In the face of changing climatic conditions and climate change, climate-smart agriculture (CSA) is an approach to transform and reorient agricultural systems to ensure food security ( Chavula, 2021). Climate change disrupts food markets, putting food supply and production at risk. Strengthening farmers’ adaptive capacity and enhancing the mitigation potential and efficiency of agricultural production systems can mitigate these risks. Smallholder farmers who are aware of climate change or perceive it as a reality are more likely to adopt climate-smart agricultural practices. These farmers aim to achieve the three pillars of CSA: improving household income and productivity, enhancing resilience and adaptation, and reducing greenhouse gas emissions. The adoption of climate-smart agriculture to attain these principles is influenced by institutional, cognitive, and socioeconomic factors ( Figure 1).

2.1.1 Location

The research was carried out in Nyimba district of Eastern Province Zambia. The district is situated 334 kilometers East of Lusaka Zambia’s national capital. In the South the district borders with Mozambique, North with Muchinga province, West with Lusaka province, and East with Petauke district. The district lies between latitude (13°30′1019″ and 14°55′81426″ South) and longitude (30° 48′5047″ and 31°48′20252′′East) ( Figure 2).

2.1.2 Climate, soil and topography

Zambia is divided into three agro-ecological zones: Zone I, Zone II (subdivided into IIa and IIb), and Zone III. Nyimba district falls within Zone I, which covers the Southern and Eastern rift valleys of the Zambezi and Luangwa River basins. This zone also extends to parts of the Western and Southern provinces in the south of Zambia. The average annual rainfall in Nyimba district ranges from 600 to 900 mm, with the wettest months being December to February and a distinct dry season from May to November. The annual mean temperature is 24.2°C, while the daily temperature range is from 10.3°C to 36.5°C ( Figure 3). Topographically, the district comprises hills and plateaus, with soils characterized as Lithosol-Cambisols, while in the valleys, the soils are classified as Fluvisol-Vertisols. The elevation varies from 450-1000 m at the bottom of the Luangwa River valley, extending to the plateau near the Nyimba district center and reaching even higher altitudes on the mountain tops in the western part of the district.

2.1.3 Vegetation type

The Miombo woodland is the most dominant formation and habitat type in Southern Africa ( Gumbo and Dumas-Johansen, 2021; Montfort et al., 2021). Miombo woodland is also the major forest type in Zambia itself, covering approximately 45% of the entire land surface ( Kalinda, 2008). Nyimba is located in the middle of the Miombo Ecoregion, a biome with a variety of flora types that is dominated by tree species from the Caesalpinioiae subfamily of leguminous plants ( Timberlake et al., 2010). Depending on the climate, soil, landscape position, and degree of disturbance, the ecoregion’s vegetation varies in composition and structure ( Timberlake et al., 2010; Halperin et al., 2016). Nyimba is located in the arid ecozone and is characterised by four types of plants: Dry miombo woodland (i.e., Brachystegia spiciformis, Brachystegia boehmii and Julbernardia globiflora), Mopane woodland ( i.e., Colophospermum mopane), Munga woodland ( i.e., Vechellia sp., Senegalia sp., Combretum sp., and trees associated with the Papilionoideae subfamily) and Riparian Forest ( i.e., mixed tree species).

2.1.4 Land use and farming systems

Nyimba district total land area is about 10,500 square-kilometers according to population and housing census of 2010. Therefore, 82% of the district population is agrarian with average household income. These households are farmers who are into mixed agriculture practices dominating the agricultural scene in the district. However, local smallholder farmers in the district practice some sort of shifting cultivation. Under this agricultural system, crops are grown in mounds and/or ridges in most cases maize. The major crops grown in include banana ( Musa sp.), maize ( Zea mays), finger millet ( Eleusine coracana), groundnuts ( Arachis hypogaea), haricot bean (Phaseolus vulgaris), cowpeas ( Vigna unguiculata spp.) and soybean ( Glycine max). Multiple cropping system is common among the farming households were cultivated land is on gently and moderately on steep slopes. The topography of the land in the district makes the agricultural or cultivation pattern different from other areas. Therein, cropping system is alongside livestock production such as cattle, goat, chickens, ducks and doves. Besides agricultural activities, farmers are engaged in charcoal production, timber, firewood supply and non-timber forest products (NTFPs) from the miombo woodland for household economic gain ( Policy Brief, 2016).

