1 Introduction

Agriculture is threatened by 20⁰C mean temperature at the global level in the long-term period of 2100 having impacts on food security [1]. Seasonal rainfall and temperature variability have an impact on African agriculture [2,3,4,5]. For instance, by 2050, it is estimated that 600 million people in Africa will be experiencing water stress, which puts agricultural production and, ultimately, food security at risk [6].

Ethiopia, a country in sub-Saharan Africa whose economy is dependent on climate-sensitive agriculture, is not an exception in this regard [7]. To increase the food security level by reducing the effects of rainfall and temperature variability, some adaptation strategies implemented by rural households have been documented in scholarly works [8,9,10]. Studies have given less attention to whether or not these responses help farmers increase crop productivity to ensure food security [12]. For instance, a study by [13] examined the role of knowledge about climate and adaptation strategies in Southern Ethiopia. Similarly, a study by [14] focused on farmers' perceptions of climate change and techniques for adaptation in Central Ethiopia. Of course, a study by [15] examined the impacts of adaptation strategies on food security in Muger sub-basin of the upper Blue Nile in Ethiopia. However, it is a different geographical setting from the current study area. A study by [16] focused on integrated land management and physical and biological conservation practices in the Ghibe III dam area. Similarily study by [17] focused on farmers response to climate change in low lands in Southern Ethiopia Several studies [2, 18,19,20] have not given attention to model such as propensity score matching for impact assessment of adaptation strategies on food security under climate variability which may limit estimating the impacts.

In addition, there has been limited use of adaptation strategies due to their unaffordability for poor farmers [21] and the extreme climate events have been occurring today beyond the resilience of farmers in Ethiopia [22]. Only, small farmers used technologically advanced adaptation systems in Southern Ethiopia [17]. The use of adaptation strategies and studying the effects of adaptation techniques that affect the food security of households have received less attention. In the study area, investigating the impact of household-level adaptation strategies on food security is timely and relevant. The reason behind this is that Ethiopia's economy is based on agriculture which generates around 33.88% of the nation's GDP [23] and is entirely based on seasonal rainfall which fluctuates from time to time. Second, the study area is near the site of a significant national project, the Ghibe III hydroelectric dam which was started in 2006 and may have contributed to current environmental changes that exacerbate local climate variability added to the historical scenarios of climate change in the area. Moreover, understanding the socioeconomic aspects that influence small-plot households is also necessary, as these characteristics were all found to be significantly associated with the households' decisions to adopt different adaptation techniques [11, 24,25,26,27,28,29].

This study would have significance in which policy makers for example, investments in technologies and methods that increase resilience to climate change, such as enhanced irrigation systems, drought-tolerant agricultural varieties, and agroforestry practices, might be prioritized by policymakers. Secondly, effective extension services are crucial for disseminating climate adaptation and mitigation strategies to farm households. Policymakers may focus on strengthening extension programs, providing training to farmers, and ensuring the availability of up-to-date information and resources. Thirdly, smallholders can facilitate access to credit and insurance schemes tailored to fulfill the demands of rural households. This can help farmers cope with climate-related shocks and invest in climate-resilient practices. Fourth, the study may highlight the importance of diversifying income sources for rural households.

Policymakers can support initiatives that promote income diversification, such as promoting off-farm employment opportunities, value addition to agricultural products, and the development of non-agricultural sectors. climate variability can disrupt market access for farmers, leading to reduced food availability so that policymakers can focus on improving transportation infrastructure, storage facilities, and market linkages to ensure farmers effectively sell their produce and get access to markets in the condition of climatic variability.

The international community would benefit from gaining lessons from practically effective adaption strategies halting climate variability in the study area. Thus, this study was conducted to assess the role of adaptation strategies on rural household food security in the face of climate variability in surrounding communities of the Ghibe III dam area, Middle parts of the Omo basin, Southern Ethiopia.

