Borena district is situated in the northcentral part of Ethiopia. According to Abebe (2023) and Nurye et al. (2023), the district is located between 10°34′02″ and 10°53′16”N and 38°27′39″ to 38°54′20″E (Figure 1). Its estimated area is 937 km². The district is bordered by the Abay River on the west, which divides it from the East Gojjam Zone, Wegdie district on the south, Sayint district on the northeast, Mehal Sayint district on the north, Legambo district on the east, and Sayint district on the west (Abebe, 2023). The center of its administration is Mekane Selam town.

The district is defined by its varied topography, which ranges from 500 meters to 3,700 meters above sea level. Due to its diverse topography, the district is divided into four traditionally recognized agroclimatic zones: Kola (32%), Woinadega (47%), Dega (20%), and Wurch (1%; Shiferaw and Singh, 2011; Abebe, 2023). The area is distinguished by a summertime rainfall maximum from June to September and intermittent showers from March to May. Mean annual rainfall in the district ranges from 889 to 1,500 mm (Shiferaw and Singh, 2011; Nurye et al. 2023). The district’s mean temperature varied between 14 and 19°C, while its annual maximum and minimum temperatures ranged from 13 to 31°C and −0.5 to 16°C, respectively. Shiferaw and Singh (2011) state that the district’s upper northwest region has the lowest temperature and is primarily Wurch-type, while the southwest region has the highest temperature and is classified as Kolla climatic. There is minimal natural cover in the area due to heavy deforestation. Other native tree species are almost non-existent, except for a few scattered acacia trees, farmer-planted eucalyptus patches, and a few bushes and shrubs (USAID, 2009). Vertisols, rich in clay and have deep cracks when dry, make up most of the soil types in Borena. Even though these soils are suitable for farming, erosion is frequent (Shiferaw and Singh, 2011).

The district’s overall population was projected to be 187,915 people in 2017. There were 94,756 females and 93,159 males among them. The district’s population density is 153.96 persons per km², greater than the South Wollo zone (143.58 persons per km²). Borena’s economy is mainly focused on agriculture, where the cultivation of crops and livestock rearing constitute the primary economic pillars of the local communities in the area (Asfaw, 2014; CSA, 2021). Cultivation of both long and short-cycle crops is entirely rainfed, except in a few select locations that use small-scale, traditional irrigation practices. Alongside growing crops, the communities have a higher livestock asset due to their significance for cash income, milk production, land plowing, and food (Mekonen and Berlie, 2021). Cattle, Sheep, goats, and equines are the primary livestock types raised in the area and contribute significantly to household earnings. However, various factors, including severe frost, hailstorms, limited input availability, soil erosion, water scarcity, and low and irregular rains with late-onset and early cessation, contribute to low agricultural productivity, resulting in chronic food insecurity in the area (USAID, 2009; Mekonen and Berlie, 2021). Borena’s native vegetation is primarily composed of grasslands and woodlands. However, excessive grazing and deforestation have seriously degraded the vegetation cover. Several perineal rivers drain the district, and the Abbay River, which forms its western boundary, offers a sizable potential water supply for irrigation (Shiferaw and Singh, 2011; Asfaw, 2014).

Though a total of 388 household heads were considered as a sample for this study, Due to incomplete responses and some technical problems, the responses of 8 household heads (2.1%) were not included in the analysis. As a result, the responses 380 (97.9% response rate) were included in the analysis. Out of the total respondents, only 27.1% have engaged in SSI, while the remaining 72.9% did not participate in irrigation activities during the study period. The finding implied that the SSI adoption rate in the study area is low compared to its potential.

