The Azuari watershed, located between 10°42′26.399^′′N to 11°9′13.462^′′N latitude and 37°52′6.066^′′E to 38°6′43.817^′′E longitude, spans ~67,523 hectares between Addis Ababa and Bahir Dar. This watershed exemplifies the agroclimatic diversity and varied environmental and farming contexts of the Ethiopian highlands (Hurni et al., 2016). It is representative of policy frameworks supporting sustainable land management and farmer livelihoods in Ethiopia's northwest highlands (Moges et al., 2020). The elevation ranges from 1,270 to 4,043 meters, dividing the watershed into four agroecological zones: low-altitude (Kola), mid-altitude (Woina Dega), high altitude (Dega), and Wirch. These zones reflect a range of climates from warm and dry sub-humid (Kola) to cool, humid, and sub-humid (Woina Dega), temperate and humid (Dega), and humid and cold highlands (Wirch). Thus, the Azuari watershed represents the diverse agroecologies and farming systems of Ethiopia's northwest highlands.

The study area experiences a single rainy season (“Kiremt”) from June to September, with the dry season prevailing from October to March. Annual rainfall ranges from 900 mm in semiarid areas to 2,000 mm in cold and humid areas.

The watershed's major soil types include Eutric Nitosols, Cambisols, Lithosols, Luvisols, and Vertisols (FAO/IIASA/ISRIC/ISS-CAS/JRC, 2012; Fischer et al., 2008). Vertisols dominate the middle part, while Luvisols are prevalent in the upper part (Haregeweyn et al., 2017). In the lower part, dry-subhumid climates feature low-fertility, sandy Regosols (Fischer et al., 2008). Severe soil erosion in croplands, especially in the lower and higher parts, is driven by population pressure, steep slopes, and deforestation (Mequanient and Kebed, 2023). Water-induced soil erosion is the primary cause of soil degradation, impacting crop yields.

The watershed's main land uses include cultivated land, grasslands, shrublands, forests, plantations, and wetlands. Over three decades, significant areas of grasslands, forests, and shrublands have been converted to cultivated and bare land (Yimam, 2022). This conversion increases surface runoff and decreases base flow (Mekonnen and Disse, 2018). Monthly streamflow increased by 2 m3/s in wet months and decreased by 0.7 m3/s in dry months between 1990 and 2004 due to vegetation loss (Yimam, 2022). Forest degradation and agricultural encroachment have led to a 5.81% increase in wet-period streamflow and a 3.34% decline in dry-period flow (Bitew and Kebede, 2024).

Smallholder mixed farming, combining crop cultivation and livestock husbandry, is the predominant agricultural system. This system helps mitigate crop failure risks amid climate variability. Climate extremes, land degradation, and population pressure on limited resources contribute to food insecurity, increasing the sensitivity of crop-livestock mixed production to climate change, especially in lower elevations (Abeje et al., 2019).

This study used a cross-sectional survey, where data were gathered at a particular moment in time. Four kebeles (small administration units), one from the high-altitude (Selam-Abebe), two from the mid-altitude (Laytachimichael and Shegekeraniyo), and one from the low-altitude (YequaraArasema) agroecologies, were randomly selected after taking into account the elevation, farming system, climate condition, and crop pattern over the three agroecological zones (Figure 1). The smallest administrative division in Ethiopia is called a kebele, and inside a kebele is a sub-kebele. Sub-kebeles that implemented terracing practices through project initiatives or local community-based campaigns were given preference for determining the representative sample's target population. Household units can decide on composting, fodder strips for terrace stabilization, and other CSAPs. Since household heads usually make the ultimate decisions on farming techniques and resource utilization, the study looked at issues associated with using households as the unit of analysis over the sub-kebeles. The sample size is computed using Cochran (1977) formula:

Where n is the required sample size, z is the confidence level, p estimated prevalence of farm household attributes in the area, and m- is the margin of error. With a standard value of 1.96 (z) and a 95% confidence level, the study's margin of error (m) is 5% or 0.05. Assume there is a large population but difficult to know the variability in the proportion that could adopt climate-smart intensifications in the crop-livestock mixed farming system, therefore, assume p = 0.2 (maximum variability), n = (1.96)² (0.2(0.8)/(0.05))² = 246 household head. A systematic random sampling technique was used to identify 246 of the 1,230 household heads in the sub-kebeles proportionately to the corresponding agroecology. A household head survey was conducted with 130 respondents from mid-altitude, 59 from low-altitude, and 57 from high-altitude agroecologies.

