Determinants of Smallholder Farmers’ Decisions to Use Multiple Climate-Smart Agricultural Technologies in North Wello Zone, Northern Ethiopia

Abstract

Climate change presents significant challenges to agriculture in developing nations, affecting farmers’ livelihoods and food security. In Ethiopia, agriculture is crucial to the economy and the well-being of millions. This study focuses on analyzing the determinants that affect smallholder farmers’ adoption of climate-smart agriculture (CSA) technologies in the North Wello administrative zone, northern Ethiopia. Through multivariate and ordered probit econometric models, data from 411 rural household heads were analyzed. Results reveal the synergy among climate-smart agricultural practices in rainfed farming systems, essential for crafting a comprehensive CSA package within an agroecological framework. The multivariate probit estimation results show that education, membership in local organizations, farm size, tropical livestock unit (TLU), irrigated land ownership, plot number, steep slope, farmland distance to home, and access to a local media source are positive determinants for the decision to use the type and several climate-smart agricultural practices. Lack of credit, large family size, distance from extension services, and proximity to the nearest market were all associated with lower adoption of all CSA technologies. The findings suggest that blanket recommendations for climate-smart agricultural technologies for smallholder farmers can be avoided. The complementarities that exist between CSA technologies may require further investigation into how such complementarities have benefits in terms of improving land productivity and food security and reducing climate-related risks for smallholder farmers in Ethiopia and other contexts. Moreover, by emphasizing an agroecological framework, the study promotes environmentally sustainable and socially equitable agricultural practices that are economically viable, contributing to broader environmental sustainability and development goals.

1. Introduction

Rainfed agriculture provides over 95 percent of Africa’s food, yet only 5 percent of irrigation supports it [1,2]. In Sub-Saharan Africa, rainfed farming covers about 97 percent of cropland, making it highly vulnerable to climate change due to its fragile environment, economic dependence, and low adaptive capacity [3,4,5,6]. Africa has experienced a rapid temperature increase of 0.5 degrees Celsius per century, exceeding the global average [7,8], with rainfall patterns becoming increasingly unpredictable [9]. Climate change has adversely affected agricultural yields and food security in Africa [10,11,12].

Ethiopia’s economy heavily relies on agriculture, contributing around 42–45 percent to the GDP, driving 90 percent of exports, and employing over 80 percent of the population [13,14]. However, this sector is highly vulnerable to climate-related shocks, with rising temperatures and erratic rainfall patterns threatening land productivity and household food security [14,15,16,17,18,19,20]. Extreme weather events also negatively impact land suitability for crop production [21,22]. The World Bank estimates that land degradation costs Ethiopia approximately 2 to 3 percent of its agricultural GDP annually [23].

Despite its high potential for economic growth in Ethiopia, agriculture requires transformation to address the problems of land degradation and low productivity caused by climate change. Therefore, it is crucial to provide guidance for agricultural management in the face of climate change [24,25,26,27]. Given these considerations, the Ethiopian government recently unveiled a vision for establishing a climate-resilient green economy by 2025 [28]. Climate-smart agriculture is an approach for steering agricultural management in a changing climate [27]. Climate-smart agriculture is a new concept developed to address climate-related risks, promoting sustainable land productivity and food security, enhancing ecosystem resilience, and reducing greenhouse gas source activities [29,30,31]. For FAO [31], climate-smart agriculture is neither a new agricultural system nor a set of practices, but rather a new approach for guidance in transforming agriculture to address food insecurity and climate change challenges. This involves addressing social, economic, and environmental factors central to transforming agriculture, aiming to enhance productivity and resilience for food security, while also mitigating risks associated with climate change [26,32].

Climate-smart agriculture practices are generally highly knowledge-intensive and site-specific [27,33], making them essential to address food security and land productivity challenges in the changing climate. The effectiveness of CSA practices is determined by the context in which they are applied [34]. Thus, climate-smart agriculture is a pathway towards sustainable development and food security and is built on three pillars: increasing agricultural productivity (crops, livestock, and fisheries) and income; enhancing the resilience or adaptation of livelihoods and ecosystems towards climate extremes; and reducing and removing GHG emissions from the atmosphere [31]. An agricultural technique or practice contributing to these pillars’ achievements can be considered climate-smart [35]. Ethiopian smallholder farmers have implemented several CSA practices, primarily to boost their farmland productivity and crop production [20,36]. However, due to differences in socioeconomic, plot, institutional, and social factors, the implementation of CSA practices varied among smallholder farmers in most developing countries including Ethiopia [37]. CSA practices are highly context-dependent, and evidence-based interventions are needed [38].

The empirical evidence revealed that CSA practices, alone or in combination, provide greater net benefits than business-as-usual practices [39,40,41,42,43,44,45,46]. As an example, Tadesse et al. [44] demonstrated that implementing climate-smart agriculture (CSA) practices resulted in crop yields that were 30–40% higher compared to conventional practices. Additionally, the total carbon stored at a soil depth of 1 m was three- to seven-fold higher with CSA practices than with conventional methods. Similarly, Asrat and Simane [39] revealed that farmers who apply CSA practices experienced higher productivity by 22.2% compared to those who were under conventional practices.

Despite its novel benefits, CSA practices in Ethiopia remain low and uneven among smallholder farmers [25,47,48,49,50]. Several studies revealed that households’ asset bases such as social, natural, and physical capital, as well as institutional factors, influence choices and adoptions of CSA practices in Ethiopia [40,51,52,53,54,55,56,57,58,59]. Nevertheless, the impact of these factors typically varies depending on the specific context, including the location and the technologies being evaluated [60].

Most of the previous studies in Ethiopia have focused mainly on the adoption of single technologies by employing a binary choice model between adopters and non-adopters and the analysis of factors that affect the adoption. Many recent studies have highlighted that farmers have the option to adopt various combinations of technologies, which may either supplement or complement each other, to address multiple climate-related risks, such as soil moisture stress, soil fertility loss, pest infestations, drought, and floods [51,59,61,62,63]. Geographically, most of the earlier studies were delimited to the south and northwest of Ethiopia. For instance, Ewunetu et al. [60], Nigussie et al. [62], and Tekelewold et al. [58] focused on the upper Blue Nile River basin, while the rest, Bedeke et al. [52] on the Wolaita administrative zone and Negera et al. [61], also studied the Bale eco-region. Though these studies provide crucial information on a variety of CSA interventions in the wet and moist parts of the country, given the broader variation topography, micro-climate, and socio-economic factors, the available literature do not provide a full perspective of the practice and adoption of CSA. Particularly, empirical evidence is scant in northern Ethiopia, where frequent drought and food insecurity problems are prevalent. Ethiopian agriculture exhibits diversity across climatic zones, crop production systems, and socioeconomic conditions [25]. Therefore, it is essential to carefully select suitable combinations of climate-smart agriculture (CSA) interventions tailored to specific geographical scales and livelihood requirements.

Farmers in northern Ethiopia including the North Wello administrative zone are known to have experienced the worst famine in history due to climate change-related risks. To overcome this problem, various modern agricultural technologies (improved crop varieties, minimum tillage, agroforestry, crop diversification, rainwater harvesting, organic fertilizer, inorganic fertilizer, and small-scale irrigation) have been introduced to farmers with the support of the government and non-government institutions. However, not all farmers are using these technologies as much as they should. The question of why smallholder farmers in this area do not apply these technologies to the required extent is not yet known. Hence, our study seeks to fill this void in the literature by systematically examining the factors influencing the adoption of climate-smart agriculture (CSA) technologies through a comprehensive multivariate choice framework. The results of this analysis aim to provide policymakers with valuable insights into the decision-making process of smallholder farmers regarding technology adoption, facilitating the design and implementation of targeted policy measures and interventions to promote the wider uptake of CSA technologies.