Prior to the actual data collection, an exploratory study was conducted to gather essential information about the research area. The information gathered included the following: the distance between villages, the number of farming households per village, contact details for lead farmers, households adopting climate-smart agricultural (CSA) practices, the location of croplands, and the identification of central meeting points for focus group discussions (FGDs).

2.2.1 Sampling technique

This study utilized a multi-stage random sampling approach to recruit participants from smallholder farming communities in Zambia. These farming communities, known as agricultural camps, are designated by the Ministry of Agriculture of the Republic of Zambia. The camps group smallholder farmers’ residences within a district, facilitating easy access to agricultural extension services. From the eight agricultural camps in Nyimba District, four camps – Ndake, Central Camp, Lwende, and Ofumaya – were randomly selected for the study. These four camps collectively house 10,700 farmers. To determine an appropriate sample size, the study employed Slovin’s formula. Additionally, three villages from the four selected camps were randomly chosen for data collection. These villages were Sikwenda, Sichipale, Mawanda, Elina, Katumbila, Sichalika, Malalo, Mwenecisango, Mulivi, Lengwe, Mofu, and Yona. With a margin of error set at 0.05, the initial calculated sample size was 386 participants. However, to optimize resources, the researchers increased the margin of error to 0.1, resulting in a smaller sample size of 99 participants. Achieving this reduced sample size required additional time and financial investments, but it was a trade-off deemed necessary for the successful completion of the study.

Sample size formula: Slovin’s (1960) formula. n = N 1 + N e 2 n = 10700 / ( 1 + 10700 ( 0.1 2 ) n = 10700 27.75 n = 99.07

Ultimately, the researchers opted for a compromise, settling on a sample size of 194 participants, which fell between the initial calculations of 99 (with a 0.1 margin of error) and 386 (with a 0.05 margin of error). With the assistance of agricultural camp officers, farmer registers from each selected village were utilized to randomly identify potential participants using an Excel spreadsheet.

2.2.2 Focus group discussion

In-depth information regarding the factors influencing the adoption of climate-smart agricultural practices, crop productivity, perceptions of these practices, the use of CSAPs, and perspectives of climate change was gathered through focused group discussions (FGD). An open-ended FGD study tool that was designed was used to achieve this. Due to the fact that delicate topics are revealed during the execution, focus groups are thought to be superior than one-on-one interviews. Individual farmers frequently find it difficult to completely express themselves in one-on-one interviews. Four (4) focus group discussions (FGDs) involving village headmen, women, men, and youth were conducted within the study region. The FGD meetings were arranged in convenient locations for individual farmers to attend. FGDs were mostly used to augment data that was gathered through household questionnaires.

2.3.3 Household interviews

There were both closed-ended and open-ended questions on the survey. The questionnaire was pre-tested six times for appropriateness (e.g., clarity, adequacy, and question sequence) prior to conducting the household interviews. The questionnaire was then modified based on the results. The questions were pre-tested on smallholder farmers’ homes who did not take part in the study.

For data collection, the study employed three enumerators who underwent training and supervision from the principal researcher, an experienced professional in the field. The enumerators meticulously verified and corrected the gathered data after each day of fieldwork. Subsequently, the validated information was securely backed up on the CSPRO Cloud platform ( https://www.csprousers.org/help/CSDeploy/deployment_options.html).

Outcome variables

The outcome variable for this study is the impact of adopting climate-smart agriculture practices on crop productivity among smallholder farmers’ households. Additionally, the study examines the factors influencing both crop productivity and the adoption of climate-smart agriculture practices (CSAPs) among smallholder farmers in the Nyimba district.

Dependent variables

Smallholder farmers’ household decision to adopt CSAPs

The dependent variable was whether a smallholder farmer’s household adopted Climate-Smart Agricultural Practices (CSAPs) or not, taking a value of 1 if the household adopted CSAPs and 0 if it did not. The main objective was to identify the factors influencing the adoption of CSAPs among smallholder farmers’ households in the Nyimba district of Zambia.

