Smallholder Farmers’ Climate Change Adaptation Strategies in the Ethiopian Rift Valley: The Case of Home Garden Agroforestry Systems in the Gedeo Zone

aberhamkebedom@du.edu.et


Introduction
Climate change has arisen as a global concern with catastrophic consequences for agricultural livelihoods and rural people, ranging from droughts to floods, altering the supply, availability, quality, access, and stability of food systems [1][2][3][4]. By 2050, climate change is anticipated to result in millions of people facing starvation, malnutrition, and poverty [5], with developing countries and small-scale farmers in Sub-Saharan Africa, such as Ethiopia, being particularly vulnerable due to their socioeconomic situations [6,7]. Despite agriculture being Ethiopia's most important sector, contributing to 32.5% of the country's GDP [8], the sector has low productivity [7,9,10], and 84% of smallholder farmers depend on rain-fed agriculture for their economic sustenance [11,12].
To mitigate the effects of climate change in poor countries like Ethiopia, it is necessary to build adequate and effective adaptive responses [1,[13][14][15]. To achieve this, strategic

Background of the Study Area
This study was conducted in the escarpment of the rift valley system of Ethiopia, and, as Figure 1 shows, the study area is located at 6 • 15 N latitude and 38 • 0 to 38 • 15 E longitude. The elevation of the district ranges from 1350 to 3000 m above sea level (MASL) [34], in which the interval of 500 to 2700 MASL accounts for 88% of the total landscape [24]. In the study area, the total population is 1,251,620, of whom 49.86% are men and the remaining 50.14% are women [35]. Further, the same source stated that 62% of residents live in rural areas. The study area receives an average annual rainfall of about 1129 mm [36].

Background of the Study Area
This study was conducted in the escarpment of the rift valley system of Ethiopia, and, as Figure 1 shows, the study area is located at 6°15′ N latitude and 38°0′ to 38°15′ E longitude. The elevation of the district ranges from 1350 to 3000 m above sea level (MASL) [34], in which the interval of 500 to 2700 MASL accounts for 88% of the total landscape [24]. In the study area, the total population is 1,251,620, of whom 49.86% are men and the remaining 50.14% are women [35]. Further, the same source stated that 62% of residents live in rural areas. The study area receives an average annual rainfall of about 1129 mm [36]. The Gedeo plain experiences both equatorial and monsoon trade winds [37], and its rainfall pattern is bimodal. The short wet season, which occurs from March to May, is followed by a brief dry spell before the main rainy season from July to December, which is crucial for coffee flowering, fruit initiation, and early development [38]. The average monthly maximum and minimum temperatures are 25 °C and 18 °C, respectively, resulting in an average annual temperature of 20.6 °C [36]. As per the same source, the widely varying agro-climatic conditions of the region are favorable for the cultivation of a range of crops, such as coffee Arabica, ensete, avocado, mango, banana, papaya, bean, ginger, koroma, cabbage, potato, sweet potato, sugar cane, maize, yam, cassava, kale, and taro, which are all produced in large quantities. The Gedeo plain experiences both equatorial and monsoon trade winds [37], and its rainfall pattern is bimodal. The short wet season, which occurs from March to May, is followed by a brief dry spell before the main rainy season from July to December, which is crucial for coffee flowering, fruit initiation, and early development [38]. The average monthly maximum and minimum temperatures are 25 • C and 18 • C, respectively, resulting in an average annual temperature of 20.6 • C [36]. As per the same source, the widely varying agro-climatic conditions of the region are favorable for the cultivation of a range of crops, such as coffee Arabica, ensete, avocado, mango, banana, papaya, bean, ginger, koroma, cabbage, potato, sweet potato, sugar cane, maize, yam, cassava, kale, and taro, which are all produced in large quantities.

Sampling Procedure and Sample Size
To select sample respondents, a three-stage sampling process was used. Firstly, three districts (Dilla Zuria, Wonago, and Yirgachefe) were picked at random from the 6 in the study area due to their similarity. Secondly, 6 kebeles were randomly picked from each district to remove any bias and ensure accuracy, resulting in 18 rural kebeles from 3 districts. Eventually, 384 households were selected randomly from each kebele based on likelihood proportional to population size. Taking into account the homogeneity of the rural kebeles in the sample, the sample size for the study was chosen using the approach provided by [39]: Where n = the desired sample size, z = the inverse of standard cumulative distribution that relates to the degree of trust with a worth of 1.96, e = the preferred degree of accuracy, p = the estimated proportion of respondents who chose adaptation strategies to climatic change in the population, and q = 1 − p.
Assuming a large population but with no information on the variability in the proportion of the chosen adaptation strategies to climate change, p = 0.5 (maximum variability) was considered, as suggested by [39], to get a representative sample size at 95% confidence level and 5% precision. The total sample size for this study was 384 respondents.

