Climate variability perceptions and adaptation response to ensure household food security in Borana zone, Southern Ethiopia

Climate variability and extremes pose considerable challenges to pastoral and agro-pastoral households’ food security, as it disrupts their livelihood system (Birkmann et al., 2022). In response to these challenges, adaptation measures have been developed to address the adverse effects of climate variability and ensure food security (Yared et al., 2022; IPCC, 2019).

Ever-rising temperature, erratic precipitation behavior and recurrent drought have implications for pastoral and agro-pastoral households’ food security (Ayal et al., 2017). Pastoral and agro-pastoral communities, whose livelihoods depend on livestock rearing and semi-nomadic or nomadic practices, are particularly vulnerable to the adverse effects of climate variability. Climate variability and extremes could compromise pastoralists’ and agro-pastoralists’ access to water and pasture for their livestock (Birkmann et al., 2022; Radeny et al., 2019). These changes could lead to the proliferation of livestock diseases, extending Calvin at first and Calvin in between, reduced milk production and increased vulnerability to food insecurity (Alhamshry et al., 2020; Ayal et al., 2017; Conway et al., 2004).

The effect of climate variability and extremes on pastoral and agro-pastoral households extends beyond their livestock. Crop failures due to erratic rainfall patterns and prolonged droughts further exacerbate food insecurity. Additionally, changes in temperature and precipitation patterns could trigger rangeland degradation, prevalence of livestock diseases and incur extra resources for the livestock management (Leal Filho et al., 2020).

Perceptions of climate variability and extremes among pastoral and agro-pastoral households could be influenced by socio-cultural contexts (Ayal et al., 2017). Cultural beliefs, customs and traditional practices shape their understanding of climate patterns and changes. For instance, indigenous knowledge systems may include specific rituals or practices to mitigate the impacts of climate variability, such as rainmaking ceremonies during droughts. Recognizing and incorporating local knowledge and perceptions of climate variability is crucial for developing context-specific adaptation responses.

Pastoral and agro-pastoral households use various adaptation responses to ensure food security in the face of climate variability and extremes. These strategies are multifaceted and span several dimensions, including livelihood diversification (e.g. petty trade, charcoal selling and wage labor), water harvesting, use of social networks for mutual support, adoption of drought-tolerant crops and improved livestock breeds. Mobility to access better grazing areas and herd diversification are also common practices aimed at enhancing resilience (Thornton and Herrero, 2014; Opiyo et al., 2015; Gebre et al., 2021). However, the effectiveness of these strategies is often limited by institutional and infrastructural barriers such as inadequate market access, limited climate information and weak extension services.

Livestock management remains at the core of these adaptation efforts. Households frequently adjust herd size, diversify livestock species and practice seasonal mobility to optimize resource use and reduce exposure to climate-induced shocks. Such practices contribute to maintaining livestock productivity, minimizing environmental degradation and strengthening household resilience to recurrent droughts (Kemal et al., 2022).

Pastoral and agro-pastoral households use strategies such as cultivating drought-tolerant crops and practicing intercropping to reduce the risk of crop failures and increase agricultural production (Yadav et al., 2018; Rufino et al., 2013). Similarly, pastoral communities also implement water resource management strategies, including the construction of water harvesting structures and the adoption of water-saving techniques. These practices are aimed at ensuring consistent access to water for both livestock and crop cultivation (Ayal et al., 2017).

In addition to livestock and crop agriculture-based strategies, pastoral households often engage in income diversification activities. This includes off-farm employment, small-scale businesses and the sale of non-livestock products. Income diversification reduces their reliance on livestock-based livelihoods and provides alternative income sources to cope with climate variability shocks (Little et al., 2001; Ellis and Freeman, 2004; Tache and Oba, 2010). These strategies not only help buffer against income losses during droughts or market disruptions but also contribute to long-term livelihood resilience.

Integrated approaches that combine multiple adaptation responses are crucial for enhancing food security in pastoral communities. The combination of livestock management, crop diversification, water resource management and income diversification strategies can help pastoral households build resilience, reduce vulnerability and ensure sustainable food production systems (Morton, 2007; Herrero et al., 2016; FAO, 2018). Such holistic strategies address the complex and interlinked challenges faced by pastoralists under climate change, allowing for more adaptive and flexible livelihood systems.

The conceptual framework guiding this paper is based on the vulnerability and adaptation theory. This theory posits that vulnerability to climate change is influenced by both exposure to climate hazards and the capacity to adapt. The theory further states that household food security is determined by the interaction between climate variability perceptions, adaptation responses and socio-economic factors within the Borana Zone context.