2. Research Methodology

2.1. Study Area Description

This study was conducted in the North Wello administrative zone found in northern Ethiopia. The study area lies between latitude 11°30′0″ to 12°30′0″ N and longitude 38°30′0″ to 40°0′0″ E (Figure 1) with an elevation range from 916 to 4197 m.a.s.l. The study area is about 12,212.3 km² spanning nine rural administrative districts and four administrative towns. The North Wello administrative zone is characterized by unique rainfall patterns, featuring a bi-modal distribution with the Belg season occurring in April–May, followed by the primary wet season, Kermit, from July to September [64]. This region exhibits diverse topography, including towering mountains, deeply carved canyons, gorges, valleys, and plateaus. The significant relief disparities have resulted in widespread degradation due to human activities. Agriculture predominantly relies on rainfall, with various crops cultivated across distinct agroecological zones.

2.2. Research Design and Sampling Procedure

In this study, we employed a cross-sectional research design along with a mixed research approach to mitigate the limitations associated with relying solely on a single method. A multistage sample design technique was utilized to select the kebeles (the lowest administrative level) and household heads. Initially, the North Wello zone was divided into similar agroecological areas, followed by grouping the nine rural administrative woredas within the zone into potential agroecological systems. Subsequently, three representative districts (Habru, Gubalafto, and Gidan) representing lowland, midland, and highland climate zones were randomly selected. This approach aimed to focus the research on agronomic and livelihood activities and adaptation strategies specific to each climate zone. Kebeles (the lowest administrative unit next to districts) sharing dominant features with each sampled district were identified through discussions with agronomic experts in the area, thus excluding those not representative of the sampled district. Finally, six kebeles, two from each sampled district, were selected using a stratified sampling technique. The number of sampled households was determined using the probability proportional to size (PPS) method, following Kothari’s [64] standard.

$$\mathrm{Mathematicaly}\mathrm{presented}:\mathrm{n}=\frac{Z^2\times p\times q\times N}{e^2\times (N-1)+Z^2\times p\times q}$$

where, n represents the desired sample size, Z is the upper point of the standard normal distribution at a 95% confidence level (equal to 1.96), p is the proportion of households (set at 0.5 or 50%), q is 1 minus p, e is the allowable error or precision (0.05 or 5%), and N is the total number of households in all kebeles. Given the total number of households across all kebeles and using the aforementioned formula, the minimum sample size required for reliable results was calculated to be 366. Furthermore, to ensure a more representative sample and accommodate a stated level of precision, the calculated sample sizes were adjusted to incorporate a nonresponse rate of 15%. Consequently, based on the formula mentioned above, a total sample size of 421 households was selected and proportionally distributed among the six kebeles using the following formula:

$$\mathrm{n}i=\frac{\mathrm{n}\ast \mathrm{N}\mathrm{i}}{\mathsf{\Sigma }Ni},$$

where n = determined sample size, ni = the required sample household size in the ith kebele, and Ni = total household in the ith kebele. Sample households were selected using a systematic random sampling technique. As shown in

Table 1. Displays the total number of households in the study areas along with the corresponding sample sizes.

Agroecological Zones	Kebele	Major Crop Type	Major Livestock	Total HH	Sample HH
Midland (Woyina Dega)	Gedober	Teff, Sorghum, Millet	Cattle, sheep, goat	1515	87
	Lay_Aluwuha	Teff, Sorghum	Cattle, camel, goat	1600	91
Lowland (Kolla)	Humo	Sorghum, Teff	Camel, cattle, goat, chicken	718	41
	Dodota	Sorghum, Pepper	Camel, cattle, goat	847	48
Highland (Dega)	Worka-Worku	Barely, wheat, lentil, bean, pea	Sheep, cattle, mules,	1674	95
	Kanbat	Barely, wheat, lentil, bean, pea	Sheep, cattle, mules,	1031	59
Total				7385	421

Source: sampled kebele administration office, 2021.

, the list of all household heads was received from the kebele offices and used to create sampling frames for each kebele. Hence, 421 household heads were included in the survey during the period November to March 2020.

2.3. Data Sources and Collection Methods

The authors of this study collected cross-sectional data through survey questionnaires held in February and March 2021. We had assured the reliability and validity of the questionnaire based on the experience in previous studies, feedback from local development agents, and pretesting on 15 randomly selected farm households in a similar district. Some questions were avoided, modified, and reorganized based on the feedback received. Under the strict supervision of the first author, the questionnaire had been administered by well-trained enumerators. Moreover, to enhance the quantitative data, a set of focus group discussions and in-depth interviews were conducted. A total of six Focused Group Discussions (FGDs) were conducted, that is, one in each sampled kebele. The discussants were farm household heads selected in consultation with agricultural extension experts working in each kebele, considering their farm experiences, active participation in farmers associations in the kebele, and implementing agricultural technologies introduced in the kebeles. Each focus group discussion (FGD) convened a limited number of 6–8 participants. Across all six kebeles, a total of 45 individuals participated in these discussions, with only three being women, indicating a male-dominated composition. The first author led each focus group session, employing a checklist as a guiding tool.

2.4. Descriptions of Variables

2.4.1. Dependent Variable: CSA Practices

In this study, CSA refers to the knowledge-based decision to implement agronomic, biological, and physical measures on cultivated land to boost land productivity, food security, and ecosystem services. Smallholder farmers were asked to describe the major climate-related risks encountered over the last ten years. Following the experience of farmers with the changing climate, they were asked to list the practices they adopt to maintain their agricultural land productivity under the changing climate conditions. Agricultural practices which fit into the Ethiopia climate-smart agriculture scoping study [37] and Ethiopia climate-smart agriculture road map [25] were identified for the study. Improved crop varieties (pest resistance, high-yielding, tolerant to drought, and short season), crop diversification, agroforestry, minimum tillage, small-scale irrigation, rainwater harvesting, and the efficient use of inorganic and organic fertilizers have been the most widely used CSA practices identified in the study area (

Table 2. Definitions of the CSA practices adopted by sampled smallholder farmers.

CSA Practices	Definitions
Crop diversification (CD)	Growing more than one improved crop variety that survives in adverse climatic conditions across farmland or time [58,65].
Improved crop varieties (CV)	Adopting improved seed varieties such as pest resistance, high-yielding, drought tolerant, and early maturing especially for the major staple that could improve crop yield [66,67].
Minimum or reduced tillage (MT)	Reducing soil disturbance and allowing crop residue to remain on the ground, increasing soil organic carbon, vital for soil fertility and structure [37,68,69].
Agroforestry (AF)	Cultivating trees that can be woody perennials and deliberately integrated with crops [70,71,72,73,74,75]
Small-scale Irrigation (IR)	Year-round cropping uses both surface and sub-surface water resources. Effective means of smoothing out yield instability in rainfed systems [76,77,78,79]
Rainwater harvesting (RW)	Intentional collection of rainwater from a surface known as a catchment and its storage in physical structures (micro-catchments, broad beds, and furrows) or within the soil profile [80,81].
Compost/Organic fertilizer (OF)	Using organic materials, such as animal waste, weeds, farm waste, straw/hay leftovers, dried leaves, and ash [82,83,84].
Chemical fertilizers (Diammoniate phosphates (DAP) and Urea) (CF)	Using chemical fertilizers with proper timing and amount to boost yield and yield components of crops [45,47,85,86]

). Eight climate-smart agricultural practices were employed as an outcome variable. The adoption intensity refers to the number of climate-smart agricultural practices adopted by smallholder farmers per their farm plot.

