Smart Urban Agriculture Needs Better Governance, Lower Costs and Wider Global Reach

Smart urban agriculture is rapidly moving from experimental sensor kits to AI-assisted production systems, but a new review suggests the sector's biggest challenge is no longer technological capability; it's whether cities can turn digital farming into an affordable, inclusive and genuinely urban development strategy.

The study, "Smart Urban Agriculture in Transition: A Systematic Review of ICT Integration, Applications, and Challenges," was published in the journal Agriculture. Authored by Ruhang Wei, Dan Wu, Wulijiang Mulati, Qifeng Hou and Shengxi Xin, it reviews 143 English-language studies published between 2016 and 2026 and maps how information and communication technologies are being applied across vertical farms, hydroponic systems, plant factories, greenhouses, rooftops, community gardens and peri-urban land. Its central finding is striking: digital innovation is advancing fastest where food production is enclosed, measurable and technically controllable, while the urban governance, equity and community dimensions of agriculture remain far less developed.

The smart-farming boom is real, but highly concentrated

The review shows that research on smart urban agriculture accelerated sharply after 2021. The field is now organized around a recognizable cluster of technologies and production systems: the Internet of Things, wireless sensors, machine learning, hydroponics, vertical farming, controlled-environment agriculture and plant factories.

The growth reflects a powerful technical logic. Enclosed systems allow growers to monitor temperature, humidity, light, pH, electrical conductivity, dissolved oxygen and nutrient concentrations in real time. Once these variables are converted into data, machine learning can support yield forecasts, disease recognition, energy optimization and automated control. In vertical farms and plant factories, researchers are increasingly moving beyond simple monitoring towards integrated "sensing-prediction-control-optimization" systems.

The study's ICT-by-scenario matrix illustrates the imbalance. IoT and sensor networks were most frequently linked to soilless cultivation and vertical farming, while automation and AI were also heavily concentrated in controlled environments. Robotics, digital twins and computer vision are growing, but remain less mature and more selective.

This concentration is understandable. Technologies are easiest to test where conditions can be standardized and performance measured. But it also creates a narrow picture of what "smart" urban agriculture means. A highly automated facility may be technically advanced, yet still be energy-intensive, capital-heavy and disconnected from the wider needs of the city. On the other hand, a low-cost sensor system helping a community garden conserve water may deliver greater public value even if it appears less sophisticated.

Urban agriculture is not governed as a system

The review diagnoses what current research overlooks. Smart urban agriculture is strong in production engineering but weak in urban integration. The keyword analysis is dominated by terms such as IoT, hydroponics, machine learning, vertical farming and plant factories. Governance, planning, public participation, food justice and social equity are far less visible. In other words, the literature largely treats urban agriculture as a production unit that can be monitored and optimized, rather than as part of a complex food system shaped by land rights, infrastructure, regulation, affordability and access.

Urban agriculture serves different purposes in different contexts. A commercial vertical farm may seek predictable year-round output close to affluent consumers. A peri-urban farm may help protect agricultural land from urban expansion. A community garden may support nutrition, education and social cohesion. A rooftop system may contribute to building performance, stormwater management and local food production. Each requires a different form of "smartness."

The review identifies at least two pathways:

• The first is a production-optimization pathway, led by vertical farming, soilless cultivation and plant factories.
• The second is an enabling and spatial-governance pathway, in which GIS, remote sensing, basic sensors, mobile applications and digital platforms support site selection, land monitoring, irrigation, coordination and public participation.

Policy should not privilege the first pathway simply because it is more technologically visible. Cities need digital tools that help answer where food production should take place, who benefits, how systems will be maintained and how urban agriculture connects with water, energy, buildings and public space.

Global South evidence gap could distort investment decisions

The review also finds a pronounced geographic imbalance. The United States accounted for 25 of the 143 studies, followed by China with 12 and South Korea with seven. Large parts of Africa, Central Asia, Eastern Europe and Latin America were weakly represented or absent. This is not merely an academic gap. It affects the relevance of policy and investment decisions. Many cities in low- and middle-income countries face acute food insecurity, rapid urban expansion, water scarcity, informal land tenure and unreliable infrastructure. Yet many digital agriculture systems are being developed in settings with stronger research funding, more stable electricity, better data systems and greater technical capacity.

A model that performs well in a laboratory greenhouse or capital-intensive plant factory may not survive in a community-managed site where replacement parts are difficult to obtain. A cloud-based monitoring platform may be inappropriate where connectivity is unreliable. Automated systems may reduce labor in one context but exclude local growers in another if training, maintenance and ownership arrangements are ignored.

For governments, development agencies and investors, the implication is clear: transferability cannot be assumed. Smart urban agriculture projects should be assessed against local energy costs, water availability, crop preferences, maintenance skills, land tenure and market demand. Low-cost, open-source and repairable systems may often offer a stronger development case than cutting-edge platforms with high recurring costs.

This is especially important for the Global South, where the most promising applications may involve combining affordable sensors, solar power, efficient irrigation, spatial mapping and community-based management rather than replicating the most capital-intensive models found in wealthier cities.

The next frontier is not more automation; it is better evidence

The review follows PRISMA 2020, draws from Scopus and Web of Science, and combines bibliometric mapping with structured thematic coding. The authors screened 2,017 initial records and produced a final corpus of 143 studies. They also double-coded a subset to improve consistency.

But the study identifies major weaknesses in the underlying evidence. Many technical papers rely on a single crop, one location or a small experimental setup. Evaluation tends to emphasize prediction accuracy, sensing stability and control precision, while broader indicators such as energy use, carbon emissions, affordability, labor accessibility, user acceptance and food access receive less attention.

The review itself is also limited to English-language publications in two databases and does not conduct a formal quality or risk-of-bias assessment. It can therefore show where technologies are being used more confidently than it can determine which systems are consistently effective.

The next phase of research should be more comparative and more urban. Multi-city trials are needed across different climates, infrastructure conditions and income levels. Evaluation frameworks should combine technical performance with lifecycle energy use, water efficiency, commercial viability, governance fit and social outcomes. Digital twins and decision-support tools may help integrate these dimensions, but they should not become ends in themselves.