0.438	<0.001
High elevation	Tmin	1983–2020	0.405	<0.001

The Kendall’s Tau is a number between −1 and +1 with positive values indicating an increasing trend and negative values indicating a decreasing trend.

Table 3. Responses of surveyed households on climate change and variability indicators along elevation gradients in south-eastern Ethiopia (n = 351).

Variables		Likert scale					Elevation zone (aggregated SI%)				Elevation zone (mean on Likert scale)				F	p-value
		SDA (0)	DSA (1)	N (2)	A (3)	SA (4)	High (n = 127)	Mid (n = 138)	Low (n = 86)	Total (n = 351)	High (n = 127)	Mid (n = 138)	Low (n = 86)	Total (n = 351)
Rising temperature	NRS	2	8	6	178	157	74.02	88.04	93.02	84.19	2.96	3.52	3.72	3.37	46.72	<0.001
	PRS	0.6	2.3	1.7	50.7	44.5
Number of hot days increased	NRS	1	10	7	247	86	72.64	78.44	89.24	78.99	2.91	3.14	3.57	3.16	35.85	<0.001
	PRS	0.3	2.8	2.0	70.4	24.5
Number of warm nights increased	NRS	30	63	27	206	25	54.92	60.33	64.83	59.47	2.20	2.41	2.59	2.38	3.36	0.036
	PRS	8.5	17.9	7.7	58.7	7.1
Increased the coldness of cold seasons	NRS	102	186	60	3	0	17.52	21.38	31.40	22.44	0.70	0.86	1.26	0.90	18.30	<0.001
	PRS	29.1	53.0	17.1	0.9	0.0
Increased rainfall	NRS	170	133	23	25	0	30.91	5.80	18.90	18.09	1.24	0.23	0.76	0.72	58.33	<0.001
	PRS	48.4	37.9	6.6	7.1	0.0
Decreased rainfall	NRS	19	25	8	193	106	55.51	83.33	87.79	74.36	2.22	3.33	3.51	2.97	73.98	<0.001
	PRS	5.4	7.1	2.3	55.0	30.2
Uncertainty of the rainfall distribution	NRS	5	20	5	192	129	71.46	84.42	85.17	79.91	2.86	3.38	3.41	3.20	17.84	<0.001
	PRS	1.4	5.7	1.4	54.7	36.8
Cessation of rainfall becomes unpredictable	NRS	9	37	14	202	89	64.96	79.71	74.71	73.15	2.60	3.19	2.99	2.93	13.32	<0.001
	PRS	2.6	10.5	4.0	57.5	25.4
Late onset of rainy season	NRS	11	46	8	156	130	54.92	87.50	83.72	74.79	2.20	3.50	3.35	2.99	75.05	<0.001
	PRS	3.1	13.1	2.3	44.4	37.0
Total grand mean											2.21	2.62	2.80	2.51
F statistics															56.68	<0.001

NRS, PRS, SDA, DSA, N, A, SA, and SI indicate the Number of Respondents, Percentage of Respondents, Strongly Disagree, Disagree, Neutral, Agree, Strongly Agree, and Severity Index, respectively.

3. Results

3.1. Trends of rainfall and temperature

We observed an increasing trend of historical annual rainfall data (1983–2020) in all three studied elevations (Figure S1). Likewise, an increasing trend was also detected in both Tmax and Tmin across the whole study sites (Table 2).

3.2. Farmers’ perception of climate change and variability

In total, 97.7% of the respondent households perceived the impacts of CC in the last 30 years (Table 3). Based on the recorded SI and the aggregated mean scores, the five most strongly perceived features of CC indicators were as follows: rising temperature, the uncertainty of the rainfall distribution, an increasing number of hot days, late onset of the rainy season, and reduced amount of rainfall. The SI values of the majority of CC indicators ranged between 73.15% and 84.19% (Table 3).

