Barriers to Climate-Smart Agriculture Adoption in Northeast China’s Black Soil Region: Insights from a Multidimensional Framework

Abstract

Climate change threatens global food security, highlighting the necessity for Climate-Smart Agriculture (CSA) to enhance agricultural resilience and sustainability. Yet low adoption among farmers highlights gaps in understanding adoption barriers. Existing models often overlook the dynamic, multi-layered nature of farmers’ decisions. This study introduces the Multidimensional Dynamic Decision Analysis Framework (MDDAF), which integrates Sustainable Livelihoods Framework, Diffusion of Innovations, and Behavioral Economics, and applies it to conservation agriculture in Northeast China’s black soil region. We conducted 125 semi-structured interviews (100 farmers, stage-mapped into six groups; 20 leaders of agricultural socialized service organizations; 5 technical experts) and analyzed transcripts in NVivo using a hybrid deductive–inductive approach. Findings show stage-specific barriers: superficial knowledge and fragmented perceptions in awareness; traditional norms and social stigmatization in evaluation; biosecurity risks, ecological mismatches, and land tenure disputes during decision-making; economic constraints and policy inconsistencies during implementation; and operational failures, incomplete practices, and climate-driven volatility at confirmation. Priority implications are as follows: professionalize service provision; safeguard bundle fidelity and manage climate risk; reduce context and tenure risks; and counter misbeliefs via complement-focused demonstrations, diverse opinion leaders, and targeted training. MDDAF thus links dynamic, stage-specific barriers to actionable interventions, supporting more effective CSA scale-up.

1. Introduction

Climate change threatens global food security, imperiling agricultural productivity and destabilizing crop production worldwide. Research shows that climate-related factors have already reduced global crop yields by approximately 30% [1]. For example, in China, the corn and soybean sectors have sustained losses of roughly USD 820 million over the past decade, with projected yield declines of 3–12% for corn and 7–19% for soybeans by 2100 [2]. Rising temperatures intensify water scarcity, impairing crop growth and reducing yields [3]; these increases also accelerate pest proliferation [4]. Prolonged droughts and soil erosion degrade arable land, further lowering agricultural output [5].

Climate-Smart Agriculture (CSA) provides an integrated strategy to address these risks. It pursues three objectives: (i) to sustainably increase productivity to improve farm income, strengthen food security, and foster development; (ii) to enhance the resilience of food systems to climatic variability; and (iii) to reduce greenhouse gas emissions from agriculture where feasible [6]. By combining productivity, adaptation, and mitigation, CSA provides a coherent approach to sustainable agriculture amid increasing climate uncertainty and contributes to carbon peaking and neutrality through carbon sequestration and resource efficiency [7,8]. Yet adoption among farmers in many developing contexts remains limited due to intertwined financial, technical, and socioeconomic constraints [9,10]. Understanding how and why these constraints operate across the adoption process is therefore essential for effective policy and intervention design.

Existing empirical studies have advanced this understanding but often rely on binary econometric models (e.g., probit and logit) that reduce the adoption process into a binary choice of “adoption” or “non-adoption” [11,12,13,14]. Such static treatments obscure the dynamic nature of technological uptake and the evolving motivations, feedbacks, and shocks that shape farmer behavior over time [15]. Analytical tools are needed to capture stages, feedbacks, and heterogeneity in decision-making.

Single-theory perspectives also yield incomplete explanations. Diffusion of Innovations (DOI) theory highlights how technology attributes and social networks shape staged diffusion and feedbacks [16], but typically treats adoption as an endpoint rather than a process integrated into livelihood strategies. The Sustainable Livelihoods Framework (SLF) situates technological choices within farmers’ portfolios of natural, human, financial, physical, and social capital, and emphasizes how vulnerability contexts and institutions condition decisions [17,18,19,20], but it gives limited attention to psychological drivers. Behavioral Economics (BE) addresses this by illuminating bounded rationality, loss aversion, and other cognitive biases that lead to departures from fully rational choice [21]. Synthesis studies therefore argue that explanatory power is context-dependent and support integrated, rather than single-theory, approaches to agricultural technology adoption [22,23].

To address these gaps, this study proposes the Multidimensional Dynamic Decision Analysis Framework (MDDAF), which integrates SLF, DOI, and BE to diagnose barriers to CSA adoption. SLF anchors a multidimensional appraisal of farmers’ capitals, constraints, and vulnerabilities, allowing technology attributes to be assessed within real livelihood contexts. BE theory enhances SLF by challenging its implicit rational assumptions, increasing the framework’s capacity to account for behavioral diversity among farmers. Meanwhile, DOI introduces a temporal dimension, delineating the evolutionary trajectory of adoption behaviors across different stages. Together, these components enable an integrated diagnosis of what constraints exist, when they bind, and why farmers respond as they do.

This study demonstrates MDDAF through a case study of Conservation Agriculture (CA) in the black soil region of Northeast China. CA is a key CSA practice that uses techniques such as crop straw mulching, reduced or no-tillage, and subsoiling to reduce the impact of droughts, heavy rainfall, and intense winds on crop production [7]. These techniques improve soil health, enhance water retention, and support carbon sequestration, while concurrently reducing greenhouse gas emissions [24]. Given growing heterogeneity among agricultural operators, many farmers now rely on agricultural socialized service organizations (ASSOs) to implement CA [25]. As long-term intermediaries, ASSO leaders and technical experts in CA possess complementary knowledge of on-the-ground barriers. We therefore employ a triangulated validation design that synthesizes perspectives from farmers (demand side), ASSO leaders, and technical experts in CA (supply or extension side).

Guided by MDDAF, we avoid treating all non-adopters as homogeneous. Instead, we decompose non-adoption into six dynamic subgroups (Groups A–F) along the adoption behavior chain: Awareness–Evaluation–Decision–Implementation–Confirmation (partial failure, Group E)–Confirmation (complete failure, Group F). This approach aims to

• Precisely identify barriers operating at each distinct stage (e.g., awareness gaps vs. implementation discontinuation).
• Reveal the technological synergy effect by differentiating partial adoption failures (Group E: abandonment after implementing one or two principles) from complete adoption failures (Group F: abandonment after implementing all three principles), thus validating the necessity of adopting a comprehensive, synergistic CA package for sustained implementation.
• Isolate underlying causal mechanisms by controlling for the confounding effect of technological incompleteness in Group E, thereby focusing on Group F to identify systemic failures (e.g., institutional mismatches or economic disincentives) leading to full abandonment despite technical fidelity.

Using MDDAF and multiple stakeholder perspectives, this study addresses three questions: Which barriers across the micro (farmer psychology and capital access), meso (service systems and institutions), and macro (policy and market) levels impede CA adoption in the black soil region? How are these barriers distributed along the adoption stage chain, and which are most salient at each stage? To what extent do partial versus complete adoption failures reflect missing technological synergy versus deeper institutional or economic misalignments? By integrating multiple theories and stakeholder perspectives, MDDAF offers a structured, transferable diagnostic for CSA adoption. Applied to CA in Northeast China’s black soil region, it clarifies where and why adoption falters, thereby informing targeted interventions to raise uptake. More broadly, the framework advances beyond binary and static models toward a stage-aware, behaviorally informed, and livelihood-grounded analysis of farmers’ technology adoption.