Types, Sources and Metahods of Data Collection
In this study, both primary and secondary sources were used to acquire pertinent data. The primary sources comprised 384 farm households, which were interviewed using a semistructured questionnaire delivered by experienced and culturally competent enumerators. The enumerators underwent three days of training on data collecting processes, with a focus on interviewing strategies and questionnaire content, and the household data collection was completed between December and January 2021. To verify the accuracy of the obtained data, the questionnaire was validated, translated into the local language, and pre-tested, with improvements made depending on the feedback received.
Secondary data sources, including literature studies and statistical data, were used to assess farmers' adaption options to climate change. By integrating these sources with primary data collection methods, issues driving farmers' decision-making processes could be completely appreciated, leading to more successful adaption tactics. Furthermore, to validate the collected data, the study conducted 18 focus group discussions (FGD) for each selected rural kebele, with each group consisting of 8-10 people, such as women, men, district-level agricultural specialists, agricultural extension workers, and elders.

Analytical Methods
In this study, descriptive statistics and a multivariate probit model were employed. Descriptive statistics tools such as frequency, percentage, mean, and standard deviation were employed. The dependent variables in the multivariate probit (MVP) model included upscaling home garden agroforestry farming, improved agronomic measures, physical soil and water conservation practices, livelihood diversifications, and multiple coping strategies. "Coping strategies" in this study are defined as short-term actions farmers take to address food or income shortfalls in unusual agricultural seasons or years [40]. Additionally, this study was driven by the "social learning theory", which implies that individuals learn from one another through interaction [41], and the "utility theory", which refers to determining the adaptation techniques that maximize their value. Thus, these theories were employed to accomplish the research purpose and address in detail the question of the link between adaptation techniques and rural households.
Different methodologies have been employed in different studies to investigate the elements that affect farmers' decisions to adapt to climatic hazards. Binary probit or logit [20,21,28], multinomial probit or logit [26,29], and the Richards model [30] have been employed in earlier investigations. Nevertheless, none of these models takes into consideration the various interrelationships between different methods. Numerous studies have demonstrated that the interdependencies and dependencies across technologies are not well controlled, resulting in an under-or overestimation of the effects of various factors on the analysis of adoption decisions [10,28,29,31,42]. To address these constraints, our study used a multivariate probit approach (MVP). As per [43], the model utilized in this study is characterized by a set of n binary dependent variables: where X is a vector of explanatory variables; β 1 , β 2 , . . . , β n are denoted confirmable parameter vectors to be estimated; and ε 1 , ε 2 , . . . , ε n are random error terms distributed as a multivariate normal distribution with zero means, unitary variance, and an nˆn contemporaneous correlation matrix R = [ρij], with density Φ (ε 1 , ε 2 , . . . , ε n ; R). The likelihood of contribution for an observation is the n-variate standard normal probability.
. . , ε n ; Z RZ dε n . . . dε 2 dε 1 where, Z = [diag. 2y i − 1 . . . 2y n − 1], which is the maximum likelihood estimation maximized for the sample likelihood function, which is the product of probabilities across sample observations. Computation of the maximum likelihood using multivariate normal distribution requires multidimensional integration, and a number of simulation approaches have been put forward to approximate such a function with the GHK simulator to be widely used [42,[44][45][46]. Further, to address the likely heteroscedasticity in the model, this study estimated a robust model that computes a robust variance estimator based on a variable list of equation-level scores and a covariance matrix (Stata 15 help robust).