Data were generated after the respondents had been informed of the objectives of the research. A structured survey questionnaire was used to conduct face-to-face interviews to gather data from the participants. Well-trained four enumerators who are familiar with the local farming system delivered the questionnaire under the direct supervision of the researcher. The procedure of gathering data was conducted through two phases. In the first phase, a pre-test survey was held outside the sample areas to tailor the data collection method and issues to the research area setting. Data were collected on socioeconomic, ecological, and institutional variables, as well as the intensity of CSAPs. Plot level data included the soil fertility status and slope class. From the socioeconomic perspective, factors such as gender, farming experience, dependency ratio-induced labor, literacy level, farmland size, livestock diversity, income, and off-farm activities were taken into account. The institutional variables taken into account were access to credit associations, extension contacts, selected seed availability, distance to the nearest market, and farmland ownership status. Key informant interviews and firsthand observations provided qualitative data that was used to supplement the narrative with a variety of themes to complement the quantitative data.

The main six climate-smart agriculture practices (CSAPs)—crop rotation, crop residue retention, improved grazing, composting with inorganic fertilizers, stabilized terracing through agroforestry, and improved seed varieties—must be put into practice to increase agricultural productivity and sustainability in the Ethiopian highlands (Jirata et al., 2016; Teklu et al., 2022). The combination of these CSAPs with climate information systems is anticipated to reduce yield loss due to climate variability and farm expenses, and increase household income, food security, and resilience (Tesfaye et al., 2021). Direct observations made during the transect walks of the preliminary and main surveys provided evidence to support the primary CSAPs outlined at the national level, indicating that these practices are most common in the research watershed. Therefore, farmers were asked to rate how well their homestead adhered to the identified CSAPs with 0 representing “no” and 1 representing “yes.” The implementation of four to six major CSAPs on a particular homestead is considered intensified. For example, many farmers frequently cultivate improved maize and wheat seeds, compost/manure plus inorganic fertilizers, rotate between cereal and legume crops while retaining residue, and use agroforestry stabilized soil bunds and Fanya juu (Table 1). Summer and autumn are the best times for farmers to implement enhanced grazing systems. The average values (range from 0.17 to 1) of the number of practices implemented at the homestead were used to determine the intensity of CSAPs. The multiple linear regression (MLR) model can incorporate the values of the intensity level of CSAPs (e.g., 0.17, 0.33,…, 1) as long as these values reflect a continuous variable. The intensity values are continuous and provide measured levels of intensity. MLR can be used to model the association between these values of intensity level and other predictor variables, including covariates and categorical. Covariates (continuous) and factors (categorical) were the independent variables. For example, farmers' assessments of soil fertility status and slope steepness class were taken into account based on plot-level parameters. Socioeconomic variables were taken into consideration, such as gender, farming experience, dependency ratio-induced labor, education level, size of farmland holding, animal diversity index, income, and off-farm activities. Institutional variables such as distance to the nearest market, availability of agricultural inputs, insecurity of land tenure, training and advising possibilities through extension contacts, and access to credit associations were included.

Where n = the total number of livestock, Si = proportion of the ith animal in total livestock number. Hence, Si is the individual livestock proportion relative to the total livestock herd in a household. The Herfindahl index (HI) is calculated for LDI as = 1-HIL, where LDI is the livestock diversity index and HIL is the Herfindahl index livestock. The range of HIL values was 0.03–0.98. A value of zero denotes specialization, whereas a number greater than zero implies some degree of diversification.

The number of dependents—individuals who are normally not in the labor force, such as children under 15 and adults over 65—compared to the working-age population—those who are typically between the ages of 15 and 64—implies how the dependency ratio affects the availability of labor. The dependency ratio calculates the proportion of working-age household members (those between the ages of 15 and 64) to dependent household members (those between the ages of 0 and 14 and over 65) to determine the prospective labor supply for the labor cost of CSAPs.