2.4.2. Explanatory Variables Employed in the Model

The selection of explanatory variables incorporated into the model specification was guided by both theoretical frameworks and empirical evidence derived from the existing literature [17,40,45,48,87,88,89,90,91,92]. The explanatory variables encompass demographic aspects (such as education, age, gender, and household size), plot attributes (including plot number, distance from the farm, slope, soil quality, and land tenure), asset-related factors (such as farm size, tropical livestock units (TLU), and irrigable land), as well as other variables associated with climate-smart agriculture (CSA) training, extension services, infrastructure, and institutional support. A detailed description of these selected variables along with descriptive statistical measures can be found in

Table 3. Descriptive statistics of the explanatory variables used in the MVP model.

Variable’s Name	Variable Description (Coding/Units)	Mean	SD
Demographic and Social factors
Gender	Household head gender type, 1 = Male, Female = 0	0.87	0.33
Age	Farm household head’s age (years)	46.90	9.74
Education	1 = if the literate household head	0.36	0.48
Family size	Number of family members (count)	5.57	2.02
Dependency ratio	Household members aged below 15 and above 64 (count)	2.08	1.52
Social Membership	=1 if farmers belong to at least one social membership, 0 otherwise	0.88	0.32
Farm characteristics
Plot number	Number of plots to be cultivated as land (count)	2.94	1.18
Farm distance	Walking distance of the plot from home, minutes	37.73	19.38
Steep slope	The slope of farmland is perceived as very steep, hectare	0.12	0.24
Moderate slope	The slope of farmland is perceived as moderate, hectare	0.37	0.35
Gentle slope	The slope of farmland is perceived as gentle, hectare	0.60	0.40
Tenure	=1 if the household head owns the farmland ownership certificate, 0 otherwise	0.99	0.09
Poor soil fertility	=1 if farmland soil status is perceived as poor fertility, 0 otherwise	0.11	0.31
Cropland Rent	=1 if the household head shared cropland with others, 0 otherwise	0.31	0.46
Farm resources
Irrigable land	=1 if the household heads own irrigable land, 0 otherwise	0.27	0.44
Farm size	Area of cultivated land, in Timad2	1.10	0.57
TLU	Livestock size (tropical livestock unit; TLU)	4.01	2.19
Credit access	=1 if the household received credit when they needed it, 0 otherwise	0.70	0.46
Institutional and infrastructural-related factors
Information on climate	1 = if extension experts give climatic information to farmers, 0 otherwise	0.25	0.43
Visits of extension agent	Number of yearly visits for extension agents	2.95	1.02
Distance to market	Market distance to the residence (minutes of walking)	99.64	125.24
Distance to extension service	Extension service office distance to the residence (minutes of walking)	47.62	36.91
Radio	=1 if household heads own radio/mobile, otherwise	0.81	0.40

According to Asfaw et al. [93], the tropical livestock unit (TLU) conversion factors are as follows: camel = 1, cattle = 0.7, horse = 0.8, mule = 0.7, donkey = 0.5, sheep/goat = 0.1, and chicken = 0.01. Additionally, local farmers use a measure called ‘2 Timad’ to describe farm size, where 1 Timad is equivalent to 0.25 hectares (source: key informant interview).

below.

2.5. Analytical Model

Model Specifications: Multivariate Probit (MVP) Model

A farmer is more inclined to embrace a particular CSA technology when the advantages it brings outweigh the decision not to adopt it. Estimating the determinants of farmers’ decisions to choose CSA practices can be achieved using multinomial probit or logit models for each CSA practice and similar covariates distinctly [87,94]. However, adopting a discrete choice model (multinomial probit or logit) may lead to ignoring the simultaneous use of decision behavior and failure to capture complementarities and trade-offs associated with implementing individual climate-smart agricultural practices. Ignoring simultaneous adoption overlooks potential unobserved variables that may influence the adoption of multiple CSA practices, as noted by Kassie et al. [91]. To overcome the drawback of the multinomial probit or logit models, the multivariate probit model (MVP) is far better. There is a broad application of this method in climate change adaptation studies [14,59,91,95] using cross-sectional multiple plot data. The selection of MVP models for a specific practice might rely on the existence of another option, influenced by either their complementary nature or their ability to substitute each other, as indicated by Tekelewold et al. [58]. Our MVP model is composed of eight equations, each representing a binary choice: organic fertilizer, inorganic fertilizer, irrigation, crop diversity, improved crop variety, minimum or reduced tillage, agroforestry, and rainwater harvesting practices.

The MVP model is specified as follows:

$${\gamma }_{\mathcal{i}m}^{*}={\beta }_m+Xim+{\epsilon i}_m\mathrm{m}=1,2,3,\dots 8$$

(1)

$${\gamma }_{\mathcal{i}m}^{*}=1if{\gamma }_{\mathcal{i}m}^{*}>0and0otherwise$$

In this context, ${\gamma }_{\mathcal{i}m}^{*}$ represents a latent variable encompassing the underlying preferences related to the selection of a particular practice (m = 1,2, 3, …8, representing the considered technologies). This latent variable is presumed to be a linear composite of observed attributes, denoted as Xim (comprising demographic, social, economic, farm characteristics, and institutional factors), along with unobservable characteristics encapsulated by the stochastic error terms, ${\epsilon i}_m$. The parameter vector to be estimated is represented by ${\beta }_m$.

In the context of the multivariate probit model encompassing the choice to adopt multiple CSA practices, the error terms collectively adhere to a multivariate normal distribution (MVN) characterized by a zero conditional mean and variance normalized to unity. This distribution is denoted as MVN (0, Ω), with Ω representing the systematic variance–covariance matrix. Specifically, the diagonal elements of Ω have a value of 1, while the off-diagonal elements signify non-zero correlations.

The subsequent segment of our econometric methodology delves into investigating the determinants impacting the degree of CSA practice adoption using an ordered probit modeling (OPM) technique. Here, the degree of adoption is quantified by a count variable indicating the total number of practices implemented. Given that farmers have gained prior exposure and experience with CSA practices along with an understanding of their benefits, the process of adopting the initial practice may vary from subsequent practices.

3. Results of the Study

3.1. Respondent’s Demographic and Socioeconomic Characteristics

Table 3 presents descriptive statistics pertaining to the variables incorporated within the models. About 87.3 percent of household heads were male, and the average age was 46.9 years. The average number of people in a household between the ages of 15 and 65 was three. The average family size of the study area was 5.57, which is higher than the rural national average of 5.1 [96]. About 64.5% of household heads were uneducated and could not read or write. A proportion of 88.3% of households were members of local social organizations like local farmers’ associations, “Idir”, and “Equib”. About 70.3% of farmers in the study area needed credit but were unable to obtain it, which could constrain both the intensity and rate of agricultural technology adoption.

The average distance to the nearby local market was 1.66 walking hours, and only 26.7% of the respondents had reported access to irrigable land. The majority of households within the study area are identified as small-scale, typically owning an average of 1.1 hectares of land, supporting an average household size of approximately 5.57 individuals, and possessing an average livestock holding of 4.01 TLU. The average total landholding of the study area was low compared with the average landholding of the national average, 1.17 ha [17]. Farmers in highland areas held significantly more land than farmers in lowland and midland areas, while lowland farmers owned more livestock compared to highland and midland areas.