Demographic, Socioeconomic, and Institutional Characteristics of Smallholder Farmers
According to Table 1, male-headed households composed 94% of all households, whereas female-headed households accounted for only 6%. Around 48% of respondents got support in cash or kind from family, friends, and neighbors, and a majority (93%) were members of farmers' associations or cooperative organizations. In addition, 52% of respondents gained financing from credible microfinance institutions, and approximately 45% of households received farmer-to-farmer extension services, offering training and information on improved agricultural techniques from diverse sources. The survey's findings also suggested that 69% of participants encountered infertility concerns with their cropland. The survey results, shown in Figure 2, indicate that 82% and 66% of households had access to climate-change-related information and perceived climatic changes, respectively. Moreover, around 61% of households experienced crop failures due to adverse meteorological conditions. Although there is a significant potential for irrigation water, only 13% of households had access to additional technologies utilizing river water and rainfall harvesting. Furthermore, in the last three decades, approximately 60% and 74% of the respondents reported recurring drought every ten years and freeze occurrence every five years, respectively. tively. Moreover, around 61% of households experienced crop failures due to adverse meteorological conditions. Although there is a significant potential for irrigation water, only 13% of households had access to additional technologies utilizing river water and rainfall harvesting. Furthermore, in the last three decades, approximately 60% and 74% of the respondents reported recurring drought every ten years and freeze occurrence every five years, respectively.  Table 2 illustrates the results that the households in the research area had an average age and farming experience of 46.32 and 23.76 years, respectively, with an adult equivalent (AE) of 4.04. The average educational attainment of the household head was grade 6, ranging from zero (no formal education or literacy) to grade 12. Moreover, households held an average of 0.81 hectares of cultivated land, ranging from 0.15 to 2.96 hectares. Livestock holding is also shown to be an essential source of income, contributing to an average of 2.49 tropical livestock units (TLUs). The survey calculated household income from farming, non-farming activities, livestock systems, and natural resource collecting, indicating an annual range of ETB 8000 (USD 180.51) to ETB 189,500 (USD 4275.72), with an average of ETB 59,496 (USD 1342.42) with standard deviation ETB 42,103 (USD 949.98). In terms of market proximity, smallholder farmers walked an average of 19.94 min on foot to reach the closest market center. The survey also indicated that households had an average of 2.94 contacts with agricultural extension workers (AEWs).   Table 2 illustrates the results that the households in the research area had an average age and farming experience of 46.32 and 23.76 years, respectively, with an adult equivalent (AE) of 4.04. The average educational attainment of the household head was grade 6, ranging from zero (no formal education or literacy) to grade 12. Moreover, households held an average of 0.81 hectares of cultivated land, ranging from 0.15 to 2.96 hectares. Livestock holding is also shown to be an essential source of income, contributing to an average of 2.49 tropical livestock units (TLUs). The survey calculated household income from farming, non-farming activities, livestock systems, and natural resource collecting, indicating an annual range of ETB 8000 (USD 180.51) to ETB 189,500 (USD 4275.72), with an average of ETB 59,496 (USD 1342.42) with standard deviation ETB 42,103 (USD 949.98). In terms of market proximity, smallholder farmers walked an average of 19.94 min on foot to reach the closest market center. The survey also indicated that households had an average of 2.94 contacts with agricultural extension workers (AEWs).

The Adaptation Techniques Chosen by Smallholder Farmers to Climate Change
This study utilized a participatory approach to investigate the adaptation strategies of farmers against climate change. The approach involved surveys and focus group discussions with farmers. The majority of household heads (384) in the study areas adopted a single or a mix of adaption strategies. However, this study focused on the five major adaptation strategies, namely up-scaling home garden agroforestry (UHGAF), agronomic measures (AMs), physical soil and water conservation practices (PSWCPs), livelihood diversifications (LDs), and multiple coping strategies (MCS), as they were widely embraced by farmers in the study area. These tactics were employed as a dependent variable, coupled with explanatory variables selected based on a literature review.
The data in Table 3 shows that households had a better probability of adopting all five adaptation alternatives mutually, with a probability of 9.90%, rather than failing to select any of these strategies, with a probability of 0.83%. Additionally, the predicted likelihood of adopting UHGAF, AMs, PSWCPs, LDs, and MCSs as adaptation strategies increased by 0.78, 0.82, 0.52, 0.69, and 0.42, respectively. Table 3. Predicted probabilities of each adaptation strategy in study areas of southern Ethiopia (from MVP model).

Description of Adaptation Strategies Expected Likelihood References Mean SD
Household (HH) who intentionally upscales indigenous home garden agroforestry farming (UHGAF) by using fruit trees, disease-resistant coffee varietals, and multipurpose tree species as adaptation strategies. It takes a value of 1 if households have adopted it.
0.785 0.280 [47][48][49][50] HH who consciously adopts improved and more effective agronomic measures (AMs) as the main means of coping with climate change. 0.820 0.214 [47,[51][52][53][54][55] Deliberately implemented adaptation strategies, particularly physical soil and water conservation practices (PSWCPs), involve using practices such as terracing, bunding, and water harvesting to conserve soil and water resources in the farm. According to sources [59,62], up-scaling home garden agroforestry can boost both farmers' livelihoods and the agroecosystem by preventing climatic adversity and helping them manage intense and unpredictable weather. The study's findings in Table 3 reveal that the majority (78%) of respondents reacted to existing and future climate problems by consciously adjusting their subsistence-based home garden agroforestry techniques. This entailed incorporating fruit trees and applying recommended coffee technology packages, such as spacing, pruning, stamping, disease-resistant coffee plants, multipurpose tree species, and shading control. These findings correspond with earlier studies [47,[49][50][51]. Moreover, [20] demonstrates that scaling up agroforestry is a sustainable approach that can mitigate hazards associated with climate change.

