Assessment of implemented physical designs and determinant factors of soil and water conservation measures: Wenago district, southern Ethiopia

Abstract

Soil erosion and its consequences is one of the major serious problems in Ethiopia. Even though adoption of soil and water conservation (SWC) measures has been underway for the past three decades, the implementation and use of introduced technologies were below the expectation and the problem is still has continued in the country. The study was aimed at assessing the implemented physical designs of soil and water conservation structures in respect to the standards and identifying the major adoption determinant factors in Wenago district, southern Ethiopia. The data for this studywas collected from a survey of 262 total household farmers selected through simple random sampling techniques in the year 2020/21 and the datawas analyzed using descriptive statistics, chi-square and logistic regression model via SPSS and Stata soft wares. Focus group discussion, key informant interview and personal observation were also undertaken to gather data having qualitative nature. (i) About 55.6% of the implemented physical design failed to meet the standards (ii) Adoption of SWC measures were determined by 47.4% of the tested variables (iii) 55.5% of the variables were significantly associated at 5% probability level between adopters and non-adopters in terms of adoption of SWC measures in the study area. Overall, we conclude that construction of conservation structures should be focused on minimizing the observed mismatch of the implemented physical designs against the standards. This study is expected to contribute in achieving sustainable land management schemes, agricultural productivity and smallholder farmers livelihood improvement in international, national, regional and local level and it is strongly recommended that policy makers and technical institutions should address SWC related issues to ensure rural farmers food security.

1. Introduction

Agriculture in Ethiopia provides job for nearly 85 million smallholder farmers and land less workers in Ethiopia [4,10]. Land degradation, soil erosion and loss of organic matter from top soil in the Ethiopian high lands are extremely linked to population pressure and continuous cultivation and the absence of appropriate land resource [4,16,18]. Loss of organic matter again leads to deterioration of soil quality and agricultural productivity [15]. In Ethiopia, soil erosion has been a problem ever since man started to cultivate the soil and domesticating animals [15,25].

According to Ref. [28] about 1.5 billion tons of topsoil is lost each year from 30,000 ha of farmland due to water erosion. The latest land degradation estimates indicate that out of the 52 million hectares of land makingup the highlands of Ethiopia, 14 million hectares are severely degraded, 13 million hectares are moderately degraded and 2 million hectares have practically lost the minimum soil cover needed to produce crops [4,8,14]. According to Refs. [22,24], the average annual rate of soil loss in the country is estimated to be 42 t/ha/year which results to 1–2% of crop loss and it can be even higher on steep slopes and on places where the vegetation cover is low.

Even in the current 21st century, soil erosion has been distinguished as a serious threat in Ethiopia which causes to loss of organic nutrients, declining agricultural productivity, income of farmers decreases, in turn leads to food insecurity [12,24]. Although Ethiopia is rich in different natural resources, agro ecologies, tourism sites and topography, it has been affected by lack of appropriate land management practices. Although the failure of soil conservation intervention can have many causes, it resulted mainly from the fact that planners and implementing agencies ignored local level biophysical and socio economic realities and there is evidence of such problems particularly in the highlands of the country [16,28]. Recognizing all these serious threats of soil erosion problems and declining agricultural productivity, the government of Ethiopia began a program of natural resources conservation activities since 1973 supported by donors and nongovernmental organizations [2,4,8,22], argued that there are various SWC measures and applications of technologies used to control soil erosion, land degradation, reduce the flow of water so as to increase soil fertility, soil moisture and crop yield in Ethiopia. According to Refs. [2,26] Physical structures (stone terracing, soil bunds, cut-off drains, water ways, water harvesting, contour ploughing); agronomic/vegetative/(agro forestry, crop rotation, intercropping, mulching, mixed cultivation, fallowing) and other community based land management strategies are some of the most implemented SWC practices particularly in the high lands of Ethiopia.

The role of these improved physical structures and agronomic SWC measures have a significant contribution to farmers in increasing crop productivity, income and improve livelihood status. However, as this determines the growth of crop productivity, there is a need for careful selection of appropriate SWC technologies, understanding the use of introduced SWC technologies and their adoption determinant factors has received very little attention in Ethiopia including the study area. Smallholder rural farmers have different attitude towards introduced/improved structures so that they use different physical and biological methods depending on their own level of perception. Thus, it is crucial to identify and use the suitable SWC measures in accordance to the local communities cultural context [21,25,27].

Theoretically, smallholder farmers wish to increase income by adoption of improved/introduced/SWC technologies. However, smallholder farmers till remain very low in Ethiopia due to lack of awareness, shortage of finance, poor institutional arrangement and remain insecure of sustainable land management [5,9,23]. The achievement was also not as expected and the threats of soil erosion problems are still widespread in the country [17,18]. Evidences by Refs. [24,27] showed that most empirical researches in Ethiopia have been focused mainly on supply of agricultural inputs, use of traditional conservation methods and area coverage instead of addressing technical aspect of structural designs, choice of best modeling, adoption determinant factors and identifying constraints. As a result, improved SWC practices discontinued, crop yields decreased, food insecurity prevailed in Ethiopia [21,26,28].

Therefore, this study was undertaken in order to bringing the gap. The objective of this research study is (i) identifying adoption determinant factors and use of smallholder household survey data from adopters of improved SWC practices and non-adopter households, in Wenago district, southern Ethiopia. The rest of the paper is organized as: materials and methods in section two, results and discussion in section three, and conclusions and policy implication in section four. The findings of this study are very crucial to use the most suitable SWC measures and standardized structural designs. Thus, it is important to conduct benefit-cost analysis of various SWC measures so that policy makers, professional experts; users and stakeholders should apply appropriate land management and technologies.