2. Theoretical Framework for Analyzing CSA Adoption

2.1. Components of the MDDAF

The MDDAF is a robust analytical tool that systematically explores the complexities of agricultural technology adoption behaviors (Figure 1). It integrates seven interconnected modules: livelihood capital, psychological factors, technology attributes, decision-making processes, livelihood outcomes, policy and institutional contexts, and vulnerability contexts. The following bullets describe each module and its role within the framework:

• Livelihood Capital: This module encompasses human, natural, financial, social, and physical assets. These collectively form the foundational conditions of farmers’ livelihoods. This capital directly shapes a farmer’s capacity to adopt innovative technologies [20]. By systematically analyzing these resources, MDDAF describes how asset endowments either facilitate or limit technology adoption, providing a basis for describing variations in farmers’ capabilities.
• Psychological Factors: This module explores psychological drivers, such as cognitive biases, risk aversion, social and time preferences, and prevailing social norms. These shape farmers’ perceptions and decision-making processes. These elements account for the heterogeneity and occasional irrationality in adoption behaviors [21]; this increases the framework’s explanatory depth.
• Technology Attributes: This component focuses on attributes such as relative advantage, complexity, observability, trialability, and compatibility. These assess the intrinsic qualities that govern technological diffusion and uptake among farmers [16], providing key insights into the adoptability of agricultural innovations.
• Technology Adoption Process: This module delineates a dynamic, iterative decision-making pathway with five stages: awareness initiation, evaluation and screening, decision optimization, adaptive implementation, and feedback confirmation. These stages cover the evolving nature of adoption decisions [16], offering a structured process-based lens.
• Livelihood Outcomes: This component evaluates the diverse impacts (outcomes) of technology adoption, which include economic benefits (e.g., increased agricultural income) and non-economic benefits (e.g., enhanced food security, improved well-being, reduced vulnerability, and greater resource sustainability). These outcomes highlight the broader implications of adoption for farmers’ livelihoods [26].
• Policy and Institutional Contexts: This module outlines the external conditions shaping technology adoption, including national and local government policies, legal frameworks, property right systems, and avenues for democratic participation [27,28]. It highlights the role of institutional structures in facilitating or constraining farmers’ choices.
• Vulnerability Contexts: This component addresses exogenous factors, including long-term trends (e.g., climate change), abrupt shocks (e.g., natural disasters), and seasonal fluctuations. These elements dynamically interact with other modules, amplifying or mitigating their influence on adoption behaviors [19,20].

2.2. The Operational Mechanisms of the MDDAF

The MDDAF considers farmers to be agents who are pursuing sustainable livelihoods through a complex, non-linear process of agricultural technology adoption. This process is constrained by livelihood capital and involves a comprehensive evaluation of technology attributes, a careful weighing of risks and benefits, and ongoing psychological adjustments. The adoption pathway occurs across five distinct stages [29,30]. The process is shaped by policy-institutional frameworks and vulnerability contexts. The following sections describe the operational mechanisms underpinning the five stages.

(1)

The awareness stage

The awareness stage represents the initial phase of the technology adoption process. Where individuals first encounter new technology and gain a fundamental understanding of it. This stage emphasizes the formation of “exposure” and “awareness,” whereby individuals learn of the new technology’s existence through institutional channels (e.g., extension services, mass media) or social networks and begin to grasp its functions, operation, and potential benefits.

Both institutional channels and social networks interact to increase awareness of new agricultural technologies. Government extension services leverage their established credibility and serve as a trusted conduit, delivering reliable information about cutting-edge technologies to farmers. Mass media sources have an extensive reach, broadly disseminating this information and collectively forming the institutional foundation for technology awareness [31]. Social networks further refine this process: strong ties that are rooted in close interpersonal relationships cultivate emotional trust; weak ties that stem from more distant connections provide diverse information and fresh perspectives [32]. Bounded rationality moderates farmers’ interpretations of this information, introducing a dual filtering mechanism. Selective attention may lead farmers to prioritize information that aligns with existing production practices, and confirmation bias may lead them to dismiss technologies perceived as unconventional or unfamiliar [16].

(2)

The evaluation stage

The evaluation (or persuasion) stage represents a critical phase where individuals form attitudes and emotional responses toward new technologies. During this stage, individuals shift from awareness to subjective assessment, actively seeking more information through interpersonal communication (such as discussions with peers) to weigh the new technology’s advantages and disadvantages. The attitude formation influenced by the new technology’s attributes: relative advantages (e.g., high cost-effectiveness) and compatibility (alignment with existing practices) tend to foster positive evaluations, while high complexity often leads to negative perceptions.

This stage unfolds as a dynamic process involving both emotional and cognitive dimensions, such as risk perception and uncertainty reduction. Farmers refine their risk perceptions over time, based on individual experiences and external influences. For example, extreme weather events or past losses can increase the perceived relevance of a technology; this reinforces risk memory and amplifies loss aversion and caution [33]. The resources available to the household, including social and human capital, further inform this assessment. Social capital facilitates a collective understanding of risks through shared insights, while human capital increases the precision of cost–benefit analyses [34]. The ability to demonstrate trial technologies mitigates risk-related concerns. Local trials offer opportunities for experiential learning, helping farmers adjust their risk perceptions based on firsthand evidence. Similarly, community demonstrations lower the costs of acquiring information, increase trust, and may expedite technology acceptance [35].

(3)

The decision stage

The decision stage is the pivotal turning point in the adoption process, where individuals choose to adopt or reject the new technology after weighing its advantages and disadvantages. The decision is a dynamic process shaped by a complex interplay of opportunities and constraints. Farmers evaluate key technological attributes alongside household endowments. These collectively influence their inclination and capacity to accept innovations.

Technology attributes define critical parameters for adoption. Compatibility increases feasibility by aligning with existing practices. In contrast, complexity introduces cognitive and cost-related barriers, increasing perceived risks and reducing the likelihood of adoption [16,36]. Household endowments further modulate this decision-making process. Livelihood structures shape risk and time preferences. For example, households that rely heavily on agriculture and that have limited risk buffers tend to prioritize short-term gains. Households with diversified income sources tend to be more flexible with respect to experimentation [20,37]. Having larger farm size leverages economies of scale and risk dilution, amplifying economic incentives and responsiveness to policy support [32]. Additionally, there is heterogeneity based on age, with older farmers often showing increased caution due to increased learning costs, shorter investment return horizons, and reduced labor availability [38]. Social capital significantly reshapes the decision-making landscape. Group adoption behaviors provide validation, mitigating individual risk perceptions. Community identity pressures can counteract inertia, motivating collective action [32].

(4)

The implementation stage

The implementation stage is the process of translating the decision into practical action, where individuals integrate the new technology into their daily routines. Effective implementation requires integrating household capabilities with institutional support and addressing micro-level (individual) and macro-level (systemic) factors to ensure successful implementation.

At the micro level, several household resources shape agricultural technology implementation. Household composition influences labor availability and opportunity costs, creating resource constraints [20]. Education and training increase technical comprehension and application, improving implementation efficiency [34]. Financial resources are critical, as initial investments and needs for ongoing cash flow create budget limits. Access to credit sustains long-term investment capacity [39].