The Borana people are predominantly pastoralists who inhabit arid and semi-arid environments where the climate is variable and the frequency and intensity of droughts and floods are increasing. The livelihood resources of the livestock production mainly depend on natural assets that, in turn, are affected by climatic impacts. Climate-related risks, such as increased incidence of recurrent drought, high temperature and low and erratic rainfall, affect the livelihood of the communities (Alemu et al., 2019). Droughts and other climate-related shocks can have severe impacts on the livelihoods of the Borana people, as the people are highly dependent on livestock and natural resources for their survival. These changes have severe implications for agricultural productivity and food availability in the region.

Alemu and Adugna, found that agro-pastoralist farmers in Borana perceived changes in temperature and rainfall patterns and associated these changes with negative impacts on crop yields. Found that both agro-pastoralists and pastoralists perceived increased rainfall variability, linking it to negative impacts on agriculture, pasture availability and livestock productivity. Deressa et al. also reported that pastoralists perceived changes in temperature and rainfall, which influenced their decision-making regarding herd management and mobility.

Consequently, pastoralist and agro-pastoralist households have adopted various adaptation responses to ensure food security. Among agro-pastoralists, crop diversification emerged as a common adaptation strategy, with a transition from traditional crops to more drought-tolerant and resilient varieties (Alemu et al., 2019). Livestock management practices, such as adjusting herd size and providing feed supplementation, have also been implemented to reduce climate change-related risks. Conversely, pastoralists have implemented diverse adaptation responses. Mobility stands out as a commonly used strategy, enabling pastoralists to access grazing areas in response to pasture scarcity. Additionally, herd diversification and destocking have been identified as adaptation responses to manage feed scarcity during droughts. The literature review underscores that pastoralists’ perceptions of climate variability and their adaptive strategies are influenced by their dependence on livestock and mobility as primary livelihood strategies.

However, studies on climate variability perceptions and adaptation responses in pastoral communities in Ethiopia, including the Borana zone, remain scarce. Besides, the existing studies in Ethiopian pastoral communities (e.g. Alqeer et al., 2023; Bekele et al., 2020; Alemu et al., 2019; Gurmu, 2018) have primarily focused on the determinants of food security and/or the economics of climate variability. To fill this gap, this study aims to explore pastoral and agro-pastoral households’ perceptions of climate variability and their adaptation responses in the Borana Zone. By generating empirical evidence, this study seeks to inform policymakers, development practitioners and communities in developing context-specific interventions to enhance climate change resilience and food security among pastoral and agro-pastoral households in the Borana Zone.

Understanding how pastoral and agro-pastoral households perceive climate variability is vital for crafting effective adaptation responses. Additionally, comprehending the impact of these adaptation responses on pastoral households’ food security is crucial for designing effective interventions and policies, as they heavily rely on livestock and agriculture for their livelihoods, rendering them particularly vulnerable to disruptions caused by climate variability. Thus, by identifying effective adaptation responses, policymakers, researchers and practitioners can formulate targeted interventions and programs to support pastoral households in building resilience and improving their food security.

The study was conducted in the Borena Zone, Ethiopia. Borena Zone is located in the southern part of the Oromia regional state in Ethiopia. Geographically, the Borena Zone is situated between 4˚ 3’ to 5˚ N latitude and 37˚ 4’ E to 38˚ 2’ E longitudes. It shares boundaries with the Guji zone in the east, the Somali regional state in the southeast, the Southern Nations Nationalities and Peoples of Southern Ethiopia in the west and has an international boundary with Kenya (see Figure 1).

The landscape of the zone is characterized by slightly undulating terrain with an altitude of 1,126 meters (3,694 feet) above sea level. The Borena Zone comprises 13 districts and covers an area of 48,743 km2. The climate in the zone is predominantly arid and semi-arid, with a small area covering sub-humid zones. The Borena Zone comprises 13 districts and covers an area of 48,743 km², with a mean altitude of 1,500 m above sea level. The climate in the zone is predominantly arid and semi-arid, with sub-humid zones. Droughts have had a significant impact on the lowland areas of Borena, and the region experiences fluctuating rainfall patterns.

Approximately 50% of the annual rainfall in the lowland areas falls during the long rainy season, which occurs from March to May, while around 30% falls during the short rainy season, which takes place between September and November. Pasture availability in the area is limited and highly variable due to the fluctuating rainfall patterns (NAPA, 2007).