2. Materials and methods

2.1. Location and description of the study area

The research was conducted in Wenago district, Gedeo Zone, Southern Ethiopia, 375 km south of the capital city Addis Ababa. The study area is located between 60 12ʹ 30ʺ and 60 22ʹ 30ʺ N latitude and 38015ʹ0ʺ and 38020ʹ0ʺ E longitude Figure 1. Based on the recent [2020] projection, the study district has a total population of 152,000 with approximately an area of 248 square kilometer giving a population density of 613 persons/km2. The majority of the populations of the study area (90.8%) live in rural area and the rest (9.2%) live in small town centers. The study is bordered in the west by Abaya District (Oromiya); in the south by Yirga Cheffie district; in the north by Sidama Zone and in the east by Dilla Zuria Wereda. The study area has a favorable climate for agroforestry dominated agricultural activities. Rainfall ranges from 799 to 1512 mm while mean annual temperature varies from 12.5 °C to 28 °C. Nitosols are dominant soil types covering highest proportion of the study area. The soils are in general derived from volcanic rocks which are important for coffee growing areas [1].

Fig. 1.

Location of the study area: Source: Ethio-Geospatioal data GIS-arc [11]. ¹The name“Kebele” refers to the lowest administration units (sub-division of a district) in the administrative structure of Ethiopia. ²“SNNPR” on the map Figure 1 stands for South Nation Nationalities and People Region which means southern Ethiopia.

2.2. Research design and site selection techniques

Multistage sampling techniques and both qualitative and quantitative approaches were employed for this study. In the first stage of sampling, Gedeo zone where improved SWC practices has been implemented was purposively selected. Then of the eight districts of Gedeo zone, Wenago district was randomly selected for the study. In the second stage, among the seventeen rural Kebeles of Wenago district, three kebeles: namely Tumata-chirecha, Kara soditi and Dobota from which sample households were purposively selected on the basis of their high rate of soil erosion problem, active human SWC intervention and presence of both SWC adopter and non-adopter household groups. In the third stage, a total of household heads from the fresh list of local administers office were taken and these total household heads were again stratified in to SWC adopter and non-adopter groups. Finally, individual household head respondents for the quantitative survey study were selected using simple random sampling technique assuming 94% confidence level.

The sample site choice was carried out based on the severe soil erosion problems and active intervention of SWC activities. The farming household heads are the main responsible body for making day to day decisions on land management practices (soil and water conservation). Thus, survey was conducted at household level to assess and investigate currently implemented physical designs of SWC practices and adoption determinant factors. Before starting the actual data survey and data collection, a preliminary field visit and physical observation in order to get the general view of the study area and identify the sample site was conducted from January 7, 2021 to January 20, 2021 with the help development agents and the actual survey and data anlysis was conducted from February 12, 2021 to May 27, 2022.

2.3. Data collection methods

For the sake of validity, reliability and acceptability of the data and achieve the intended ojectives, the author used both primary and secondary data sources. The primary data were generated from 262 sample households using a structured questionnaire. A questionnaire was administered by 3 Development Agents (DA's), 2 office experts, 3 kebele leadiers, 4 local elders and 5 model farmers. The questionnaire was pre-tested for more clarity and related issues. To strengthen the structured questionnaire and to generate additional information regarding implementation level of improved SWC structure and adoption determinant factories, focus group discussions (FGD) with 24 persons and interview with 33 key informants was made. To have a general image about the existing physical land scapes, personal observations using checklists was also held. Secondary data were gathered from published materials (books, journals, magazines, and newspapers), unpublished sources (unpublished manuscript of thesis, government reports and documents), online access (internet). The target population of the study district were about 4463 household head farmers (2237 SWC adopters and 2226 non adopters). This data was taken from the study district's agriculture and natural resource office annual report in February 2021. The survey sample size was determined using the formula proposed by Ref. [13] stated as: n = N/1 + N (e) ², where n is the required sample size, N is the total target population (4463), e is the error limit (0.06). Therefore, a total of 262 sample households heads were selected to represent the entire study of the district of which 132 were SWC adopters and 130 were non-adopters.

2.4. Data analysis methods

2.4.1. Descriptive data analysis

Demographic, Socio-economic characterstics, social, institutional and biophysical factors, implemented physical designs of SWC structures were analyzed using descriptive statistics such as mean, frequency, percentage, graphs and chi-square through IBM SPSS version 20 statistics.

2.4.2. Econometric data analysis

With regards to econometric methods of data analysis, parametric and semi–parametric models were used. Propensity Score Matching (PSM) was used to analyze the impact of adoption of SWC practices on the livelihoods' of smallholder farmers. Furthermore, bivariate logit was used to identify factors influencing farmers’ choice of adoption of SWC practices in the study area. Details of the methods of data analysis and their model specification are described hereunder.