3.2. Types and Spatial Variations of CSA Practices

Figure 2 illustrates the particular types of CSA practices implemented and their spatial distribution. The findings indicate that, notably, inorganic fertilizers (chemical fertilizers) emerged as the predominant practice adopted by respondent households across all agroecological zones. A total of 54.5% of household heads used chemical fertilizers to boost crop production and productivity among the total respondents. It was used more in the highlands (72.48%) than in the midlands (53.45%) and the lowlands (26.14%). With regard to the improved crop varieties used, better performance was observed in the midland (50.57%) than in the highland (36.24%) and lowland (18.18%). More differences were observed in irrigation practices. It was more implemented in the midland (37.93%) than in the highland (16.78%) and lowland (6.82%) agroecological zones. Similarly, crop diversification practices were implemented more in the midland (22.41) than in the highland (18.12) and lowland (15.91). Organic fertilizer (compost and manure) application was used by 43.1% of the respondents in the midlands, 33% of the highlands, and 29.55% of the lowland areas on at least one plot. Agroforestry practices were higher in the lowland (21.59%) than in the highland (12.75%) and midland (4.02%) parts of the study area. Rainwater harvesting was also primarily used in lowland areas (12.5%) rather than in midland (10.34%) and highland (4.7%) agroecological zones. In contrast, minimum tillage was implemented in the highland (26.85%) rather than the lowland (18.18%) and midland (6.9%) parts of the study area. In general, the sample respondents used 54.5% of chemical fertilizers, 38.44% of improved seeds, 36.5% of organic fertilizers, 23.6% of irrigation, 19.46% of crop diversification, 16.55% of minimum tillage, 10.95% of agroforestry, and 8.76% of rainwater harvesting in their overall agroecological zones (Figure 2).

According to FGDs and key informants’ reports, the low adoption rates of rainwater harvesting and agroforestry among smallholder farmers are primarily attributed to the fact that these practices often necessitate initial investments in infrastructure and resources. For instance, in the case of rainwater harvesting, farmers require suitable catchment areas, storage tanks, and irrigation systems, all of which can entail significant costs, particularly for smallholder farmers with limited financial resources. Similarly, the establishment of agroforestry systems demands both time and resources for procuring seeds, saplings, and implementing sustainable land management practices. Another constraint arises from technical challenges. Both rainwater harvesting and agroforestry necessitate technical expertise for proper implementation and management. Farmers must comprehend various factors such as soil type, climate patterns, and tree-crop interactions to effectively integrate these practices into their farming systems. The weak extension services and technical support can impede successful adoption and result in suboptimal outcomes.

3.3. The Intensity of CSA Technologies Implementation across Different Agroecology

Figure 3 illustrates the varying levels of adoption of CSA technologies within different agro-ecologies in the study region. As shown in Figure 3, 18.98% of farmers in the study area did not use any climate-smart agricultural practices, while 73.49% used one to four climate-smart agricultural practices. The remaining 7.54% of farmers used more than five climate-smart agricultural practices. Only 8% of smallholder farmers in the midland area did not implement any of the CSA practices across all of their farm plots, whereas 86% used one to four practices. Moreover, 6.3% of the respondents implemented five to eight CSA technologies in the midland agroecological zone (Figure 3). More than 70% of the respondents in the highland and 54.5% in the lowland agroecological zone used one to four practices. A significant number of respondents (38.6%) in the lowland and (20.13%) in the highland areas did not use the CSA technologies considered in this study (Figure 3).

3.4. Model Results

The outcomes of the MVP and OPR models are outlined in

Table 4. MVP joint covariance matrix of regression results between CSA practices.

	CV	CD	OF	IR	AF	RW	MT	CF
CV	1.000
CD	0.444 (***)	1.000
OF	0.425 (***)	0.291 ***	1.000
IR	0.190	0.169	0.394 ***	1.000
AF	0.123	0.485 ***	0.416 ***	0.304 **	1.000
RW	−0.046	0.200	0.418 ***	0.176	0.582 ***	1.000
MT	0.085	0.396 ***	0.356 ***	0.468 ***	0.370 ***	0.332 **	1.000
CF	0.407 ***	0.424 ***	0.208 **	0.254 **	0.127	−0.102	0.296 ***	1.000

Likelihood ratio of rho CD,CV = rho OF,CV = rho AF,CV = rho RW,CV = rho MT,CV = rho CF,CV = rho OF,CD = rho IR,CD = rho AF,CD = rho RW,CD = rho MT,CD = rho CF,CD = rho IR,OF = rho AF,OF = rho RW,OF = rho MT,OF = rho CF,OF = rho AF,IR = rho RW,IR = rho MT,IR = rho CF,IR = rho RW,AF = rho MT,AF = rho CF,AF = rho MT,RW = rho CF,RW = rho CF,MT = 0: chi²(28) = 178.715 prob > chi² = 0.0000; *** and ** refers significance at 1% and 5% levels, respectively.

and

Table 5. Coefficient estimates of the multivariate probit (MVP) model and Ordered Probit Model (OPM) (standard errors in parentheses).

Explanatory Variables	(MVP) Model of Climate-Smart Agricultural Practices (Dependent Variables)
	ICV	CD	OF	IR	AGF	RWH	MT	CF
Demographic and Social factors
Gender	−0.075 (0.242)	0.437 (0.285) a	−0.283 (0.224)	0.420 (0.413)	0.477 (0.419)	0.560 (0.585)	−0.540 (0.247) b	0.094 (0.240)
Age	0.008 (0.009)	0.014 (0.009) a	0.009 (0.008)	0.015 (0.015)	0.031 (0.011) c	0.041 (0.014) c	0.003 (0.010)	−0.008 (0.009)
Education	0.437 (0.164) c	0.202 (0.164)	0.228 (0.150) a	−0.031 (0.250)	0.213 (0.207)	0.611 (0.253) b	0.055 (0.181)	0.296 (0.160) a
Family Size	0.023 (0.053)	0.037 (0.055)	−0.004 (0.050)	−0.202 (0.099) b	−0.047 (0.070)	−0.063 (0.090)	0.046 (0.059)	0.057 (0.054)
Age Dependency Ratio (ADR)	−0.010 (0.062)	−0.083 (0.064)	−0.021 (0.060)	−0.091 (0.109)	0.111 (0.079)	0.183 (0.098)	0.018 (0.071)	−0.256 (0.065)
Social Membership	0.382 (0.271)	0.374 (0.256) a	−0.312 (0.233)	1.246 (0.515) c	−0.272 (0.366)	0.557 (0.586)	−0.281 (0.294)	0.168 (0.255)
Farm resources
Farm size	−0.781 (0.675)	−0.184 (0.677)	0.553 (0.580)	0.407 (1.149)	1.285 (0.725) a	2.115 (0.852)	−1.154 (0.886)	1.281 (0.550) b
TLU	−0.039 (0.042)	−0.018 (0.045)	0.031 (0.042)	−0.016 (0.062)	0.086 (0.051) a	0.123 (0.064) b	−0.033 (0.047)	0.135 (0.044) c
Irrigable land	0.566 (0.253) b	0.575 (0.258) b	0.191 (0.244)	2.472 (0.346) c	−0.060 (0.348)	0.038 (0.417)	0.117 (0.354) a	0.199 (0.282)
Credit	0.033 (0.175)	−0.135 (0.184)	−0.176 (0.167)	−0.786 (0.308) c	0.095 (0.234)	0.346 (0.335)	−0.790 (0.189) c	−0.167 (0.185)
Farm characteristics
Plot number	0.052 (0.077)	−0.074 (0.084)	0.002 (0.078)	0.006 (0.135)	0.063 (0.098)	0.246 (0.113) b	0.217 (0.091) b	0.271 (0.085) c
Gentle slope	0.917 (0.679)	0.776 (0.691)	−0.065 (0.588)	−0.435 (1.182)	−0.779 (0.739)	−2.176 (0.884)	1.806 (0.902) b	−0.933 (0.546)
Moderate slope	0.661 (0.669)	0.571 (0.669)	−0.170 (0.585)	−0.754 (1.136)	−0.502 (0.723)	−2.579 (0.900)	1.473 (0.883) b	−1.117 (0.551)
Steep slope	1.237 (0.735) a	0.678 (0.727)	−0.293 (0.636)	0.520 (1.217)	−0.822 (0.777)	2.393 (0.944) c	1.121 (0.902)	0.938 (0.633) a
Soil fertility	0.069 (0.250)	0.338 (0.264	0.135 (0.238)	−0.134 (0.445)	0.240 (0.367)	0.537 (0.402)	0.359 (0.319)	0.039 (0.245)
Farmland rent	0.322 (0.162) b	−0.006 (0.180)	0.290 (0.155) a	0.468 (0.266) a	0.586 (0.219) c	0.763 (0.256) c	−0.443 (0.207) b	0.542 (0.170) c
Farm distance	0.008 (0.004) b	0.010 (0.004) a	0.002 (0.004)	0.006 (0.007)	0.003 (0.005)	−0.004 (0.007)	0.004 (0.005)	0.010 (0.004) c
Institutional and infrastructural-related factors
Radio	0.191 (0.229)	0.216 (0.245)	0.207 (0.212)	0.487 (0.424)	1.287 (0.464) c	1.328 (0.571) b	−0.122 (0.252)	0.266 (0.224)
Distance from home to extension	0.001 (0.003)	−0.004 (0.003) a	−0.004 (0.002) a	−0.002 (0.004)	0.000 (0.004)	−0.004 (0.005)	0.001 (0.003)	−0.003 (0.002)
Distance from home to the nearest market	−0.002 (0.001) b	0.000 (0.001)	−0.004 (0.002) a	0.001 (0.001)	−0.001 (0.001)	−0.002 (0.003)	0.001 (0.001)	−0.001 (0.001) b
Extension contacts	0.065 (0.083)	−0.065 (0.089)	0.077 (0.079)	−0.095 (0.142)	−0.010 (0.105)	0.124 (0.131)	0.354 (0.096) c	−0.209 (0.082) c
Climate-related information	0.786 (0.178) c	−0.343 (0.209) a	0.472 (0.173) c	−0.069 (0.302)	−0.340 (0.256)	−0.233 (0.294)	0.044 (0.225)	−0.680 (0.181) c
Constant	−2.060 (0.616) c	−2.150 (0.632) c	−0.369)0.621)	−2.815 (1.047) c	−6.147 (1.003) c	−7.556 (1.386) c	−2.309 (0.671) c	0.729 (0.574)
Wald chi2 (200) = 600.64; Prob > chi2 = 0.0000