To evaluate the impact of Climate-Smart Agricultural Practices (CSAPs) on crop productivity, this study employed the Propensity Score Matching (PSM) method, comparing adopters and non-adopters. PSM is a statistical technique that adjusts for confounding variables across a sample population, enabling more accurate estimation of treatment effects. As outlined by Caliendo and Kopeinig (2008), the implementation of PSM involves several key steps: first, estimating propensity scores using a binary model; second, selecting an appropriate matching algorithm; third, verifying the common support condition; and fourth, assessing the quality of the matching between the treatment and control groups. The following are the steps involved in PSM:

Step 1: Model Specification

The logistic model was chosen for this study due to the robustness of its parameter estimations, which stems from the assumption that the error term in the equation follows a logistic distribution ( Baker and Melino, 2000; Ravallion, 2007). Therefore, the Logit model was used to estimate the probability of smallholder farmers’ adoption of CSAPs allotted to socio-economic, agro-ecological and institutional characteristics. Therein, a dependent variable considered a value of 1 for CSAPs adoption and 0 for non-CSAPs adopters. P i = P ( Y = 1 | X ) (1)

In line with Pindyck and Rubinfeld (1981), the cumulative logistic probability function is specified as follows; P i = F ( Z i ) = F [ a + ∑ i = 1 m β i X i ] = [ 1 1 + e − ( a + ∑ β i X i ) ] (2) where e represents the base of natural logs, X _i represents the i ^th explanatory variable, P _i the probability that a household adopted CSAPs, α and β _i are parameters to be estimated.

Expressing the logistic model in terms of odds and log-odds aids in interpreting the coefficients ( Gujarati, 1995). The odds ratio quantifies the relative probability of an individual participating ( P _i) versus not participating (1- P _i). The probability of non-participation can be calculated as: ( 1 − P i ) = 1 1 + e zi (3) ( P i 1 + P i ) = [ 1 + e zi 1 + e − zi ] = e zi (4)

Alternatively, ( P i 1 + P i ) = [ 1 + e zi 1 + e − zi ] = e [ a + ∑ B i X i ] (5)

Taking the natural logarithms of equation (3.5) will give the logit model as indicated below. Z i = ln ( P i 1 − P i ) = a + B 1 X 1 i + B 2 X 2 i + … B m X mi (6)

If consider a disturbance term, μ _i, the logit model becomes Z i = a + ∑ t = 1 m B t X ti + μ i

So, the binary logit will become: Pr ( pp ) = f ( X ) (7)

Where pp is CSAPs adoption, f( X) is the dependent variable project participation and X is a vector of observable covariates of the households. The dependent variable will take a value of 1 for CSAPs adoption and 0 for non-adopters.

In addition to the estimated coefficients, the marginal effects of the change in the explanatory variables on the probability of CSAPs adoption are also estimated. The interpretation of these marginal values will be dependent on the unit of the measurement for explanatory variables.

When an explanatory variable is binary, the marginal effects provide a reasonable estimate of the change in the likelihood of the outcome variable (Y = 1) evaluated at representative values of the regressors, such as their means.

Step 2: Identifying the Common Support Region and Conducting Balancing Tests

The region where the propensity score distributions of the treatment and comparison groups overlap is known as the area of common support. This needs to be identified. However, if there is a systematic difference in the observed characteristics between the dropped non-adopters of CSAPs and the retained non-adopters, sampling bias may still occur. These differences should be closely monitored to aid in the interpretation of the treatment effect. Furthermore, balancing tests can be conducted to check if the mean of the covariates (X) and the average propensity score are equal within each quantile of the propensity score distribution.

For PSM to be effective, the treatment and comparison groups must be well-balanced based on their observed covariates (X), as similar propensity scores are derived from similar observed characteristics. Achieving balance necessitates that the covariate distributions of the treated and untreated units are indistinguishable. Formally, this requires verifying that the conditional distributions P(X|T=1) and P(X|T=0) are equivalent. Here’s a version with reduced similarity:

Step 3: Matching adopters to non-adopters

Selecting an available data matching algorithm is the third stage. Selecting control subjects who are matched to treated subjects based on context factors that the investigator feels should be tracked is a standard process known as matching. A different one might use comparable criteria. according to propensity score, divide adopters against non-adopters. Kernel-based matching (KBM), radius matching (RM), and closest neighbour matching (NN) are the most often used matching algorithms.