To estimate the Logit model, the dependent variable was SWC practices participation status, which takes the value of 1 if a household participated in the SWC practices and 0 otherwise. According to Ref. [5] the Logit model is specified as:

$$Pi=\frac{e^{Zi}}{1+e^{Zi}}$$ (1)

Where, P_i is the probability that the household participated in SWC practices and (1-P_i) is the probability that a household did not participate in the SWC practices [6] which is given by:

$$1-P_i=\frac{1}{1+e^{Zi}}$$ (2)

The odds ratio $\left(\frac{P_i}{1-P_i}\right)$ is defined [6]as:

$$\frac{P_i}{1-P_i}=\frac{1+e^{Zi}}{1+e^{-Zi}}=e^{Zi}$$ (3)

$$L_i=\ln \left(\frac{P_i}{1-P_i}\right)=Z_i={\beta }_1+{\beta }_2{x}_i+.........{\beta }_i+e_i$$ (4)

Where, P_i is the probability of participation, e = 2.71828, i = 1, 2, 3, - --, n, β₁ is intercept, β₂ is regression coefficients to be estimated and e_i is an error term [14].

After the participation equation (Logit) is estimated, the predicted values of participation (T) from the participation equation are derived. The predicted outcome represents the estimated probability of participation or propensity score. Every sampled participant and nonparticipant had an estimated propensity score, P (X |T = 1) = P(X) = P (X |T = 0) [19].

2.4.3. Defining the region of common support and balancing tests, matching participants with non participants, estimating the average treatment effect on the treated (ATT) and factors influencing farmers’ choice of adoption of SWC practicesand

The region of common support needs to be defined where distributions of the propensity score for treatment and comparison control group overlap [19]. Since some of the nonparticipant observations were dropped because they fall outside the common support. Therefore, it is important to remove all observations out of the overlapping region, whose propensity scores were below the minimum and larger than the maximum, of the treatment and no treatment groups, respectively [5].

Balancing tests was conducted to check whether, within each quantile of the propensity score distribution, the average propensity score and mean of X are the same. For PSM to work, the treatment and comparison groups must be balanced in that similar propensity scores are based on similar observed X. The distributions of the treated group and the comparator must be similar, which is what balance implies. Formally, one needs to check if P (X |T = 1) = P (X |T = 0) [19].

To estimate the Average Treatment Effect, estimation of the propensity score, defining the region of common support and balancing tests are not enough. This is because propensity score is a continuous variable and the probability of observing two units with exactly the same propensity score is in principle zero [14]. In addition, matching smallholder farmer based on the propensity scores alone reduces what is called the curse of dimensionality [19]. Thus, among different matching algorithms used to match participants and nonparticipants, in this analysis Nearest-Neighbor Matching (NNM) method was used in order to control for bias and to ensure robust results. This is useful for estimating the possible impact of making a voluntary intervention compulsory, or extending the intervention beyond the current eligible group. To obtain Average Treatment Effect (ATE) using PSM, the treated were used to estimate the counterfactual for the untreated. Once common support was enforced, treatment on the treated and treatment on the untreated was estimated. Thus, the control group was matched on the probability (propensity score) of adopting given a set of observable characteristics from a logit regression [6].

To understand the potential estimated intervention effect, the treatment impact across different observable characteristics, such as position in the sample distribution of explanatory variable was examined. ATT was computed as the weighted average of the two groups [6].

The ATT is given by:

$$ATT=E\left(Y^T-Y^C/T=1\right)=\frac{1}{NT}\left[\sum _{i\epsilon {\rm T}}^T{Y}_i^T-{\sum }_{j\epsilon c}\phi \left(i,j\right)Y_i^c\right]$$ (5)

Where, N^T is the number of observations treated, Y^T is the outcome with treatment, Y^C is the outcome without treatment or control group, and $\Phi$_ij is the weight factor used in the matching. PSM constructs a statistical comparison group that is based on a model of the probability of participating in the treatment T conditional on observed characteristics X, or the propensity score: P (X) = Pr (T = 1|X). According to Ref. [19], the reliability of propensity score matching estimators was based two necessary assumptions: The conditional independence assumption and presence of a common support or overlap condition. Lastly, using predicted propensity score, by using alternative methods of matching estimators match pairs were constructed. The effect estimation is the difference between simple mean of outcome variable for adopter and non-adopter smallholder households. Therefore, in this study to analyze the impact of SWC practices on livelihoods of the smallholder farmers by using the PSM method.

Bivariate probit model was employed to identify the factors that affect farmers’ decision of adoption of SWC practices in the study areas. The Bivariate probit is a generalization of the probit model used to estimate two correlated binary outcomes jointly. The dependent variable has two categories representing the existing improved soil and conservation practices on a given of land of the sampled households. These are improved soil bund and stone bund/terraces. The dependent variables were selected based on how frequently they were used or number of SWC technologies received by farmers on their plots. The covariates include demographic, socio-economic characteristics, institutional and environment characteristics. According to Ref. [19], the bivariate probit model is generally specified as:

$$Y_{j1}^{*}={{\beta }^{\prime }{X}^{\prime }}_{j1}+{\epsilon }_{j1}$$ (6)

$$Y_{j2}^{*}={{\beta }^{\prime }{X}^{\prime }}_{j2}+{\epsilon }_{j2}$$ (7)

In the data Y_j1 andY_j2 are only observable through the two discrete choice responses [19]such that:

$$Y_{j1}=\left\{\begin{array}{l}1if{\mathrm{Y}}_{\mathrm{j}1}^{*}>{\mathrm{Y}}_1\\ 0otherwise\end{array}\text{and}{Y}_{j2}=\begin{array}{l}1if{\mathrm{Y}}_{\mathrm{j}2}^{*}>{\mathrm{Y}}_{\mathrm{j}2}^{*}\\ 0otherwise\end{array}\right\}$$ (8)

$$E\left[{\in }_1\right]=E\left[{\in }_2\right]=0$$

$$Var\left[{\in }_1\right]=Var\left[{\in }_2\right]=1$$

$$E\left[{\in }_1\right]=E\left[{\in }_2\right]=0$$

$$Cov\left[{\in }_1,{\in }_2\right]=\rho$$

As the second equation is based on the first question response, the error terms are correlated and hence the two equations can be estimated jointly using the model developed by Refs. [14,19] which assumes Bivariate Normal distribution for the two valuations, BVN (β₁X₁, β₂X₂, σ₁², σ₂², ρ) or BVN (0,0.1.1, $\rho$). As mentioned earlier the two questions have four possible pairs of response: (Y_ji, Y_j2) = (1, 1), (1, 0), (0, 1) and (0, 0). Combining the associated probabilities of all possible responses in the likelihood function, Equation 10 can be estimated using the Bivariate Probit or the seemingly unrelated bivariate probit Model.