The economy of the Borana Zone is predominantly pastoral with limited industrial and commercial activities. Livestock plays a central role in the livelihoods of the Borana people, providing food, income and social status. Livestock is used for milk, meat, hides and as a form of savings and insurance against risks such as drought and disease. The Borana Zone is facing several challenges, including poverty, food insecurity, water scarcity, frequent conflict, limited access to social services and environmental degradation. Droughts and other climate-related shocks can have severe impacts on the livelihoods of the Borana people.

The study used both primary and secondary data sources. A cross-sectional research design was used to obtain primary quantitative data from pastoralists and agro-pastoralists through structured questionnaires, while qualitative data were collected using 12 focus group discussions, 12 in-depth interviews, 2 case studies and observations. In addition, secondary data were collected from reports of the Borena Zone office, farmers’ cooperative, central statistical agency and published and unpublished documents. During primary data collection, well-trained enumerators who have good experience in the household survey were employed and deployed.

A multi-stage sampling technique was used to select pastoralist and agro-pastoralist households. At the first stage, representative districts were selected using simple random sampling. The second stage followed the proportional and random selection of kebeles within the specified districts. The study has chosen pastoralists and agro-pastoralists from each kebele by referring to the kebeles’ register as the study sampling frame.

The sample size was determined using Cochran’s (1977) formula by taking the following assumptions: the estimated proportion of an attribute that is present in the population of 50% (to get maximum sample size) and a 5% margin of error at 95% confidence level. The calculated sample size was 384, and 10% non-responses and incomplete responses, i.e. 38 pastoral and agro-pastoral households were added. Hence, the final estimated sample size was 422 households. The nonresponse was 5 people, which resulted in the final sample size being 417 households (98.8% actual response rate).

The analysis was performed using SPSS version 25, STATA 17 and R statistical software packages. Both descriptive and econometric analyses were used. Mean, standard deviation, percentages, ratios and frequency distributions were used to portray the characteristics of the pastoral and agro-pastoral households. The Multivariate Probit Model (MVP) and Endogenous Switching Regression (ESR) Model were used to further examine the effects of perceptions and adaptation on the outcome variable. Moreover, the Chi-square test, F test and t-test were used to test the existence of any bivariate statistically significant association between different adaptation responses and household food security status.

A multivariate probit (MVP) model was applied to identify adaptation responses adopted based on the perceived climate variability. The dependent variable of this study was a binary variable indicating whether or not a particular adaptation response had been used. The variable takes a value of 1 if the household has adopted the specified climate variability adaptation response and takes a value of 0 otherwise.

The MVP model is used to analyze the influence of various climate variability perceptions on the adopted adaptation responses by the households, while allowing for the correlation of unobserved and unmeasured factors (error terms). When households decide to adopt certain responses based on their perceptions of climate variability, correlations may arise due to synergies (positive correlation) and tradeoffs (negative correlation) among these perceptions. The essence of using MVP stems from the fact that smallholder farmers, by observation, have multiple climate variability perceptions. Some of these perceptions produce a synergistic (complementary) effect, while others produce a trade-off effect (substitutes). As a result, failing to account for these unobserved factors and effects among the perceptions will result in biased and inefficient estimates (Greene, 2003).

The MVP econometric model is characterized by a set of binary dependent variables (Y_ij), such that:

And:

where i = 1,2,3 denotes the climate variability perceptions such as 1 = increasing temperature 2 = decreasing precipitation, 3 = increasing drought; and j = 1,…, n and n denote the sample size. The equation (1) assumption is that a rational jth household has a latent variable, Y^*_ij, which captures the unobserved preferences derived from the ith climate variability perception. This latent variable is assumed to be a linear combination of adaptation responses adopted (X_ij), as well as unobserved characteristics captured by the stochastic error term ε_ij. The vector of parameters to be estimated is denoted by βi. Given the latent nature of Y^*_ij, the estimations are based on observable binary discrete variables Y_ij, which indicate whether or not a household perceived the ith climate ntvariability perception. If the specific perceived climate variability is independent of another climate variability perception, then equations (1) and (2) specify univariate probit models where information on households’ climate variability perceptions does not alter the prediction of the probability that they have another perceived climate variability. As we assumed that a household can have multiple climate variability perceptions, the error terms in equation (1) jointly follow a multivariate normal (MVN) distribution, with 0 conditional mean and variance normalized to 1. Where (⁠$ρ1$⁠,$ρ2$⁠, $ρ3$⁠) distributed MVN (0, Ω˝) and the symmetric variance-covariance matrix ˝ is given by:

where (⁠$ρim)$ denotes the pairwise correlation coefficient of the error terms corresponding to any two perceived climate variability equations to be estimated in the model.

