The study defines the integration of the digital economy and agricultural industry through two linked processes. Agricultural digital industrialization, or ADI, covers the commercialization of digital infrastructure, technologies, products, and services for agriculture, including infrastructure construction and the large-scale application of digital solutions. Agricultural industrial digitalization, or AID, describes the use of modern information technologies by farmers, enterprises, and other operators to transform the agricultural value chain, encompassing production, circulation, and sales. Data are central to both processes: digital enterprises supply infrastructure and technical services, while agricultural operators generate practical demand and application scenarios for those services. The authors argue that this arrangement creates a feedback loop in which digital-economic development enables agricultural transformation and agricultural transformation stimulates the commercialization of digital products.

To quantify this integration, the researchers constructed an indicator system from data for all 31 provincial-level administrative regions in China. The indicators covered rural communications, broadband and mobile-phone access, fiber-optic networks, telecommunications services, electronic-information manufacturing, information-sector employment and income, rural e-commerce participation, agricultural machinery power, facility agriculture, postal services, and online purchasing and sales. They weighted the indicators with the CRITIC objective weighting method, which accounts for the variation, contrast, and correlations among measures, and calculated weights from the aggregated sample so scores remained comparable across years. Missing observations were filled through linear interpolation. This composite approach is important because digitalization is not represented by a single technology: it includes physical infrastructure, commercial digital services, the capabilities of agricultural actors, and changes across the wider industry chain.

The temporal results showed a fluctuating but sustained increase in the national integration index. ADI rose from 0.073 in 2013 to 0.192 in 2022, while AID increased from 0.093 to 0.150. Across the four major economic regions, eastern China advanced from 0.224 to 0.399, the central region from 0.141 to 0.324, the western region from 0.141 to 0.320, and the northeastern region from 0.122 to 0.275. The study also divided the country into three functional grain zones. The main grain-consumption area increased from 0.239 to 0.416, the main grain-production area from 0.150 to 0.331, and the balanced production-consumption area from 0.139 to 0.309. These scores describe the study’s constructed index rather than an official measure of agricultural digitalization, and their absolute level was characterized as relatively low even after the decade-long increase.

Spatial analysis found persistent clustering but evidence of gradual convergence. Guangdong, Zhejiang, Jiangsu, Beijing, and Shanghai were identified as relatively high-integration provinces, while the northeast, parts of central China, and the mountainous southwest generally lagged. The authors linked the eastern advantage to stronger digital-technology capacity and more favorable economic conditions, whereas weaker infrastructure and high-tech industries constrained some western and southwestern areas. Tibet and Qinghai improved substantially, which the study partly attributes to relatively abundant per-capita public facilities and services, lower population pressure, and participation in China’s “East Data, West Computing” initiative. The spatial autocorrelation of the index declined: Moran’s I fell from 0.436 in 2013 to 0.296 in 2022, indicating that clustering remained significant but weakened. The calculated center of gravity moved 45.7 kilometers northwest, consistent with faster relative growth in central and western provinces.

Regional inequality declined overall, with the national Gini coefficient for the integration index falling from 0.152 to 0.098. The decomposition identified differences between regions, rather than differences within them, as the principal source of overall disparity. For the four economic regions, the mean intergroup contribution was 70.78%; for the three grain zones, it was 65.57%. Within the eastern region, the Gini coefficient averaged 0.108 and showed a fluctuating decline, while the central, western, and northeastern averages were 0.045, 0.053, and 0.041, respectively. Among grain functional areas, the main consumption area had the highest average coefficient, 0.095, the main production area averaged 0.072, and the balanced area averaged 0.058. The pattern therefore combines a narrowing national gap with continuing internal variation in some economically advanced or production-oriented regions.

A dynamic panel model indicated that ADI and AID reinforce each other, but not at the same speed. ADI lagged by two periods had a significant positive association with AID, with a coefficient of 0.735, suggesting that the enabling effect of digital industrialization on agricultural digital transformation may emerge with a delay. The authors relate this lag to technical barriers, investment costs, and infrastructure constraints. AID, in turn, had an immediate positive association with ADI, with a coefficient of 0.164, consistent with agricultural digitalization creating demand and application opportunities for digital products and services. In the one-period lag terms, the coefficient for ADI was 0.746 in the AID equation, while the coefficient for AID was 1.067 in the ADI equation, emphasizing that improvements accumulate over time. Diagnostic tests supported the model specification: first-order autocorrelation was present, while second-order autocorrelation and the Hansen test indicated no evidence of the respective problems.