As shown in Table 3, 380 smallholder household heads (86.1% male-headed while female-headed households were 13.9%) participated in the survey. The great majority are married (83.4%). Crop production (46.8%) and mixed farming (44.5%) are the principal sources of livelihood. In contrast, non-farm sources of livelihood, like petty trading and wage labor, are negligible. As displayed in Table 4, the average age of the surveyed household heads was 50.08 years, while the mean years of schooling was 2.93 years. The average family size, mean land size, and livestock assets were 5.56 individuals, 0.73 ha, and 3.92 TLU, respectively. Sampled respondents should travel 1.91 and 1.54 h on average to get to the nearest market and communication centers, respectively. An average of 1677.4 Ethiopian birr is earned from irrigation activities annually. Being in a productive age and having a relatively large family size might be a good opportunity to participate in SSI if facilities are fulfilled.

Relatively, most of the respondents have better contact with development agents. Around 96.8% of the respondents have their land. The great majority (67.6%) do have training regarding crop production and SSI, while those having training access to climate change adaptation and early warning were found to be less (29.2%). As a means of information, radio and mobile ownership is worth mentioning as one major factor. Around 45.6 and 46.8% of the respondents do have a radio and mobile phone currently working. In the study area, 48.2% of them utilize financial institutions either for credit or saving; 27.9% have income sources from remittance; around 33.2% are members of cooperative associations. Non-farm sources of income were found to be negligible (Figure 2), but around 46.8% replied that income sources besides agriculture might be the income earned from safety net programs and direct humanitarian supports. Around 73.1% of the respondents had faced either partial or total crop failure, 62.6% had received food aid at least once, 72.6% had information regarding climate change, and 62.5% believed climate change could be adapted. The Phi correlation coefficient (ɸ = 0.395, p < 0.001) revealed that having information regarding the impacts of climate change has a moderate effect on believing in its adaptability. Thus, dissemination of information by whatever means would enable smallholder farmers to believe in the possibility of adapting to the impacts of climate change. It would also allow them to undertake feasible adaptation strategies to overcome the effects of climate change and harness opportunities.

As depicted in Table 4, an independent sample t-test was employed to examine whether there is a statistically significant difference between irrigation users and non-users regarding different covariates. The statistically significant mean age difference was found (t = 3.548, p < 0.001) where irrigation users were found to be younger (M = 47.37, SD = 9.22) as compared with non-users (M = 51.08, SD = 9.01). On the other hand, the mean years of schooling for irrigation users were found to be (M = 4.98, SD = 4.45), which is higher than their counterparts (M = 2.17, SD = 3.38), and the difference is statistically and significantly different (t = 6.589, p < 0.001) at 95% level of significance. Another variable where a significant difference was observed in total land holding, where households having relatively better land holding (M = 0.8648, SD = 0.344 for users and M = 0.6735, SD = 0.28 for non-users) were found to be irrigation users with a statistically significant mean difference in land size at 95% significance level (t = 5.611, p < 0.001). Similarly, households that have a better size of TLU assets were found to be irrigation users than their counterparts (M = 5.38, SD = 2.89 for users and M = 3.37, SD = 2.17 for non-users), and the mean TLU difference was statistically significant (t = 6.421, p < 0.001). Distance from nearby markets was another variable that showed a statistically significant mean difference where those having closest proximity to market centers were found to be more irrigation users (M = 1.639, SD = 0.951 for users and M = 2.017, SD = 1.33 for non-users) and (t = 2.644, p < 0.005). A statistically significant difference in terms of family size was also observed (M = 5.99, SD = 1.32 for users and M = 5.04, SD = 1.27 for non-users) and (t = 6.43, p < 0.001). Generally, the younger, those with better educational status, those households with larger family sizes, those with relatively better land holding and livestock assets, and those residing close to market areas were found to be irrigation users more than the other way round. The Chi-square result revealed that the likelihood of participating in SSI activities is higher for those who are microfinance beneficiaries, those who have remittance income sources, cooperative members, heads having responsibility in kebele, those who believe that climate change is adaptable, those who reside in upper and middle-lands, those having radio (source of information), having training exposure on SSI and having non-farm income sources. The proportion difference was statistically significant at a 95% significance level (see Table 5 for details).