Multiple Linear Regression (MLR) analysis was used to examine the relationship between the independent variables and farmers' decisions to intensify CSAPs. It can manage continuous dependent variables, regardless of the frequency at which a value occurs (Mohr et al., 2021). Verifying that the model assumptions can support the appropriateness of MLR is important in situations where the frequency of each intensity level may have an impact. The linear relationship, the lack of multicollinearity, the independence of errors, the homoscedasticity of the data, and the normality of the residuals were all confirmed before the analysis was conducted (Mohr et al., 2021; Dawson et al., 2021). MLR assumes that each predictor variable and the response variable have a linear relationship. The points in the scatter plots generally formed a straight line, suggesting a linear relationship. MLR assumes that predictor variables don't strongly correlate with one another. For this study, there was no noticeable multicollinearity because all of the VIF values were <2 and tolerance scores were above 0.6. Furthermore, a correlation matrix revealed that all correlation coefficients were <0.3, indicating that none of the predictor variables were substantially associated. The residuals (errors) must be independent for the MLR results to prove valid and robust. The value of the Durbin-Watson test in the model summary indicates 2.12, which lies between 1 and 3 implying that the residuals were independent or uncorrelated. The residuals should have constant variance at every level of the predictor variables, according to the homoscedasticity assumption of the MLR model. The assumption test result shows that homoscedasticity was indicated by the residuals' random distribution around zero. The model's predicted standardized values are plotted against the obtained standardized residuals using a random array of dots on the graph. Results demonstrate that the homoscedasticity assumption has been met since the residual variation is similar as the expected values increase along the X-axis. The model assumes that the residuals have a normal distribution. Plot values (P-P) show that the residuals were normally distributed since the points nearly followed a straight line.

The cause-effect relationship between the intensity of CSAPs with socioeconomic, institutional, and plot level variables was analyzed by a multiple linear regression (MLR) model. The model is a good choice for controlling other covariates, such as household socioeconomic, demographic, and institutional variables (Mohr et al., 2021). The equation of the regression model: y_i= β₀+ β₁x₁+ β₂x₂+…. + β_kx_k+ £, with the mean value of y given as:

where, y is the random response variable and μ_y is the mean value of y, β₀, β₁, β₂, and β_k are the parameters to be estimated based on the sample data, x₁, x2,…,x_k are the predictor variables that are assumed to be non-random or fixed and measured without error, and k is the number of predictor variables, and ε is the random error, which allows each response to deviate from the average value of y. The errors are assumed to be independent, have a mean of zero and a common variance (σ2), and are normally distributed.

The farmer's decision to use intensive CSAPs, such as forage strips or other agroforestry stabilized terracing is influenced by the homestead's steepness. The decision of farmers to use vegetative strips and terracing is significantly influenced by the steepness of the farmland (Alemayehu et al., 2024; Zeleke et al., 2024). Farmers in hilly regions could find vegetative stabilized terracing more appealing since the advantages of the terrace for changing the steep slope gradient often outweigh its expensive and labor-intensive constraints (Hilger et al., 2013).

Campaigns to construct terraces on farmlands with moderate to severe slopes were frequently started by the local community during the dry season, which lasted from January to March. However, the improperly designed campaign-based terrace, which was constructed without considering the plot owner's practical experience, did not meet the requirements and was not implemented sustainably in the cultivated field (Debie et al., 2019; Mersa et al., 2023). It was observed that terrace structures were frequently modified with plot tillers by incorporating native drainage channels and modifying slope gradients.

The result indicated that the homestead with moderate to severe slopes has maintained vegetative stabilized Fanya Juu and soil bunds because farmers are accustomed to their many advantages, which include controlling soil, water, nutrient, and crop losses; producing fodder for sheep and goats; and improving soil nutrients and crop output. Farmers were able to determine the impact of terracing on the reduction of slope gradient by observing the accumulation of deposits in the lower portion of croplands located between terraces. Natural and planted vegetative strips are effective in reducing the length of moderate and steep slopes and act as barriers to slow down runoff, thereby reducing soil erosion (Haddaway et al., 2018). The authors further noted that vegetative strips make them a popular choice among farmers due to the request being easier and less costly to implement compared to terracing. To optimize the potential for reducing soil erosion and increasing productivity, terraces should be utilized in conjunction with perennial vegetation, moisture retention, and a continuous presence of soil cover during the rainy season (Debie et al., 2019; Desta et al., 2021). Numerous ecological services are provided by stabilized terraces, such as higher soil moisture content, reduced runoff, and silt, pH, availability P, available K, and organic carbon, and enhanced grain yields (Debie et al., 2019; Tolesa et al., 2021; Deng et al., 2021). Terrace stabilizers including shrubs, annual herbaceous plants (like legumes), and trees like Faidherbia albida are examples of nitrogen-fixing plants that are essential to most CSAPs.