The perception of smallholding farmers differed (p < 0.05) among the three elevation zones (F_{2, 348} = 56.68; p < 0.001) and decreased with increasing elevation. The SI results with KIs and FGDs confirmed that weather events seemed to be more variable and less predictable compared to the past three decades, particularly in low elevations. For instance, the SI value for rising temperature is higher for low elevation (SI = 93.02%), followed by mid (SI = 88.04%) and high (SI = 74.02%) elevations (Table 3). Of the surveyed households (n = 351), 50.7% agreed, and 44.5% strongly agreed with rising temperature. Also, the result showed 54.7% agreed and 36.5% strongly agreed with the uncertainty of the rainfall distribution. Further FGDs on CC and variability revealed that farmers were concerned with the frequency and severity of extreme weather and significant changes that they perceived in weather patterns. Moreover, farmers emphasized the difficulties in recognizing the start of rainy seasons, which is critical for planting new coffee plants and other crops.

3.3. Perceived impacts of climate change and variability on coffee production

The perceived changing climate was reflected in the observed impacts on coffee production. Based on the mean scores and SI results, the five most reported impacts of CC on coffee production in the order of importance were loss of coffee berries (falling of coffee fruit), late-ripening, higher incidence of coffee pests and diseases, decreased coffee yield and death of coffee plants (planted young seedlings and saplings) (Table 4). The perceived impacts of CC on coffee production differed among farmers in different elevations (F_{2, 348} = 346.76; p < 0.001) and CC was perceived to have a more substantial impact on coffee production in low elevation, followed by mid, and, however, nearly no impact in high elevation (Table 4). The aggregated SI values of climate indicators were higher for lower elevations, followed by mid and high elevations. Of the respondent households (n = 351), the farmers agreed with decreased coffee yield (27.6%), late ripening (44.2%), loss of berries (43.3%), and increased coffee pests and diseases (52.7%) (Table 4). The results of FGDs and KIs interviews corroborated findings obtained from the questionnaires. Farmers who participated in FGDs highlighted that the coffee producers in the three elevations already experienced the impacts of CC. The KIs confirmed that rising temperatures and erratic rainfall distribution affect coffee yield. Informants also stated that coffee production depends on optimal rainfall distribution. While the rising temperatures and rainfall uncertainty have already started to reduce coffee yields in the low elevation, the KIs from high elevation stated that the areas under coffee production were increasing but also shifting to higher elevation regions, where farmers did not grow coffee 30 years ago.

Table 4. Responses of surveyed households on the perceived impact of climate change and variability on coffee production along elevation gradients in south-eastern Ethiopia (n = 351).

Items	Likert scale						Elevation zone (aggregated SI%)				Elevation zone (mean score on Likert scale)				F	p-value
		SDA (0)	DSA (1)	N (2)	A (3)	SA (4)	High (n = 127)	Mid (n = 138)	Low (n = 86)	Total (n = 351)	High (n = 127)	Mid (n = 138)	Low (n = 86)	Total (n = 351)
Decreased yield	NRS	35	130	21	97	68	27.76	62.86	71.80	52.35	1.11	2.51	2.87	2.09	79.43	<0.001
	PRS	10.0	37.0	6.0	27.6	19.4
Late ripening	NRS	22	91	6	155	77	31.89	79.53	79.94	62.39	1.28	3.18	3.20	2.50	198.69	<0.001
	PRS	6.3	25.9	1.7	44.2	21.9
Loss of berries	NRS	18	78	0	152	103	35.63	84.42	86.92	67.38	1.43	3.38	3.48	2.70	250.96	<0.001
	PRS	5.1	22.2	0.0	43.3	29.3
Death of coffee tree	NRS	58	139	51	95	8	21.06	46.01	57.27	39.74	0.84	1.84	2.29	1.59	66.87	<0.001
	PRS	16.5	39.6	14.5	27.1	2.3
Increased coffee pest and diseases	NRS	15	92	14	185	45	38.98	72.64	74.42	60.90	1.56	2.91	2.98	2.44	89.79	<0.001
	PRS	4.3	26.2	4.0	52.7	12.8
Decreased coffee bush density	NRS	97	142	14	77	21	12.20	40.58	57.85	34.54	0.49	1.62	2.31	1.38	85.99	<0.001
	PRS	27.6	40.5	4.0	21.9	6.0

NRS, PRS, SDA, DSA, N, A, SA, and SI indicate the Number of Respondents, Percentage of Respondents, Strongly Disagree, Disagree, Neutral, Agree, Strongly Agree, and Severity Index, respectively.