Figure 3 depicts the top three most commonly practiced irrigation types during data collection: river or stream diversion, night storage, or using nearby springs, and underground water or hand-dug wells. Rainwater harvesting, small-scale dams, and motor pumping are also practiced to a lesser extent. Most of the irrigation types being implemented are not water efficient. Drip irrigation and similar water-efficient irrigation technologies are not practiced at all. An interviewed irrigation expert from the Borena Woreda agricultural office stated that most of the irrigation practices practiced in the Woreda are traditional and not water-efficient by nature. During field observation, we observed a furrow irrigation (Figure 4) where a significant amount of water was wasted. Such irrigation modalities are not recommended where water scarcity is among the significant challenges partially caused by climate change.

An interviewed farmer agreed that such irrigation modalities are not water efficient, and according to the respondents, modern irrigation technologies are not affordable. Besides, for small plots of land, securing modern irrigation technologies at a high cost is not profitable from the point of view of economies of scale. As shown in Figure 4, the irrigation practice in the study was constrained by water scarcity, shortage of land, shortage of labor, and shortage of modern inputs. As shown in Figure 5 and observed during field observation, most SSI products are marketable fruits and vegetables. According to one interviewed farmer, sales of fruits and vegetables are crucial in helping cover various costs.

Interviewed farmers engaged in irrigation cited water scarcity as their main critical problem. They added that most water sources have decreased in volume, and farmers are forced to apply rationing methods for their irrigation fields. Another critical constraint mentioned by the district agricultural office’s interviewed farmers and experts is a land shortage. The district is among the highly populated areas of the region where land per capita is minimal. As a result, devoting a certain proportion of land to irrigation is challenging for most farmers. Furthermore, the traditional irrigation system is labor-demanding. So, the shortage of active labor force was another challenge in the study area (Figure 6). In addition, one of the critical issues raised during the interview was the lack of technical expertise regarding the installation, utilization, and maintenance of modern pumping. Almost all key informant participants agreed that the lack of a market for their products is not a problem (most constraints are from the supply side, not the demand side).

We employed a logistic regression model to estimate the factors that determine farmers’ participation in SSI. Multicollinearity diagnosis for continuous explanatory variables and degree of association for discrete variables were inspected using VIF, tolerance statistics, and the contingency coefficient. The maximum VIF and contingency coefficient found were 1.602 and 0.32 (see Annex 1), respectively (both are below the required threshold to consider the multicollinearity problem), which evidenced the appropriateness of the independent variables for running binary logistic regression. Correlation tests and contingency coefficients were used to screen out variables strongly associated with the dependent variable. Five variables with a correlation coefficient below 0.3 were omitted from the analysis. Accordingly, 17 variables (out of 22 candidate variables), assumed to be associated with the participation of smallholder farmers’ irrigation strategies, were included in the model (see Table 6). Among the variables, eight were statistically significant at 0.05 alpha level.

The model classification ascertained the model’s goodness-of-fit, where 84.8% of respondents were correctly classified by the model (72.6% in the constant-only case). The Omnibus Tests and −2 log-likelihood ratio test revealed that the estimated model, including a constant and the set of explanatory variables, fit the data better than the model containing only the constant. This suggests a stronger correlation exists between the odds ratio (or log of odds), the likelihood that a particular factor will influence a farmer’s decision to participate in irrigation, and the explanatory variables in the model. These factors significantly contribute to explaining smallholder farmers’ influence in adopting SSI. The overall estimated model has remarkable explanatory power, although some coefficients were insignificant when considered individually. The highly significant Chi-Square result, the overall percentage of correct prediction, and the Pseudo R² value (Cox and Snell’s of 0.366 and Nagelkerke of 0.53) all support this claim.

Moreover, a non-significant result of the Hosmer-Lemeshow goodness-of-fit test (χ² = 12.741, p = 0.121, df = 8) revealed the model’s appropriateness (well–fitting). Using the VIF and tolerance statistics, multicollinearity diagnosis for continuous explanatory variables was also examined. The outcome showed no issues with multicollinearity among the explanatory variables. Furthermore, the degree of association between each dummy/discrete variable was also assessed using the contingency coefficient. The estimated coefficients of the logit model are presented in Table 6. As indicated by the χ2 statistic, the likelihood ratio statistics is significant at 1%. This suggests that all the variables in the logit model are jointly significant in influencing smallholder farmers’ decision to partake in irrigation activities. Finally, adoption probability is estimated using the Adjusted Odd Ratio (AOR).