Household factors lay the groundwork. However, macro-level institutional and policy frameworks are needed to overcome broader implementation barriers. Institutional frameworks reduce costs through a “market-credit-service” model. Market signals optimize resource allocation; credit mechanisms increase resilience against uncertainty; and localized technical services provide continuous support [29,40]. Policy systems reinforce adoption through adaptive measures. Innovation policies simplify adoption with modular designs [41]. Environmental regulations and fiscal incentives balance constraints with incentives [40]. Consistent policies increase long-term confidence in technology uptake [39]. This interplay between household capabilities and institutional backing makes technology feasible and sustainable.

(5)

The confirmation stage

The confirmation stage integrates short-term and long-term evaluations, shaped by cognitive, policy, livelihood, and environmental factors. This stage ensures that technological adoption remains attuned to the evolving needs and challenges that farmers face.

This stage involves an iterative process to validate a technology’s effectiveness and adaptability. Short-term economic benefits deliver immediate outcomes, enabling rapid assessments of a technology’s viability. In contrast, long-term ecological resilience requires sustained observations over multiple production cycles to confirm the durability and environmental fit [39]. Cognitive biases, such as shifts in reference points driven by climate change, may distort assessments. This may lead to incorrect judgments of a technology’s effectiveness [42]. Policy support and institutional coordination increase decision continuity through intertemporal benefit discounting and risk mitigation [43]. However, evolving livelihood strategies can prompt technology substitution, altering confirmation trajectories [34,40]. Vulnerability contexts moderate this process, potentially requiring adaptive substitutions when environmental shifts exceed a technology’s operational limits; this may result from factors such as extreme climate variability [44].

2.3. Cross-Scale Interactions of Livelihood Capital in CSA Adoption Using MDDAF

The MDDAF reveals the crucial role that livelihood capital plays in fostering CSA technology adoption, with a particular emphasis on the nested distribution of capital across multiple scales. Specific examples follow.

At the agricultural plot level, natural capital directly influences the suitability of CSA technologies for specific locations; examples include soil quality, water resources, and microclimate [42]. At the household level, human capital (e.g., labor, education, and health) and financial capital (cash, savings, and access to credit) determine a household’s capacity to adopt innovative technologies [17,18]. Within the community, social capital (social networks, organizational involvement, and trust) and physical capital (roads, markets, and irrigation systems) facilitate the diffusion of technologies through information sharing and collective action [45]. At the regional level, institutional capital provides essential external support for adoption of technology; this is manifested through agricultural markets, credit systems, and insurance programs [46]. At the national level, capital allocations are shaped by policies related to land, environmental protection, and agricultural technology subsidies; these influence the promotion and application of CSA technologies [6]. On a global scale, factors such as climate change, trade variability, and market fluctuations create a vulnerability context that impacts forms of livelihood capital across all scales [45].

The MDDAF models the interdependent roles of different forms of livelihood capital when adopting CSA technologies. For example, effectively implementing CSA technologies relies on integrating plot-specific soil quality (natural capital), household skills and labor (human capital), and community-based information sharing mechanisms (social capital).

2.4. MDDAF’s Dynamic Adaptation Through Recursive Feedback in Technology Adoption

Unlike conventional linear transmission models, the MDDAF uses recursive feedback mechanisms across its iterative stages. Each stage incorporates exit thresholds and reevaluation processes, enabling flexibility in decision-making through the dynamic interplay of environmental stimuli, capital restructuring, and cognitive changes. The framework’s key pathways are delineated below.

Technology adoption directly affects livelihood outcomes by improving production efficiency, increasing economic returns, or optimizing risk structures. This process drives the subsequent multidimensional reconfiguration of livelihood capital. This includes the accumulation of technical skills to build human capital, the transformation of increased profits into financial capital, the expansion of social networks to build social capital, and the enhancement of ecological efficiency in natural capital [9].

As livelihood capital increases, farmers’ perceptions of technology and their decision-making frameworks cognitively restructure. This includes increased recognition of a technology’s value, a recalibrated risk tolerance, and improved learning capacity. These shifts increase farmers’ adaptive capacity, helping them better navigate exogenous environmental challenges [9]. A self-sustaining feedback loop emerges, whereby technology adoption fosters capital accumulation. This, in turn, refines cognitive capabilities, strengthens adaptive resilience, and encourages further adoption. This positive cycle sustains momentum in decision-making and resource optimization, reinforcing the framework’s effectiveness.

In contrast, negative feedback may arise when technology adoption leads to unintended consequences, such as ecological degradation or the inequitable distribution of benefits. These adverse effects may prompt farmers to abandon the technology or disrupt its adoption trajectory. This highlights the framework’s sensitivity to detrimental outcomes and its ability to facilitate corrective adjustments.

External factors are critical disruptors within complex systems; these include market volatility, climate change, natural disasters, and policy shifts (e.g., subsidy adjustments or property rights reforms) [47]. These variables influence the system through two primary mechanisms. Directly, they modify capital accumulation or the suitability of technological applications. Indirectly, they reshape decision-making thresholds by recalibrating cognitive benchmarks or reallocating resources. As such, these dynamics may precipitate pauses, exits, or accelerations in adoption. This dynamic feedback mechanism illuminates how external conditions can initiate, accelerate, or obstruct technology adoption.

3. Materials and Methods

3.1. Selection of Case Study Area

This study applies the decision analysis framework to CA adoption in Northeast China’s black soil region to evaluate MDDAF’s explanatory power for CSA uptake. The National Soil and Water Conservation Plan (2015–2030) notes that the black soil region spans 1.09 million km², encompassing the entirety of Heilongjiang and Jilin provinces, and portions of Liaoning and Inner Mongolia (Figure 2). The climate is temperate monsoon with a short, cool growing season. Farming is primarily rainfed, with maize, soy and rice. In practice, CA mainly follows three locally adapted modes: residue-mulched no-till, wide-and-narrow-row mulched no-till with band rotation, and residue-retaining strip tillage. These modes are typically paired with a single-pass operation combining no-till seeding, fertilization, and pressing. Socioeconomically, holdings are fragmented, and the labor force is aging; ASSOs provide mechanized CA services but face dispersed demand and tight seasonal windows.

Recognized as one of the world’s four principal black soil regions, this area underpins national food security owing to its exceptional fertility and productivity. However, degradation has reduced soil organic matter by 30–40% [48] and thinned the tillage layer from 60–80 cm to 20–40 cm [49]. In response, black soil protection has become a national priority, reflected in recent policies such as the Notice on Further Strengthening the Protection of Black Soil Cultivated Land (2022), the National Black Soil Protection Project Implementation Plan (2021–2025), and the Black Soil Protection Law of the People’s Republic of China (2022). Theory and evidence identify CA as a core pathway for restoring fertility and improving resource use [50], and the 2020–2025 Action Plan mandates its wider promotion where suitable. Despite this policy push, substantial barriers to broader CA adoption persist [50]. Against this backdrop, MDDAF is used here to identify mechanisms shaping farmers’ CA decisions and to inform policy optimization.

3.2. Data Collection and Analysis

3.2.1. Sampling Strategy and Sample Size Sufficiency

This study employed qualitative methods to capture practical knowledge and lived experiences that quantitative models often overlook. Semi-structured and in-depth interviews were conducted with three stakeholder groups: technical experts in CA, leaders of ASSOs, and farmers. Using a multistage composite sampling strategy, we recruited 125 participants (5 technical experts in CA experts, 20 ASSO leaders, 100 farmers). ASSO leaders were purposively sampled across Inner Mongolia, Heilongjiang, Jilin, and Liaoning, while all farmers were sampled in Jilin Province (Figure 3).