3. Results and discussion

3.1. Descriptive statistics results

3.1.1. Implemented SWC measures in the study area

The major types of SWC structures dominantly implemented in the study area were both traditional and introduced SWC measures and are categorized in to two as physical structures and biological measures Figure 2. Traditional SWC practices are implemented by non-adopters while both traditional and introduced SWC technologies were implemented by SWC adopters. The difference between the two groups is that introduced stone terraces and soil bund practiced are practiced only by SWC adopters in the absence of non-adopters Figure 2. However, all SWC measures have been contributing in the process of controlling soil erosion and increasing soil fertility, moisture, crop yield, income and better livelihood status of the smallholder farmers. The field observation also revealed that the common SWC structures are soil bund and stone terraces which are applied 80% on individual farm plot and 20% on communal lands.

Fig. 2.

Implemented SWC measures by adopters (left) and non-adopters (right). Source: Wenago district agriculture natural resources development office [2021].

3.1.2. Physical design of the implemented SWC structures in the study area

As shown in Figure 3 (A-D), the physical designs of each structures implemented mainly on farm plots in the study area were compared against to the standard as recommended by Refs. [7,29]. The technical parameters selected for evaluation level of soil bund and stone terraces were length, spacing, vertica interval (VI), depth and width. Spacing refers to the ground distance between two consecutive bunds and vertical interval (vi) is the height of the bund [7]. For more information about dimension of each constructed structures refer Table 1 and Figure 3 (A-D).

Fig. 3.

Typical physical designs of implemented structures, stone terraces on hillside (A), Soil bunds and Fanya Juu on communal land (B), Traditional drainage ditches and Grass strip on farm plot (C) and Micro basins inside enclosure (D). Source: Field observation and survey result. Photograph by the researcher February [2021].

Table 1.

Comparison of implemented dimensions against the recommended standards by slope and sample sites.

Sites	Slopes (%)	Dimension	Stone bund embankment & spacing		Deviation from the standard
			Implemented (m)	Standard(m)
Sample site 1 (Kara Soditi kebele)	14	Height(h)	0.37	0.55	−0.18
		Length	52	55	−3
		Top width	0.50	0.50	Standard
		Bottom width	0.9	1.5	−0.6
		Spacing	15	14	+1
Sample Site 2 (Tumata Chirecha Kebele)	12	Dimension	Stone bund embankment & spacing		±
			Implemented (m)	Standard(m)
		Height(h)	0.65	0.55	+0.1
		Length	50	55	−5
		Top width	0.4	0.50	−0.1
		Bottom width	1.5	1.5	Standard
		Spacing	10	14	−4
Sample Site 3 (Dobota Kebele)	14	Dimension	Stone bund embankment & spacing		±
			Implemented (m)	Standard(m)
		Height (h)	0.39	0.55	−0.16
		Length	65	55	+10
		Top width	0.5	0.5	Standard
		Bottom width	0.9	1.5	−0.6
		Spacing	12	14	−2
Sample Site 1 (Kara Soditi kebele)	13	Dimension	Soil bund embankment & spacing		±
			Implemented (m)	Standard(m)
		Height (h)	0.3	0.5	−0.2
		Length	82	70	+0.03
		Top width	0.5	0.47	+0.03
		Bottom width	1.0	1.3	−0.3
		Spacing	12	15	−3
Sample Site 2 (Tumata Chirecha Kebele)	10	Dimension	Soil bund embankment & spacing		±
			Implemented (m)	Standard(m)
		Height(h)	0.5	0.5	Standard
		Length	65	70	−5
		Top width	0.42	0.47	−0.05
		Bottom width	1.3	1.3	Standard
		Spacing	12.5	15	−2.5
Sample Site 3 (Dobota Kebele)	12	Dimension	Soil bund embankment & spacing		±
			Implemented (m)	Standard(m)
		Height(h)	0.46	0.5	−0.04
		Length	80	70	+10
		Top width	0.2	0.47	−0.27
		Bottom width	1.3	1.3	Standard
		Spacing	11	15	−4
Sample Site 1 (Kara Soditi kebele)	11	Dimension	Traditional D.ditch embankment & spacing		Deviation
			Implemented (m)	Standard(m)
		Depth	0.45	0.5	−0.05
		Width	0.4	0.4	Standard
Sample Site 2 (Tumata Chirecha) Kebele)	11	Dimension	Traditional D.ditch embankment & spacing		±
			Implemented (m)	Standard(m)
		Depth	0.42	0.5	−0.08
		Width	0.42	0.4	+0.02
Sample Site 3 (Dobota Kebele)	12	Dimension	Traditional D.ditch embankment & spacing		±
			Implemented (m)	Standard(m)
		Depth	0.44	0.5	−0.06
		Width	0.48	0.4	+0.08
Sample Site 1 (Kara Soditi kebele)	8	Dimension	Micro basin embankment & spacing		±
			Implemented (m)	Standard(m)
		Depth	0.55	0.55	Standard
		Width	0.5	0.77	−0.77
		Spacing	4.5	3.5	+1
Sample Site 2 (Tumata Chirecha)	7	Dimension	Micro basin embankment & spacing		±
			Implemented (m)	Standard(m)
		Depth	0.3	0.55	−0.25
		Width	0.6	0.77	−0.17
		Spacing	4.0	3.5	+0.5
Sample Site 3 (Dobota Kebele)	9	Dimension	Micro basin embankment & spacing		±
			Implemented (m)	Standard(m)
		Depth	0.45	0.55	−0.1
		Width	1.0	0.77	+0.23
		Spacing	3.5	3.5	Standard