Step 4: Matching quality

Matching quality tests could be conducted in the fourth step. Whether or whether the matching method can balance the distribution of different variables is determined by checking for matching, irrespective of quality. If there are discrepancies, it could be a sign of insufficient matching, and corrective action is advised ( Caliendo & Kopeinig, 2008). The next action is to determine if the treatment caused a difference in the impact indicators.

The difference between the mean outcome of matched adopters and nonadopters with common support conditional at the propensity score provides the average treatment effect at the treated (ATT).

Step 5: Sensitivity analysis

Lastly, to verify the strength of the conditional independence assumption, a sensitivity analysis will be performed. Sensitivity analysis will also be used to examine whether the influence of an unmeasured variable on the decision-making process is significant enough to compromise the matching strategy ( Ali & Abdulai, 2010). The sensitivity analysis (r-bounds test) will be performed using the Rosenbaum bound sensitivity test.

The study was conducted by the researcher and two supervisors, adhering to principles of integrity, objectivity, openness, respect for research participants, respect for intellectual property, confidentiality, informed consent, fidelity, and honesty. The researcher and supervisors take full responsibility for their actions and publications, ensuring that all agreements are intended to be upheld. The study was approved by the University of Zambia Directorate of Research and Graduate Studies (NASREC) on March 18, 2024, and is registered under NASREC IRB No. 00005465 (IORG No. 0005376). The principal researcher provided a written consent form for research participants to participate in the household survey, conforming to the research ethics guidelines of the University of Zambia and Haramaya University.

Characteristics of the participant smallholder farmers

The household survey consisted of 194 randomly selected smallholder farmers from the research area. These farmers were interviewed about their crop production and the application of various Climate-Smart Agriculture (CSA) practices. The study presents the findings of the household survey, beginning with the demographic characteristics of the participants ( Table 1), followed by sections on crop production and productivity, adoption of CSA practices, constraints on the adoption of CSA practices, effects of CSA practices on crop productivity, and factors affecting crop productivity. A total of 339 field plots of various crops were surveyed from the 194 farmer participants. Table 2 provides detailed demographic and socio-economic data on the respondents. The mean age was 46 years (standard deviation: 14.59), and the majority of households (62.18%) were male-headed, with 69.43% of respondents being married. The mean years of formal education were 5.49 (standard deviation: 3.5), and the mean years of farming experience were 26.22 (standard deviation: 15.55). Respondents had lived in the area for an average of 30.92 years (standard deviation: 18.68). The average family size was 5.42 (standard deviation: 2.14). The average total annual income was K5,472.68 (USD 331.68, at an exchange rate of K16.5 per USD), with a standard deviation of 7,626.52. Additionally, 57.51% of respondents reported participating in off-farm activities. Regarding agricultural practices, 78.76% used improved seed varieties, and the average farm size was 3.396 hectares (standard deviation: 3.363). All land was under customary tenure (100%), with the mean cultivated land being 1.83 hectares (standard deviation: 1.45). On average, smallholder farmers grew two different crops, with a standard deviation of 0.930.

Crops grown by smallholder farmers

Regarding the crops that the farmers grew, the study discovered that the most common crop was maize, which was recorded in 194 crop plots. Groundnuts were reported in 99 plots, sunflower in 69 plots, and soy beans in 16 plots. ( Table 2). It was said that the other crops—cowpea, sweet potatoes, millet, cotton, and bambara nuts—were produced in small plots.