• Technical Experts in CA: Selected based on authoritative expertise, specifically requiring more than a decade of research and development experience in CA. Interviews explored the complete CA package; critical operating and timing parameters and step integration; site suitability (climate, topography, soils); complementary implements, and minimum service standards and training/certification requirements. Drawing on these expert interviews, this study defines CA for Northeast China’s black soil region as an integrated technical system based on three core principles: straw mulching, no-till (or minimal tillage), and subsoiling. Farmers implementing one or two principles were classified as partial adopters; those implementing all three were classified as full adopters.
• ASSO Leaders: Selected through purposive sampling, guided by a spatial matrix accounting for soil types (e.g., black soil, chernozem, sandy soil) and topography topographical variations (e.g., hills, gentle slopes, depressions), yielding 20 regionally representative organizations. As both implementers and service providers, they described operational bottlenecks, service logistics and economics, contract and liability gaps, subsidy and payment frictions, and sources of farmers’ reluctance. Leaders were purposively sampled using a spatial matrix spanning soil types.
• Farmers (Primary Practitioners): In-depth interviews explored the sage-specific decision-making logics, cognitive biases, livelihood priorities, and implementation constraints, straw-allocation trade-offs, and climate-contingent experiences. Given the study’s primary focus on identifying adoption barriers, non-adopters and discontinuers were purposively targeted and, guided by MDDAF, were categorized into six decision-stage groups:
  • Group A (Awareness Stage): limited exposure to CA knowledge; examine how knowledge exposure shapes awareness.
  • Group B (Evaluation Stage): exposed but exhibiting negative attitudes; identify reasons for persistent negativity.
  • Group C (Decision Stage): interested but without adoption intent; identify blockers to intent formation.
  • Group D (Implementation Stage): intent but without implementation; identify implementation constraints.
  • Group E (Confirmation Stage: Partial Adoption): partial adopters who exited; test the importance of complete-package use.
  • Group F (Confirmation Stage: Full Adoption): full adopters who exited; isolate abandonment factors beyond incomplete implementation.

A triangular validation design was employed to integrate these perspectives and characterize interactions among groups, informing a regionally adaptive CA implementation mechanism. In lieu of a separate cohort of current adopters, adopter practices and enabling conditions were derived from ASSO leaders and CA technical experts with direct service experience, and from systematic analysis of exit rationales in Groups E–F.

The farmer survey was conducted in two phases. Initially, five villages were selected based on their soil-topography profiles and the presence or absence of CA demonstration sites. Within each village, an initial group of farmers was randomly sampled and interviewed using semi-structured techniques to assess their CA technology adoption status and categorize them into the six MDDAF groups. Subsequently, snowball sampling was employed, with initial respondents recommending fellow villagers exhibiting adoption patterns aligning with Groups A–F. This approach is widely used in qualitative research and facilitates access to hard-to-reach populations, such as those resistant to technology adoption [51].

Sample size sufficiency was justified using an information-power and saturation logic: a narrow, theory-informed aim (testing MDDAF), a specific stage-defined sample, strong interview quality via stakeholder triangulation, and cross-case analysis anchored to a priori framework. Saturation was monitored at two levels and tracked in a saturation table. Within-subgroup saturation was defined as no new first-order codes in two consecutive interviews: cross-stakeholder saturation as no new second-order themes in five consecutive interviews. Both thresholds were reached before the final interviews; remaining cases were used for confirmatory probing and negative-case checks.

3.2.2. Interview Procedures

Interview guides were derived from MDDAF, piloted during early fieldwork, and refined for clarity and probing depth. With consent, interviews were audio-recorded and transcribed verbatim. Identifiers were anonymized; each respondent received a letter code used consistently in the Section 4 (Figure 3). Figure 3 includes only the most representative interviewees from Groups A–F. All interviews were conducted with participants’ consent and the audio was recorded. Semi-structured interviews lasted an average of 20 min; in-depth interviews averaged 40 min, ranging from 15 to 90 min.

Table 1. Summary of main interviews.

Interaction of Constraining Dimensions	Interaction of Subdimensions	Original Interview Transcript	ID
Technical Attributes and Psychological Factors	Technical Compatibility and Cognitive Biases	“With all that straw mulch, seeds land on the top and can’t touch the soil. How are they supposed to germinate? No seedlings mean no yield!”	K13
		“Without plowing, the ground gets rock hard. How can the crop roots even penetrate and grow properly?”	C3
	Technical Complexity and Cognitive Biases	“Covered in straw? The planter will get jammed where it gets piled up, making normal sowing impossible.”	K4
	Technical Compatibility and Social Norms	“I was the first in the village to try no-till planting. Over 80% of the villagers laughed at me back then, calling it nonsense to sow into the messy straw without clearing or plowing.”	J5
		“I used to clear the straw to keep my land neat. If everyone cleans theirs and I leave mine messy, people will think I’m not serious about farming.”	C11
	Technical Compatibility and Social Preferences	“Even when I encouraged CA, my own father stuck to the old ways. After I sowed his field with the new method, he wouldn’t even go look; he said the straw-covered field was an eyesore.”	J1
	Technical Complexity and Risk Aversion	“You can’t weed mechanically with straw mulch blocking it. Herbicides are the only option, but the straw barrier reduces their effectiveness.”	S15
		“Straw can harbor diseases and insect eggs. Leaving it unburned or uncleaned just spreads them when you mulch.”	D6
Livelihood Capital and Psychological Factors	Natural Capital and Risk Aversion	“No-till fields are colder in spring. Seeds struggle to sprout, and young seedlings grow weakly.”	K15
		“Corn straw mulch can take a year or two to fully decompose. Sowing no-till into last year’s straw? Barely possible. How do you seed into two-year-old straw?”	X9
		“In low-lying fields prone to waterlogging, straw mulch slows down evaporation and keeps the soil temperature low.”	C9
		“Using no-till on our clay soil makes it not loosen up enough. The crops grow poorly, and yields are noticeably lower than traditional tillage.”	C13
	Physical Capital and Risk Aversion	“Encroachment is almost guaranteed where fields bordered by conventional tillage meet CA plots. Many farmers would rather accept lower yields using the old method than risk losing land to disputes over the new one.”	L2
Vulnerability Context and Livelihood Capital	Natural Disasters and Physical Capital	“Our area is rainfed, and is usually drought prone. But during heavy rains in 2022, our flat-planted fields with no ridges or ditches just drowned. The corn rotted.”	K6
Technical Attributes and Livelihood Capital	Technical Compatibility and Physical Capital	“If a neighbor encroaches 10 cm on a flat-field system over a long border, I could lose 0.03 hectares of land by year-end without even noticing.”	X3
Interaction Between Livelihood Capital	Physical Capital and Social Capital	“Farms here are fragmented, causing disputes over encroached boundaries due to CA every year. It’s not just our town; it’s happening all around.”	J1
		“My 2 hectares of land are scattered across 11 plots, none is larger than 0.33 hectares. It’s really hard to get a machinery operator to cover them all during the narrow planting window.”	D8
	Human Capital and Financial Capital	“I’m getting older, and finding off-farm work is tough. Farming’s my livelihood. Doing it myself saves money; otherwise, I’d barely scrape by.”	S12
	Physical Capital and Financial Capital	“I have my own gear for traditional farming. Switching to CA means paying for services while I sit idle.”	C17
		“CA wants straw left as mulch, but I feed cattle, I need that straw.”	K20
		“Baled corn straw sells for over 200 CNY per ton now.”	D10
	Physical Capital and Natural Capital	“Strong winter and spring winds here mean one field fire could easily sweep the whole area.”	S1
Feedback Mechanism of Livelihood Outcomes	Decreased Sustainability of Natural Resources	“After just two years of continuous no-till, the soil compacted so badly by the third year we had to rotary till it.”	D5
	Decline in Yield and Income	“That straw cover shades the soil, keeping it colder and leading to lower yields.”	K7
		“Trying to place seeds in moist soil, the operator sowed too deep. The seeds just rotted.”	S3
		“Fertilizer was placed too shallow; it volatilized. The ears were undeveloped with bare tips, and the kernels weren’t full.”	C2
		“The operator didn’t adjust the planter spacing properly; fertilizer burned the seedlings.”	K9
Vulnerability Context and Livelihood Outcomes	Climate Change and Reduced Vulnerability	“During dry years, CA outyields conventional by 1500–2000 kg per hectare.”	J9
		“After the extreme drought of 2018, conventional yields were nearly halved and CA held up better, so more than half the village switched over in 2019.”	H1
		“The no-till with straw mulch holds moisture incredibly well. Even if it’s not dry at planting time, you’re still guaranteed a solid stand.”	J5
	Natural Disasters and Increased Vulnerability	“Last year with all that continuous rain, the conventional fields managed to produce something, even if yields were down. But the CA fields? They got completely wiped out.”	J6
	Climate Change and Decreased Income	“After last year’s excessive rains wiped out the CA crops, about a third of the farmers in the village ditched the practice this year.”	J10