The bold coefficients, as well as those labeled with a, b, and c, signify significance at the 10%, 5%, and 1% levels, respectively. The number of observations for this analysis is 411, and the data source is attributed to field survey data processed by the author.

, respectively. The Wald test statistic, indicating the overall significance of the model (p = 0.000), reinforces the suitability of the MVP model for the dataset. Additionally, the significance of the likelihood ratio test (p = 0.0000) further supports the choice of the MVP model. Regarding the OPR model, its results indicate a good fit to the data (p = 0.0000; log-likelihood ratio (X² (24) = 147.180); pseudo R² = 0.0993).

3.4.1. Interdependence of Multiple CSA Practices

Table 4 presents the binary correlations among the error terms of the eight adoption equations. Across 19 out of 28 combinations of the eight CSA practices observed in the study area, the pairwise correlations between the error terms were statistically significant at levels of p = 0.01 and 0.05. Notably, most practices exhibited positive correlations, indicating that farmers commonly employed a blend of CSA technologies, thus supporting the premise of the researchers. Rainwater harvesting and agroforestry had the highest positive correlation (58.2 percent) among positive and significant correlations (p < 0.01). A positive and significant (48.5 percent) correlation was also found between agroforestry and crop diversification (p < 0.01). Minimum tillage or zero tillage shows a positive correlation with irrigation (46.8%), which demonstrated the complementarities between the two technologies (p < 0.01). Another interesting result was a correlation (44.4%) between improved crop variety and crop diversification (p < 0.01), showing the synergies of the two practices. Organic fertilizers like compost and manure and improved crop varieties (p < 0.01), chemical fertilizers, and crop diversification (p < 0.01) showed positive and significant associations. The positive correlations among the practices affirm the idea that farmers typically embrace a variety of CSA technologies concurrently. These outcomes provided validation for our use of the multivariate probit model (MVP) over the independent estimation of eight probit equations.

3.4.2. Factors Affecting the Intensity of Practicing CSA Technology

Table 6. The estimated coefficients of the ordered probit model (OPM).

	Coef.	Std. Err.	Z	p > z
Demographic and Social factors
Gender	0.012	0.173	0.07	0.94
Age	0.016	0.006	2.53	0.01 ***
Education	0.066	0.116	0.57	0.57
Family Size	−0.011	0.039	−0.27	0.79
Social Membership	0.017	0.183	0.09	0.93
Farm resources
TLU	−0.039	0.030	−1.29	0.20
Farm size	0.174	0.439	0.40	0.69
Irrigable land	0.871	0.193	4.51	0.00 ***
Credit	−0.261	0.128	−2.04	0.04 **
Farm characteristics
Plot number	0.170	0.060	2.86	0.00 ***
Gentle slope	0.304	0.444	0.68	0.49
Moderate slope	−0.089	0.444	−0.20	0.84
Steep slope	0.285	0.483	0.59	0.56
Cropland sharing	0.386	0.122	3.17	0.00 ***
Poor soil fertility	0.234	0.184	1.27	0.20
Institutional and infrastructural-related factors
Radio	0.312	0.161	1.93	0.05 **
Distance to extension service	−0.003	0.002	−1.58	0.12
Distance from home to the nearest market	−0.001	0.000	−1.32	0.19
Visits of extension agent	0.042	0.060	0.70	0.49
Information on climate	0.163	0.134	1.21	0.23
Age Dependency Ratio (ADR)	−0.054	0.045	−1.19	0.23

Number of observations = 411, LR, chi2(24) = 147.18, prob > chi² = 0.0000, pseudo R² = 0.0993, log pseudo-likelihood = −667.3371; **, and *** refers to 5%, and 1% significant level. The data originate from field surveys conducted by the author and subsequently processed.

presents the estimated coefficients derived from the ordered probit model. The findings indicate a statistically significant positive coefficient for age (p < 0.01), suggesting its influence on the farmers’ adoption of CSA technology. Additionally, land fragmentation emerged as a pivotal factor in shaping farmers’ decisions regarding the intensive use of CSA technology, with a notably positive coefficient for farm plot number (p < 0.01). Cropland rent from other farmers for crop share is positively and significantly related (p < 0.01) to the intensity of CSA technology, which implies that a household head who rents land is more likely to implement different CSA technologies. A household head with irrigable land had a significant positive contribution (p < 0.01) to the intensity of practicing CSA technology.

The model estimation results revealed that farmers who needed credit but could not access it had a significant negative association (p < 0.05) with the intensity of CSA practices, suggesting that financial constraints limit the number of agricultural technologies they can adopt. The results also indicated that access to radio as a source of information is significant and positively associated (p < 0.05) with the intensity of CSA practice.

3.4.3. Demographic and Social Factors Affecting the Decisions to Use Different CSA Technologies

The results from the multivariate probit model are presented in Table 3 and Table 5. The likelihood ratio test, assessing the null hypothesis that the covariance of the error terms across equations was uncorrelated, yielded a chi-square value of chi²(28) = 178.715 with a probability greater than chi² = 0.0000, indicating significant correlation among the error terms and thereby supporting the use of the multivariate probit model over univariate probit models (Table 4). This suggests that the error terms across the adoption equations were indeed correlated. Additionally, the Wald test demonstrated a high level of model fit, with a Wald chi-square value of chi²(200) = 600.64 and a probability greater than chi² = 0.0000 (Table 5). Consequently, the null hypothesis that all regression coefficients are jointly equal to zero was rejected.

