با همکاری انجمن اقتصاد کشاورزی ایران
نوع مقاله : مقالات پژوهشی
نویسندگان
گروه مدیریت صنعتی، دانشکده علوم اداری و اقتصاد، دانشگاه اراک، اراک، ایران.
کلیدواژهها
عنوان مقاله English
نویسندگان English
Title
A Comprehensive Model of Artificial Intelligence and Internet of Things Applications in Smart Agriculture with a Meta-Synthesis Approach
Introduction
Smart agriculture is rapidly becoming a cornerstone of sustainable food systems by integrating artificial intelligence (AI), the Internet of Things (IoT), and data-driven decision support. Yet many available frameworks are not tailored to the realities of semi-arid regions, where water scarcity, heterogeneous soils, and fragmented connectivity constrain deployment. This study develops a localized Smart Agriculture Model designed for Iran, organizing technologies and tools into a coherent, three-level structure that links macro domains, enabling technologies, and operational instruments. The model emphasizes practical decision-making and real-time monitoring over purely algorithm-centric approaches, aligning technical capabilities with environmental and infrastructural constraints. In doing so, it seeks to convert dispersed knowledge into an implementable blueprint for national and regional planning.
Materials and Methods
The research employs a meta-synthesis methodology to integrate qualitative and quantitative evidence on AI/IoT in agriculture. A systematic search identified 334 potentially relevant publications. Through staged screening (title, abstract, content, and methodology), 12 high-quality studies were retained for full analysis. From these sources, 44 indicators were coded and then grouped into ten technological domains that span the intelligence layer (e.g., analytical models and decision aids), the sensing and communication layer (e.g., IoT devices, wireless sensor networks, robotics), and the processing and security layer (e.g., big data platforms, cloud/edge computing, blockchain).
Objective importance weights for each indicator were computed using Shannon entropy, capturing the dispersion and informational value of concepts across the literature set. To contextualize these objective weights with local realities, expert judgments from four national specialists in digital agriculture were elicited and combined with the entropy results via a hybrid scheme. The integration coefficient was set to α = 0.4 after sensitivity testing to balance objective evidence with contextual expertise and to minimize divergence between the two weight sets. Reliability and agreement of the qualitative coding were assessed using multiple indices: Cohen’s kappa (0.784) indicated substantial agreement; additional measures, including Holsti’s percentage agreement (0.822) and Krippendorff’s alpha (0.81), further supported coding consistency.
Results and Discussion
The resulting model clarifies priorities across the three levels. At the operational tools level, IoT-based sensing achieved the highest integrated weight (W = 0.123), followed closely by artificial intelligence tools (W = 0.116) and wireless sensor network capabilities. Aggregating to the technology level shows a clear emphasis on decision-centric intelligence and field-level data acquisition. In contrast, deep learning components and broad supply-chain-oriented blockchain applications received lower weights under current conditions, reflecting limitations in data volume, standardization, and computing infrastructure.
These priorities mirror the pressing needs of semi-arid agriculture: granular monitoring of soil moisture and fertility, microclimate tracking, and optimized water and fertilizer use. The model therefore foregrounds low-power, wide-area connectivity; interoperable sensor gateways; and edge-aware analytics that support timely farm management actions. At the same time, it maintains a pathway for growth by including big data platforms, cloud/edge hybrids, and secure data stewardship, which become increasingly important as data flows expand.
A policy-relevant insight is the model’s “human–technology hybrid” orientation. Rather than replacing human expertise, AI and analytics are positioned to augment agronomists and farmers with transparent recommendations, scenario analysis, and early-warning capabilities. This orientation not only enhances adoption but also fits the realities of varying digital readiness across regions. Sensitivity analysis around α confirmed the stability of the weighting scheme (standard deviation < 0.007), indicating that the prioritization is robust to reasonable shifts in the balance between evidence and expert input.
Overall, the distribution of weights suggests a pragmatic modernization pathway: first, build reliable sensing and communications; second, couple these data streams to actionable decision support at farm and regional scales; third, progressively expand data management, cloud/edge resources, and advanced modeling as data coverage and quality improve.
Conclusion
This study proposes a localized, data-driven framework that integrates AI and IoT within a structured three-level model for smart agriculture. Using a hybrid weighting method (α = 0.4) to merge entropy-based evidence with expert knowledge, the results demonstrate that artificial intelligence tools and IoT-enabled sensing are the most impactful levers for immediate advancement in semi-arid contexts. The framework translates technological possibilities into implementable priorities: expand low-power connectivity and interoperable sensor networks; establish a national backbone for secure, standardized agricultural data; and elevate AI literacy so that farmers and extension agents can act on timely, explainable insights. While the synthesis draws on published research rather than field trials, the robustness of the weighting and agreement metrics provides confidence in the model’s near-term usefulness. Future work should validate and regionalize the weights with real-world datasets, incorporate scenario-based dynamic weighting under climate and market uncertainty, and couple the framework to simulation tools to estimate system-level impacts on productivity, resilience, and resource efficiency.
Keywords
Smart Technologies; Shannon Entropy; Hybrid Weighting; Decision Support Systems.
کلیدواژهها English