Agronomical Measures
The findings in Table 3 show that a large majority (81.96%) of smallholder farmers in the study area have adopted improved agronomic techniques as a means of adapting to climate change. These techniques include intercropping, adjusting planting schedules, using cover crops, shifting to mono-cropping, employing conservation tillage, diversifying crops, adhering to suggested spacing, and utilizing organic fertilizers. Similar research conducted in Ghana [53] and Eastern Ethiopia [55] has also shown that smallholder farmers are adopting these practices to mitigate the effects of climate change and safeguard their agricultural production. Furthermore, farmers in Nigeria have implemented a crop management system that involves changing the farming calendar, crop diversification, multiple cropping, intercropping, crop specialization (mono-cropping), and utilizing improved crop varieties to address environmental variability and the economic risks associated with climate change, as noted in another study [51].

Physical Soil and Water Conservation Practices
Smallholder farmers in the study area utilize various soil and water conservation practices, including establishing physical structures such as terraces on slopes, soil bunds, micro trenches, micro-basins, and humidity conservation holes, to actively control risks and mitigate the adverse effects of global warming. Given that more than half of households (51.99%) in the research area employed one or more physical soil and water conservation practice as an adaptation measure to address climate-change-related hazards, the results are not surprising (Table 3). Current research conducted by [47,55] indicates the heavy application of such approaches by farmers in Eastern Ethiopia. Likewise, research reveals that farmers in Ethiopia [1,63] and Nigeria [19] also utilize analogous strategies.

Livelihood Diversification
According to previous research by [10] in the Ethiopian Nile basin, [50] in Nigeria, and [49] in Vietnam, livelihood diversification is an efficient adaptation strategy that can offset the detrimental implications of climatic shocks. To assess the most common kind of livelihood diversification other than farming as a risk management approach related to climate change, respondents in the research area were polled. The survey's findings suggest that a majority of households (68.25%) undertook various income-generating activities, such as farming fattening, beekeeping, timber production, handicrafts, commerce, and transportation (motorbikes) ( Table 3). Our results accord with [51], which demonstrates that embracing non-farm employment enhances smallholder farmers' adaptive capacity in southwest Nigeria and helps them overcome the harmful impacts of climate shocks. Similarly, Ref. [54] discovered that farmers in Ethiopia's Eastern Tigray adopt non-farm livelihood alternatives, boosting their ability to react to the unfavorable consequences of climate shocks.

Multiple Coping Measures
The results in Table 3 show that 42.24% of respondents in the study area adopted numerous coping methods to handle the impacts of climate change. These measures included selling household possessions; borrowing money from friends, family, or village moneylenders and loan associations; reducing food consumption; pulling children out of school; selling seed banks; borrowing food (particularly Kocho, a local food) from relatives, neighbors, or friends; purchasing Kocho from the local market; and selling small livestock. As a result, households are adapting to climatic adversity. As stated by scholars, smallholder farmers in Ethiopia employed various coping mechanisms, including borrowing money from family members, friends, or local savings and loans groups [15] and selling livestock and adjusting their consumption habits [64] as adaptation methods to manage the negative implications of climate change. Furthermore, prior studies indicate that farmers in Ghana have used a variety of coping mechanisms, including selling off household possessions [65,66], cutting back on daily mealtimes [60], and selling firewood, animals, fruits, and vegetables [67] to manage the risks connected with climate change. Table 4's correlation coefficients of adaptation strategies in response to climate change were utilized to test their independence in the MVP model. The likelihood ratio test (Chisquare = 52.937, p-value = 0.000) of independence of error terms rejected the null hypothesis for the interdependence of predicted variables in the model. This finding reveals that the five adaptation measures were mutually interdependent, making the MVP model excellent for discovering the factors that influence farmers' choices of adaptation techniques, as it captures broader effects than individual univariate probit models. Notes: *** and ** refer statistically significant at 1% and 5% level, respectively. Likelihood ratio test of rho21 = rho31 = rho41 = rho = 51 = rho32 = rho42 = rho52 = rho = 43 = rho53 = rho54 = 0, Chi2 (10) = 52.937, Prob. > chi2 = 0.0000.
Furthermore, seven of the ten pairwise correlation coefficients, notably PSWCPs and UHGAF, LDs and UHGAF, MCS and UHGAF, LDs and AMs, MCS and AMs, MCS and PSWCPs, and MCS and LDs, were found to have significant values (Table 4). The negative coefficient sign implies both a negative correlation (substitutability) and an interaction correlation between PSWCPs and UHGAF, LDs and UHGAF, MCS and UHGAF, LDs and AMs, and MCS and AMs, as well as MCS and PSWCPs. In contrast, the positive coefficient sign shows a complementary (positive) and interactive link between MCSs and LDs. Before applying the empirical model, the existence of multi-collinearity among the repressors investigated. Consequently, all independent variables' variance inflation factors (VIF) were between one and less than two (Table S1).