The educational level of the household head is among the factors determining the adoption of SSI (Ngango and Hong, 2021). As predicted, it was found that the head of the household’s education has a positive and significant effect at the 0.05 alpha level, suggesting a positive relationship with the adoption of irrigation. This indicated that, compared to their peers, household heads with greater levels of education are more inclined to practice SSI. If all else remains equal, an additional year of education could raise the likelihood that smallholder farmers will use irrigation by 1.239% (Table 6). This result is consistent with Belete and Melak (2018), Zegeye et al. (2022), and Gemeda et al. (2023). If they receive education, smallholder farmers may be better equipped to process and apply vital information from various sources. Higher-educated people are more inclined to adopt new technology since they understand their relevance better. Besides, they can look at market opportunities and have a better likelihood of taking calculated risks. Higher educated farmers are more likely to choose better adaptation strategies in response to climate change; they have a greater degree of knowledge and are therefore more able to comprehend and respond to changes that are anticipated; they are more adept at predicting future scenarios; and, overall, they typically have access to opportunities and information that may encourage climate change adaptation (Kumasi et al., 2017).

The sex of the household head is worth mentioning as a determinant factor in technology adoption, favoring male-headed ones. In this study, the likelihood of female-headed households participating in SSI activities is statistically and significantly lower than (24.6%) male-headed ones at a 95% significance level (Table 6). The finding coincides with Teha and Jainjun (2021) and Zemarku et al. (2022). This might be due to a difference in information, extension, and asset access, where males are better than females. Besides, women are more engaged in household chores like cooking and childcare activities; in most cases, they are endowed with fewer resources and less exposed to new information, which hampers their adoption rate of new technologies. Gender plays a role in adaptation decisions (man-headed households often implement some adaptation strategies). Since women are more involved in household chores and men perform a significant portion of outdoor farming activities in developing nations, male-headed households have greater access to farming experience and knowledge of various adaptation strategies (Nigussie et al., 2014). A study in Ethiopia found that women handle almost all household chores, leaving little time for repairs or other financial activities (Nigussie et al., 2014; Mare, 2017).

Moreover, Mare (2017) has shown that women in Ethiopia encounter obstacles when obtaining loans, inputs, extension services, and information—all essential when undertaking adaptations. The female participant in KII, age 52, holds a distinct perspective. This respondent says that having a male or female head of household is not a significant obstacle; instead, because traditional irrigation methods are labor-demanding, it is more important to have an active labor force in the family. Irrigation activities are common among families with more productive labor forces.

By its nature, SSI practice is a laborious activity where younger ones are expected to be engaged more. The impact was statistically significant at a 95% significance level. As predicted, the probability of SSI declines with household age. The likelihood of participation decreases by about 9.45% when the age increases by 1 year (Table 6). A similar finding was reported by Teha and Jianjun (2021), Zemarku et al. (2022), and Gemeda et al. (2023). The reason for the inverse relationship between adoption and age is that older farmers are less likely than younger farmers to adopt new agricultural technologies because they are risk-averse and conservative. However, our finding differs from Feyisa (2020), where a positive association was reported.

Location is another covariate considered a possible factor affecting the probability of households’ engagement in irrigation, mainly due to differences in the availability of resources. In this study, location was found to significantly affect the likelihood of participation in SSI, where households in middle and upper lands participated more in irrigation than those residing in the lower lands due to better availability of water resources and rainfall amount. The probability of households in the middleland and upper-land participating in SSI was 13.929 and 7.605% more likely than those living in the lower-land, respectively, at a 95% significance level (Table 6). Our finding corroborates with Geddafa et al. (2021), Zemarku et al. (2022), and Zegeye et al. (2022). This negative association implies that the farther the plot of land, the lesser would-be farmers’ initiative to participate in SSI due to the opportunity cost of training to and from an irrigation plot, and proper surveillance of the farm activities might be challenging when distance increases from the homestead.