3.4. Farmers’ adaptation strategies

The adoption of agroforestry/tree planting (WAI = 3.30), application of organic manure/compost (WAI = 3.12), soil conservation (terrace construction) (WAI = 2.82), modification of farming calendar (WAI = 2.46), and crop diversification (WAI = 2.38) were the five most important adaptation practices implemented by the farmers to overcome the impacts of CC across the three studied elevations (Table 5). The households moderately practised other adaptation practices such as change of crop varieties, growing drought-resistant crops, fodder tree planting, mulching, and water harvesting. Migration, application of inorganic fertilizer and insecticides, and irrigation were positioned as the least common adaptation practices employed by the farmers in the study region (Table 5). No difference was detected in the adaptation strategies applied by farmers in different elevations (F_2.84; p = 0.07) (Table 5). In the low and mid-elevations, organic manure/compost application ranked in the second position (after AFS) while for high elevation area in the third place. In the low elevation, mulching and replacing coffee with growing drought-resistant crops such as Khat (Catha edulis Forsk) and eucalyptus species (Eucalyptus spp.) (Figure 3) were the main adaptation practices. Terrace construction is the most important adaptation practice reported for mid and high elevations.

Figure 3. The replacement of coffee with Khat (Catha edulis Forsk) (a) and Eucalyptus spp. (b) in low elevation areas.

Table 5. Adaptation strategies to climate change adopted by surveyed households along elevation gradients in south-eastern Ethiopia (n = 351).

Adaptation strategies	Total (n = 351)		Elevation zone (WAI)
			High (n = 127)		Mid (n = 138)		Low (n = 86)
	WAI	Rank	WAI	Rank	WAI	Rank	WAI	Rank
Agroforestry/tree planting	3.30	1^st	3.00	1^st	3.46	1^st	3.48	1^st
Application of organic manure/compost	3.12	2^nd	2.71	3^rd	3.32	2^nd	3.41	2^nd
Soil conservation (terrace construction)	2.82	3^rd	2.91	2^nd	2.73	4^th	2.85	7^th
Modification of farming calendar	2.46	4^th	1.76	7^th	2.78	3^rd	2.88	6^th
Crop diversification	2.38	5^th	1.89	5^th	2.49	6^th	2.93	5^th
Change crop varieties	2.36	6^th	2.22	4^th	2.28	7^th	2.71	8^th
Growing drought resistant crop varieties	2.23	7^th	1.32	9^th	2.54	5^th	3.08	4^th
Fodder tree planting	2.11	8^th	1.87	6^th	2.01	9^th	2.64	9^th
Mulching	1.97	9^th	1.45	8^th	2.06	8^th	3.31	3^rd
Water harvesting	1.31	10^th	1.29	11^th	0.96	11^th	1.90	10^th
Selling of livestock	1.15	11^th	0.54	12^th	1.44	10^th	1.57	12^th
Migration	0.85	12^th	0.40	15^th	0.93	12^th	1.41	13^th
Application of inorganic fertilizer	0.77	13^th	0.53	13^th	0.28	15^th	1.90	10^th
Application insecticides	0.72	14^th	0.44	14^th	0.43	13^th	1.60	11^th
Use irrigation	0.69	15^th	1.32	9^th	0.35	14^th	0.30	14^th
F test statistics
N	15
Sum of Squares	5.196
F statistics	2.84
p-value	0.07

WAI = Weighted Average Index.

3.5. Determinants of perceptions and adaptation to climate change

The studied determinants altogether explained 50% and 67% of the variation of smallholding farmers’ perceptions and adaptation strategies to CC, respectively (Table 6). The results revealed that elevation, education, farming experience, membership in the coffee cooperatives, radio ownership, access to agricultural extension, and access to weather information significantly and positively impacted perceptions (p < 0.05), while gender and family size significantly and negatively affected the perception of CC, respectively (p < 0.05).

Table 6. Determinants affecting the farmers’ perception of climate change and the adoption of adaptation strategies.