Determinants for Smallholder Farmers' Choices of Adaptation Strategies
Out of 19 explanatory variables included in the MVP model, smallholder farmers' choices of adaptation strategies, such as UHGAF, AMs, PSWCPs, LDs, and MCSs were significantly affected by 9, 9, 7, 6, and 4 explanatory variables, respectively. The results of the MVP model in Table 4 indicate that the explanatory power of independent variables on the dependent variables was satisfactory because the authors failed to accept the null hypothesis because the likelihood function from an MVP model was significant (Wald Chi-square (95) = 415.52 with p-value < 0.001).
According to this study, the age of the household head had a significant negative impact on the adoption of LDs (p < 0.1). As the age of the head increased, the likelihood of pursuing various livelihood strategies decreased. Specifically, a one-unit decrease in the age of the head increased the likelihood of using LDs by 0.4%, holding all other factors constant (refer to Table 5). Older farmers may find it difficult to launch new businesses or grow current ones due to age-related physical strength and health difficulties and their ability to interact with various markets and benefit from sources of income. The general decline in productivity that occurs with advancing age may further constrain farm households' capacity to diversify their sources of income. Previous studies by [18,56] have also found a negative correlation between the age of the household head and the likelihood of using adaptation options to address climate risks. Similarly, research conducted by [10] suggests that younger farmers are more innovative, productive, and more likely to adopt strategies that enhance their adaptability. In contrast, recent research conducted by [68][69][70][71] has found a positive and strong correlation between aged households' capacity to diversify their sources of income.
This study supports prior expectations by finding a positive correlation between the likelihood of using UHGAF (p < 0.01) and AMs (p < 0.01) and the contribution of the household workforce to agricultural production. Larger labor pools in households are more likely to adopt UHGAF and AMs compared with their counterparts. Specifically, utilizing family labor in farming operations increased the likelihood of using UHGAF and AMs by 2.9% and 3.3%, respectively, all else being equal (Table 5). Households with larger family sizes are more likely to adopt labor-demanding adaptation strategies that are resilient to climate risks when their household capacity increases due to higher labor productivity and decreased dependency ratios, resulting in less appeal to have a portion of the labor force participate in non-farm income generating activities. Previous studies [1,50,55] support this suggestion by finding that larger-family-size households use labor-intensive adaptation strategies to lessen the impact of climate change. However, Refs. [57,72] found a negative significant association between labor-intensive adaptation strategies such as UHGAF and AMs and larger-family-size households.
Households with more farming experience are more likely to adopt UHGAF, AMs, and LDs, as the positive connection and statistically significant role of their coefficient demonstrate (p < 0.01 for UHGAF and AMs and p < 0.05 for LDs). The positive sign indicates that they are more inclined to implement these tactics than households with less farming experience. Table 5's marginal effects demonstrate that a one-unit increase in farming experience improves the chance of adopting UHGAF, AMs, and LDs by 4.5%, 3.6%, and 1.9%, respectively, holding all other factors constant. This is likely due to their better access to skills, technologies, and financial resources, as well as their expertise in the effects of climate change and adaptation measures. These findings are consistent with earlier research conducted in Africa that indicates a positive connection between farming experience and the adoption of climate change adaptation strategies [20,52].
Farm households with higher levels of education are more likely to adopt climate change adaptation techniques, particularly UHGAF (p < 0.05) and PSWCPs (p < 0.01), as found by the study. The study reports a positive correlation between education and the use of these techniques. Table 5 shows that an increase in the school level of households leads to a 7.8% and 9.0% increase in the chance of utilizing UHGAF and PSWCPs, respectively, keeping all else constant. This is because higher education levels provide greater accessibility to knowledge, higher earning potential, enhanced labor efficiency, and a better understanding of current farming practices. Thus, such households are better equipped to embrace novel production techniques, obtain technical support from extension professionals, administer their farms more successfully, and apply suggested practices. Studies conducted by [10] in Ethiopia and [20] in Kenya have proven a strong positive link between education level and the adoption of adaptation techniques in response to climate change. Thus, households with higher education levels are better equipped to embrace novel production techniques, obtain technical support from extension professionals, administer their farms more successfully, and apply suggested practices, boosting their ability to cope with climatic conditions.  Note: ***, **, and * refer to being statistically significant at 1%, 5%, and 10%, respectively. Multivariate probit (MSL, # draws = 5) Log pseudo likelihood = −804.21004 Wald chi2 (95) = 415.52 Prob. > chi2 = 0.0000.