The off-diagonal elements in the covariance matrix, ρ_im, which represent the unobserved correlation between the stochastic component of the ith and mth perceived climate variability, are important. This assumption means that equation (2) tests whether an MVP model was appropriate for the analysis or whether the univariate probit model suffices for the analysis.

To determine the effect of independent variables on climate variability perceptions against adaptation response, the final analysis contains marginal effect analysis results based on equation (3) (Greene, 2012). Therefore, the marginal effect of adaptation responses (X_ij) was calculated because marginal effects measure the impact that a specific adaptation response has on the perceived climate variability of households while all other variables are held constant:

The second objective of the study was focused on the impact of climate variability adaptation responses on agro-pastoral and pastoralists food security. To analyze the impact of climate variability adaptation responses on agro-pastoral and pastoralists food security the study used an endogenous switching regression (ESR) model. The model accounts for potential endogeneity and self-selection bias and allows interactions between climate change decisions and other explanatory variables in the household food security outcome function. Two instrumental variables, climate change training and access to weather information, were used in the endogenous switching regression model to address potential endogeneity issues by providing exogenous variation that influences the treatment variables, ensuring a more accurate estimation of the model’s parameters. Both instruments passed the falsification tests, demonstrating that they are efficient and consistent for use in the model. The model consists of two equations: a selection equation (adoption of adaptation responses) and an outcome equation (food security). The selection equation models the probability of selecting into the group that adopts the outcome, while the outcome equation models the relationship between the binary outcome and the explanatory variables (Wooldridge, 2003). The selection equation and the outcome equation are estimated jointly using maximum likelihood estimation.

To account for selection biases, an ESR model was used for the outcome variables (food security) in which farmers face two regimes: (Regime 1) to use adaptation responses and (Regime 2) not to use adaptation responses, as defined as follows:

Here $Gi*$ is a latent variable that determines the utility obtained whether the household i used adaptation responses or not; $Uji$ is the outcome variable value of a household i who used adaptation responses and j = Regime 1 and Regime 2; $Zi$ is a vector of characteristics that influences the decision to use adaptation responses but not the outcome variable value. $Xi$ is a vector of household characteristics that are thought to influence the decision to adopt the innovation, β₁, β₂ and γ are vectors of parameters, and $ui$⁠, $ϵ1i$ and $ϵ2i$ are the error terms?

The regression model coefficient of adaptation responses, which measures the impact of adaptation responses, should be random. But in the case of adaptation responses, farmers freely choose the particular adaptation responses they want to adopt with their consent. Hence, there is the problem of self-selection, which leads to selection bias. The decision to use adaptation responses is likely to be affected by unobservable characteristics that may be correlated with the outcome variables (food security). Finally, the error terms in equations (4), (5) and (6) are assumed to have a tri-variate normal distribution(⁠$υ$⁠, ε₁, ε₂) $∼$n (0,$Σ$⁠):

where $συ2$ is the variance in the adoption equation (6), which is equal to 1, since the coefficients are estimable only up to a scale factor, $σ12$ and $σ22$ are the variances of the error terms in the outcome variable functions (4) and (5), and $σ1υ$ and $σ2υ$ represent the covariance of $υ$_i and ε_1i and ε_2i. Since equations (4) and (5) are not observed simultaneously, the covariance between ε_1i and ε_2i is not defined (reported as dots in the covariance matrix. An important implication of the error structure is that, because the error term of the selection equation (6)$ui$ is correlated with the error terms of the outcome variable functions (4) and (5) (ε_1i and ε_2i), the expected values of ε_1i and ε_2i conditional on the sample selection are nonzero:

where $ϕ$ (.) is the standard normal probability density function, $Φ$(.) the standard normal cumulative density function, and λ_1i=$ϕ(Ziα)Φ(Ziα)$ and λ_2i= − $ϕ(Ziα)1−Φ(Ziα)$⁠. If the estimated covariances $σ1υ$ and $σ2υ$ are statistically significant, then the decision to adopt and the outcome variable are correlated, that is evidence of endogenous switching was found and reject the null hypothesis of the absence of sample selectivity bias. An efficient method to estimate endogenous switching regression models is full information maximum likelihood estimation (Lee and Robert, 1984). The logarithmic likelihood function, given the previous assumptions regarding the distribution of the error terms, is as follows:

where θ_ji =$Ziα+εijσjρj(1−ρj2)$⁠, j = 1, 2, with ρ_j denoting the correlation coefficient between the error term $ui$ of the adaptation response equation (6) and the error term ε_ji of equations (7), respectively. The ESR model can be used to compare the expected outcome variable of the pastoralists and agro pastoralist households that used a particular adaptation responses to the farm households that did not used adaptation responses, and to investigate the expected outcome variable result in the counterfactual hypothetical cases that the pastoralists and agro pastoralist households that did not used adaptation responses, and that if the pastoralists and agro pastoralist household who did not used adaptation responses what is they used:

In addition, the effect of the treatment “to use adaptation responses” on the treated (ATT) was calculated as:

which represents the impact of adaptation responses on the outcome variable result of the pastoralist and agro pastoralist households that actually used a particular adaptation response. Similarly, the effect of the treatment on the untreated (TU) for the pastoralist and agro pastoralist households that actually did not use to be calculated as:

Food security was measured using proxy analysis using household food insecurity access scale, household dietary diversity score, food consumption score. Household food insecurity access scale (HFIAS) consists of two types of related questions. The first question type is called an occurrence question. There are nine occurrence questions that ask whether a specific condition associated with the experience of food insecurity ever occurred during the previous four weeks (30 days). Each severity question is followed by a frequency-of-occurrence question, which asks how often a reported condition occurred during the previous four weeks. The HFIAS score is a continuous measure of the degree of food insecurity (access) in the household in the past four weeks (30 days). First, a HFIAS score variable is calculated for each household by summing the codes for each frequency-of-occurrence question. The higher the score, the more food insecurity (access) the household experienced. The lower the score, the less food insecurity (access) a household had experienced. The HFIAS indicator categorizes households into four levels of household food insecurity (access): food secure and mild, moderately and severely food insecure. Households are categorized as increasingly food insecure as they respond affirmatively to more severe conditions and/or experience those conditions more frequently (Coates et al., 2007).

Table 1 shows pastoral and agro-pastoral households’ perceptions about the situation of temperatures, rainfall and drought in their locality. Specifically, 94.72% and 83.93% of respondents perceived an increase in temperatures and drought frequency and magnitude, respectively, while 87.77% felt rainfall in their locality has been decreasing. Key informants and focus group discussion participants emphasized that climate variability and extremes are seriously affecting the rangeland and, hence, eroding the adaptive capacity of pastoral and agro-pastoral households. However, pastoral and agro-pastoral households were not passive victims to the adverse effect of climate variability and extremes (Ayal et al., 2017). Table 1 shows that pastoral and agro-pastoral households practice adaptation responses, including feed storage (61.9%), destocking (39.6%), rainwater harvesting (67.4%), haymaking (59.0%), growing feeder (31.2%), index-based livestock insurance (27.34%) and drought-resistant varieties (68.4%) in the livestock sector. Likewise, adjusting planting and harvesting dates (59.5%), water and soil conservation (71.5%) and growing early maturing varieties (61.8%) were adaptation responses implemented by agro-pastoral households in the crop sector. Pastoral and agro-pastoral households also implement coping responses such as using media (14.9%), borrowing from a credit union (36.5%), livelihood diversification (50.0%), saving crop seed or money (59.8%) and receiving free support (76.8%).

The specification tests for the multivariate Probit model confirm its validity. The Wald test X2 = 132.70, p = 0.0000) demonstrates that the covariates jointly have a significant impact on the outcomes, indicating strong explanatory power. The likelihood ratio test for ρ21 = ρ31 = ρ32 = 0 rejects the null hypothesis (p = 0.0000), showing significant correlations between the error terms of the equations. This result validates the use of the multivariate Probit model over independent Probit models. The log likelihood value of −244.04 further supports the overall fit of the model (Table 2).

Table 2 shows the correlation between climate variability perceptions among pastoral and agro-pastoral households and their adoption of various climate change adaptation responses, resulting from a multivariate Probit model. The result reveals a statistically significant negative relationship between the adoption of the feeder growing strategy by pastoral and agro-pastoral households and their perceptions of temperature change in their locality. This suggests that these households may be hesitant in acknowledging the adverse impact of rising temperatures on pasture scarcity and livestock feed intake. Similarly, there is a statistically significant negative association between the adoption of index-based livestock insurance and temperature change perception. This implies that households may not be inclined to participate in livestock insurance programs aimed at mitigating the effects of temperature change. This reluctance could be attributed to a lack of awareness among pastoral and agro-pastoral households regarding the direct and indirect consequences of temperature change, such as reduced pasture availability, decreased livestock feeding intake and an increased prevalence of diseases.