For the analysis of factors associated with integration, the researchers used geographically and temporally weighted regression, a method designed to allow relationships to vary across location and time. Five influences were examined: technological innovation, represented by high-tech research-and-development expenditure; market demand, represented by rural online retail sales of physical goods; government guidance, represented by fixed-asset investment in agriculture; financial services, represented by the Digital Inclusive Finance Index; and scale operation, represented by the number of agricultural cooperatives per 10,000 rural residents. The model’s adjusted R-squared was 0.930, compared with 0.823 for ordinary least squares, and its corrected Akaike information criterion was −156.690, compared with −4.550. These results favor the geographically and temporally weighted specification within this dataset, although they do not establish that the measured factors cause the observed changes.

The strongest and most consistent positive association involved technological innovation. In the 2022 example, its mean regression coefficient was 0.439, and high-value clusters shifted from southeastern toward southwestern areas over the study period. Market demand was also positively associated with integration, with its mean coefficient rising from 0.095 to 0.115 and strong clusters expanding from the southeast toward central and northeastern China. Government guidance showed a positive relationship, with its coefficient increasing from 0.022 to 0.175; stronger clusters were concentrated in the northwest and northeast, regions that have received attention through national development strategies. Financial services initially reinforced integration but weakened after 2018 and became nearly neutral in the 2022 mean estimate, falling from 0.089 to −0.001. The authors suggest that regulation, risk aversion, reduced effective financing demand, and pandemic-related balance-sheet pressures may have contributed, but the observational design cannot isolate these explanations.

Scale operation was negatively associated with integration in the baseline model, with its mean coefficient moving from −0.020 to −0.094. The authors caution that this relationship may be nonlinear and do not interpret the result as evidence that larger farms or agricultural organizations generally reduce digitalization. Instead, they note that China remains dominated by small-scale farming, while digital transformation involves high fixed costs, technical barriers, long return periods, and limited profitability. The study also identifies important limitations: it focuses on the development path of the integrated system rather than fully analyzing the separate trends and coupling of ADI and AID; it treats IDEAI mainly from an industrial perspective, leaving macroeconomic and enterprise-level influences less developed; and its proxy indicators, constrained by available data, may introduce measurement error. The results suggest that closing the geographic divide will require more than infrastructure alone, with technological capability, market demand, public guidance, cross-regional cooperation, and effective financing all showing geographically and temporally varying associations with digital agricultural integration.

Subject of Research: Digital economy and agricultural industry integration across Chinese provinces

Article Title: Spatiotemporal patterns, regional differences, and influencing factors of the integration of digital economy and agricultural industry in China

Article References: Wu, D., & Ai, W. (2026). Spatiotemporal patterns, regional differences, and influencing factors of the integration of digital economy and agricultural industry in China. Frontiers in Sustainable Food Systems, 10, Article 1985406. https://doi.org/10.3389/fsufs.2026.1985406

Image Credits: AI Generated

DOI: 10.3389/fsufs.2026.1985406

Keywords: digital agriculture, digital economy, China, regional disparities, agricultural industrialization, geographically weighted regression, rural infrastructure, technological innovation, rural e-commerce, agricultural policy, Spatiotemporal, patterns

The study is anchored in Mission Antyodaya, the Indian government’s nationwide rural survey programme designed to aggregate village-level data on infrastructure and services in support of poverty-free gram panchayats. Rather than relying on anecdote or single indicators, the researchers built a composite model from 15 distinct parameters drawn from the survey: paved roads, anganwadi childcare centres, drainage facilities, higher secondary schools, high schools, primary schools, telephone connectivity, markets, the public distribution system, electricity, common service centres, all-weather road connections, automated teller machines, banks, and public transport. Each of these parameters captures a different dimension of what makes a village functionally developed, from the ability to get produce to market to whether children can pursue education beyond the primary level without leaving home.