The outcomes reveal notable variations in the influence of explanatory variables on the likelihood of implementation across different types of CSA practices. To facilitate interpretation, Table 5 presents a summary of the primary determinants of multiple CSA adoption, focusing solely on variables that exhibit significance for two or more CSA practices, along with their positive and negative associations. Household heads’ gender positively affected crop diversification, but its influence was negative on the implementation of minimum or zero tillage practice. The age of the household head had a significant positive impact on crop diversification, rainwater harvesting, and agroforestry practices (p < 0.01). This demonstrates that older farmers are more likely to adopt climate-resilient agricultural technologies. As indicated by the results in Table 5, education significantly and positively influenced the use of improved crop variety and organic fertilizer (p < 0.01), rainwater harvesting (p < 0.05), and chemical fertilizers in at least one of their plots (p < 0.1). This may be because social networks and memberships of farmers’ associations affected the decisions to implement some of the CSA practices considered in the study. The variable social capital positively affected the decision to use crop diversification (p < 0.1) and irrigation practices (p < 0.01). Thus, there is a need for knowledge-oriented processes for implementing agricultural technologies.

3.4.4. Household Resource Variables

Livestock size (TLU) was positively associated with the decision probability of agroforestry, the use of chemical fertilizer (p < 0.01), and rainwater harvesting (p < 0.05) (Table 5). Livestock size is frequently used as an indicator of an asset in rural households, and it is one of the important asset bases that facilitate the adoption of farm technology in the study area. Likewise, farm size was positively and significantly correlated with the decision to use the practices of agroforestry (p < 0.1) and inorganic (chemical) fertilizer (p < 0.05). Household head to larger farm sizes is preferred using agroforestry practices and inorganic fertilizer applications.

Having irrigable land had a positive and significant correlation with the decisions to use improved crop variety and crop diversification (p < 0.05), irrigation application (p < 0.01), and minimum or reduced tillage practices (p < 0.1). Therefore, a farm household head who had irrigable land was more likely to use various agricultural technologies on their farm plot through irrigation than others who had not. As expected, a lack of access to credit when a household needed it had a significant and negative impact on decisions to use irrigation and minimum or reduced tillage in the study area (p < 0.01).

3.4.5. Plot Characteristics

As it was hypothesized, the decisions to use CSA technology differ for a different number of farm plots. A household that had a large number of farm plots was more likely to practice rainwater harvesting, minimum or reduced tillage (p < 0.05), and use inorganic (chemical) fertilizers (p < 0.01). According to the information obtained from key informant interviews and FGDs, a fragmented farm plot increases farm diversity (including soil type and the fertility of farm plots), resulting in a diversity of crops grown and production seasons, thereby improving food security. The farm plot with a gentle and moderate slope positively influenced minimum or reduced tillage practice (p < 0.05) and was negatively associated with the decisions to apply rainwater harvesting and inorganic fertilizer. The steep slope of land positively associated with the decision to use improved crop variety (p < 0.1), rainwater harvesting (p < 0.01), and inorganic fertilizer (p < 0.1).

Cropland rent for crop share purposes was linked to improved crop variety (p < 0.05), organic fertilizer and irrigation practices (p < 0.1), agroforestry, rainwater harvesting, and inorganic (chemical) fertilizers (0.01), implying that the household head who shared farmland with others was more likely to intensify CSA technologies. However, it was inappropriate for minimum or reduced tillage practices (p < 0.05). The distance of farmland from home was expected to have an inverse relationship with the probability of the adoption of technology. However, surprisingly, it had a significant positive relationship with the decisions to use improved crop varieties (p < 0.05), crop diversification (p < 0.1), and inorganic fertilizer application (p < 0.01).

3.4.6. Infrastructural and Institutional Factors

The result revealed that as the distance between homes and the extension service increased, so did the likelihood of farmers’ decisions to practice crop diversification and organic fertilizers (p < 0.1). The distance between home and the nearest market makes it challenging to use improved crop varieties, inorganic (chemical) fertilizers, and organic fertilizers (p < 0.05). A household head who has access to media (radio) was more likely to decide to practice agroforestry (p < 0.01) and rainwater harvesting (p < 0.05). Contact with development agents (DAs) is the other vital institutional factor influencing decisions to implement CSA technologies. The result shows that frequent extension contact with farm household heads increased the probability of increasing the practice of minimum or reduced tillage (p < 0.01), but it is inversely associated with inorganic fertilizers (p < 0.01). The result also shows that farmers who received agronomic and climate-related information were more likely to use improved crop varieties and organic fertilizers, but it had a negative relationship with inorganic fertilizers (p < 0.01).

4. Discussion

The purpose of this study was to look into the interdependence of various CSA practices and the factors influencing farmers’ decisions to implement those practices among smallholder farmers in the study area. The findings indicate a robust complementarity among the CSA practices examined in the study. The interrelation among these practices suggests significant policy implications; a policy alteration affecting one practice may trigger spillover effects on the adoption of others.

The most interesting finding was that multiple CSA practices showed a positive correlation, which indicates farmers implemented a combination of CSA practices on a single farm to mitigate multifaceted climate-related risks such as moisture stress, soil erosion, and pests. The results show that strong correlations were found between rainwater harvesting and agroforestry practices. In the study area, the effectiveness of cultivating trees that can be woody perennials integrated with crops is determined by the availability of moisture. As a result, intentional rainwater collection as a reservoir of moisture in storage like broad beds in a catchment and physical instruments like geomembranes may be required. This finding is supported by a study by Teshome et al. [79], which found that agroforestry increases crop productivity while decreasing soil erosion [74], which is an important approach for in situ water conservation practices. Evidence also showed that agroforestry with water harvesting had a more significant positive effect on soil nitrogen stocks than other crop management systems [97], so combining both could synergize.

Another important finding was the strong positive correlation between agroforestry and crop diversification practices. Farmers commonly use crop diversification to increase farm income and the agricultural system’s resilience to weather variability-related problems [58,98]. Agroforestry has been widely used as an alternative income source, while improving the adaptive capacity of the livelihood system to climate change-related risks [70]. As a result of the mutual benefits of increasing crop productivity and soil quality, the two CSA practices demonstrated positive and significant correlations.

In the study area, there was also a strong complementarity between minimum tillage or zero tillage and small-scale irrigation practices. As a result, in small-scale irrigation practices, farmers frequently used perennial crops that do not require frequent plows, allowing crop residue to remain on the ground until the following irrigation season. Empirical evidence showed that farmers who practiced small-scale irrigation could grow crops more than once a year, ensuring increased and stable production, income, and consumption and improving their food security [77,78]. During the field survey, we observed that soil disturbance was reduced, and the plow was avoided on irrigated farmlands, which play a vital role in soil fertility and structure. The most interesting finding was a correlation between improved crop variety and crop diversification. This implies that farmers who have used different improved crop varieties in one crop season could also have used crop diversification on their farm plots [14]. The positive correlation observed among these practices supports the notion that farmers commonly adopt a blend of CSA technologies.

A surprising observation emerged, revealing a positive association between age and the degree of CSA adoption. The positive link between age with the intensity of adopted CSA practices implies that a household head with a higher age was more likely to intensify CSA technology adoption. Key informant interview data backed up those model results, as older farmers used various methods to increase the productivity of their land, implying that older people are more likely to have relatively large farm sizes, stronger ties to agriculture, and more agricultural experience. However, because young people frequently believe they do not have enough land, they frequently leave their farms and engage in off-farm activities. As a result, the technologies mentioned in this study are less likely to be used. Consistent with this result, Atinkut and Mebrat [99] noted that older farmers exhibited higher adoption rates, likely due to their increased familiarity and experience with weather forecasting. Tekelewold et al. [58] revealed that the number of CSA practices adopted increases with the age of the household head.