Source: Technical evaluation of dimensions (2021).

3.1.3. Technical evaluation of implemented physical SWC practices

The most common and widely used physical conservation practices in the study area are soil bund and stone terraces which has been implemented since 1980's. Implementation of these structures considers topography, socio-economic, land use and preferences by farmers. The field measurement on all of the indicated parameters on three selected sample sites of the study district is shown in Table 1. As shown in Table 1, the field measurement result revealed that 55.6%, 24.4% and 20% represents the negative, positive and standard value of implemented dimensions deviated from the recommended standard respectively. This result revealed that about 44.4% of the total implemented SWC structures was carried out according to the recommended standard. The remaining 55.6% of the structural dimensions failed to meet the standards Table 1. The possible reason of implementing dimensions below the recommended standard might be due to shortage of technical skill, lack of training opportunity, poor institutional arrangements, weak coordination between farmers and experts and negative attitude of farmers. For instance, of the 33 total interviewed key informants, 22 (66.6%) of them reported that construction of structures on farm land reduces farm plot size and causes to harbor redones indicating some farmers have a negative attitude towards improved/introduced/SWC technologies.

3.2. Result of the chi-square test in association to adoption category and SWC practices

The details of socio-economic variables associated with adoption categories in creation to soil conservation measures was analyzed using chi-squre test Table 2. As showed in Table 2, educational level, off-farm activities, extension service, TLU, and farm land size holding with chi-square value 6.65, 57.88, 163.96, 0.04 and 239.05 respectivelly were found to be statistically significant at (p < 0.05) indicating there were weak association between adopters and non-adopters in terms of adoption of soil conservation technologies for agricultural systems in the study area. This result was contradictory with the finding of [3,20] who reported that these variables have a strong relationship among small holder farmers in terms of adoption of soil conservation technologies in Jeldu district, western Shewa, Oromia region.

Table 2.

Association between adoption categories in relation to soil conservation measures.

Explanatory variables	Adoption category (%)		Chi-square results
	adopters	Non-adopters	X2-value	p-value
Sex	0.68	1.07	4.240	1.110 ns
Age	3.75	3.61	7.180	0.410 ns
Marital status	1.14	1.24	6.180	0.100 ns
Education level	1.35	1.22	6.659	0.004**
Family size	3.61	1.06	11.562	0.170 ns
Farming experience	2.13	1.94	7.641	0.170 ns
Off-farm activities	0.80	0.94	57.881	0.000**
Extension service	2.00	0.98	163.96	0.000**
Market access	1.09	1.07	0.580	0.440 ns
TLU	3.62	3.44	0.041	0.040**
Farm land size holding	2.90	0.51	239.05	0.000**

Source: survey data computed result [2021]. Notes:** means significant, ‘ns’ not significant at 5% significance level.

As indicated in Table 2, the remaining variables like sex, age, marital status, family size, farming experience and market access with chi-square value 4.24, 7.18, 6.18, 11.56, 7.64 and 0.58 respectivelly were found to be statistically had no significant difference at (p < 0.05) which revealed that there is a strong association between adopters and non-adopters in terms of adoption of agricultural systems. This result of the survey implies that there is a positive relationship between social characterstics of household farmers and adoption of soil conservation technologies in the study area. This result was inline with the finding of [2] who reported that these variables have a strong relationship between adopters and non-adopters in terms of adoption of soil conservation technologies in the high lands of Ethiopia.

3.3. Level of community participation in SWC practices

In order to gather additional information regarding soil erosion problems and its consequences, community participation level in SWC practices, technical skill of installation of improved structures, household farmers adoption constraints and benefits of improved SWC measures, a deep interview and focus group discussion with selected office experts, DA's, men, women, knowledgeable persons and elders was made Figure 4. According to their response, community participation in SWC practices is mostly dependent on their own interest and perception level of the soil erosion problems. As indicated in Figure 4, of the total 24 and 33 focus group discussion (FGD) and key informants interviewed respectively, 12 key informants and 18 FGD respectively were adopters of introduced/improved/stone and soil bunds. To assess the impact of implementing improved stone and soil bund structures on crop yield change, sample household adopters were asked to tell their crop yield harvest before and after adoption of improved SWC measures. Accordingly, of the 12 key informant adopters, 10 (83.3%) of them reported that they abled to produce an average of 1500 and 3000 kg/ha/year before and after adoption of improved stone and soil bund structures respectivelly during the harvest years of 2019/20 and 2020/21.

Fig. 4.

Focus group discussion with selected groups (a) and interview with SWC adopter and non-adopter key informants (b) (Photograph by the author, February [2022].