Climate-smart agriculture practices adopted by smallholder farmers

The study’s findings shed light on the adoption of various conservation agriculture techniques across the surveyed field plots. Among these practices, pot-holing (basin) emerged as the most widely implemented method, with 61 plots (17.99%) employing this technique. Closely following was multi-cropping, which was practiced on 50 plots, accounting for 14.75% of the total. Minimum tillage, a soil conservation approach, was utilized on 34 plots, representing 10.03% of the survey sample. The ripping technique, which involves creating furrows in the soil, was observed on 32 plots (9.44%). Furthermore, 18 plots (5.31%) incorporated crop rotation as a means of maintaining soil fertility and controlling pests and diseases. The application of manure, a natural fertilizer, was recorded on 11 plots (3.24%), while alley cropping, a system that combines crops with trees or shrubs, was adopted on 9 plots (2.65%) ( Table 3). It is noteworthy that the remaining conservation agriculture techniques were tested on fewer than 10 plots each, indicating their relatively limited implementation within the surveyed area.

Number of climate smart agriculture practices adopted by smallholder farmers

The study’s findings, as presented in Table 4, shed light on the extent of conservation agriculture (CSA) practice adoption among the surveyed plots. Notably, a substantial number of plots, 167 (49.26%), did not incorporate any CSA techniques whatsoever. This highlights a significant gap in the implementation of sustainable agricultural practices within the surveyed area. On the other hand, a considerable portion of the plots, 123 (36.28%), had adopted at least one CSA practice, indicating a positive step towards embracing more environmentally friendly farming methods. Additionally, 4 plots (12.68%) had implemented two different CSA practices simultaneously, demonstrating a more comprehensive approach to sustainable agriculture. Interestingly, the data revealed that a smaller number of plots had adopted multiple CSA techniques concurrently. Specifically, four plots had incorporated three distinct CSA practices, while one plot had implemented an impressive four different CSA practices. Furthermore, another plot stood out by adopting a remarkable five separate CSA techniques. Despite these instances of multiple CSA practice adoption, the overall findings suggest that the majority of farmers within the surveyed area were either not implementing any CSA techniques or had adopted only a single practice. This observation underscores the potential for further education and awareness-raising efforts to encourage the widespread adoption of multiple sustainable agricultural practices, ultimately contributing to improved soil health, crop productivity, and environmental conservation.

Quantities harvested for various crops (kg)

The survey data presented in Tables 5, 6 offers valuable insights into the crop cultivation patterns and yield outcomes among the surveyed farming community. Notably, the crops that emerged as the most prevalent choices among farmers were soybeans, maize (corn), groundnuts, and sunflowers. Across all crop types, the average harvest weight recorded was 1223.51 kg, accompanied by a substantial standard deviation of 1442.82. This variation in yields highlights the diverse factors that can influence agricultural productivity, including soil conditions, farming practices, environmental variables, and access to resources. When examining the individual crop yields, maize stood out as a prominent crop, with an impressive average harvest weight of 1766.57 kg. However, the high standard deviation of 1594.23 suggests significant variations in maize yields among farmers, potentially attributable to differences in cultivation techniques, seed quality, or localized environmental conditions. Groundnuts, a staple crop in the region, exhibited an average harvest weight of 511.08 kg, with a standard deviation of 605.07. This relatively lower yield, coupled with the substantial variation, may indicate challenges faced by farmers in optimizing groundnut production, such as pest or disease pressures, or limitations in access to appropriate inputs and knowledge. Sunflowers, a valuable oilseed crop, yielded an average harvest weight of 609.67 kg, with a standard deviation of 513.02. While the average yield appears moderate, the substantial variation observed could be attributed to factors like soil fertility, water availability, or sunflower variety selection. Notably, soybeans emerged as a crop with significant yield potential, boasting an average harvest weight of 1007.5 kg. However, the remarkably high standard deviation of 1835.615 points to substantial disparities in soybean yields among individual farmers. This variability may stem from differences in cultivation practices, access to quality seeds, or the adoption of specific agricultural techniques tailored for soybean production. These findings not only underscore the crop preferences of the surveyed farmers but also highlight the need for targeted interventions and support measures to address the observed yield variations. By identifying and addressing the underlying factors contributing to these disparities, efforts can be made to enhance agricultural productivity, promote sustainable farming practices, and ultimately improve the livelihoods of smallholder farmers in the region.