The degree of land fragmentation also played a crucial role in influencing farmers’ decisions regarding the intensive adoption of CSA technology, as evidenced by a significantly positive coefficient for the number of farm plots. This finding contrasts with studies by Ewunetu et al. [60] and Sileshi et al. [100], which indicated that a high level of land fragmentation discourages the likelihood of adopting SLM technologies. However, Cholo et al. [101] reported a positive impact of land fragmentation, aligning with the results observed here. They attributed this phenomenon to the increased farm diversity associated with land fragmentation, which leads to diversity in crops cultivated and production seasons, ultimately enhancing resilience to climate change risks.

Cropland leasing, or crop share, denotes an arrangement between the property owner (lessor) and the lessee outlining the terms of land usage for farming purposes. The study’s identification of a positive correlation between cropland leasing and the level of CSA technology adoption suggests that household heads who lease land are more inclined to implement various CSA technologies. This finding aligns with research conducted by Leonhardt et al. [102], which highlighted that farm management effectiveness is enhanced by the personal relationships established between renters and landowners. However, Jones et al. [103] convincingly reported that long-term or reliable tenure arrangements create a more conducive environment for adaptation. The result suggests that the tenants were better managed and invested more inputs for farm management practices, as they needed to boost crop productivity and production.

Having irrigable land promotes agricultural intensification due to the strong commitment of different actors, like extension experts and political actors, to the intensification of boosting crop income and productivity. Smallholder farmers who needed credit but could not access it were negatively associated with the intensity of CSA practices, suggesting that financial constraints limit the number of agricultural technologies they can adopt. The result is consistent with Mutyasira et al. [104] and Teklewold et al. [58], who showed that farmers who accessed agricultural credits increased the probability of adopting more than two sustainable agricultural practices. Media access and intensity of CSA practices were shown positive correlations, implying that access to information through media increases the use of more than two CSA practices. This could be because most agricultural technologies are advertised on local FM radio stations.

The decisions to use crop diversification and minimum or zero tillage were gender-sensitive. The result revealed that males were more likely to implement crop diversification and less likely to use a minimum or reduced tillage in at least one of their farm plots. Parallel to this, a study conducted by Deressa and Hassan [87], Beyene et al. [89], and Kassie et al. [105] also found that male-headed households are more likely to apply crop diversification and plant trees. This is probably due to labor constraints and the cultural taboo against women plowing in Ethiopia [105]. Asfaw and Neka [106] show that women in Ethiopia, which is also in the study area, are involved in taking care of their children, preparing food, and other related tasks at home. The negative coefficient on the decision to use minimum tillage was probably because the male household heads were more likely to practice over-cultivation. The evidence from key informant interviews also shows that overcultivation was not practiced on farmland owned by women-headed households in the study area. This is because women take care of their children, prepare food, and complete other related tasks at home [106]. Hence, the previous empirical evidence supported the current result that women’s decision on using land for CSA practices was limited [88].

The positive link of the age of household head on the adoption of crop diversification, rainwater harvesting, and agroforestry practices is related to better farm experiences and land ownership. In favor of farm experience, Asrat and Simane [39] suggested that farmers with long farm experience are well-aware of climate change and the options to adapt to climate change risks. This finding is consistent with that of other research by Amsalu and De Graaff [107] and Deressa and Hassan [87], which found that as a farmer’s age increases, the likelihood of adopting soil and water conservation and changing crop varieties increases.

The educational level of the household head had a positive influence on the decisions to adopt improved crop varieties, organic fertilizers, rainwater harvesting, and inorganic fertilizers in at least one of their plots. This may be because of the need for knowledge-oriented processes for implementing agricultural technologies. The findings indicating a direct relationship between the educational attainment of farmers and the adoption of CSA technology align with earlier research [17,40,47,94], implying that individuals with higher levels of education are better equipped to recognize climate change threats, consider adaptation strategies, and exhibit greater openness to novel ideas and technological advancements.

The unanticipated finding was that family size negatively influences the decision to implement irrigation practices. It was hypothesized that irrigation practices are labor intensive and that family size has a positive effect on irrigation practices as a source of human labor. However, the information obtained from the FGD participants and key informant interviews revealed that irrigation practices are a capital-intensive agricultural activity, but we have no other capital sources other than crop income. Unfortunately, the revenue generated from crop sales was inadequate to adequately provide for the children’s nutritional needs. This outcome echoes the conclusions drawn by Amsalu and De Graaff [107] and Asfaw et al. [17], who observed that larger households incur greater expenses in meeting their basic needs, thereby reducing the likelihood of investing in land-augmented technologies.

The influence of social capital on the decisions to use crop diversification and irrigation practices was positive. This result confirms the previous study by Kassie et al. [91], which reported social network variables to show that farmers who organized in groups were more likely to adopt crop diversification in Ethiopia. The information obtained from key informant interviews shows that irrigation practice is not just an individual farmer’s decision to use irrigation but requires a team decision and participation. Therefore, farmers’ associations and organizations play an important role in irrigation development. The result was also similar to the study by Adela et al. [75], which reported that access to networks through local organizations significantly affected the farmers’ decision to irrigate.

This study’s findings indicate the positive influences of livestock assets (TLU) on the decisions to adopt agroforestry, the applications of inorganic fertilizers, and rainwater harvesting. This implies that livestock size is frequently used as an indicator of an asset in rural households, and this could imply that a household with a more extensive asset base is more likely to use farm technology because they may be able and willing to bear more risk than their counterparts and may have preferential access [105]. As a result, TLU positively impacted the adoption of inorganic fertilizer, rainwater harvesting, and agroforestry, indicating the ability to purchase inputs and materials to cope with the increased risk [58]. This finding was consistent with the findings of Kassie et al. [105], Zeleke and Aberra [108], Balew et al. [94], and Kassie et al. [105], who found that livestock ownership has a significant and positive impact on the adoption of agricultural technologies such as chemical fertilizer.

Farm size was found to positively influence the decision to adopt agroforestry and inorganic fertilizers. The positive result of agroforestry also conforms to the result found by Beyene et al. [109] and Tafere and Nigussie [110], who reported that land size has a significant positive effect on the probability of adopting agroforestry. Household heads to larger farm sizes preferred using agroforestry practices. The positive correlation between large farm size and agroforestry is most likely since agroforestry takes up proportionally more space on small plots. Similarly, Yirga and Hassen [111], Zeleke and Aberra [108], and Pender and Gebremedhin [92] found a positive and significant relationship between farm size and fertilizer use in Ethiopia. This implies that larger farm sizes are more likely to use farm inputs. A possible justification is that a larger farm size increases farm investment by increasing the asset base, making it possible to purchase inorganic fertilizers.

Farm plots with irrigable potential, on the other hand, facilitated decisions to adopt improved crop variety, crop diversification, irrigation application, and minimum or reduced tillage practices. The qualitative sources stated that a household head with irrigable land has a better opportunity to receive training on various agricultural technologies and access to credit services. Therefore, a farm household head who had irrigable land was more likely to use various agricultural technologies on their farm plot through irrigation than others who had not. Credit services were also found to be very instrumental in constraining the practices of irrigation activities. Empirical evidence shows that farmers who have access to credit services are more likely to purchase agricultural inputs [90]. This finding is also in line with that of Deressa and Hassan [87], who discovered that having access to credit enhances the likelihood of choosing conservation agriculture, crop calendar, and irrigation strategies.