2 Materials and methods

2.1 Description of the study area

The study area is located between 6048′0" to 6058′0"N latitude and 37020′0"E to 37 36′0" Ghibe III Dam area, Middle Omo basin (Kindo koisha Woreda from wolaita zone in Southern Ethiopia (Fig. 1). The materials used in this study included XLstat software, Stata version 15, and Arc GIS for making study area maps, questionnaires, and interview papers as well as field notes during the preliminary survey and the actual survey from 2020–2021.

Fig. 1
figure 1

Map of the study area

Full size image

2.2 Research method

2.2.1 Design and approaches of the study

This study used both cross-sectional in Kindo-koisha and Loma Bosa Woredas were specifically chosen as their relative location closest to the Ghibe III dam reservoir and due to the frequent occurrence of climatic extremes and the presence of a large number of safety net beneficiaries taken from the two woredas disaster risk and prevention offices. Primary survey data were collected using a mix of qualitative and quantitative approaches.

2.2.1.1 Sampling and sample size determination

The representative households from sample kebeles in Kindo koisha and Loma Bosa woreda must be fixed by taking the sample from each kebele administration. The complete list of households obtained from sample kebele was 3761, of which 3157 from kindo koisha and 604 from Lom Bosa. Thus, by using [30] formula as specified below sample size was determined for the current study. The Cochran formula was used because married boys live with their parents, but they do not own land that has been officially registered and confirmed by the kebele—Ethiopia's smallest administrative unit. Although they are households managing their established families, they are not recognized as such at the kebele level. As a result, the following formula Eq. (1) was used to calculate the sample size for this unidentified population in this specific Ethiopian scenario.

$$\frac{{Z}^{2}p.q}{{e}^{2}}$$
(1)
$$\frac{{1.96}^{2}0.5.0.5}{{0.05}^{2}}=384$$

where,\({n}_{0}\) = the sample size, z = the selected critical value of the desired confidence level, the estimated proportion of an attribute that is present in the population. q = 1-p and e = the desired level of precision. A representative sample size was selected using multi-stage sampling techniques; firstly, seven Woredas bordering Omo River namely: Boloso Bombe, Kindo koisha and Kindo Didaye from Wolaita zone and Zaba, Gena, Loma Bosa, and Disa Woredas from Dawro Zone were selected. Secondly, two Woredas namely Kindo Koisha and Loma Bosa were chosen purposively as the area has been seriously subjected to land use land cover change due to Ghibe III Dam project since the 2000s because of their location in proximity to the dam and their areas high amount of temperature and frequent drought that left millions in hunger as reported by different local and international organizations. Thirdly, six Kebele administrations such as Fajena Mata, Mundena, Manara, and Chereche from Kindo Koisha and Zimawaruma and Chicho Hayo from Loma Bosa were selected purposively based on the history of the food insecurity situation and road accessibility for data collection. Fourthly, total households were selected from each kebele proportionally, and finally, simple random sampling was used to choose both the adapter and non-adapter households.

2.2.1.2 Data types, sources, and tools of collection

The adapters and non-adapters were identified during the data collection based on the field visit and report from agricultural extension workers employed by the government A structured questionnaire consisting of socioeconomic and institutional factors and questions that were expected to measure the farmers’ experience of climate variability, the role of adaptation practices implemented in their localities, and their impact on household food security were collected from both from adapters and non-adapters by trained enumerators and the researcher.

A total of 384 households (175 adapters and 209 non-adapters) were selected from six Kebeles in Kindo Koisha and Loma Bosa Woredas by making them proportional to the sample population in each Kebeles. For instance, (Fajena Mata = 26 with male 19 and Female 7, Mundena = 69 with male 48 and female 21, Manara = 128 with male 77 and female 51 and Chereche = 99 with male 94 and female 5 from Kindo Koisha and Zimawaruma = 36 with male33 and female 3 and Chicho Hayo = 26 with male 24 and female 2). supplemented with 6 focus group discussions and 6 key informant interviews consist of heterogeneous groups concurrently across six kebeles from April to May 2021. Focus group discussions were made with six independent groups per kebele at each group consisting of 8 people which coincides with [31] in which 8–10 individuals were recommendable.