Productivity of various crops (Yield (kg) per hectare)

The study’s findings shed light on the crop productivity levels observed among the surveyed farming communities. Considering the overall yield across all crop types, the average yield per hectare stood at an impressive 1316.60 kg, although the substantial standard deviation of 1214.13 suggests significant variations in individual yields ( Table 7). Focusing specifically on maize, a crucial staple crop, the mean yield per hectare was calculated to be 1682.52 kg. However, the high standard deviation of 1325.87 indicates substantial disparities in maize yields among farmers, potentially influenced by factors such as soil fertility, farming practices, and environmental conditions. Groundnuts, another important crop in the region, exhibited an average yield of 822.90 kg per hectare, accompanied by a standard deviation of 547.88. This variation in yields could be attributed to challenges faced by farmers in optimizing groundnut production, including pest or disease pressures, access to quality inputs, or knowledge gaps. Sunflower cultivation yielded an average of 962.79 kg per hectare, with a relatively lower standard deviation of 437.38, suggesting more consistent yields among farmers. This could be a result of better-adapted cultivation practices or more uniform environmental conditions suitable for sunflower growth. Soybeans, a valuable crop with growing demand, recorded an average yield of 808.40 kg per hectare. The standard deviation of 426.74 indicates moderate variations in soybean yields, which could be influenced by factors such as seed quality, planting techniques, or soil management practices. These findings not only provide insights into the productivity levels of various crops but also highlight the need for targeted interventions and support measures to address the observed yield variations. By identifying and addressing the underlying factors contributing to these disparities, efforts can be made to enhance agricultural productivity, promote sustainable farming practices, and ultimately improve the livelihoods of smallholder farmers in the region.

Impact of climate-smart practices on crop productivity among smallholder farmers

The study examined the impact of Climate-Smart Agriculture (CSA) techniques on the crop yields of smallholder farmers. It was found that farmers who adopted CSA practices experienced a 20.20% higher crop yield compared to those who did not adopt these practices ( Table 8). The difference was statistically significant, with a p-value of 0.027 ( p < 0.05). These results suggest that the adoption of CSA practices leads to increased crop yields.

Impact of climate-smart practices on maize productivity among smallholder farmers

The study utilized propensity score matching analysis to specifically assess the impact of CSA on maize productivity ( Table 9). The findings indicated that CSA adoption led to a 21.50% increase in maize yield compared to non-adoption. This significant increase in maize yield, with a p-value of 0.035 ( p < 0.05), demonstrates the positive effect of CSA practices on maize productivity.

Factors affecting smallholder farmers’ adoption of climate-smart agricultural

The study employed logistic regression analysis to examine the factors influencing the adoption of Climate-Smart Agricultural (CSA) practices among farmers. The findings revealed that age played a pivotal role, with older farmers being more inclined to embrace sustainable agricultural methods. Specifically, the results indicated a statistically significant positive relationship between age and the adoption of CSA practices, with a p-value of 0.0000 ( p < 0.001). Farmers within the age groups of 40-55 years and above 55 years were found to have a higher rate of adopting CSA practices in the study area. Interestingly, the research uncovered an inverse correlation between prior farming experience and the willingness to implement climate-smart farming techniques. Contrary to expectations, farmers with more years of experience in the agricultural sector were less likely to adopt CSA practices, a finding that was statistically significant at a p-value of 0.0000 ( p < 0.001). Income level was also identified as a significant factor, with a positive relationship between higher income and the adoption of CSA practices. This relationship was statistically significant at a p-value of 0.0640 ( p < 0.1) ( Table 10). This finding aligns with Zakaria et al. (2020), who demonstrated that factors such as experience in rice cultivation, access to media and training, and perceived decrease in rainfall quantity positively influenced the adoption of climate-smart agricultural technologies. Conversely, larger farm sizes, greater distances between farmers’ homes and farm sites, location, and reported temperature rises were found to hinder the adoption of these technologies. While the present study highlighted age and income as facilitators of CSA adoption, it contrasted with previous research by Saha et al. (2019), which identified factors like education, occupation, family size, farm size, climate change adaptation techniques, cattle ownership, market accessibility, information access, training, organizational affiliations, and climate change perceptions as influential determinants. This suggests that while higher income levels facilitate the adoption of CSA practices, other factors such as farm size, accessibility, and perceived climatic changes can pose challenges to the adoption of these technologies.