Similarily, of the 18 FGD adopters, 11 (61%) of them forwarded that they abled to produce an average of 1200 and 2500 kg/ha/year before and after adoption of improved stone and soil bund structures respectivelly during the same harvest years indicating that adoption of improved stone and soil bund has a significant contribution in increasing crop yield, income, grass for animal feed and improving household's livelihood. This implies that SWC adopter households are more aware of and are more beneficiary from the adoption of improved SWC measures than those of non-adopter households. During the field survey, the researcher realized that the basic source of livelihood of the under study community is agroforestry dominated agriculture mainly coffee and Enset plant followed by fruits and animal husbandry. It is also important to note that, according to the reflection of both adopter and non-adopter key informants Figure 4(b), they engage in an other income generating activities like daily labor, selling fuel wood, charcoal production, carpenter and local trades.

In the study area there are also community based land management practices which included all residentials. Of the 33 total interviewed key informants, 26 (78.78%) of them reported that local communities participate in two programs of land management and soil conservation practices that is applied either in an individual farm plots or communal lands. The first program was household participation under Safety Net program. According to the office experts, an average of 2123 people participate every month per year and the program has continued for 6 months during the dry season (December–April). Participants of the FGD Figure 4(a) also responded that this type of program have a monthly payment in cash budgeted by the Ethiopian government and technically supported by NGOs.

According to the key informants Figure 4(b), the second program was a free service of community campaign on SWC practices. FGD participants Figure 4(a) explained that in the campaign the local communities have participated based on the order of local leaders without their willing and incentives. According to the office experts, an average of 1422 people participate every month per year and the program had lasted only for one month work term. During the field observation, the researcher practically realized that there are community based land management activities in the study area. Interviewed key informants of the SWC adopters Figure 4(b) again explained that they do not participate on the planning and decision of work projects to implement improved stone and soil bund structures on individual farm plots. This report implies that SWC adopter households are passive in decision making for sustainable land management activities in their local area.

Generally, 56%, 79% and 82% of the surveyed households, FGD groups and interviewed key informants respectively responded that their major constraints which hider implementation of improved/introduced/SWC practices are: (i) limitation of technical skill (ii) small farm land size (iii) lack of incentives for adoption of SWC structures (iv) shortage of training opportunity and (v) the negative attitude of non-adopter farmers towards the newly introduced SWC technologies.

3.4. Logistic regression model for propensity score matching (PSM)

To measure the impact of SWC practices in logistic regression model for propensity score matching model, data collected from the two groups, namely: households adopting SWC practices and not adopting SWC practices were pooled such that the dependent variable takes a value 1 if the household participated in SWC practices and 0, otherwise Table 3. According to Ref. [6], the positive coefficient encourages households to continue use of SWC structures whereas the negative coefficient discourages households not to perform SWC activities. As indicated in Table 3, of the 11 total hypothesized variables intered in the logistic regression model, 8 (72.7%) of hypothesized variables including age, education, family size, farm land size, farming experience, off-farm activities, extension contact and market access of households was found to be positively associated with continuous use of soil and water conservation structures. On the other hand 3 (27.2%) of the variables consisting Sex, marriage and owing Tropical live stock was found to be negatively related with not continuous use of SWC practices which affects their activities Table 3.

Table 3.

Results of the logistic regression model for propensity score matching.

Variables	Coef.	Std. Err.	Z	P > z
SEX	−0.463	0.389	−1.19	0.234
Age	0.036	0.059	0.62	0.534
Marriage	−0.498	0.344	−1.45	0.147
EDU	0.040	0.147	0.28	0.783
FS	0.153	0.120	1.27	0.205
FLS	0.218	0.137	1.60	0.011**
FEX	0.028	0.016	1.77	0.077*
OFFARM	0.944	0.253	3.74	0.000***
EXCONTACT	0.951	0.123	7.70	0.000***
MARK	0.131	0.221	0.59	0.553
TLU	−0.627	0.843	−0.74	0.457
_cons	−3.849	1.180	−3.26	0.001***
Number of obs	262
LR chi2 (11)	162.47
Prob > chi2	0.0000
Log likelihood	−100.36096
Pseudo R2	0.4473

Source: Own survey results [2021]. Note: ***, ** and × means significant at the 1%, 5% and 10% probability levels respectively.

The logistic regression result in Table 3, also depicted that, out of the eleven hypothesized variables which were entered to the multiple linear regression models, four variables (farm land size, farming experience, off-farm activities and extension contact times) were found to be significant at 1%, 5% and 10% probability level.

3.5. Matching SWC adopter and non-adopter households

The main reason of imposing a common support region is to check any combination characterstics observed in the treatment and comparison group [5]. In setting the common support region in this study, minima and maxima comparison was made Table 4. The basic criterion for determining the common support is to delete all observations whose propensity scores smaller than the minimum of the program and larger than the maximum in the opposite group Table 4. Accordingly, the estimated propensity scores as shown in Table 4, vary between 0.072 and 0.997 (mean = 0.535) for treatment household while for the control households vary between 0.010 and 0.977 (mean = 0.493). This needs the use of common support region under the distribution of propensity scores lie between the two groups. Therefore, in this study our common support region Table 4 would lie between 0.010000 and 0.997000. Thus, households outside this range were not included under the matching process due to their contribution to bias in the estimation effect.

Table 4.

Distribution of estimated propensity scores by household type.

Group	Obs	Mean	Std.	Minimum	Maximum
Total households	262	0.504	0.362	0.010	0.997
Treatment households	132	0.535	0.265	0.072	0.997
Control households	130	0.493	0.235	0.010	0.977

Source: Own survey result [2021].