Innovative adaptation strategies, household demographic and socioeconomic characteristics, household food balance sheet model questions were included in the data sets to identify the factors influencing the adaptation options and their effects on rural household food security. Data on temperature and rainfall were obtained from ERA 5 data for 9 locations (Table 1) were used in this study. At some locations from nine grids, ERA5 estimates were compared to the meteorological data, and found that they were in good agreement.

Table 1 Grids for ERA5 around Ghibe III dam (1990–2021)
Full size table

ERA 5 is a replacement for ERA-Interim that performs better and provides time series data on temperature and rainfall with high spatial and temporal resolutions [32]. Thus, ERA5 opted for the current study as a gap filling in the shortage of meteorological data.

Using the daily rainfall and temperature data of Bele (1990 to 2005) station, the ERA 5 data were examined and bias-corrected before the analysis.

2.2.1.3 Variables of the treatment in propensity score matching model

The treatment variables are based on the questions such as what are the types of adaptation strategies and what are the roles of adaptation strategies on food security? Thus, the common adaptation types identified in the area and used as treatment variable in this study are soil and water conservation, planting of different crop varieties livelihood diversification, and planting climate-resilient crop.

2.2.1.4 Outcome variable

The outcome variables are the dependent variables on which adaptation strategies can have an impact positively or negatively accelerated by the different covariates of socio-economic and in some cases biophysical factors. The food calorie consumed by a household was the indicator of food security obtained by calculating the steps used in the food balance sheet model [33] based on the food item data collected during the household survey from smallholder farmers.

2.2.1.5 Quality control

A pilot study was carried out on 31 randomly selected households in Molticho kebele in the Kindo Koisha Woreda. The main study did not incorporate the data from the pilot test. The pilot study's sample was so small that its data and analysis could not be generalized to the larger group. The reliability was found to be 0.72. Based on the pilot study results, the questionnaire was checked and modified. The enumerators were given training on conducting household surveys with the assistance of the investigators. The missing values were corrected by revisiting the questionnaire during the entry of data into STATA software version 15. The missing values in the climate data were filled with the inverse distance weighing method.

2.2.1.6 Data analysis

Descriptive statistics (frequency distribution that gives count or percentages, measures of central tendency especially mean, median, the middle value in case of rainfall anomaly index calculation used, range and standard deviation and checking assumption such as normality tests, whether the data follows a normal distribution, skewness, left skewed or right skewed and kurtosis, how the data is flat) used to describe the characteristics of the sample households age, gender, family size, education status, and family income. Descriptive statistics lay a foundation for inferential statistics which help us conclude a larger population.

2.2.1.7 Propensity score matching (PSM)

PSM was used in this study because it fills in the gaps created by the lack of experimental data and accounts for sample selection bias[15]. PSM compares the observable properties of treated and non-treated groups(control) [34].

The outcome of this study is food security, the treatment variables are the adaptation strategies (soil and water conservation, planting of different crop varieties livelihood diversification, and planting climate-resilient crops) and the treatment groups are not randomly assigned. As it is known the quasi-experiment is applied in case when the true experiment cannot be used for ethical or practical reasons. Since the impact assessment of adaptation strategies on food security is a quasi-experiment, the subjects are not assigned to the group randomly but rather based on non-random criteria.

In the binary-treatment (t) scenario, where t = 1 when a person receives treatment and t = 0 otherwise, two distinct possible outcomes for a person can be denoted as y¡¡ and yi0 which are the realizations of the random variables – y1 and y0, respectively. Given these notations, the parameters of interest can be defined as follows: (1) ATE is the average effect of treatment in the population—which is expressed as ATE = E (yh 11- y0i); where E[1] stands for expected value, y1 is the outcome (food security) if the practices implemented and y0 is the outcome for the same household in the absence of adaptation. (2) ATET is the average treatment effects of those who received the treatment (t = 1) and is written as ATET = £(y1i,- y(),1t = l) [35].