provides the quote–code matrix to support traceability from findings to verbatim evidence.

3.2.3. Coding Reliability and Analysis Transparency (NVivo)

NVivo was used with a hybrid deductive and inductive strategy. A codebook was seeded from MDDAF (stages, capitals, technology attributes, psychological factors, institutional and policy, vulnerability) and expanded through open coding. Two coders independently coded approximately 20% of transcripts, compared interpretations, and refined definitions; Cohen’s κ of at least 0.75 on major parent nodes was treated as acceptable, with disagreements resolved by discussion and senior-author adjudication. The revised codebook was then applied to the full corpus. Axial coding grouped first-order codes into second-order themes (for example, partial package failures, norm enforcement, service standardization bottlenecks, climate memory) using constant comparison across MDDAF stages and stakeholder roles. NVivo matrix queries cross-tabulated themes by stage and stakeholder, generating the evidence matrix summarized in Table 1. An audit trail (memos, decision logs, versioned codebooks) and negative case analyses supported transparency and reproducibility. Theoretical saturation was declared when matrix queries produced no new node combinations and axial links remained stable across two successive coding cycles, consistent with the saturation criteria in Section 3.2.1.

3.2.4. Cross-Stakeholder Integration and Validation

Triangulation matrices were used to align (i) farmers’ behavioral narratives, (ii) ASSO operational constraints, and (iii) expert technical standards with the MDDAF stages. When accounts conflicted, evidence was weighted by domain: farmer reports for behavior and decision logic, expert input for technical feasibility and package completeness, and ASSO narratives for service delivery and policy frictions. Discrepancies were logged, probed in follow-up interviews, and adjudicated through team discussion and negative case analysis. Joint code matrices and theme maps synthesized signals across groups, ensuring systematic integration and strengthening construct validity.

4. Results

Stage-specific findings are supported by exemplar interview quotations; see Table 1 for the coding matrix and quote IDs cited below.

4.1. The Awareness Stage: Knowledge Gaps and Cognitive Biases to CA

Farmers’ limited comprehension of CA substantially contributes to their cognitive biases against it. An analysis of the survey data indicates that 13% of respondents in Group A possess only superficial awareness of CA. These farmers lack a thorough understanding of the integrated practices essential to CA, such as scientific straw management, seedbed preparation, and subsoiling. The effectiveness of no-till systems depends on complementary practices such as straw mulching, no-till seeding, and soil-conservation measures. Farmers in this group often reduce the multifaceted CA framework to isolated practices such as “no-till” or “straw mulching.” This fragmented perception fosters misconceptions about CA’s feasibility and applicability, leading to skepticism and resistance.

Cognitive bias manifest in two primary forms. First, some farmers believe that straw mulching impairs seeding quality through poor seed–soil contact and residue blockages (K13, K4), overlooking complementary implements such as press wheels and double-disc openers. Second, some associate no-till with soil hardening (C3), contrary to evidence that no-till combined with straw mulching improves friability as organic matter accumulates. If unaddressed, these gaps seed skepticism in subsequent stages and weaken downstream adoption.

4.2. The Evaluation Stage: Reasons for Farmers’ Negative Attitudes Towards CA

4.2.1. Traditional Beliefs, Confirmation Bias, and Kinship Trust as Barriers to CA

Group B farmers (16%) retain negative attitudes toward CA despite knowledge exposure, primarily due to long-standing “tillage benefits” mental models, especially among older farmers, which emphasize aeration, temperature regulation, and weed control. These beliefs conflict with CA’s principle of minimal soil disturbance. Confirmation bias further reinforces this conflict. Decades of conventional practice lead farmers to privilege evidence favoring straw removal and tillage. At the same time, they discount the benefits of CA (e.g., improved soil quality and erosion control), creating an “experience cocoon” that lowers receptivity to innovation. A concise illustration comes from an ASSO leader who noted that even his father rejected a CA-treated field as “messy with straw everywhere” (J1). Trust structures deepen these patterns: 65% rely mainly on kinship or geographic networks for technical information, in which high trust and shared experiences filter out dissonant messages and entrench bias.

4.2.2. Kinship Networks as Information Channels and Enforcers of Traditional Norms Against CA

Kinship and geographic networks simultaneously diffuse information and enforce social norms; infused with mutual aid and risk-sharing values, these networks heighten normative pressure. Because CA challenges the ethos of “meticulous cultivation,” it encounters moral judgments and community scrutiny that foster implicit rejection. Survey evidence shows that 21% of respondents view straw removal and rotary tillage as signs of “diligence,” while no-till is misconstrued as “lazy farming”; conformity concerns and stigma (e.g., being seen as “not living properly”—C11) push farmers toward conspicuous conventional practices. Early adopters face reputational costs, as one interviewee explained: “I was the first in the village to try no-till seeding, and over 80% of the villagers ridiculed me” (J5), revealing a paradox: these networks excel at dissemination yet, due to homogeneity and consensus around tradition, operate as structural barriers to innovation. Thus, at the evaluation stage, social norm externalities, not information scarcity, are the dominant constraint.

4.3. The Decision Stage: Farmers’ Multidimensional Concerns to CA

4.3.1. Biosecurity Risks and Environmental Mismatch Drive Farmers’ CA Disinterest

Group C farmers (18%) are aware of CA but remain uninterested, chiefly due to perceived biosecurity risks and doubts about ecological fitness. Mulch is seen as impeding physical weeding and reducing herbicide efficacy, prompting fears of higher doses, repeated applications, and crop phytotoxicity. This anticipated escalation amplifies their skepticism toward CA. Uncomposted straw is also viewed as a reservoir for pathogens (e.g., Fusarium) and pest eggs (e.g., corn borer). Moreover, the moist, sometimes anaerobic conditions resulting from directly covering soil with straw are thought to heighten infestation risks.