To combine these indicators into a single, defensible measure, the team turned to the Analytic Hierarchy Process, a structured decision-making technique introduced by mathematician Thomas Saaty in 1980. AHP works by breaking a complex problem into a hierarchy of criteria, then eliciting pairwise comparisons that assign relative weights to each factor. In this case, the 15 infrastructure parameters were weighted according to their contribution to overall rural development, and the weighted scores were aggregated for every panchayat in the subdivision. The result is a continuous development index that can be mapped, classified, and statistically interrogated, converting a sprawling table of survey responses into a spatially explicit picture of advantage and deprivation.

A crucial strength of the study lies in its validation. The researchers tested their model using the Receiver Operating Characteristic curve, a statistical tool borrowed from diagnostic medicine and machine learning that plots the true positive rate against the false positive rate across all possible classification thresholds. The area under the ROC curve came out at 85 percent, a level generally interpreted as indicating good discriminatory power. In practical terms, this means the model’s classification of villages into development categories is far more reliable than chance, giving policymakers a quantitative reason to trust the resulting maps rather than treating them as illustrative sketches.

The spatial results are revealing. According to the AHP model, only 6.36 percent of the study area falls into the highest development category, with a further 26.6 percent classified as high development. At the other end of the spectrum, 22 percent of the area falls into the low development zone and 17.4 percent into the very low category. Taken together, nearly four in ten hectares of this predominantly rural subdivision sit in the two most deprived classes, a distribution that underscores how unevenly India’s rural infrastructure investments have landed even within a single administrative subdivision.

Geography explains much of the pattern. The panchayats surrounding Tarapith, one of Bengal’s most important religious centres and a magnet for pilgrims, have experienced significant development, which the authors attribute to the economic potential generated by religious tourism. Temples draw visitors, visitors demand roads, hotels, banking, markets, and transport, and the resulting service ecosystem lifts the surrounding villages’ infrastructure scores. It is a textbook example of how a single node of economic activity can radiate development outward into its rural hinterland, and it suggests that tourism-linked investment can be a genuine engine of rural transformation when the conditions are right.

The contrast at the other extreme is equally instructive. Panchayats along the borders of the Chotanagpur plateau, where the terrain rises into the rugged, forested margins that separate the Bengal plains from the Deccan, show minimal to negligible development. These remote and inaccessible locations suffer a compounding disadvantage: difficult terrain raises the cost of building roads and extending utilities, poor connectivity then deters banks, markets, and transport operators from establishing services, and the absence of services in turn deepens isolation. The study’s maps make this feedback loop visible, offering planners a precise target list of where interventions are most urgently needed.

Beyond the composite index, the researchers deployed factor analysis, a multivariate statistical technique that identifies hidden dimensions underlying a set of observed variables. The first factor extracted from the data loaded heavily on fundamental infrastructure: the road network, school education, anganwadi centres, and electricity supply. This is a meaningful analytical finding, because it suggests that these basic services travel together as a coherent package, and that they constitute the dominant axis along which villages differ. In other words, the single most important thing separating developed from underdeveloped panchayats in Rampurhat is not any one amenity but the presence or absence of this foundational cluster of connectivity, education, childcare, and power.

The implications extend well beyond one subdivision. India’s economy remains fundamentally shaped by its agrarian society, and the country’s growth trajectory depends substantially on how its rural areas develop. Evidence-based mapping of this kind speaks directly to persistent national challenges: tackling poverty, raising agricultural productivity, improving rural infrastructure and services, promoting sustainable resource management, strengthening governance, and fostering social inclusion. Because Mission Antyodaya data covers the entire country, the AHP-plus-validation framework developed here is potentially replicable at scale, giving district and state planners a standardised method for identifying which gram panchayats need investment first and which parameters matter most in each locality.

What makes the study particularly timely is the growing global recognition, echoed in recent World Bank and OECD rural development reports, that infrastructure is not a uniform lever but a spatially differentiated one. Roads alone do not create development, as economic research on India’s rural roads programme has shown, but the right bundle of services in the right place can. By quantifying that bundle, validating the measurement, and locating the gaps with map-level precision, this research offers a template for turning the vast streams of village survey data that India now collects into actionable intelligence, and a reminder that the distance between a thriving temple town and a neglected plateau hamlet can be measured, mapped, and, in principle, closed.