PSM is based on two underlying assumptions, that is: the common support and the conditional independence assumption [36]. The first step the PSM procedure was to measure each respondent’s propensity score. This can be done by assessing the probability of the farmers adopting the adaptation strategies. In the PSM method, the probability is indented as propensity score and can be formulated as follows (Xi) = Prob(Ai = 1/Xi) where P(Xi) is the propensity score which is assessed by the probit model that regresses the adaptation strategies on farmer’s characteristic variables. The second is choosing a matching algorithm that is used to perform the matching process between covariates. The third is measuring the differences of each pair of units matched and eventually obtaining the Average Treatment Effect on the Treated (ATT) as the average of all these differences identifying overlap and common support [37].

The potential confounders measured in the propensity score matching were age, gender, education, family six = ze, and institutional factors (variables) such as access to credit, extension services(providing inputs and technical advice to farmers to increase productivity), farm input expenditure, debt of fertilizer and seed varieties and total livestock unit. As there are advantages to using propensity ratings, there are also drawbacks. It is important to highlight that the propensity score matching model performs best in large data sets with a reasonable distribution of baseline characteristics across treatment groups. Furthermore, the range of propensity scores between treatment groups must overlap significantly to allow for matching or regression modeling. Furthermore, the quality of the propensity score analysis is based on the appropriate specification of the propensity score model.

The propensity score will only account for the impact of observed confounders that are included as predictors in the logistic regression that generated it. There will be no adjustment for the impact of baseline characteristics that are not included in the propensity score model, such as unknown or unmeasured covariates [38]. Propensity score matching (PSM) has been widely used to reduce confounding biases in observational studies [39]. In non-randomized selection, there may be potential biases so that propensity score matching minimizes the biases in the selection.

2.2.1.8 Mann Kendall test and rainfall anomaly index

The statistical significance of precipitation and temperature was ascertained by the Mann–Kendall test with reference to climate change of the world [40]. There are two hypotheses considered in the Mann Kendall test namely the Alternative Hypothesis (Ha) and the Null Hypothesis (HO). The HO states the absence of a trend in time series and the HA states the existence of a trend in times series. The probability value (p-value) is used in hypothesis testing. If the P > 0.05 the null hypothesis is accepted and if the P < 0.05) the null hypothesis is rejected and contrastingly the alternative hypothesis is accepted indicating that the presence of trends in the data [41].

From ERA reanalysis data of the precipitation, the Annual Rainfall Anomaly Index (RAI) was calculated to analyze the frequency and intensity of the dry and wet years around the Ghibe III dam area. It is not only yearly but also the monthly RAI can be calculated for certain years over time to show the distribution of rainfall in years of greatest anomaly [42].

RAI has the following equations being developed by Roy 91,965) and later adapted by Freitas (2005) for positive anomalies Eq. (2) and negative anomalies Eq. (3).

$$RAI=3\left[\frac{N-\overline{N} }{\overline{M }-\overline{N} }\right]$$
(2)
$$RAI=-3\left[\frac{N-\overline{N} }{\overline{X }-\overline{N} }\right]$$
(3)

Where: N = current monthly/yearly rainfall(mm)̵; \(\overline{\text{N} }\)= monthly/yearly average rainfall of the time series (mm); \(\overline{\text{M} }\)= average of the ten highest monthly/yearly rainfall of time series (mm); \(\overline{X }\)= the average ten lowest monthly/yearly rainfall of time series (mm); positive anomalies have their values above average and negative anomalies have their values below average.

The values of the rainfall anomaly index are evaluated using the following rainfall anomaly index range (Table 2).

Table 2 Classification of rainfall Anomaly Index Intensity.
Full size table

Statistical analyses like the MK test and RAI contribute significantly to informed decision-making in water resource planning and management [43].

2.2.1.9 Food balance sheet model

Food security has availability, acess, utilization and stability dimension [44]. However food availability was selected for this study. As agriculture determines Ethiopia’s economy, analyzing the availability dimension explains the food security status in Ethiopia. Of course, this model does not take into all dimensions of food security and this can be taken as a limitation.