CA’s effectiveness is context-dependent across climate, topography and soil, and infrastructure. Some constraints are intrinsic (e.g., delayed soil warming), while others are surmountable with appropriate practices and resources; this variability creates uncertainty about CA’s generalizability.

(1) Climate Adaptation. Straw mulching and no-till techniques are foundational components of CA, but they delay spring soil warming, impeding seed germination (K15). This problem is particularly acute in cold and arid regions like northern China, where slow straw decomposition, often one to two years (X9), shifts sowing schedules and increases uncertainty.

(2) Topography and Soil Sensitivity. In low-lying areas, straw mulching obstructs water drainage after heavy rainfall. This heightens risks of waterlogging and crop lodging (C9). Additionally, no-till practices in clayey soils can increase soil compaction. Another farmer noted “soil compaction leads to significant yield reductions” (C13).

(3) Infrastructural Limitations. In rainfed regions that do not have adequate drainage infrastructure, straw mulching and flat planting increase flooding risks during extreme rainfall (K6). CA’s vulnerabilities interact with ecological risks, creating a feedback loop that undermines its practicality. A lack of supportive infrastructure amplifies the difficulties of effectively implementing CA.

4.3.2. Land Fragmentation Risks and Short-Term Disincentives Undermine CA Appeal

The shift from ridge to flat cultivation under CA delivers ecological gains but introduces risks to property rights across three dimensions: blurred boundary demarcation, fragmentation-driven spillovers, and governance frictions. These institutional barriers emerge at the decision stage as salient deterrents.

First, flat cultivation weakens the visual cues that anchor plot boundaries. Without ridges and furrows, gradual encroachment (e.g., “boundary shifts of 10 cm per year”—X3) becomes harder to detect, heightening perceived land-loss risk and dampening willingness to adopt CA.

Second, fragmented landscapes amplify these risks: interspersing flat-cultivated plots among ridge-cultivated plots produces indistinct borders and increases exposure to encroachment (L2). Faced with a trade-off between potential yield gains and risks to property rights, many farmers choose conventional methods to safeguard land claims, even at the cost of lower productivity.

Finally, the incompatibility between CA practices and the existing rural land system has increased the incidence of property-rights disputes across the region, creating significant governance challenges. Land encroachment conflicts are particularly pervasive in areas where this technology is applied. One ASSO observed, “Each year, disputes over land encroachment due to flat cultivation occur, with similar issues evident in adjacent townships.” (J1) This cascading effect increases farmers’ reluctance to adopt CA and undermines the sustainability of its implementation.

Beyond property rights, the long-term nature of CA benefits increases farmers’ perception of risk. CA’s benefits accrue over a three-to-five-year soil-improvement cycle or longer, while early yields may stagnate or decline. Learning costs, such as operating specialized seeders, and uncertain short-term returns further deter risk-averse farmers who prioritize near-term income maximization.

4.4. The Implementation Stage: Constraints Despite the Willingness for Adoption

4.4.1. Interlocked Barriers Hinder CA Implementation for Willing Farmers

Among surveyed farmers, 26% of respondents fall into Group D; they are willing to adopt CA but face constraints. Their ability to implement CA was hindered by the interplay of three key constraints: small farm size, low income, and an aging labor force.

(1) Economic pressures from limited land and financial resources. Group D farms average 1.56 ha, with 53.85% at or below 1.33 ha. Initial adoption requires equipment costing about CNY 50,000, which is equivalent to 1.79 times the study group’s average annual agricultural income of CNY 27,900. Even with a 30% government subsidy, out-of-pocket costs remain approximately 125% of annual income. An older workforce (mean age 51 years) and limited schooling (mean 8 years) further constrain the capacity to operate advanced machinery, collectively undermining feasibility.

(2) Livelihood vulnerability reinforces path dependence. In 76.92% of Group D households, production relies on just one to two laborers, and agriculture provides more than 85% of total income. This highly concentrated income structure, coupled with a fragile livelihood system, results in minimal tolerance for technological risk. One farmer (S12) remarked, “We can only rely on farming to make a living and cannot save money.” The upfront costs of existing traditional machinery lead to an additional lock-in effect: farmers who have invested in such equipment tend to extend its service life where possible to avoid additional service costs (C17).

(3) Cost sensitivity conflicts with operational logic. Implementing CA requires machinery service fees of CNY 1200–1500 per hectare. Based on Group D’s average land size, this is an annual incremental cost of CNY 1871–2339, or 6.7–8.4% of their agricultural income. Given thin margins, households substitute family labor for capital; when CA service fees threaten those margins, willingness does not translate into action, resulting in approval without adoption.

4.4.2. Constraints on Standardized Mechanized Services for CA Implementation

An insufficient supply of standardized mechanized services constrains CA implementation, and is driven by dispersed demand, high supplier entry barriers, and policy inefficiencies.

On the demand side, fragmented cultivated land leads to inefficient service delivery. Surveys indicate that a typical farmer (e.g., D8) manages two hectares of land divided into 11 plots; each plot is at or below 0.33 ha. This fragmentation means that machinery must be frequently moved between plots. This significantly increases scheduling costs and raise the risk of missing critical windows such as sowing. Consequently, ASSOs face diminishing marginal returns, thereby reducing their willingness to supply services. Additionally, heterogeneous cropping patterns (crop mixes, row spacings) further limit standardization and economies of scale, thereby reducing ASSOs’ willingness to supply services.

On the supply side, entry costs and skill requirements are high. CA-specific machinery, such as no-till seeders, costs approximately four to six times as much as traditional equipment and is tightly coupled to CA systems. Precision operations (seeding depth control, residue handling) increase technical demands, while generally low educational attainment among rural labor (average 8 years) constrains the local skills base.

At a policy level, subsidy design weakens provider participation. In some regions, payments arrive after 12 to 18 months, straining ASSO cash flow. Additionally, subsidy calculations depend on farmers’ self-reported data and manual spot checks. This system is vulnerable to dishonest behavior, such as exaggerated claims about work areas and subsidy fraud. Unclear policy guidelines create uncertainty, making service providers pessimistic about long-term profits. Faced with these frictions, providers default to traditional tillage services, slowing CA expansion.

4.4.3. Challenges of Straw Resource Competition and Policy Discontinuity in CA Implementation

Farmers face two implementation-stage constraints: competing uses of straw and inconsistent policy implementation. Survey data indicate that 57.69% of Group D farmers report having to choose between retaining straw to meet technical standards and using it for feed or sale. Livestock producers prioritize feed security (K20), while non-livestock households are pulled by market prices of roughly CNY 200 to 240 per ton (D10). As a result, retention frequently falls short of agronomic requirements, reflecting a micro-level portfolio trade-off and meso-level market incentives that impede CA uptake.