On the moderate to steep homestead, legume shrubs that are planted along the homestead's boundary and terrace segment are primarily pruned when they reach a productive age to provide fodder. Farmers stated that elephant grass is preferable to feeding cattle and that the best plants to feed sheep and goats are Rahmnu sprinoides, Sesbania sesban, and Acacia decurrens. One of the most widely used shrubs for fodder in Ethiopia; sesbania sesban improves the intake and digestibility of the basic diet for sheep and goats (Oosting et al., 2011; Debie, 2022). Productive natural grass developed on strips, terrace segments, homestead boundaries, and streams is commonly utilized for feeding cattle, especially oxen and cows, through the cut-and-carry grazing technique. Shrubs and trees must provide shade to shield animals from midday sunlight. Beekeeping productivity also benefits from the growth of flowering bushes along the homestead boundary and terrace segment. Shrub plants on the terrace segment may need to have deeply ingrained roots to stabilize the terrace and supply nutrients to the soil via root nodules. Farmers have observed that lowering the slope gradient significantly lowers crop damage during the early stages of growth. Consequently, farmers have developed an understanding of the detrimental effects of slope steepness on agricultural productivity, enabling them to sustain stabilized terraces over an extended period. Farmers choose to spend a significant amount of labor in upholding the practices and guarding them against free grazing because of the numerous advantages of doing so. Compost and livestock dung are applied by farmers to farmland areas situated in between terraces stabilized by grass or shrubs.

The results indicate that the husbandry of various livestock kinds at farm households leads to an increase in the amount of manure applied, which is made up of the dung of various livestock types, including horses, chickens, sheep, goats, cattle, and donkeys, to improve the fertility of their farm soil. In association with this, total livestock holding size was one of the major determinants of the adoption of CSAPs (Alemayehu et al., 2024; Zeleke et al., 2024). Cattle dung is regularly utilized to make dung cake, which is used as fuel for energy during the autumn, winter, and spring seasons. The dung of chickens, horses, donkeys, sheep, and goats is not used as fuel energy. To maximize the use of manure to improve soil fertility and be able to replace cattle dung used for fuel energy, it is essential to herd chickens, horses, donkeys, sheep, and goats. The best-integrated degree of decomposition and productivity benefits of crop grain and biomass yield were seen in co-composting or mixed use of cow, pig, and wheat straw (Fan et al., 2023). Various farming methods employ a blend of manure from pigs, cattle, and poultry (Rayne and Aula, 2020). Thompson et al. (2023) state that well-kept livestock is vital for agricultural systems and can yield significant benefits. More carbon sequestration is encouraged by manure soils (Washaya and Washaya, 2023).

According to the findings, a lower dependency ratio denotes fewer dependents per person of working age, which increases labor availability and may motivate the household to intensify CSAPs on homesteads and other farmlands to enhance livestock and crop yields. A low dependency ratio boosts labor availability, which can help with enhanced cattle husbandry practices and terrace stabilization through agroforestry and composting on a particular homestead (Debie, 2024). Having more people in the working age range in the household can help manage different CSAPs in an efficient and long-lasting way (Sharma et al., 2024). Household labor is required at large levels for CSAPs, such as manure application, composting, distribution of crop straw across the entire homestead, improved grazing management, and agroforestry management. Farmers stated that the labor cost of spreading the collected manure of various livestock on the land requires less labor cost than that of composting. During the dry season, women and young people were mostly responsible for distributing the collected manure at the homestead. Famers reported that carrying manure to scatter on a homestead needs less work compared to transporting manure to distant agricultural areas using manpower and donkeys. As per the training provided by the local administration unit's development agents, farm households with a low labor dependency ratio can devote more time to making compost from green manure, ash, and livestock manure. During the end of winter and throughout the spring, farmers can distribute manure or compost and mix it in the soil by plowing. Despite the increased labor costs associated with its production, farmers stated that compost overtakes manure in terms of improving soil fertility and crop yields for the forthcoming year. The application of compost generally had a significant and positive impact on plant nutrients, such as soil pH, EC, SOM, total N, and accessible P (Kavvadias et al., 2023). For many reasons, including low quality, inconsistent supply, and difficulties with manure collection caused by inadequate facilities and great distances between farms, manure must be properly dealt with before it can be used (Washaya and Washaya, 2023). Activators can be added to the basic constituents of compost to increase its nutritional value. Adding fungicides, viricides, nematode inhibitors, and antibacterial treatments for plants or organic materials can also improve the quality of compost (Ayilara et al., 2020). Generally, manure or compost is used to cultivate cereal crops (primarily sorghum and maize), followed by legume crops. Rotating crops between legumes and grains is a fundamental practice for managing soil.