A household that had fragmented farmlands was more likely to practice rainwater harvesting, minimum or reduced tillage, and inorganic (chemical) fertilizers. According to the information obtained from key informant interviews and FGDs, a fragmented farm plot increases farm diversity (including soil type and the fertility of farm plots), resulting in a diversity of crops grown and production seasons, thereby improving food security. Studies show that fragmented land with different soil types, slopes, and altitude increases farm diversity, leading to crop diversification and production seasons [101,112].

The farm plot with a gentle and moderate slope positively influenced minimum or reduced tillage practice. This finding is similar to the result of Marenya et al. [113] and Teklewold et al. [58], who confirmed that farmers who perceived a gentle and moderate slope were more likely to use a minimum or reduced tillage. The steep slope of land had positively associated with the decision to use improved crop variety, rainwater harvesting, and inorganic fertilizer. This result confirmed the findings of several previous studies in Ethiopia: Kassie et al. [105], Beyene et al. [109], and Amsalu and De Graaff [107], which reported that adopting land management practices is less likely on flat to moderate slope plots. This implies that households in hilly or rugged land are more likely to adopt adaptive strategies.

Land rent for crop share was linked to improved crop variety, organic fertilizer and irrigation practices, agroforestry, rainwater harvesting, and inorganic (chemical) fertilizers, implying that the household head who shared farmland with others was more likely to intensify CSA technologies. Farmland rent from other farmers had a negative and significant relationship with the decision to use inorganic fertilizers. When renting land to another farmer, renting land to tenants is based primarily on their ability to use technology and their past farm care history. Therefore, this leasing requirement will force tenant farmers to use a variety of technologies to acquire leased land in the future.

Far from the expectations, the distance of farmland from home positively influences the decisions to use improved crop varieties, crop diversification, and inorganic fertilizer application. As we observed in field observation during the data collection, their irrigated farms were far from home, but farmers use different technologies in irrigated agriculture. This implies that the distance of the farmland to the home does not limit the decision to use agricultural technology; instead, the net benefit of the CSA technologies, once they are implemented, would influence the farmer’s decision to adopt CSA technologies. In line with this finding, previous studies reported that the influence of farm income on the decision to adopt different farm management practices was positive and significant [87]. This result is also supported by Kassie et al. [105], who found that the distance of plots from home positively and significantly influences the adoption of conservation tillage and chemical fertilizer. Plots near the homestead may be more fertile than plots further away because homestead plots may benefit from the addition of manure, compost, and other crop residue materials.

The distance of the home to the nearest market negatively influences the decisions to adopt improved crop varieties, inorganic (chemical) fertilizers, and organic fertilizers. Several climate change adaptation studies [58,114,115,116] have discovered a negative relationship between home distance to the nearest market, extension service office, and overall climate change adaptation strategies.

Access to information plays a pivotal role in enhancing climate change adaptation efforts by enabling the transportation of inputs and outputs, acquiring timely market updates, and accessing relevant information regarding products and evolving climatic conditions [58]. In this context, a household head who has access to media (radio) was more likely to decide to practice agroforestry and rainwater harvesting. The outcome suggests that utilizing FM radio for information dissemination correlates with a rise in the adoption of rainwater harvesting and agroforestry practices. The reason seems to be that since the area is drought-prone, rainwater harvesting practices and agroforestry are promoted in the local media.

Another crucial institutional aspect impacting decisions to adopt CSA technologies is the engagement with development agents (DAs). The result shows that frequent extension contact with farm household heads increases the probability of increasing the practice of minimum or reduced tillage, but it is inversely associated with inorganic fertilizers. The positive influence of extension contact on minimum tillage is consistent with the findings of Saguye [117], Moges and Taye [115], Tesfahun and Chawla [118], and Destaw and Fenta [119], who reported that the frequency of extension contact significantly and positively affects the adoption of land management practices. The inverse relationship between extension contact and chemical fertilizer could be attributed to the blanket recommendation against chemical fertilizers [105]. Unreliable moisture is a common problem in this study area, so farmers frequently resist applying chemical fertilizers.

Farmers who received agronomic and climate-related information were more likely to use improved crop varieties and organic fertilizers, but it had a negative relationship with inorganic fertilizers. This result is similar to the findings of Deressa and Hassen [87], Zeleke and Aberra [108], Kassie et al. [91], Marie et al. [57], and Mihiretu et al. [116], who found that access to climate information is a significant adoption determinant of climate change adaptation strategies. This result implies that a household with access to climate-related information is also more likely to have agricultural technology on their farmland.

5. Conclusions of the Study

This study aimed to look at the factors that influence farm household head decisions to implement different climate-smart agriculture (CSA) technologies in the North Wello Zone, taking into account household and plot-level characteristics using multivariate and ordered probit models. Crop diversification, improved crop varieties, inorganic fertilizers, organic fertilizers such as manure and compost, agroforestry, irrigation, minimum or reduced tillage, and rainwater harvesting were all considered. The results show that inorganic fertilizers and improved crop varieties are less common in lowland agroecology, whereas it is more common in highland and midland agroecology. Rainwater harvesting and agroforestry practices appear to be well-suited for overcoming the critical constraints of low rainfall patterns and warmer climatic conditions in lowland agroecology.

Multivariate probit regression (MVP) results show a strong complementarity among CSA practices. The highest complementarity was observed between rainwater harvesting and agroforestry, agroforestry and crop diversifications, minimum tillage and irrigation, crop diversification, and improved crop variety. This implies that the practice of rainwater harvesting may positively influence the practice of agroforestry and vice versa. The factors that exhibit a positive correlation with a higher probability of adopting at least one CSA technology encompass age, level of education, participation in local organizations, farm size, total livestock units (TLU), ownership of irrigable land, number of plots, steepness of slope, distance of farm, and access to local media outlets.

Meanwhile, lack of credit access, family size, the distance of home to extension services, and the nearest market were negatively associated with adopting all CSA technologies. Others, including gender, moderate and gentle slope of the plot, sharing farm plot, and climate-related information, had a mixed effect on the decision to adopt CSA technology. The probability of a household adopting multiple CSA technologies was higher for elderly household heads, a higher number of plots, leasing farmland from others, more significant irrigable land, and radio ownership. Lack of credit access was associated with a decreased likelihood of adopting multiple CSA technologies

6. Policy Implications and Recommendations

The findings of this study have several important implications for policy. First, different climate-smart agricultural land crop management practices are identified in the study area. The most common are improved drought-resistant and high-yielding crop varieties, crop diversifications, organic fertilizers (compost and manure), small-scale irrigations, minimum tillage, agroforestry, and rainwater harvesting practices. However, smallholder farmers only implemented some of these practices due to different biophysical constraints of farm plots, demographic characteristics of household heads, and institutional gap constraints. Therefore, it is crucial to provide effective agricultural extension services, a well-designed irrigation infrastructure, and economically viable crop varieties to increase the farm households’ adoption rates of climate-resilient agricultural technologies. The result also suggests that blanket recommendations for climate-smart agricultural technologies for smallholder farmers can be avoided. Second, the spatial differences in the adoption effects of different CSA technologies at the farm level suggest avoiding blanket recommendations of CSA technologies. For example, improved crop varieties and inorganic fertilizers were less likely in lowland agroecology but high in midland and highland agroecology. Therefore, the careful design of agroecological-based technology diffusion and promotion of climate change adaptation strategies is critically needed. Third, the complementarities between CSA technologies might need further investigation into how such complementarities have co-benefits in improving land productivity, food security, and reducing climate-related risks for smallholder farmers in Ethiopia and anywhere with similar contexts.