As presented in Table 2, pastoral and agro-pastoral households were practicing climate change adaptation responses in the livestock sector, including feed storage, destocking, rainwater harvesting, haymaking and keeping drought-resistant livestock to address challenges related to the unpredictability of rainfall behavior and drought in their locality. The results suggest that these adaptation strategies have positive and statistically significant associations with the decreasing rainfall and increasing of drought frequency and magnitude perceptions of pastoral and agro-pastoral households. This indicates that these households are well aware of the impact of decreased rainfall and recurrent drought on their food security. Conversely, the adoption of drought-resistant breeds shows a positive and statistically significant effect on rainfall perception. This suggests that the declining trend of rainfall in their locality has encouraged households to select drought-resistant breeds to mitigate the adverse effects of water scarcity.

However, index-based livestock insurance exhibits a negative and statistically significant relation with decreased rainfall and increasing drought frequency and magnitude perceptions, suggesting a lack of interest among households in participating in such insurance programs. Moreover, water and soil conservation as a crop adaptation response show a positive but insignificant association with rainfall perception. This implies that the reduction of rainfall in their locality did not motivate households to implement water and soil conservation practices to address soil moisture stress. The adoption of soil and water conservation responses by pastoral and agro-pastoral households shows a statistically significant relationship with their perceptions of increasing drought. Despite the relevance of soil and water conservation technology in capturing water and enhancing rangeland moisture, these households appear to lack interest in adopting this response. This reluctance may stem from a lack of emphasis by the government and NGOs in encouraging them to implement such practices. Additionally, the adoption of early-maturing varieties as a crop adaptation response exhibits a negative statistically significant relation with pastoral and agro-pastoral rainfall perception. It appears that pastoral and agro-pastoral households were not inclined toward early-maturing crop varieties to mitigate the effects of reduced rainfall. This could be attributed to a lack of input and technical support from development agents. According to key informants, crop production is a relatively new livelihood option, typically practiced by pastoralists affected by drought.

Table 2 shows that pastoral and agro-pastoral households employ nonagricultural climate change adaptation responses, such as borrowing from credit unions, receiving forecast information from media, saving crop seeds or money, receiving support from clean and family members and livelihood diversification. The adoption of borrowing from credit unions and receiving forecast information from media exhibits a negative and statistically significant association with pastoral and agro-pastoral perceptions of temperatures, drought and rainfall. This suggests that their inclination to adopt borrowing from formal institutions and rely on media forecast information may be weak, possibly due to a lack of awareness and access to financial institutions. The lesser dependence of pastoral and agro-pastoral households on media forecast information implies a heavy reliance on traditional weather forecasting sources.

Conversely, saving crop seeds or money, receiving support from clean and family members and livelihood diversification show positive and statistically significant associations with households’ perceptions of increasing temperatures, drought frequency and magnitude and decreasing rainfall. This indicates that the practice of sharing risks among pastoral and agro-pastoral communities persists to support victims of climate change and extremes. Key informants also reported that households may postpone cultural festivals, marriages, etc., based on traditional forecast outputs to conserve resources during unfavorable seasons. The pastoral and agro-pastoral households’ interest in engaging in livelihood diversification confirms that their asset background is eroded by climate change and extremes. Households in the study site tend to pursue non-pastoralism livelihoods only when their livestock assets are weakened. However, if these livelihood options are properly managed and viable alternatives are identified, they could serve as a means to alleviate pressure on the rangeland and ensure food security.