This study indicated that the size of cultivated land had a significant and favorable impact on the chance of employing UHGAF (p < 0.01), consistent with expectations. This shows that households with bigger cultivated land holdings had a higher probability of adopting UHGAF than those with smaller holdings. The marginal effects in Table 5 demonstrate that an increase in cultivated land size by one unit improves the chance of adopting UHGAF by a factor of 0.124. Variables such as enhanced labor productivity, longer planning horizons, higher earning potential, market orientation, and a willingness to experiment with innovative techniques to handle climatic risks, all of which are associated with larger landholdings, explain the stronger capacity for adaptability and coping with environmental issues in households with larger holdings. Earlier studies conducted by [18,47,73] have corroborated the positive correlation between land size and the implementation of climate change adaptation tactics, including production-smoothing choices, while [19,56] found that households' cultivated land sizes positively and significantly influenced the adoption of production smoothing options. Conversely, landowners with substantial properties, such as [20,74], have demonstrated an aversion to investing in costly adaptation techniques in the face of climate change.
Larger livestock holdings significantly and positively influence the likelihood of adopting AMs (p < 0.05), LDs (p < 0.1), and PSWCPs (p < 0.01). The analysis shows that households with more animals are more inclined to adopt these strategies than those with fewer animals. The computed marginal effect shows that a unit increase in livestock holdings leads to a 1.6%, 3.3%, and 1.6% rise in the probability of using AMs, LDs, and PSWCPs, respectively, while keeping other variables constant. This is because households with larger livestock holdings have greater access to resources, such as land and water; modern and improved agronomic practices, such as the use of fertilizers, pesticides, and high-yielding crop varieties; and higher incomes, which enhance their economic freedom and enable them to purchase necessary materials and start up new businesses to adopt capital-intensive adaptation measures as AMs, LDs, and PSWCPs. This finding is aligned with prior research [19,56,73].
Household income had a positive correlation with the likelihood of employing UHGAF (p < 0.05), AMs (p < 0.05), and LD (p < 0.01), as per the prior expectation, and the correlations were statistically significant. Table 5 shows that UHGAF had a marginal effect of 1 × 10 −6 , AMs had a marginal effect of 1 × 10 −6 , and LS had a marginal effect of 2 × 10 −6 , indicating their significant positive influences. High-income households endowed with additional income are more likely to use UHGAF, AMs, and LD, resulting in a 0.1%, 0.1%, and 0.2% increase, respectively, ceteris paribus. This is because they can avoid climate risks better. The study found that households not solely dependent on agriculture have higher earning potential, asset accumulation, and willingness to invest in financially needed activities, such as fattening, honey production, coffee collector businesses, and commercial-tree plantations. This enhances their adaptive capacity toward climate risks. The research in [57,75] supports the idea of introducing a new financial flexibility program to allow them to explore different markets and find lucrative income-generating activities.
Households' perceptions of climate change have a significant and positive relationship with their adoption of UHGAF (p < 0.01), PSWCPs (p < 0.05), and MCSs (p < 0.01), as demonstrated by the statistics showing an increase in the likelihood of uptake by 12.7%, 9.2%, and 26.3%, respectively. Those households who experience climate change are more likely to adjust to the risk of it compared with households that do not notice climate change. When their understanding of the altering climate is greater, the probability of these households implementing UHGAF, PSWCPs, and MCS is higher. This may be because those who perceive the effects of climate change would be more willing to invest in protective and productive resources to defend against any adversities. This study found that farmers in the Dabus watershed in Ethiopia's northwest are more willing to invest in adaptation measures when they have access to climate information [18]. Additionally, research in Nigeria's southwest highlighted that the availability of climate information plays a critical role in a farmer's determination to employ adaptation techniques, such as UHGAF, PSWCPs, and MCS, increasing the likelihood that they would take up adaptation measures [76].
The impact of previous experience with crop failure on the probability of implementing LDs and MCS was studied. Our results revealed a significant positive relationship (p < 0.01), indicating that households with prior experience of crop failure had a higher probability of implementing these methods. Furthermore, this study found that for each additional household with prior agricultural failure, the likelihood of embracing LDs and MCS increased by 28.2% and 14.0%, respectively, when other factors held constant. Households that rely on their crops for sustenance and income face significant consequences from climate-related threats, including financial losses and food insecurity. Thus, prior crop failures could serve as a strong incentive for households to adopt adaptation measures to reduce their risk of future crop failure. Other studies, such as those conducted by [1] in Ethiopia and [50] in Nigeria, reported similar findings.