A parallel macro-level barrier stems from the shift from blanket burning bans to seasonal permits, which introduces ambiguity and weakens compliance. First, in regions with high wind speeds, environmental risks can compromise the effectiveness of CA practices. One farmer (S1) offered an example: “A fire in one place can spread rapidly.” This highlights the environmental challenges that impede the uniform application of CA technical standards. Second, the issuance of seasonal burning permits has led some farmers to perceive that adhering to CA practices is discretionary. This perceived policy leniency incentivizes intermittent burning as a cost-reduction strategy. These signals erode straw retention norms and soil organic matter objectives, discouraging sustained implementation and indicating systemic misalignment rather than purely technological shortcomings.

4.5. The Confirmation Stage: Reasons for Farmers’ Discontinuation of CA

4.5.1. Incomplete Technology Implementation and the Negative Cycle of CA Discontinuation

Farmers’ discontinuation decisions are strongly associated with incomplete implementation of the CA package. Among those who abandoned CA, Group E (partial adopters who later exited) and Group F (full adopters who later exited) account for 19% and 8% of the sample, respectively; and 70.37% of exits are directly attributable to incomplete implementation.

In Group E, two recurrent partial-implementation patterns drive failure. First, no-till seeding without straw mulching and subsoiling leads to progressive compaction, restricted rooting, and yield decline, prompting reversion to traditional rotary tillage (D5). Second, no-till with straw mulching but without subsoiling delays soil warming during seeding (K7) and reduces soil permeability, raising risks of waterlogging and seedling loss during heavy rains, thereby eroding confidence in CA.

Overall, CA’s benefits depend on the synergistic package of no-till, straw mulching, and subsoiling. When farmers simplify operations or selectively apply technology components, they disrupt the integrity of the technology chain. This disruption creates an imbalance between production efficiency and ecological objectives, placing farmers in a negative feedback loop of “trial-failure-discontinuation.” The core misalignment is between CA’s systematic, multi-year demands and fragmented, risk-averse practices shaped by short-term economic logic, ultimately undermining the successful adoption and sustained use of CA.

4.5.2. How Operational Failures in Four Dimensions Drive CA Abandonment

The cumulative effect of operational errors substantially increases farmers’ likelihood of discontinuing CA. This outcome can be systematically analyzed across four critical dimensions:

(1) Technical execution deviations. Deviations in technical execution are a primary driver of CA failure. Errors in machinery operation, such as excessive seeding depth that causes seed rot (S3), inadequate spacing between seeds and fertilizer that leads to seedling burn (K9), and insufficient fertilization depth that results in nutrient deficiency (C2), directly erode expected agronomic gains and trigger abandonment.

(2) Inappropriate operational timing. The improper timing of agricultural operations increases the technical risks associated with CA. Even when farmers have adopted CA, they may be influenced by traditional farming habits, such as “rushing to sow early.” As a result, they may overlook CA’s specific timing requirements. For instance, straw mulching delays soil warming, warranting a slight delay in sowing to ensure adequate germination. Overburdened ASSOs during peak seasons may be unable to service all fields within the optimal technical window. This challenge is compounded by service providers’ insufficient technical knowledge, such as mistakenly postponing autumn subsoiling until the seedling stage.

(3) Disconnections in operational integration. Breaks in sequential steps create a “broken chain”: overly long straw at harvest can block seeders, while delaying harvest to meet corn moisture standards can preclude timely autumn subsoiling and the associated soil moisture benefits.

(4) Human capital and accountability deficiencies. Structural deficiencies in human capital and the absence of robust accountability mechanisms are deep-seated obstacles to CA adoption. Farm machinery operators frequently lack specialized training, resulting in an imprecise understanding of critical technical parameters. Moreover, the inertia of traditional farming practices causes their operations to diverge from established standards. Compounding this, ambiguous service contracts that fail to assign technical responsibility thereby shifting error costs to farmers, eroding trust and willingness to persist with CA.

4.5.3. Decision Changes Driven by Climate Change

Climate change reshapes farmers’ perceptions of CA benefits, establishing a dynamic feedback mechanism that links risk assessments, realized outcomes, and subsequent technological decisions.

(1) Climate scenario dependence. Farmers’ evaluations of the value of CA strongly depend on climate conditions. In drought years, CA’s moisture retention can increase yields by about 1000 to 1500 kg per hectare (J9); while in excessively wet years, straw mulching heightens waterlogging risk and can cause complete crop failure (J6).Under moderate conditions, the yield gap relative to conventional tillage narrows. These climate-contingent results produce polarized outcomes, ranging from yield gains to total losses, and weaken the consistency of CA adoption.

(2) Climate memory and lagged effects. Historical climate events shape farmers’ expectations of CA benefits through “climate memory,” generating lagged effects. After the 2022 heavy rains, abandonment rose in some villages (about one-third, J10), whereas the 2018 drought was followed by broader uptake in 2019 (more than half, H1). These asymmetric, lagged responses complicate promotion strategies and sustain volatility in adoption.

(3) Climate Conditions at Planting. Real-time conditions at sowing are decisive. Prolonged drought favors CA for moisture conservation and seedling protection (J5), while periods of adequate rainfall prompt reversion to conventional tillage to avoid perceived absolute losses, reinforcing climate-contingent decision-making.

5. Discussion

5.1. Stage-Specific Barriers Along the Adoption Pathway: Awareness to Confirmation

In our sites, the uptake of CA is shaped by layered constraints at three levels. Micro-level barriers include thin cash buffers, an aging labor force, and bounded rationality. At the meso level, agricultural socialized service organizations face skill and standardization gaps given dispersed, heterogeneous demand. At the macro level, subsidy-liquidity constraints, straw-burning regimes, and property-rights frictions during ridge-to-flat transitions further impede adoption. This configuration is consonant with syntheses that highlight knowledge, finance, and institutional capacity in smallholder systems [52,53] and with European evidence that diffusion is slowed by both supply- and demand-side frictions [54], not merely by awareness deficits. It also accords with meta-analytic findings that simultaneous constraints, not single bottlenecks, best explain low adoption in developing-country contexts [55]. Our contribution is to specify CA-specific mechanisms that translate these broad categories into operational chokepoints (e.g., boundary ambiguity following ridge-to-flat transitions; service standardization hurdles).

Barriers cluster by stage in ways binary models cannot reveal: awareness (misconceptions), evaluation (kin-network norm enforcement), decision (ecological fit, biosecurity, tenure risk), implementation (cost–skill–standardization), and confirmation (partial-package traps, weak accountability, climate shocks). This sequencing operationalizes DOI’s temporal logic [16] and behavioral reappraisal under uncertainty [21,56], answering calls to map where and when constraints bind along the innovation pathway [57]. Our evidence also clarifies why information volume alone seldom shifts attitudes: homophilous kin networks both transmit information and enforce “meticulous cultivation” norms that stigmatize residue retention and no-till [17,32]. Two refinements extend prior work: (i) infrastructure contingency (micro-drainage can flip mulching from risk-reducing to risk-increasing) and (ii) climate memory (extreme seasons trigger lagged over- and under-corrections), consistent with vulnerability-centered CSA research [44,45,47] and the context dependence of CA performance.

We identify two exit pathways. Package incompleteness dominates: partial adoption aligns with global evidence that no-till alone often depresses yields, whereas the combined implementation of no-till, residue retention, and crop rotation mitigates losses and can increase yields in dry, rainfed environments [58]. System design failures also matter: unclear liability, training deficits, and delayed subsidies push discontinuation despite technically faithful use, echoing evidence that durable CSA requires coordinated institutional and service architectures, not awareness alone [53,54]. Complementarities we observe are consistent with multi-practice correlations in India [59], reinforcing a “teach complements, not components” extension logic.