Variables	Perception of CC		Adaptation strategies
	Coefficient	P-value	Coefficient	p-value
Elevation	0.122	<0.001	0.251	<0.001
Gender	−0.154	<0.001	0.193	<0.001
Education	0.032	0.043	0.050	0.001
Farming experience	0.062	0.050	0.086	0.004
Family Size	−0.022	0.042	0.029	0.004
Household head age	0.000	0.994	0.022	0.590
Membership in coffee cooperatives	0.168	<0.001	−0.021	0.538
Radio ownership	0.139	0.001	0.063	0.094
Area under coffee production	−0.149	0.139	0.321	0.001
Average annual income	−1.685E-007	0.920	2.498E-006	0.113
Average annual income of coffee	−2.347E-007	0.293	−4.729E-006	0.024
Access to agricultural extension	0.221	<0.001	0.169	<0.001
Access to farmer-to-farmer extension	0.047	0.163	0.111	<0.001
Access to credit service and sources	−0.026	0.467	0.151	<0.001
Access to weather information	0.097	0.007	0.053	0.115
Constant	2.771	<0.001	1.171	<0.001
Adjusted R^2	0.50		0.67
F statistics	22.670	<0.001	47.288	<0.001

The results revealed that elevation, gender, education, farming experience, family size, the area under coffee production, access to agricultural extension, access to farmer-to-farm extension and access to credit services and sources (p < 0.05) significantly and positively affected the adoption of adaptation strategies. The average annual income from coffee production (p = 0.02) also affected the adoption of adaptation strategies in the study area.

3.6. Barriers of climate change adaptations

Based on the WAI, the top five recorded barriers to CC adaptation across the three studied elevations included poor soil fertility (WAI = 2.93), land shortage (WAI = 2.78), lack of weather information (WAI = 2.61), lack of credit (WAI = 2.60), and lack of water (WAI = 2.39) (Table 7). Lack of agricultural extension services and tree seedlings were reported as minor challenges in implementing adaptation strategies. There were common challenges, but slightly different levels of rankings were found across the three elevations (Table 7). In the low and mid-elevations, higher importance was placed on poor soil fertility, lack of weather information and credit services. In the high elevations, farmers stated mainly land shortage, poor soil fertility, and lack of agricultural labour (in decreasing order) were the significant challenges.

Table 7. The barriers of climate change adaptation strategies adoption by surveyed households.

Barriers	Total (n =351)		Elevation zone (WAI)
			High (n =127)		Mid (n =138)		Low (n = 86)
	WAI	Rank	WAI	Rank	WAI	Rank	WAI	Rank
Poor soil fertility	2.93	1^st	2.43	2^nd	3.21	1^st	3.21	1^st
Shortage of land	2.78	2^nd	2.46	1^st	3.01	4^th	2.87	4^th
Lack of weather information	2.61	3^rd	1.77	6^th	3.12	2^nd	3.05	2^nd
Lack of credit/money	2.60	4^th	1.87	4^th	3.08	3^rd	2.94	3^rd
Lack of water	2.39	5^th	1.72	7^th	2.50	5^th	3.21	1^st
Lack of agricultural labour	2.18	6^th	2.02	3^rd	2.25	6^th	2.30	5^th
Shortage of farm inputs	1.97	7^th	1.86	5^th	1.96	7^th	2.15	6^th
Education level	1.62	8^th	1.42	8^th	1.57	9^th	1.90	7^th
Lack of tree seedlings	1.46	9^th	1.08	9^th	1.59	8^th	1.83	8^th
Lack of agricultural extension services	1.04	10^th	0.91	10^th	0.68	10^th	1.80	9^th
F test statistics
N	10
Sum of squares	3.144
F statistics	3.579
p-value	0.04

WAI = Weighted Average Index.