As indicated in Figure 5, the propensity score in common support after composition of all groups seem to overlap at 0.5 (centre). The propensity score of control households in common support before matching was 0.7 Figure 6 while the estimated propensity score of treated households in common support after matching was 0.8 Figure 6. The difference of common support before and after matching was 0.1 and this difference shows that there was no similarity between the propensity scores of control and treated households.

Fig. 5.

Kernel density of estimates distribution of households with respect to Pscore before matching all households. Source: Own survey result [2021/22].

Fig. 6.

Kernel density of estimated propensity score of treated in common support before matching (left side) and after matching (right side) matching. Source: Own survey result [2021/22].

Figure 7: shows the distribution of matched and unmatched individual groups with respect to estimated propensity scores. The result confirmed that there exists a considerable overlap in common support. Figure 7 illustrates the imbalanced distribution between treatment and control households and shows the cases excluded from the analysis to avoid bad matches. The result guarantees a sufficient overlap in the distribution of the propensity scores between adopters and non-adopters. The bottom half of the graph on the common support area stands for the propensity score distribution for the non-adopters and the upper half refers to the adopters. The densities of the scores are indicated on the Y-axis Figure 7.

Fig. 7.

Propensity score distribution between the treated and control groups both on and off the common support region. Source: Own survey result [2021/22].

3.6. Analysis results for adoption determinant factors of improved soil bund and stone bund practices

As indicated in Table 5, the choice of biprobit model consists two SWC practices, that was soil bund and stone bund. The model fit the Wald test ($\chi 2$ (38) = 61.47, þ = 0.0093 which statistically significant at 1% level, depicting that independent variables power of the factors in the applied model is satisfactory. Besides, results of likelihood ratio test in the other model (LR $\chi 2$ (1) = 77.317, prov$>\chi 2$ = 0.0000) was statistically significant at 1% level, showing that the explanatory variables of the disturbance terms is ignored and there were significant joint associations of the many estimated coefficients across the equations in the models. The simulation analysis results pointing out that the possibility of soil bund and stone bund were 64.2% and 61.5%, respectively. The joint possibilities of success and failure of the two explanatory was again recommended that it would be unlikely for farmers to choose the two market way on the same time, for their likelihood to do so was only 26.4%. As showed in Table 5 of the nineteen independent variables tested in bi-variant probit model, three of them significantly affected soil bund and six variables significantly affected stone bund at 1%, 5% and 10% probability levels.

Table 5.

Biprobit estimations for adoption determinant factors of improved soil bund and stone bund practices.

Variables	Soil bund			Stone bund
	Coef.	Std. Err.	mfx	Coef.	Std. Err.	mfx
Sex of HHHs	0.501	0.316	0.184	−0.450	0.381	−0.167
Marital status	0.227	0.457	0.087	−0.695	0.508	−0.258
Forest land access	0.021	0.177	0.007	0.020	0.173	0.007
Grazing land access	0.206	0.231	0.076	−0.195	0.286	−0.072
Farming experience	0.012	0.020	0.005	−0.015	0.019	−0.006
Perception soil erosion	−0.298	0.384	−0.105	−0.324	0.417	−0.120
Distance to farm land	0.005	0.008	0.002	−0.008	0.008	−0.003
Off-farm engagement	−1.412	0.341***	−0.504	−0.633	0.340*	−0.236
Modern Input used	0.183	0.321	0.069	−0.409	0.324	−0.152
Training	0.635	0.568	0.235	0.855	0.466*	0.318
Credit used	−0.926	0.413**	−0.340	−0.933	0.519*	−0.347
Market access	0.151	0.268	0.056	0.252	0.269	0.094
Vegetation cover	0.416	0.402	0.155	0.759	0.368**	0.282
Tropical livestock unit	0.964	0.917	0.327	2.882	0.980***	1.072
Family size	−0.035	0.038	−0.013	0.037	0.041	0.014
Education level	0.544	0.310*	0.201	0.766	0.316**	0.285
Age of households head	0.004	0.012	0.001	0.005	0.013	0.002
Farm size	−0.179	0.258	−0.066	0.283	0.323	0.105
Annual Income	−2.33 E+07	9.32E-06	−1.44E-07	8.73E-06	1.02E-05	3.25E-06
_cons	0.270	1.646		2.098	1.809
/athrho	14.142	432.628
Predicted probability	0.642	0.613
Joint probability (failure)	0.007
Joint probability (success)	0.263
Number of obs	132
Wald chi2 (38)	61.47
Log likelihood	−107.615
Prob > chi2	0.0093
Estimated correlation matrix
ρ1	1	ρ1		ρ2
ρ2	0.378***		1
LR test of rho = 0: chi2 (1) 77.317, Prob > chi2 = 0.0000

Note: ***, ** and × stands for 1%, 5% and 10% respectively. Source: Own survey result [2021].

As shown in Table 5, the result indicated that educational level of sample households affected the adoption of soil bund and stone bund positively at significant of 10% and 5%. This means educational level of household-heads enhance farmers' ability to access and use information and improves farmers' decision to adopt SWC practices. This finding is in conformity with the prior expectation. As the educational level of the sample smallholder farmers’ increased by 1 year of schooling, the adoption of soil bund and stone bund increased by 20.1% and 28.5% respectively. The possible explanation is that educated farmers have better knowledge on risks associated with soil erosion and hence tend to apply appropriate soil and water conservation measures for their individual farm land. Therefore, if education status of sample respondent farmers upgrade, the adoption status on soil bund and stone bund also increases. The result of this study indicated that education level of implementers have increased the adoption behavior of SWC practices in the study area. The result was similar with many research findings of [7,9]. But in the contrary to this result [24] found that illiterate farmer have better awareness on SWC than literate farmers that engaged in off-farm activities.