Subject of Research: Spatial assessment of rural development using Mission Antyodaya data and the Analytic Hierarchy Process in Rampurhat subdivision, Birbhum district, West Bengal, India

Article Title: Empowering villages: a study on rural development through Mission Antyodaya in a part of Rarh Bengal, Eastern India

Article References: Das, S., Khatun, S., & Das, N. (2026). Empowering villages: a study on rural development through Mission Antyodaya in a part of Rarh Bengal, Eastern India. SN Social Sciences, 6(9), Article 428. https://doi.org/10.1007/s43545-026-01703-2

Image Credits: AI Generated

DOI: 10.1007/s43545-026-01703-2

Keywords: rural development, Mission Antyodaya, Analytic Hierarchy Process, ROC curve, geospatial analysis, West Bengal, Rarh Bengal, Birbhum district, factor analysis, rural infrastructure, panchayat, India

The team gathered primary cross-sectional data from 231 cabbage producers, selected through a multi-stage sampling technique across four kebeles, the smallest administrative units in Ethiopia, and from 26 traders operating in the district. The traders were deliberately categorized into licensed and unlicensed entities working across both formal and informal market networks, a distinction that matters enormously in a setting where most produce changes hands outside officially regulated channels. By combining descriptive statistics with Cragg’s double-hurdle econometric model, the researchers were able to disentangle two separate decisions that earlier studies often blurred together: whether a farmer participates in the market at all, and, if so, how much of the harvest he or she actually supplies.

The double-hurdle framework is what gives this study its analytical teeth. In many rural economies, the decision to sell and the decision of how much to sell are governed by different forces. A farmer might be willing to enter the market but constrained in the quantity offered, or might sell only trivial surpluses even when market access exists. Cragg’s model treats these as two sequential hurdles, each with its own set of explanatory variables, and estimates marginal effects, reported as dy/dx values, that quantify how a one-unit change in each factor shifts the probability of participation or the intensity of supply. This two-stage logic reveals patterns that a single-equation approach would hide.

At the first hurdle, market participation, the results are strikingly consistent with human-capital theory. Education raises the probability of selling by 0.110, farming experience by 0.006, family size by 0.071, landholding by 0.116, and total production by 0.048. Each of these coefficients is statistically significant, and together they sketch a portrait of the archetypal market-oriented cabbage grower: literate, experienced, commanding enough household labor to tend a larger plot, and producing enough volume to make the trip to market worthwhile. The one variable that pushes in the opposite direction is distance to market, with a marginal effect of −0.013, a small number per kilometer but a meaningful drag in a landscape where roads are rough and transport options scarce.

The second hurdle, supply intensity, tells a more surprising story. Here the positive drivers include male household headship, with a marginal effect of 1.842, age of the household head at 0.025, landholding at 0.504, total production at 1.089, and, most powerfully, access to market information at 2.765. That last figure is the single largest effect in the entire model, and it carries a clear policy message: when farmers know what prices prevail in distant markets, they release dramatically more of their crop for sale. Information, in other words, is not a soft factor but a hard constraint on commercialization. Conversely, two variables suppress supply intensity: distance to market, at −0.127, and, counterintuitively, the use of credit, at −1.541.

The negative credit coefficient deserves careful reading. The authors do not claim that credit is inherently harmful; rather, the result suggests that in Bonke District credit is often obtained under terms or for purposes that do not translate into expanded marketed supply. Loans may be consumed by household needs, repaid immediately after harvest in ways that force farmers to hold back produce, or channeled into inputs whose returns arrive after the marketing window closes. Whatever the precise mechanism, the finding challenges the reflexive assumption that simply expanding rural credit will automatically commercialize smallholder agriculture. The study’s recommendation is for targeted credit utilization strategies, meaning credit designed around the specific cash-flow rhythms of perishable horticulture rather than generic agricultural lending.

Underlying all of these econometric results is a structural diagnosis the authors call the infrastructure-perishability trap. Cabbage is a fragile commodity: it loses water, wilts, and rots quickly, especially without cold storage or shade. When the roads connecting Bonke’s kebeles to district and regional markets are poor, the time and physical punishment of transport multiply losses dramatically. Severe post-harvest losses follow, and those losses in turn erode the bargaining power of farmers. A trader who knows that a farmer’s cabbage will be unsellable by tomorrow can dictate today’s price. The trap thus feeds on itself: bad infrastructure causes losses, losses weaken the farmer’s negotiating position, weak bargaining power depresses farm-gate prices, and low prices discourage the investments that might improve production and marketing in the first place.