4. Discussion

4.1. Smallholding farmers perception of climate change

The 37 years of meteorological data confirmed the rising temperatures and also increasing annual rainfall, which were partially perceived by the coffee farmers, who reported rising temperatures, but rainfall reduction. Our findings are in line with several studies in other parts of the tropics (Isa et al. 2005; Masud et al. 2017). Nearly all farmers agreed that rising temperature, increased number of hot days, and decreased rainfall were the main manifestations of CC and variability. These are consistent with existing literature that reported rising trends in temperature (e.g. Deressa et al. 2011; Ayal and Leal 2017; Asfaw et al. 2018) and an increase in the number of hot days and warm nights (Ayal and Leal 2017) in Ethiopia. Similarly, our study coincided with other studies that reported decrease in duration and amount (e.g. Zampaligré et al. 2014; Abid et al. 2015; Chepkoech et al. 2018), unpredictability (Berhe et al. 2017; Mulinde et al. 2019) and uneven distribution (Tesfahunegn et al. 2016) of rainfall in Ethiopia and elsewhere in the tropics. Similar results were also reported from other parts of Ethiopia (Meze-Hausken 2004; Bewket 2012), where the authors attributed the perception of declining rainfall to the increasing variability and unpredictability of extreme weather events. Thus, although the farmers’ perception of rainfall trends is not associated with an overall rainfall reduction, it is likely based on the lack of rainfall during crucial periods of coffee berries development (Speranza 2010). The perception of lower rainfall could also be explained by higher evapotranspiration rates resulting from rising temperatures (Slegers 2008), which also explains why the impacts of changing climate were perceived more by the farmers in the lowlands in comparison to mid- and high-elevation farmers (Table 3). A similar assumption was also asserted by Belay et al. (2005), observing a higher frequency of drought periods in the lowlands than in other areas of Ethiopia.

The result from KIs and FGDs indicates that farmers were keenly concerned about rising temperature and erratic rainfall condition and their effects on farming activities and livelihoods. Recalling the start of the rainy season was one of the bottlenecks for the farmers in the study sites. This agrees with Johansson et al. (2019) and Asare-Nuamah and Botchway (2019) studies that rainfall variability was one of the most perceived impacts of CC who depend on subsistence rain-fed agriculture in East and West Africa, respectively. The KIs and farmers during FGDs also asserted that CC impacted coffee’s sustainable production and lowered their incomes. Our field observation confirmed that the impacts of CC on coffee production and other agricultural activities were manifested more in the low elevation than in the mid and high elevations owing to high rainfall variability and loss of soil moisture in the former. Similarly, PdS et al. (2018) noted that excessive heat in warmer areas makes it unsuitable for growing coffee and causes yield reduction.

On the other hand, the KIs in the higher elevation stated an increasing rainfall trend, which might be linked with the steep topographic nature of the areas and more intense rainfall events that often resulted in strong erosion and enhanced farmers’ perception of rainfall (Deressa et al. 2011).

4.2. Climate change adaptation strategies

The adoption of agroforestry is the most common adaptation strategy among coffee-based farmers to cope with the changing climate (Ruiz-Meza 2015; Eshetu et al. 2021) due to the positive effect of shade trees on microclimate, soil fertility, and production of diversification, and likely because of its historical and cultural importance in the tropical areas. Farmers integrate commercial crop and fruit species such banana or avocado or timber shade trees (e.g. Cordia africana), similarly to farmers in Mexico (Ruiz-Meza 2015) or Guatemala (Jassogne et al. 2013). The KIs also confirmed that shade trees reduce the high intensity of direct sunlight, reducing day air temperature, maintain soil fertility, and help farmers to diversify income. The tree species in the agroforestry system also increase the resilience of coffee farming systems and buffer risks arising from CC and variability. Besides the positive effect of shade trees on coffee production, shade tree leaves and litter (along with other organic materials) are commonly used by farmers in the study area for compost preparation. In soil, compost improves water retention capacity, increases soil fertility and crop resilience to drought (2021), while enabling farmers to certify and market their coffee as organic, further increasing income. Similar practices have been observed among Ghanaian farmers where compost application is a common CC adaptation strategy in horticulture production (Fagariba et al. 2018).

Moreover, as a reaction to irregular distribution and often more intense rainfall, farmers in high elevations more frequently opt for terrace construction to reduce water runoff and erosion. Nevertheless, despite the enhanced water storage, the erratic nature of the rainfall has already forced the farmers to modify the farming calendar in the study area and elsewhere (Asfaw et al. 2018).

Mulching is another common CC adaptation strategy during high temperature and drought periods. In the study area, farmers use mulch materials such as cut grass, weed, crop residues, and tree leaves. Mulch reduces soil moisture evaporation (Jiménez et al. 2017) and improves the topsoil’s soil structure and biological activity (Zhao et al. 2017), while reducing the labour requirements compared to compost preparation. Farmers also shift towards more drought- and disease-resistant coffee varieties (Model −71110 and 71,112) with better performance under changing climate or even replace coffee with drought-resistant perennial crops such as khat. Khat is commonly grown in monoculture for its economically important leaves and tender twigs, which are chewed for their stimulating effect. Farmers claim to currently obtain better income from khat than coffee, especially in the lower elevations of the Sidama region (Mellisse et al. 2018), which is alarming given the Sidama region being one of the world’s most important arabica coffee cultivation areas.