Off-farm involvement includes selling of fire wood, charcoal, medicinal value leaves, carpenter work, daily labor, small scale local tradings, transport service using motor cycle e. t. c. like in other parts of the region and Ethiopia [24]. The survey result shows that off/non-farm incomes of the households have a negative impact on use of soil bund and stone bund SWC practices respectively. As shown in Table 5, the result showed that Off/non-farm income affected the adoption of soil bund and stone bund negatively at significant of 1% and 10% respectively. As the possibility of households’ engagement in off/non-farm activities increases by one, the possibility to adopt soil bund and stone terrace is decreased by 50.4% and 23.6% respectively. The negative relationship within off-farm activities & application of improved SWC practices implying that others income may negatively affect the impetus of investing on measure that are aimed at increasing farm income soil and water conservation practices [26].

Credit used is another factor that determines and affects the use of soil bund and stone terrace negatively at 5% and 10% significant levels, respectively. As households’ received credit from formal credit sources, the adoption of the selected SWC (soil bund and stone bund) practices decreased by 34.0% and 34.7%, respectively Table 5. The correlation of credit and selected SWC practices is not in agreement with the prior hypothesis. The negative sign indicates that households who use credit could not adopt soil bund and stone bund. Even if there was credit service opportunity, the purpose of the credit mainly meant for purchasing livestock, house construction, and fertilizer or improved seed purchasing and others that fulfill necessity inputs for household member rather than managing their farm plot. Moreover, households who had access to credit use, they favored non/off-farm activities to improve their livelihoods rather than constructing SWC practices. This study agree with finding of [23] who reported that access to credit use harm the adoption of SWC practices negatively.

As depicted in Table 5, training on soil conservation activities affected particularly the adoption of stone bund positively at 10% significant level. The survey results showed that as participation of training related to SWC practice increased, the adoption of strip cropping increased by 31.8%. The probable reason for adopting stone bund due to training could be indigenous knowledge that the farmers acquired through intergeneration. Other reasons may be that farmers applied SWC measures, mainly focused only on the technical aspects. This result disagreed with previous research findings of [15] that training influenced the adoption of check dam negatively.

Similarily, as indicated in Table 5, the number of livestock size holding has a positive and significant effect on adoption of stone bund practices at 1% level of significances. The survey results implying that livestock size owned by the household increase by 107.2%. In other words, farmers who own more livestock are more likely to adopt SWC practices than those with less number of livestock owned. This positive result in livestock holding of households might be due to the fact that farmers who own relatively more livestock size make use of animal manure as a source of organic fertilizer. In addition, the income obtained from the sale of livestock could be used for purchase agricultural inputs and hiring of the labor force to undertake various activities of soil erosion control methods. This finding is consistent with [24] studies in Ethiopia.

4. Conclusion

The purpose of this research paper was to contribute to the existing literature of traditional knowledge and scientific methods for modernizing and improving soil and water conservation technologies in Ethiopia in general and in the study area in particular. To achieve the objective of the study a carefully technical evaluation of structural dimensions, household survey including focus group discussion, key informant interview and field observation were made. Overall, the authors conclude that (i) more than 55.6% of the physical design of the implemented SWC structures failed to meet the recommended standard dimensions (ii) educational level, off-farm activities, extension service, TLU, and farm land size holding were found to be statistically significant at (p < 0.05) respectivelly indicating that there were weak association between adopters and non-adopters (iii) of the 11 total hypothesized variables intered in the logistic regression model, 8 (72.7%) of them was found to be positively associated with continuous use of soil and water conservation structures (iv) out of nineteen explanatory variables included in bi-variant probit model, three variables significantly affected soil bund and six variables significantly influenced stone bund at 1%, 5% and 10% probability levels (v) 36% of the variables negatively affected the household farmer's adoption decision of soil and water conservation measures (vi) the difference of common support before and after matching is 0.1. Non adopter household heads were also aware of these improved structures but have chosen to reject them because they may not suit their particular circumstances. Inorder to generate additional income, both adopter and non-adopter groups engage on off-farm activities. The major identified household farmers' limitation of SWC practices are severe soil erosion, steep slopes, lack of technical skill to implement standardized physical designs, ignorance of incentives, lack of awareness about the benefit of improved structures and the negative attitude of some farmers towards the newly introduced SWC technologies.

Finally we recommended that implementation of SWC practices should be focused on minimizing the observed mismatch of the designed structures against the recommended standards. Policy makers, planners, experts, stakeholders, NGOs and other concerned bodies should give priority attention in implementing standard designs of improved physical structures at local, district, zonal, regional, national and international levels. To overcome all the constraints of adoption of improved and introduced SWC technologies, shifting farmers mind to the desired level of awareness through training, practical field demonstration and scaling up technical skill aspects is very necessary.

Author contribution statement

Mengistu Meresa: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Menfese Tadesse: Conceived and designed experiments; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Negussie Zeray: Conceived and designed experiments; Contributed reagents, materials, analysis tools or data.

Funding statement

Mengistu Meresa was supported by Dilla University.

Data availability statement

Data will be made available on request.

Declaration of interest's statement

.