This study revealed several intriguing findings regarding the factors influencing the adoption of climate-smart agricultural practices among farmers. Gender played a significant role, with a p-value of 0.0660 ( p < 0.1), indicating that the gender of the farmer had a notable impact on the likelihood of embracing sustainable farming methods.

Interestingly, farm size exhibited an inverse relationship with the adoption of climate-smart practices, as evidenced by the negative significant effect observed at a p-value of 0.0050 ( p < 0.01). This suggests that farmers with larger landholdings were less inclined to implement these environmentally-friendly agricultural techniques.

On the other hand, livestock quantity emerged as a positive driving force, exerting a significant influence on the adoption of climate-smart agriculture with a p-value of 0.0180 ( p < 0.1). Farmers with more substantial livestock holdings were more likely to adopt these sustainable practices, potentially due to the perceived benefits or increased resources associated with livestock ownership.

Notably, access to climate information had an unexpected negative impact on the adoption of climate-smart agriculture, with a p-value of 0.0060 ( p < 0.01). This counterintuitive finding raises questions about the effectiveness of climate information dissemination and the potential barriers that may hinder its translation into practical implementation. These findings align with the research conducted by Kurgat et al. (2020), which highlighted female ownership of farm assets, farm location, and household resources as critical determinants of climate-smart agricultural adoption in Tanzania. Similarly, Aryal et al. (2018) concluded that various factors, including household characteristics, market access, and primary climate hazards, influenced the probability and level of implementing different climate-smart practices among smallholder farmers. However, in contrast, a study by Abegunde et al. (2019) found that marital status, education level, fertilizer use, credit access, and access to extension services did not significantly impact the adoption of climate-smart agricultural practices. This divergence in findings underscores the complex interplay of factors influencing the embracement of sustainable agricultural methods and the potential variations across different contexts and regions.

Factors affecting smallholder farmers’ crop productivity

To identify the variables influencing agricultural yield, the study conducted a Cobb-Douglas production analysis ( Table 11). The findings demonstrated that income has a positive and significant effect on agricultural productivity; specifically, productivity increases by 0.002% as farmers’ income levels rise. This relationship is statistically significant, with a p-value of 0.0040 ( p < 0.01). According to a study by Urgessa (2015), the land-labor ratio, the use of pesticides and fertilizers, the amount of manure used, and household size are the main factors affecting agricultural labour and crop productivity. WenJing et al. (2021) examined income disparities across household quartiles and found that while middle-class and upper-middle-class households may benefit more from renting out their farmland, households with high on-farm incomes are less likely to expand their farm size through farmland rental. Fertilizer use significantly improved crop productivity, with crop yield increasing by 0.12% for every unit increase in fertilizer use. This relationship was statistically significant, with a p-value of 0.0000 ( p < 0.001). Additionally, Du et al. (2020) demonstrated that long-term applications of synthetic fertilizers and manure enhance soil productivity and crop yields.