The actor network the study maps is multi-tiered, running from producers through local collectors, licensed wholesalers, unlicensed itinerant traders, and retailers, with produce flowing through both formal and informal channels. The researchers argue that regulatory frameworks should support both licensed and informal traders rather than squeezing the informal tier out of existence. In districts like Bonke, informal traders often perform essential aggregation and distribution functions that licensed intermediaries do not reach, and heavy-handed formalization risks severing the very arteries through which rural produce reaches urban consumers. The policy implication is a middle path: light-touch licensing, market infrastructure, and price transparency that formalize gradually without destroying livelihoods.

Placed in a broader context, the findings speak directly to Ethiopia’s ambition of transforming subsistence agriculture into a competitive, commercialized sector, and to the Sustainable Development Goals that frame that ambition, particularly those targeting poverty, hunger, and decent economic growth. Horticultural crops like head cabbage are often promoted as quick-win diversification options for smallholders because they have short growing cycles and strong urban demand. But the Bonke evidence shows that agronomic potential means little without the surrounding market architecture: roads that keep perishables intact, information systems that broadcast prices, credit instruments matched to crop calendars, and value chain management that treats production, logistics, and marketing as one integrated system rather than separate problems for separate ministries.

What makes this study resonate beyond Southern Ethiopia is its methodological clarity. By separating the decision to participate from the decision of how much to supply, and by quantifying the marginal effect of each driver, it gives policymakers a ranked menu of interventions. Landholding and total production matter at both hurdles, suggesting that anything that expands the production base, from irrigation to improved seed, will pay off twice. Market information access is the standout lever for supply intensity, which makes investment in price reporting, mobile phone coverage, and radio market bulletins unusually cost-effective. Distance deters both entry and volume, which places rural road investment at the top of any serious agenda. And the cautionary tale of credit reminds development practitioners that financial inclusion must be designed, not merely delivered. For the cabbage farmers of Bonke, the difference between subsistence and commercialization is not a single grand reform but the accumulation of these connected fixes, each one loosening a strand of the trap that currently holds their harvest, and their incomes, well below what the land could yield.

Subject of Research: Market participation and marketed supply of head cabbage among smallholder producers in Southern Ethiopia

Article Title: Market integration and value chain performance of head cabbage production in Southern Ethiopia

Article References: Goshime, H., Alehegn, T., Katola, G., Deginew, A., Admasu, B., & Ayaye, E. (2026). Market integration and value chain performance of head cabbage production in Southern Ethiopia. Discover Sustainability. https://doi.org/10.1007/s43621-026-04871-6

Image Credits: AI Generated

DOI: 10.1007/s43621-026-04871-6

Keywords: head cabbage, market integration, value chain, double-hurdle model, Ethiopia, smallholder farmers, post-harvest losses, market information, rural infrastructure, agricultural credit, Bonke District, sustainable development

The study, led by Jackline Juma Robert of Egerton University together with Mary W. K. Mathenge and Augustus Sammy Muluvi, examines why some smallholder cashew farmers throw themselves fully into cooperative life while others remain on the margins. The stakes are considerable. Kenya’s national cashew output has collapsed from 28,410 tonnes in 2015 to just 8,997 tonnes in 2023, and Kilifi County, which produces roughly 40 percent of the national crop and supports more than 70,000 smallholder farmers, sits at the epicenter of the decline. Weak institutional coordination, thin market linkages and low cooperative participation are identified as central culprits.

Agricultural cooperatives, the authors argue, are among the best-documented institutional fixes for smallholder market failure. By pooling produce and sharing infrastructure, they cut transaction costs, improve access to quality inputs, credit and markets, and create platforms for disseminating knowledge and new technologies. In cashew value chains specifically, cooperatives enable collective investment in storage, transport and processing, allowing farmers to capture more value from each nut. The problem, the researchers found, is not that cooperatives do not work; it is that participation intensity, meaning the number of different services a farmer actually draws from the cooperative, remains stubbornly low.

To investigate, the team surveyed 394 smallholder farmers in Kilifi North Sub-County, a designated cashew production zone where the crop is the principal cash enterprise. Using a multistage sampling design, they first purposively selected the county, then the sub-county, and finally four major cashew-producing wards: Kibarani, Matsangoni, Dabaso and Tezo. Individual respondents were drawn by simple random sampling from a register of 11,012 registered cashew farmers maintained by the county agriculture department and the Empowering Farmers Foundation, with the sample size calculated using Yamane’s formula and inflated slightly to guard against non-response.