As anticipated, farmer-to-farmer extension had a significant positive impact on the likelihood of using enhanced agronomical measures (p < 0.01). Households with access to this extension employed enhanced agronomic practices and measures to adapt to climate risks than their counterparts without access. Conversely, farmers without access to this method were more inclined to implement physical soil and water conservation practices than their counterparts with access, resulting in a negative and significant impact. Access to farmer-to-farmer extension increased the probability of implementing enhanced agronomical measures by 11.3% but decreased the probability of using physical soil and water conservation practices by 10.5%. Enhanced agronomical measures are likely simpler and less expensive to implement than physical soil and water conservation practices, which require more resources and investment. Various investigations, including [26,31,44,57,77], have consistently found this. Similarly, Ref. [1] reported that observation, model farmers, formal training methods, and experience-sharing programs improve farmers' understanding of production or risk management tactics.
Households who viewed their farmland as infertile were more likely to employ UH-GAF (p < 0.1) and PSWCPs (p < 0.01) as adaptation methods to reduce the impacts, as anticipated. Conversely, this variable negatively and significantly (p < 0.01) influenced the adoption of MCS. Therefore, those households who did not feel their farmland was infertile were more prone to take on MCS when compared with those who did. Controlling all other variables, a single unit increase in the households' outlook of their farmland as infertile improved the likelihood of utilizing UHGAF and PSWCPs by 6.1% and 37.6%, respectively, but diminished the possibility of implementing MCS by 23.4%. Households who perceive their farmland as infertile may have the monetary resources, personnel, and knowledge resources to implement new practices that may lead to improved agricultural yields. Households with restricted choices in countering agricultural disasters may be more prone to dedicating themselves to long-term approaches such as UHGAF and PSWCPs. It can be presumed that proactive and long-term tactics such as UHGAF and PSWCPs could increase agricultural productivity and defend against climate change, compared with reactive and temporary options like MCS which are used when all else fails. Studies [26,77,78] have also discovered similar findings.
Recurring droughts influenced households' inclination to adopt AMs (p < 0.1) and LDs (p < 0.05). The average marginal effects showed that the probability of households using AMs and LDs increased by 38.1% and 50.6%, respectively, with an increase in recurrent drought observation by one unit while holding all other factors constant. Farmers who were aware of droughts occurring over the past three decades were more likely to choose AMs and LDs than their counterparts who were unaware of such occurrences. This was because they understood the recurring droughts, had higher risk aversion inclinations, could plan for future risks, and appreciated the benefits of AMs and LDs. Therefore, farmers' early knowledge of droughts significantly influenced their preference for adaptation techniques in the area, which is consistent with the ideas put forward by [50,53,58]. Furthermore, Ref. [63] conducted research in Ethiopia and Ref. [79] conducted research in Pakistan, and both studies found that households' exposure to drought had a significant and constructive impact on their decision to adapt to climate change.
Access to climate change information related issues significantly and positively affected the propensity to uptake UHGAF (p < 0.1) and AMs (p < 0.01) to reduce the effects of climatic adversity. The existence of this information increased the likelihood of households employing UHGAF and AMs by 7.7% and 12%, respectively, all other factors being equal. This implies that households with access to climate-change-related information are more inclined to utilize UHGAF and AMs in comparison to those who do not. The likely explanation is that such households are more aware of the impact of climate change, possess the capacity to make feasible decisions concerning responses, and have more exposure to enhanced farming technologies and practices. This increases their capacity to handle climatic risks by using UHGAF and AMs, which aligns with findings in [55,[80][81][82]. Likewise, similar studies conducted in Ethiopia by [1,83] also determined a positive relationship between preparedness and early warning systems and the inclination to utilize adaptation strategies against climate change.
A higher number of contacts with AEWs positively affected the probability of UHGAF and AMs adoption, as demonstrated by the results (p < 0.05). This suggests that households with more interactions with AEWs are more likely to embrace these adaptation options than those with fewer AEW contacts. Furthermore, each unit increase in AEW contact led to a 2.9% increase in the probability of UHGAF adoption and a 3.0% increase in the probability of AMs adoption, with all other factors held constant. Households are more likely to learn about adaptation options and modern agricultural technologies, and to be willing to take suggestions from AEWs and substitute traditional practices with modern production technologies. This could be attributed to their higher level of awareness. These findings are supported by similar research conducted by [15,84,85], which found a positive and significant correlation between the frequency of contact and the adoption of these options.