SSI activity is labor demanding sector. At the 95% significance level, the size of the entire household influenced people’s participation in small-scale activities. Participation probability increases at 1.936% when family size increases by one unit (Table 6). The result is consistent with Geddafa et al. (2021), Zegeye et al. (2022), Zemarku et al. (2022), and (Gemeda et al., 2023) that reported a positive relationship between the labor force of a family and the propensity of adopting irrigation. Having a larger family size encourages participation in diversified activities, including irrigation. A different result was revealed by Challa and Tilahun (2014), where family size was negatively and significantly associated with the rate of adoption. Although the relationship between family size and the adoption rate of agricultural technology adoption is not clear-cut, it is reasonable to assume that a large labor force may accelerate the adoption rate of labor-intensive technologies. As SSI participation requires more labor, a household with a more significant labor force is more likely to adopt irrigation than one with a smaller labor force. In countries like Ethiopia, family labor is among the most widely utilized inputs for irrigation. It is primarily used for plot preparation, canal construction, diverting water from rivers, weeding, watering, and other purposes. As a result, households with a higher labor force are more likely to adopt irrigation than households with a smaller labor force, and SSI participation requires more labor.

Land size matters for diversifying income sources in rural areas. The participation of households with better land size was higher, and the result was statistically significant at a 95% significance level (Table 6). An increase in land size by 1 hectare increases the likelihood of engagement in irrigation by about 8.609%, citreous paribus. A similar finding was reported by Challa and Tilahun (2014), Belete and Melak (2018), Feyisa (2020), and Gemeda et al. (2023). More productive and profitable households may arise from better arable land and be more likely to embrace new agricultural technologies. Furthermore, it might be possible to distribute failure risks with increased farmland. Both interviewees concurred that the district’s meager land per capita makes it difficult for farmers to diversify their agricultural operations, including setting aside a portion of their land for irrigation. Having responsibility in local administration exposed individuals to better information regarding market and capital, which, in turn, encouraged them to participate in income-generating activities like irrigation. Keeping all other factors constant, being a member of local administration has increased the probability of engagement in irrigation by 2.069% at a 95% significance level (Table 6). Having responsibility in local administration creates social capital and trust to secure input, information, and finances, which are crucial in increasing the probability of participation in SSI. A similar finding was reported by Feyisa (2020) and Ngango and Hong (2021). The plausible reason is that participation in local leadership can help to have more access to financial resources and information, share experiences with others, and create a more social network. Our result differs from Teha and Jianjun's (2021), where a non-significant relationship between social association membership and SSI adoption was reported.

On the other hand, the incidence of crop failure due to different factors discourages farmers from engaging in crop production and might shift towards non-farm income sources. We found that the likelihood of engaging in irrigation decreased by about 38.9% whenever the household faced crop failure problems (Table 6). Our result is in line with the findings of Ngango and Hong (2021) and Gemeda et al. (2023), where individuals who faced crop failure due to climate-related hazards and those who perceived similar risks tend to shift into non-agricultural activities. This probably might be due to a shift into non-farm activities whenever there is frequent crop failure, which discourages smallholder farmers and tends to participate in non-agricultural sources of income as a risk aversion strategy. The impact of most institutional factors (training, radio ownership, distance from market, microfinance access, association, having non-farm income source, remittance, TLU) was found non-significant in the study area through hypothesized to have positive impacts in determining the likelihood of participation of farmers in irrigation activities.

Since the study depends on cross-sectional data, it did not enable the examination of the dynamism through time and limits the control of selection bias due to problems of unobserved heterogeneity. As a result, further research with panel data is recommended to better capture the effects of irrigation over time. Future research should pay particular attention to the impacts of irrigation on income, poverty level, consumption expenditures, and food security status. Besides rigorous irrigation potential assessment, GIS and remote sensing techniques are essential.