The pairwise correlation coefficients matrix presented in Table 3 indicates that temperature perception is moderately correlated with rainfall perception and weakly correlated with drought perception. Conversely, rainfall perception and drought perception exhibit a strong positive correlation. These findings imply that households’ perceptions of temperature, rainfall and drought are interconnected, with some degree of consistency in how they perceive these climate change variables. Hence, this test shows the appropriateness of the multivariate probit (MVP) model for modeling the interaction of climate variability perception and adaptation responses among pastoralist and agro-pastoralist communities

Table 4 depicts the association between pastoral and agro-pastoral households’ perceptions of temperature, rainfall and drought and their Household Dietary Diversity Score (HDDS). The analysis aims to explore how households’ perceptions of temperature, rainfall and drought are linked to the utilization dimension of food security. The findings reveal that among households perceiving an increase in drought, 47.1% have a low HDDS, 30.9% have a medium HDDS and 22.0% have a high HDDS. The chi-square test result (χ² = 5.61, p < 0.05) indicates a statistically significant association between drought perception and household dietary diversity scores. Thus, pastoral and agro-pastoral households experiencing an increase in drought frequency and magnitude are more likely to consume less diversified food compared to those who do not perceive an increase in drought. The mean dietary diversity score for households not perceiving drought is 5.1, while for those perceiving drought, the mean household dietary diversity score is 4.4. This indicates that drought perception does not appear to influence the utilization dimension of household food security.

Table 5 illustrates the correlation between perceptions of temperature increase, rainfall decrease and drought increase among pastoral and agro-pastoral households, and the accessibility dimension of their food security. Their level of food insecurity is measured by the Household Food Insecurity Access Scale (HFIAS). Notably, households that perceived a decrease in rainfall had an average HFIAS score of 15.03. The t-test results (t = 5.29, p < 0.05) indicate a statistically significant difference in HFIAS scores between households that perceived rainfall decrease and those that did not. Similarly, households that perceived higher levels of drought had an average HFIAS score of 14.86, with a significant difference (t = 6.78, p < 0.01) compared to households that did not perceive higher drought levels. There are significant differences in the scores based on rainfall perception and drought perception. Households that perceive higher rainfall levels and higher levels of drought tend to have higher mean household food insecurity access scale scores, indicating greater food insecurity. These findings highlight the potential impact of climate change perceptions on household food insecurity in pastoral communities.

Table 6 shows the effect of various climate variability adaptation responses on food security among pastoral and agro-pastoral households. The effectiveness of each adaptation response was evaluated by comparing food security outcomes between adopters and non-adopters. The decision stage indicates whether households adopted a particular adaptation response or not. The t-test was employed to assess the statistical significance of the treatment effects. For instance, among households that adopted feed storage as an adaptation response, the Average Treatment Effect (ATT) on Household Food Insecurity Access Scale (HFIAS) is 14.1, whereas the Average Treatment Effect on the Untreated (ATU) group (non-adopters) is 22.4. The t-test result (t = −8.3, p < 0.01) suggests a statistically significant difference in food security outcomes between adopters and non-adopters of feed storage. This indicates that adopting feed storage has a significant positive impact on food security among pastoral and agro-pastoral households.

Similarly, among households that adopted growing feeders as an adaptation response, the Average Treatment Effect (ATT) on Household Food Insecurity Access Scale (HFIAS) is 15.6. The ATU group has an average treatment effect of 18.4. The t-test result (t = −2.8, p < 0.01) suggests a statistically significant difference in food security outcomes between adopters and non-adopters of growing feeders. Furthermore, among households that adopted soil and water conservation measures as an adaptation response, the ATT on HFIAS is 14.2. The ATU group has an average treatment effect of 23.8. The t-test result (t = −9.6, p < 0.01) indicates a statistically significant difference in food security outcomes between adopters and non-adopters of soil and water conservation measures.

Moreover, among households that adopted saving crop seed or money as an adaptation response, the ATT on HFIAS is 14.6. The ATU group has an average treatment effect of 21.3. The t-test result (t = −6.7, p < 0.01) suggests a statistically significant difference in food security outcomes between adopters and non-adopters of saving crop seed. Furthermore, among households that received free support as an adaptation response, the ATT on HFIAS is 14.7. The ATU group has an average treatment effect of 21.7. The t-test result (t = −7.0, p < 0.01) indicates a statistically significant difference in food security outcomes between households that received free support and those that did not. Furthermore, among households that used media as an adaptation response, the ATT on HFIAS is 12.8. The ATU group has an average treatment effect of 21.5. The t-test result (t = −8.7, p < 0.01) suggests a statistically significant difference in food security outcomes between households that used media and those that did not.

The analysis demonstrates that several climate variability adaptation responses have a significant effect on pastoral household food security. Adopters of feed storage, growing feeders, soil and water conservation measures, saving crop seed, free support and the use of media tend to have better food security outcomes compared to non-adopters. These findings underscore the importance of implementing and promoting these adaptation responses to enhance food security in pastoral communities facing climate variability challenges.