Food balance sheet model was employed following [45] after Tolossa [46] Eq. (4).

$${N}_{ij}=({P}_{ij}+{B}_{ij}+{F}_{ij}+{R}_{ij})-({H}_{ij}+{S}_{IJ}+{L}_{IJ})$$
(4)

where Nij is the net food available for household i in year j, and pij is the total grain produced by household i in year j. Bij represents the total grain bought by household i in year j. Fij is the total grain acquired through food for work (FFW) by household i in year j, while rij is the total relief food received by household i in year j. Hij represents the post-harvest loss to household i in year j. Sij is the total crop used for seed from the home by household i in year j, while Lij is the total grain sold by household i in year j.

The dependent variable was the binary food security status,, which was represented by the dummy variables 1, if food was secure, and 0, if food insecure. The net grain available for a household is obtained by converting from kilograms into kilocalories based Food Composition Table for Use in Ethiopia [47]. The net available food is then computed by dividing the total number of days in a year and then dividing by the adult equivalence, which yields the total available calories per adult equivalent every day. The minimal calorie requirements for an adult equivalent per day were then set at 2100 kilocalories, in accordance with Ethiopia's 1996 Food Security Strategy (FSS). Finally, a comparison was made between available kilocalories (supply) and the required kilocalories (demand).

A household that consumed less than 2100 kilocalories per/day/adult equivalency was identified as food insecure whereas as greater than 2100 kilocalories was identified as food secure.

3 Results and discussion

The common innovative farmers adaptation practices are conservation practices (17.19%), diversification of livelihoods (9.89%), planting crop varieties (11.46%), and climate resilient crop varieties (7%). Households who did not use at least one of the adaptation options were 54.46% and those who belonged to the adapter group were 45.54% (Table 3). The average family size of the sample households is 6.4 (Table 4). This is in line with [48] in which the family size of wolaita zone is 6.7. However, it is greater than the national average of 4.6 [49]

Table 3 Adaptation strategies around Ghibe III dam, Southern Ethiopia
Full size table
Table 4 Descriptive statistics of determinants of farmers adaptation options around Ghibe III dam
Full size table

The t-test result also reveals a significant and negative difference in the number of calories consumed per day per adult equivalent between the two groups (Table 4). The mean daily calorie intake of households was 1626.55, which only meets77.45% of the minimum daily requirement (2100 kcal).

The results of the probit model estimates showed that the adoption of adaptation options was positively and significantly associated with family size, farm size, extension service, credit, and weather information at the 1% level of significance. There is an average 7% increase in the likelihood of using adaptation choices for every 24% increase in farm size (Table 5). This is in line with in which farm size positively affected the adaption option of farmers [17]

Table 5 Factors affecting the adaptation options by probit model
Full size table

Access to extension services also has a major impact on the adoption of adaptation measures. As expected, farmers’ decisions to invest in adaptation are highly influenced by their access to extension services (p < 0.01). The possibility that adaptation options will be adopted increases by 18% with access to extension services. Having access to climate warnings raises the probability that adaptation is used by 64% (Table 3). The likelihood of adopting adaptation strategies increases by 9% with each unit increase in family size. The likelihood of adopting adaptation strategies increases by 24% for every unit increase in farm size in study area. The probit model result reveals that getting access to climate warnings about drought and/or flood increases the likelihood of using adaptation option by 19%.

Debt is negatively associated with adaptation of adaptation strategies and significant at 5%. Finally, the results from the probit model showed that farmers decision to adopt adaptation options is not significantly influenced by the variables age, gender, education, farm input and TLU (Table 3). In this study age does not show any association with adaptation practices. This finding was in contrast to [36] in which age has negative association with adoption of new agricultural techniques that enhances the productivity of farmers. This is due to the fact that aged farm households may be reluctant to new and creative agricultural techniques despite more awareness. Similarly, the findings by [50] showed that age of farmers had a positive association with adaptation practices to climate variability. Getting credit service had a significant influence on the adoption of adaptation practices that negatively affected household’s decision to adopt adaptation practices at 5% probability level. This result reveals that a unit increase in a credit would result in a 18% decrease in the probability of adopting adaptation practices which is in disagreement with [51] reported that credit services increase the probability of adopting adaptation practices. The negative association is similar with [2] in which credit services is negatively associated though not statistically significant.