Moreover, the most important barriers hampering CC adaptation strategies by farmers in the study area are poor soil fertility, shortage of land especially in higher elevation, lack of weather information, lack of credit, and lack of water, which were reported to be the key barriers also elsewhere (Bryan et al. 2009). According to the farmers, poor soil fertility (or more precisely poor soil health) resulting from inadequate agricultural practices is related to the low capacity of soils to cope with changing climate. Healthy soils are capable of withstanding the increasing temperature because of their ability to hold water and regulate soil temperature (Lal 2016). However, the inadequate farming practices have resulted in soil degradation linked with deterioration of soil fertility and crop performance. Compost application is often not sufficient to increase soil health due to the labour intensiveness.

Shortage of land is also one of the bottlenecks for smallholder farmers to adapt to CC. The average landholding (0.86 ha) in the study area supports the average family size of around seven people. Feeding the large family size forced the farmers to cultivate a small plot of land from year to year and thus, reduced land productivity. Hence, for such a large family size, land availability is an important agricultural asset to diversify more products and include improved crop varieties to minimise risks related to CC. Similar observations have also been made by Bryan et al. (2009) and Abid et al. (2015), reporting that rural farmers with large land size could produce more and use improved crop varieties, which enhances their adaptation capacity to CC. Ultimately, farmers pointed out that they rely on their own perception to adapt to CC due to inaccessibility and less trust in weather forecasts. Thus, integrating farmers’ knowledge in CC perception and developing trust among farmers about weather information from meteorological agencies could improve the likelihood of implementing different adaptation options. Eshetu et al. (2021) also stated that regular access to weather information is more likely to change cropping time to adapt to the changing climate. Additionally, lack of credit access clearly impedes farmers from getting the necessary technologies (irrigation, water harvesting technologies, and others) and resources to adapt to CC in the study area and elsewhere (Bryan et al. 2009; Asfaw et al. 2018; Williams et al. 2019).

4.3. Determinants of perception of climate change and variability

Our result indicated that female household heads perceived more the impact of CC than male household heads because they are more concerned about environmental issues that threaten their families and the surrounding communities. This study coincides with Safi et al. (2012) and Ayal and Leal (2017), who stated that female-headed households are perceived more than male-headed households because they are more affected by the changing environment. The result showed that educated farmers perceived the impact of CC more than less-educated farmers because they are more aware of and know the adverse effect of CC and variability on farming activities.

Our results also showed older people perceive the impact of CC more than younger people. This is in line with Maddison (2007) and Ayal and Leal (2017), who reported that older farmers perceived the impact of CC more because older farmers understand their environment in time horizon, enabling them to perceive CC easily. On the contrary, studies argued that younger people are more insightful of CC and variability in their localities (Semenza et al. 2012). Farmers with more farming experience perceived the impact of CC more than low-experienced farmers. This is in line with Maddison (2007) and Mbwambo et al. (2021), who observed that farmers with more farming experience were more likely to have stronger perceptions of CC than farmers with lower farming experience. Moreover, those farmers who participated in the formal institutions, owning radio and access to agricultural extension, farmer-to-farmer extension, credit services, and weather information, perceived more CC and variability than other farmers. This is a similar finding to Deressa et al. (2011) and Ayal and Leal (2017).

4.4. Determinants of climate change adaptations

The result revealed that biophysical and socio-demographic factors influence CC adaptation practices. More educated farmers adopt more adaptation strategies than less educated ones because they are more likely to accept new ideas and technologies to improve their farming systems. Our results concur with the findings of Abid et al. (2015) and Masud et al. (2017) from Pakistan and Malaysia, respectively. Similarly, farmers with longer farming experience have been observed to be more capable of identifying and reacting to fluctuations in climate and optimizing adaptation decisions (Arunrat et al. 2017). Our results also agree with Deressa et al. (2011), demonstrating that larger families are usually associated with a higher labour force allowing for the implementation of various adaptation practices. Likewise, the expansion of the coffee production area, participation in coffee cooperatives, access to agricultural and farmer-to-farmer extensions, and credit services positively influence the adaptation strategies similarly to other studies in Ethiopia (Asfaw et al. 2018).