The pastoral and agro-pastoral households’ perceptions of climate variability and adaptation responses have provided valuable insights. While most households demonstrated awareness of changing temperatures, rainfall patterns and drought frequency and magnitude, variations in perceptions were observed. This finding aligns with previous research (Kemal, Mohammed and Lelamo, 2022; Leal Filho et al., 2020; Radeny et al., 2019) and suggests that demographic and socioeconomic factors may influence these perceptions. This underscores the importance of tailored awareness initiatives and the integration of local knowledge and indigenous practices in crafting effective adaptation strategies (Radeny et al., 2019).

The diverse array of adaptation responses employed by pastoral and agro-pastoral households reflects their proactive efforts to address climate variability challenges and ensure food security. Strategies such as livestock management, crop diversification and water resource management have emerged as key approaches. Identifying and promoting feasible adaptation measures could enhance adaptive capacity and mitigate the impacts of climate variability on food security (Yared et al., 2022).

These findings hold significant implications for household food security in pastoral and agro-pastoral communities facing climate variability. Rufino et al. (2013) note that the adoption of diverse adaptation responses is associated with improved food security outcomes. Livestock-based strategies have helped sustain livestock productivity and reduce vulnerability to droughts, ensuring a reliable source of food and income. Similarly, crop diversification measures have enhanced agricultural production, offered a broader array of food options and improved dietary diversity and nutrition. Water resource management efforts have contributed to increased water availability for both livestock and crop cultivation, crucial for maintaining sustainable food production systems.

This study underscores the importance of aligning perceptions of climate variability, adaptation responses and food security goals. It stresses the necessity for context-specific and multidimensional approaches that consider local knowledge, socio-economic circumstances, institutional support, social capital and external influences. Policymakers, practitioners and researchers can leverage these insights to develop targeted interventions, policies and programs aimed at enhancing climate resilience and improving food security in pastoral communities.

The study assesses climate variability perceptions, adaptation responses and their effect on household food security among pastoral households in the Borana Zone, Southern Ethiopia. It found that pastoral communities had a commendable understanding of climate variability and extremes, utilizing various adaptation responses. Adaptation strategies encompassed livestock management, crop diversification and water resource management, demonstrating resilience to climate variability. These approaches resulted in enhanced food security, with livestock-focused methods ensuring stability in food and income, crop diversification boosting agricultural production and dietary diversity and water management increasing availability for both livestock and crops.

Based on the study’s findings, recommendations can be proposed to strengthen climate resilience and food security among pastoral households.

Strengthen institutional support: Enhance the availability and accessibility of support services, extension programs and climate information tailored to the specific needs of pastoral communities. This will enable pastoral households to make informed decisions and implement effective adaptation measures.

Foster social capital: Promote community-based initiatives and strengthen social networks within pastoral communities. Facilitate platforms for knowledge-sharing, cooperation and collective action, which can accelerate the dissemination and adoption of adaptation practices.

Improve access to financial resources: Develop financial mechanisms such as microfinance programs and insurance schemes to provide pastoral households with access to credit and risk management tools. This will enable them to invest in adaptation strategies and cope with climate variability shocks.

Promote sustainable natural resource management: Encourage sustainable land use practices, water resource management and conservation efforts. Advocate for the adoption of climate-smart agricultural techniques and technologies that enhance productivity and resilience.

Enhance education and awareness: Invest in education and awareness campaigns to enhance climate change literacy and understanding among pastoral households. This will empower them to make informed decisions, adopt suitable adaptation measures and actively engage in policy processes.

This study has several limitations to consider. Firstly, the findings are based on a specific study area, limiting their direct generalizability to other pastoral communities. Future research could broaden the scope to include multiple regions to obtain a more comprehensive understanding of climate variability adaptation in pastoral systems. besides,

Secondly, the study primarily relied on self-reported data, which may be susceptible to recall bias and social desirability bias. Integrating self-reported data with objective measurements and remote sensing data could offer a more thorough and accurate assessment of adaptation responses and their impacts.

Finally, the study focused on adaptation responses and their implications for food security. Future research could delve into the long-term sustainability and scalability of these strategies, as well as their effects on environmental conservation and social equity.

In conclusion, this study underscores the significance of comprehending climate variability perceptions, implementing diverse adaptation responses, and their positive effects on household food security in pastoral communities. The findings offer valuable insights for policymakers, practitioners and researchers to design context-specific interventions and policies that bolster climate resilience and enhance food security in pastoral regions.