The model was statistically significant with Prob > chi2 = 0.0000, Pseudo R2 = 0.1584. According to the findings, there is a statistically significant difference between treated (n = 175) and control (n = 209) households in terms of the distributions of family size, farm size, access to credit, extension services, climatic information, and debt (Table 5).

3.1 Propensity score estimation

Following the calculation of propensity scores, the nearest-neighbor matching (NNM) method, invers distance weighting, and probability score matching were used to match the control group of individuals (non-adapters) to the treated group (adapters) based on similar propensity scores (Table 4). During the matching process, all matching algorithms eliminate unmatched non-adapters, resulting in a smaller sample size for the post-matching impact study.

The analysis found that there is a significant overlap in common support. Figure 2 depicts the distribution of propensity scores for matched and mismatched individuals from both groups. The result assures that the propensity score distributions of adapters and non-adapters overlap. The lower part of the graph depicts the propensity score distribution for non-adapters, while the top half represents adapters. The scores' densities are represented on the y-axis. Individuals within adoption group who are successful in finding a match were "treated on support," while those within the adoption group who are unsuccessful in finding a match were "treated off support." The results corroborated the substantial overlap in shared support. An additional quality of match or balance test was carried out with a shared support graph.

Fig. 2
figure 2

Propensity score distribution and common support for propensity score estimation

Full size image

3.2 Estimation of the treatment effect

The nearest neighbor matching post-matching result shows that adaptation has a favorable and significant impact on the amount of calories consumed by the household. Soil and water conservation would raise the chance of food security by 7% compared to those who did not adopt, according to the ATE estimations from the Nearest neighbor matching and this difference would be significant at the (P < 0.05) level. Similarly, the increase in crop diversification and livelihood diversification would increase the likelihood of being food secure by 9% and 8% respectively.

The likelihood of having enough food would rise with increased conservation of soil and water, use of crop variety, and livelihood diversification. However, the use of improved varieties and drought-tolerant crops did not show any increase in the probability of food security (Table 6). This is due to a lack of training on its implementation by agricultural offices in the study area.

Table 6 Treatment effects: nearest neighbor matching
Full size table

Mann–Kendall test resulted in a down and upward trend in annual temperature however, it was statistically not significant at P < 0.05 (Fig. 3)

Fig. 3
figure 3

Trends of temperature (in 0c) from 1990 to 2021

Full size image

The ATE estimates from the Propensity matching showed that soil and water conservation practices, crop diversification and livelihood diversification would increase the probability of being food secure by 10%,8% and 9% respectively and significant at 5% level (Table 7).

Table 7 Treatment effects: propensity matching
Full size table

The Potential Outcome Means (POM) estimates through inverse weighting showed that Use of soil and water conservation, crop diversification and livelihood diversification would increase the probability of being food secure by 2.2%, 2.2% and 6.5% respectively and significant at 5% level (Table 8).

Table 8 Estimating average adopters and food security status through inverse weighting
Full size table

Rainfall had shown declining trends from 1990 up to 2021 as analyzed by rainfall anomaly index (Fig. 4) in Middle parts of Omo basin, Ghibe III Dam area, southern Ethiopia.

Fig. 4
figure 4

Rainfall anomaly index in Ghibe III Dam area (1990–2021)

Full size image

Crop diversification, livelihood diversification, use climate resilient crop varieties is the dominant adaptation practices in the study area. Covariates were displayed both before and after propensity score matching using standardized bias (%). The outcome demonstrated that candidate factors, such as age, were well-matched (Fig. 5).