Because the outcome of interest, the count of cooperative services a farmer receives, is count data plagued by excess zeros, the researchers compared four regression specifications: standard Poisson, negative binomial, zero-inflated Poisson and zero-inflated negative binomial. Diagnostic tests revealed significant overdispersion, with a mean of 1.82 against a variance of 2.62, and a glut of zeros largely produced by non-members who are structurally excluded from cooperative services. The zero-inflated Poisson model, which separately models this structural-zero group from the count process among members, delivered the lowest Akaike Information Criterion at 1304.858 and emerged as the preferred specification. Multicollinearity checks using variance inflation factors showed no problematic correlation among predictors.

The descriptive portrait of the farming population is striking in its own right. Nearly 64 percent of surveyed cashew farmers were women, a pattern the authors attribute to men migrating toward alternative income activities as cashew returns have stagnated, leaving women to shoulder routine farm management, harvesting and marketing. Cooperative members were also significantly more likely to have received training in cashew production and Good Agricultural Practices, to access market information, to use pesticides and to belong to social groups, suggesting that membership is entangled with broader networks of information and support.

The regression results tell a coherent economic story. Older farmers used more services, with each additional year of age raising the expected rate of service utilization by a factor of 1.012, plausibly because experience has taught them the long-run value of cooperative support. Education mattered as well: each additional year of schooling increased service uptake by a factor of 1.023. Larger households, with their deeper labor pools and diversified risk, participated more intensively, as did farmers earning off-farm income, whose expected rate of service use was 1.264 times higher per unit increase in such income, presumably because financial slack makes cooperative engagement less risky.

Scale and market access proved decisive. Each additional cashew tree owned raised expected service utilization by a factor of 1.004, reflecting economies of scale that make cooperative services worth the trouble, while farmers with market access used services at 1.895 times the rate of those without, a finding the authors interpret as evidence that a ready market builds the trust that underpins collective action. On the flip side, geography worked against participation: every additional kilometer from the nearest market, or from the nearest tarmac road, reduced service utilization by a factor of 0.985. Remoteness, the study suggests, cuts farmers off not just from physical market access but from extension visits, cooperative announcements and the peer networks through which cooperative life actually happens.

The authors are careful to acknowledge limits. Their data are cross-sectional, gathered between July and August 2025, and thus capture a snapshot rather than the dynamics of participation over time. Conditions in other cashew-producing areas, with different cooperative structures, market access and agroecology, may diverge from Kilifi North, cautioning against overgeneralization. Still, the analytical framework, grounded in Collective Action Theory and Random Utility Theory, treats each farmer as a utility maximizer weighing the perceived costs and benefits of engagement, and the data support that framing with unusual clarity.

The policy implications are concrete. The researchers recommend targeted training to diffuse information about cooperative benefits, deliberate strengthening of social networks and market interactions, and subsidy programs channeled through cooperatives, especially for farmers with fewer trees who might otherwise be priced out of participation. They also call for public investment in rural roads and market proximity, alongside growth in the non-farm economy, to raise off-farm incomes that fuel cooperative engagement. In a county where one nut once funded households, schools and futures, the message is blunt: reviving Kenya’s cashew sector may depend less on the tree than on the roads, education and trust that carry farmers into collective action.

Subject of Research: Determinants of the intensity of participation in agricultural cooperatives among smallholder cashew farmers in Kilifi County, Kenya

Article Title: Intensity of participation in cooperatives among smallholder cashew farmers in Kilifi County, Kenya

Article References: Robert, J. J., Mathenge, M. W. K., & Muluvi, A. S. (2026). Intensity of participation in cooperatives among smallholder cashew farmers in Kilifi County, Kenya. Discover Agriculture, 4(1), Article 282. https://doi.org/10.1007/s44279-026-00755-5

Image Credits: AI Generated

DOI: 10.1007/s44279-026-00755-5

Keywords: cashew farming, agricultural cooperatives, Kilifi County, Kenya, smallholder farmers, Zero-Inflated Poisson model, cooperative participation, market access, rural infrastructure, farm productivity, collective action, coastal agriculture