Taken together, the stage-specific, cross-scale mechanisms we document integrate agronomic bundle synergies [58] with socioeconomic complementarities [59] and frictions that impede adoption on both the supply side (service quality, standardization, coordination capacity) and the demand side (knowledge, norms, risk, finance) [52,53,54,55]. The MDDAF lens thus delivers a stage-aware, behaviorally informed, livelihood-grounded diagnosis that directly underpins the targeted implications detailed below.

5.2. Contributions to Theory: Advancing MDDAF and Integrating SLF, DOI and BE

First, the findings validate MDDAF’s premise that constraints are stage-specific and cross-scale, making explicit the sequence by which cognition (awareness), norms (evaluation), institutions and ecology (decision), capabilities (implementation), and shock reappraisal and liability (confirmation) become binding, thereby enriching SLF’s asset view with a DOI-temporal structure and BE-based mechanisms [16,19,21]. MDDAF is technology-agnostic; its stage-specific, cross-scale mechanisms apply to other CSA practices where adoption likewise depends on package complementarity and service and institutional alignment.

Second, MDDAF’s separation of technology design problems (complementarity, operational parameters) from system design problems (standardization, liability, subsidy liquidity) refines classic adoption models that aggregate these as “costs and constraints” [41,56,57]. It yields a clear diagnosis-to-instrument mapping: teach complements and enforce bundle fidelity (technology design) while fixing contracts, payment lags, and training standards (system design).

Third, by incorporating climate memory and accountability into the confirmation stage, MDDAF advances dynamic, recursive accounts of adoption beyond single-shot “adopt/not adopt” outcomes [15,29,30,47] and aligns with calls for iterative, uncertainty-aware models in CSA diffusion [53,54].

5.3. Implications: Prioritized by Leverage on Observed Failure Modes

(1)

Implementation: fix the service ecosystem (highest priority)

Standard contracts and certification. Require ASSO contracts that specify technical parameters (for example, fertilizer–seed spacing) and clearly assign liability; introduce an operator certification requirement with periodic refreshers to close skills and accountability gaps identified in the field.

Aggregation and routing. Fund township-level scheduling platforms or machinery rings and promote harmonized row spacing to batch small plots and reduce setup and switching costs on fragmented fields.

Straw-retention incentives. Offer differentiated subsidies tied to measured levels of residue retention, and clarify burning rules with only rare, clearly defined emergency permits.

(2)

Confirmation: prevent partial-package traps and climate-driven exits

Bundle-conditional support. Link credit and subsidies to full conservation agriculture packages, comprising no-till, residue retention, and subsoiling or crop rotation, with two-tranche disbursement: one after a post-planting quality check and another after a post-harvest agronomic verification.

Climate-contingent advisories. Provide residue-aware sowing windows, waterlogging alerts, and rapid-response micro-drainage crews during wet springs; offer index insurance products for drought and waterlogging to buffer climate shocks that trigger exits.

Post-adoption assistance. Establish a technical hotline and guarantee field support within seventy-two hours during peak windows to troubleshoot seeding depth, residue length, and fertilizer placement, failure points frequently observed in the field.

(3)

Decision: reduce context risk and tenure frictions

Micro-drainage grants. Provide grants for surface drains or raised beds on low-lying plots, with ASSO calibration to residue conditions.

Tenure safeguards. Install field-edge GPS pegs, conduct annual drone mapping, and adopt village by-laws for boundary maintenance and dispute resolution during ridge-to-flat transitions, directly addressing CA-specific boundary ambiguity under fragmentation.

Integrated pest management under residue. Deliver extension on pest and disease management in mulched systems and herbicide regimes compatible with residue retention to address prominent biosecurity concerns.

(4)

Awareness to evaluation: correct misbeliefs and loosen norm pressure

Complement-focused demonstrations. Run two-season, side-by-side demonstration plots that explicitly showcase press wheels, double-disc openers, and residue handling to counter specific misconceptions documented in the study.

Norm-diversifying messengers. Engage mixed-status opinion leaders and early adopters to run peer-led field days, with modest input vouchers to offset participation costs, to weaken homophilous conformity pressures.

Focused training. Deliver short, modular machine-operation clinics that emphasize bundle fidelity; expand access given evidence that targeted training increases both the likelihood and the intensity of multi-practice adoption.

5.4. Limitations and Future Research

This study purposively centered on non-adopters and discontinuers; continuous adopters were represented only indirectly via ASSO leaders and technical experts. We also lacked a contemporaneous cohort of current adopters, and medium- to large-scale operators are underrepresented in the farmer sample. These design choices, while appropriate for diagnosing failure points, limit prevalence inference, external validity across farm sizes, and identification of enabling factors among successful adopters.

Future research should: (i) track farmer cohorts longitudinally to estimate transition probabilities across MDDAF stages under climate shocks [47]; (ii) test stage-specific interventions experimentally or quasi-experimentally (e.g., contract redesign, boundary marking, drainage micro-investments, bundle-conditional incentives); (iii) apply MDDAF to other CSA practices (for example, drip irrigation, agroforestry, improved varieties) to assess external validity and refine practice-neutral levers.

6. Conclusions

This study applied the MDDAF, which integrates the SLF, DOI, and BE, to diagnose low uptake of CA in Northeast China’s black soil region. Evidence comes from stage-mapped farmer interviews (Groups A–F) triangulated with ASSO leaders and technical experts.

Adoption is hindered by mutually reinforcing constraints that operate across scales and unfold across decision stages. In the awareness stage (Group A), superficial knowledge fosters cognitive biases, fragmenting perceptions of CA’s integrated practices and perpetuating misconceptions. Negative attitudes in the evaluation stage (Group B) arise from traditional norms and kinship networks enforcing conformity, creating an “experience cocoon” that resists innovation. Decision-stage concerns (Group C) encompass biosecurity risks, environmental mismatches, land fragmentation-induced property disputes, and short-term disincentives, highlighting CA’s context-dependent compatibility. For willing adopters (Group D), implementation is thwarted by interlocking constraints like small-scale farming, financial burdens, aging demographics, mechanized service deficiencies, straw competition, and policy ambiguities. Discontinuation in the confirmation stage (Groups E and F) stems from incomplete implementation triggering negative cycles, operational failures, and climate-driven volatility, where “climate memory” amplifies asymmetric responses to weather extremes.

Conceptually, MDDAF distinguishes technology design from system design problems and embeds temporal and behavioral mechanisms within a livelihoods frame. Methodologically, stage-mapped sampling and cross-stakeholder triangulation convert adoption “barriers” into actionable mechanisms. Policy and practice implications align with dominant barriers: (i) professionalize and standardize the service ecosystem (contracts, certification, reliable subsidy liquidity); (ii) protect bundle fidelity with conditional support and climate-contingent advisories; (iii) reduce context risk via micro-drainage, tenure safeguards, and integrated pest management under residue; and (iv) correct misbeliefs and norm pressures through complement-focused demonstrations, diversified messengers, and targeted training. Adopting these stage-aware, mechanism-targeted measures offers a credible pathway to durable CA uptake, improved soil stewardship, and greater climate resilience.