The finding revealed that agricultural extension advice and training services had a significant impact on farmer awareness development of many advantages and integrated use of CSAPs on homesteads and other farmland. Increasing public awareness of the advantages, practical procedures, and technical know-how of CSAPs can be essential to increasing the rate of adoption. The more intensive CSAPs, including terracing, composting, improved grazing systems, and others were influenced by frequent extension contacts. The most important aspect influencing the adoption of CSAPs was having access to extension contacts (Diro et al., 2022; Xu et al., 2022; Bwalya et al., 2023 Debie and Ayele, 2023; Ferrer et al., 2023; Tadesse and Ahmed, 2023; Hussein, 2024; Ma and Rahut, 2024). Due to a lack of training and access to agricultural knowledge, farmers were found to be ill-prepared to adopt newly introduced CSAPs (Nyairo et al., 2022). Farmers stated that agricultural extension workers helped them with the functional and technical components of several CSAPs. Agricultural extension workers can also be knowledgeable on the use of improved seed, careful terracing management, making good use of terrace stabilizers for livestock fodder, beehive and livestock husbandry, and compost preparation. Farmers are familiar with the indigenous technique of legume-cereal crop rotation or intercropping. Thus, there will be more dedicated farmers who can fill their technical and knowledge gaps about the need to apply CSAPs to fulfill multiple objectives by having access to possibilities for training and advice. Farmers who have received relevant training and skills in CSAPs should be the focus of the extension program (Kifle et al., 2022).

The findings of the study showed that farmer experience across time also considerably influences the diversity, intensity, and effective management of CSAPs. Geda et al. (2024) found that the adoption of CSAPs in wheat production was influenced by farming experience. By implementing crucial technologies early on in the context of their cultivated area, farmers can develop competence in managing projected production hazards (Ainembabazi and Mugisha, 2014). Based on their local knowledge and experience, farmers commonly intercrop modified maize on their homestead with sunflowers, faba beans, peas, and common beans. The purposes of rotation or legume-cereal intercropping are widely known to farmers. The primary perceived goals of the technique are to boost crop productivity and soil fertility. Legumes and maize intercropped improved soil fertility, total nitrogen content, and grain and biomass yield of cereal crops (Gidey et al., 2024). Cereals (maize, wheat, sorghum, and teff) interplanted with oil seeds Guizotia abyssinica (named Noug locally), Ethiopian mustard, also known as gomenzer (Brassica carinata A. Braun), and linseed (flaxseed) on homestead boundary strips that touch on pathways used by people and livestock or trees (Eucalyptus globulus). Legumes and oil seed straws or wastes are utilized on the farm to mulch the soil's surface and young plants, such as Eucalyptus globulus. Straws from maize and sorghum are mostly utilized as cattle feed and fuel. The soils can be combined with a small amount of residual maize and sorghum. For the first plowing in the dry season, the collected cereals, legumes, and oil seed straws can be incorporated into the soil. Frequent plowing during the dry and spring seasons allows the organic biomass of the residue to break down and provide the soil with nutrients for the next crop's productivity. For cattle hay fed in the spring and summer, teff straw is primarily utilized. Based on their experience, the majority of farmers employ a cut-and-carry technique in the autumn and summer to improve cow grazing strategy since these seasons may require less labor than other seasons to harvest valuable feed for cattle, such as grass and shrub leaves. Weeds that grow beneath the biomass of maize crops can be harvested for cattle fodder. Post-harvest practices frequently involve establishing agroforestry and limiting unrestricted grazing on terraced homesteads. During the post-harvest time, pruning of Acacia decurrens, Sesbania sesban, and elephant grass is done on the farm to provide feed for cattle, sheep, and goats. Therefore, a combination of farmer knowledge, experience, and abilities, as well as expert scientific observation are required to properly manage multifunctional CSAPs and sustainable production of mixed agricultural systems (Debie, 2024).