Contrary to expectations, this study found that households' inclination to use laborand resource-intensive adaptation strategies such as PSWCPs was significantly (p < 0.01) and negatively influenced by their perceptions of frost occurrence. This suggests that farmers' perceptions of frost occurrence make them less likely to adopt capital-intensive strategies such as PSWCPs due to resource constraints like financial, informational, and risk tolerance. According to [86] smallholder farmer adoption of climate-related adaptation strategies, farmers' perceptions of frost occurrence, which can have a significant and negative impact on their inclination to use labor-and resource-intensive adaptation strategies (such as physical soil and water conservation practices). In addition, farming households may lack access to financial and informational resources, making it challenging for them to accurately evaluate the costs and risks associated with these practices, which is consistent with the previous findings of [87][88][89].
The likelihood of using UHGAF and AMs (p < 0.01) is significantly negatively influenced by households with a long walking distance to the nearest market center. This suggests that these adaptation alternatives were less accessible to households further away from the nearest market center. Conversely, the research showed a statistically significant and negative adoption effect of this variable on PSWCP (p < 0.01), indicating that households closest to the nearest market center were less likely to employ this adaptation technique. The calculated marginal effects reveal that farmers are more likely to implement UHGAF and AMs when their walking distance to the market center is reduced by 0.3 and 0.4%, respectively, while they are less likely to embrace PSWCPs by 0.8%, ceteris paribus. This is because close proximity to the nearest market center increases access to information on contemporary agricultural technologies and climate-change-related matters. In addition, easier access to the market is associated with decreased transaction costs and greater risk aversion, both of which enable households to tackle climate-change-induced risks better. These results are consistent with those from [1,57].
Ultimately, the status of social networking in households influences the adoption of response measures. The research findings revealed that households with access to social networks had a significant and negative impact on farmers' decisions to use MCSs (p < 0.05) as adaptation responses. In other words, those without access to social networks are more likely to resort to MCSs, as compared with their counterpart with access. The results of the research indicate that an increase in access to social networks by one unit leads to a decrease of 11.9% in the likelihood of embracing MCSs. Households with access to social networks can attribute their ability to use their own and other people's social capital and access more practical and efficient adaptation strategies. On the other hand, those without access to social networks must rely on multiple temporary coping methods. This emphasizes the need to broaden access to social networks to help households prepare better for financial problems. According to a current study conducted by [88], households with access to social networks require fewer coping methods for economic difficulty. This finding is consistent with the results of studies conducted by [89,90], which suggest that social capital can enhance people's quality of life and promote the exchange of knowledge and understanding.

Conclusions
The findings of this study suggest important implications for policymakers and future research. The severe impact of climate change on smallholder farmers engaged in home garden agroforestry necessitates a focus on providing access to a range of adaptation strategies to mitigate these effects. This includes upscaling the agroforestry system, promoting modern agronomic measures, and implementing physical soil and water conservation practices. Additionally, policies should aim to diversify livelihoods and promote multiple coping strategies to help farmers adapt to the changing climate. The results of the multivariate probit model analysis demonstrate that certain variables significantly affect the choice of adaptation strategies for smallholder farmers. Policymakers should therefore focus on enhancing rural institutional services, increasing climate change education, and strengthening social capital to improve farmers' capacity to adapt to the changing climate. Future research should also explore the effectiveness of different adaptation strategies in mitigating the impact of climate change on smallholder farmers engaged in home garden agroforestry. This could involve evaluating the performance of different adaptation strategies in different contexts, as well as identifying potential barriers to the adoption of these strategies and interventions to overcome them. Overall, this study provides valuable insights into the determinants of adaptation strategies in Ethiopia's Rift Valley home garden agroforestry system and highlights the importance of policy interventions to support smallholder farmers in the study region.