The study unveiled a complex interplay of factors influencing crop productivity among smallholder farmers. Notably, farm size exhibited an inverse relationship, with larger landholdings leading to a 6.52% decrease in crop yields. This counterintuitive finding may be attributed to factors such as resource constraints, management challenges, or diminishing returns associated with larger land areas. In contrast, livestock quantity emerged as a significant driver of crop productivity, with each additional unit of livestock owned contributing to an 0.86% increase in crop yields. This statistically significant effect, with a p-value of 0.0001 ( p<0.001), highlights the synergistic relationship between livestock and crop cultivation. Livestock not only provide valuable manure and animal draught power but also serve as a source of income that can be reinvested in technologies that enhance crop production capabilities ( Anderson, 1989; Beedy et al., 2010). Remarkably, the adoption of climate-smart agricultural (CSA) practices was found to have a profoundly favorable impact on crop productivity. When an additional farmer implemented these sustainable farming methods, the average yield among farmers improved by a staggering 13.49%, a statistically significant effect with a p-value of 0.0720 ( p < 0.1). This finding underscores the potential of CSA practices to boost crop yields and contribute to overall agricultural productivity while promoting environmental sustainability. Corroborating these results, the study by Mujeyi et al. (2021) also demonstrated that the adoption of climate-smart agriculture significantly contributed to increased crop yields among smallholder farmers operating within an integrated crop-livestock system. This alignment across different research contexts reinforces the potential benefits of embracing sustainable agricultural practices and highlights the importance of an integrated approach. Furthermore, the study explored the impact of demographic characteristics on crop productivity. While marital status and the education level of the household head exhibited modest contributions of 0.07% and 0.098%, respectively, household size played a more significant role, contributing 0.2% to smallholder farmers’ crop productivity. These findings resonate with the research conducted by Serote et al. (2021), which highlighted the influence of smallholder farmers’ household demographics and institutional characteristics on the adoption of climate-smart agriculture and, consequently, crop productivity. This convergence of results across multiple studies underscores the complex interplay of factors that shape agricultural outcomes and the potential for targeted interventions to enhance productivity and sustainability. By addressing key drivers such as farm size, livestock integration, adoption of climate-smart practices, and demographic considerations, policymakers and development organizations can develop tailored strategies to support smallholder farmers in achieving improved crop yields while promoting environmental resilience.

3.2.1 Climate-smart practices impact on crop productivity

The study aimed to assess the impact of implementing Climate-Smart Agricultural (CSA) practices on crop productivity among smallholder farmers in the Nyimba district of Zambia. Additionally, it sought to identify the factors influencing the adoption of CSA practices and crop production among smallholder farmers in the region. The data analysis revealed that the crop yields of smallholder farmers who adopted CSA practices were 20.20% higher than those of non-adopters. Furthermore, the study found that smallholder farmers who adopted CSA practices experienced a 21.50% increase in maize yield compared to those who did not adopt these practices. These findings clearly demonstrate that the implementation of climate-smart farming methods can significantly boost crop productivity for smallholder farmers. By adopting CSA practices, smallholder farmers have the potential to substantially increase their crop yields, thereby enhancing their agricultural productivity and food security.

Several studies have investigated the impact of Climate-Smart Agricultural (CSA) practices on crop productivity, food security, and soil quality. Abegunde et al. (2022) conducted a study on the effect of CSA on household food security among 327 smallholder farmers in Nigeria, using binary and multinomial logistic regression models. Their findings revealed that the adoption of CSA practices significantly and favorably affected household food security. Additionally, agricultural revenue and income from non-farm sources positively influenced household food security. In a similar vein, Mossie (2022) assessed the impact of CSA technology, specifically wheat row planting, on productivity in Southern Ethiopia using propensity score matching. The study found that row planting had a favorable and significant effect on productivity, with adopters producing 1368 kg of wheat per hectare more than non-adopters, highlighting the positive impact of this CSA practice on yield. Tadesse et al. (2021) examined the impact of CSA practices in Ethiopia on soil carbon, crop productivity, and fertility. Their results demonstrated a 30–45% higher yield under CSA practices compared to the control ( p < 0.05). Furthermore, CSA interventions slightly increased soil pH and significantly increased plant-available total nitrogen and phosphorus by 2.2–2.6 and 1.7–2.7 times, respectively, compared to the control. This study highlights the potential of CSA practices to enhance crop yield and nutrient availability, contributing to the resilience of resource-poor farmers in the face of climate change. Kichamu-Wachira et al. (2021) also reported similar findings on the positive effects of CSA practices on crop yields, soil carbon, and nitrogen pools in Africa. Their study concluded that CSA practices represent an advanced agricultural technology compared to conventional farming, enhancing food production while promoting climate mitigation and soil quality. Moreover, Amadu et al. (2020) found that 53% of CSA adopters among 808 smallholder farmers’ households in southern Malawi experienced increased maize yields during the drought year of 2016. Similarly, Fentie and Beyene (2019) revealed a significant positive impact of CSA adoption on crop yield per hectare using a propensity score matching model. These studies collectively highlight the potential of CSA practices to enhance crop productivity, food security, and soil quality, contributing to the resilience of smallholder farmers in the face of climate change and variability.