Fig. 5
figure 5

Standardized % bias across covariates. Perceive-perception of households, whether info-weather information, Yrsscho-years of schooling, extension serv-extension service, participation in off farm activities

Full size image

This test is helpful since it shows the propensity score overlap between the treatment and control cases visually. The distribution of propensity ratings for paired and unpaired individuals in both groups is shown in Fig. 6. The outcome ensures that the propensity score distribution has a suitable overlap between adopters and non-adopters.

Fig. 6
figure 6

Distribution showing adapters and non-adopters

Full size image

The results showed that the coefficients of ATE were positive, significant, and estimated at 0.36 for Nearest Neighbor 0.37 for Propensity score, and 0.07 for Inverse probability weighting (Table 9). Similarly, the coefficients of ATET were positive and significant at 5% level for nearest neighbor matching, propensity, and inverse weighting. This showed that households that implemented strategies for adapting to climate variability were more likely to experience increases in food security of 0.4, 0.39, and 0.09 compared to households with no practices.

Table 9 Matching estimation of impacts of adaptation practices on food security
Full size table

Similarly, farmers who adopted adaptation strategies had higher levels of food security (7–9%) for ATE and ATET than those who did not adopt adaptation strategies in inverse probability weighting. This result is incongruent with [52]. The present result is also similar to [36] who noted that farmers who adopt adaptation strategies have higher levels of food security (7–8%) than those who do not adopt strategies.

Ogunpaimo and Oyetunde-usman [55] reported that farming households that adopted at least one of the adaptation options had 9% greater food security levels compared with those who did not implement any of the adaptation strategies. A study by Abate et al., [53] in Southwestern Ethiopia revealed that the likelihood of being food secure would improve by a factor of 0.8 if climate-smart crop varieties were adopted, with 59.74% of households adopting climate-smart crop types.

The ATE estimates from the nearest Niebour matching exhibited that conservation practices would increase the probability of food security by 7% than those who did not use soil and water conservation and significant at 5%. Increased soil and water conservation, livelihood diversification, and crop variety would all contribute to greater food security. However, using improved varieties of seeds and drought-tolerant crops did not increase the likelihood of being food secure. This is in line with [52] who showed that adapters increased food security level by 7to 9% than non adapters. Access to weather forecasting increases the likelihood of implementing adaption strategies by 18%. This is in agreement with [15, 54] whose study revealed a positive association between acess to climate information and adaptation option.

Study by [36] reported similar finding in which adaptation practices and food security were closely related.. Farmers adopting any of the adaptation options had higher food calorie intake per day per adult equivalent than those who did not [15].

The study by [55] showed that the probability of using adaptation options increased with farm size, extension contact, and marital status but decreased with the age of the household head, credit facilities, ownership of farmland, household size, years of schooling household assets which in turn increases food security.

4 Conclusions

In this study a binary propensity score matching estimator as part of a non-parametric approach was applied. The farm households were found to be primarily using soil and water conservation practices, crop diversification, livelihood diversification, and climate-resilient crop varieties as adaptation strategies to adverse effects of climate variability. Probit model estimates showed that family size, farm size, extension service, credit and weather information were all positively and significantly related to the adoption of adaption options. Adapters of any of the adaptation strategies had higher food security levels for both ATE and ATET compared to those that had not adopted in inverse probability weighting. This finding has limitation and weakness in some aspects concerning the dimension of food security as it was the availability dimension only used in this study even if food security studies mostly include both availability and access dimensions thus, future research should consider other dimensions. The other is. the study’s focus near the Ghibe III dam which is mostly focused on central parts of the Omo Ghibe basin thus may limit generalizability to other regions. The feature research should be conducted in the entire Omo-Ghibe basin to get a clear understanding about the spatial and temporal variability of rainfall and temperature.

This finding offers policymakers the opportunity to further encourage the adoption of adaptation practices for climate variability that appear to be especially crucial for households with the least capability for adaptation of farm households. Conservation agriculture and climate-resilient crop varieties should be implemented at large by farmers assisted by local government representatives.