4.5. Future sustainable management and policy interventions

For rainfed coffee-producing smallholder farmers, CC affects their coffee production and forces the farmers to replace coffee with other crops suited to the changing environment. In our study area, smallholder coffee producer farmers have attempted to shift traditional coffee-based agroforestry systems to cash crop khat-based farming systems, strongly relying on the input of inorganic fertilizers and pesticides. Studies in Ethiopia showed that coffee production is now being abandoned and replaced by drought-resistant cash crop khat, mostly grown as monocultures (Gebissa 2008). A study by Jara et al. (2017) also disclosed that coffee fields were abandoned and replaced by khat in the eastern and south-eastern parts of Ethiopia. The farmers in our study region have started reducing shade tree species and coffee bush density to plant khat. This concurs with Dessie and Kinlund (2008) finding that the expansion of khat on the farmlands has contributed to the removal of on-farm trees and the conversion of forestland to khat in Wondo Genet, southern Ethiopia. The authors further stated that khat production contributed to forest decline.

Our field observation confirmed that the expansion of khat significantly reduces shade trees and coffee bush density. Jara et al. (2017) also revealed that the major shift from coffee to khat eroded much of the woody species’ diversity since khat is usually grown without shade. Also, the land conversions reduce the genetic resources of arabica coffee in the study region. Arabica coffee is grown in specific climatic and biophysical conditions coupled with narrow genetic diversity (Chemura et al. 2021). Hence, there is an urgent need to identify and develop appropriate adaptive interventions in the low-elevation areas of the study region. Policy-driven actions are crucial to facilitating farmers’ long-term credit to implement improved water harvesting technologies and promoting irrigation to support farmers coping with CC. Also, the government should invest in and promote agricultural extension services and research in developing, testing, and using more drought-, pest-, and disease-resistant coffee varieties and support planting shade tree species better suited to warmer and drier conditions.

On the other hand, our KIs and field observation results indicated that coffee migrates to higher elevation areas challenged the local ecosystem management and the production of food crops, such as barley (Hordeum vulgare L.) and wheat (Triticum aestivum L.). This is in line with the finding of Chemura et al. (2016), who reported that expanding coffee plantations to higher elevation areas might increase pressure on local ecosystems and conflict with food crops. Hence, appropriate government policies are required to ensure that shifts in production locations will not affect local ecosystems and decrease food security for the local population. Moreover, as coffee planting, managing, and harvesting require knowledge and skill, the future policy aims to provide training and adequate agricultural extension services for farmers in higher-elevation areas. Furthermore, empirical research will be needed to identify and assess the synergies and trade-offs with the existing land use in the higher elevation areas.

5. Conclusion

We focused on farmers’ perception of CC and the undertaken adaptation measures along Ethiopia’s elevational gradient (1600–2000 masl). The farmers perceived the impacts of rising temperatures on coffee and the steps taken to adapt to CC. Most farmers adopt agroforestry practices, organic manure/compost, soil conservation, changing farming calendar, and crop diversification. Farmers also perceived a rainfall reduction, which is not supported by the meteorological data and is likely caused by irregular rainfall distribution. The farmers’ perceptions differed among the three elevations, but no significant difference was observed in their adaptation strategies. Farmers in the low-elevation areas perceived the higher impact of CC than in mid and high elevations because they experienced a higher frequency of drought periods. The results of KIs, FGDs, and field observations also confirmed that CC affected the coffee production systems of smallholder farmers. Moreover, it forced the farmers to replace coffee with drought-resistant crops, particularly in the low elevation areas. Education, farming experience, family size, and access to the extension are the most significant factors influencing farmers’ perception of CC and their adaptation practices. Poor soil fertility, land shortage, lack of weather information, and lack of credit access have been identified as the key challenges to adapting to CC. Hence, policymakers should design and support appropriate adaptation strategies to lessen the adverse effect of CC, such as improved agroforestry practices, farm management, farmers’ training, and increasing access to credit, market, and weather forecasting information.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was conducted with the financial support of the Cuomo Foundation through an IPCC scholarship; Hawassa University, Ethiopia thematic research project; and Internal Grant Agency of CZU Prague (grant No. 20223103)