Objectives of the service
The objective of SAT2LEAF was to establish whether a space-enabled monitoring service is technically feasible and commercially viable on the very small, fragmented plots of urban and peri-urban agriculture — plots typically between 50 and 2,000 m², for which precision-agriculture platforms designed for fields above one hectare do not work. The project developed and tested a smartphone-first architecture in which the phone's GNSS receiver and camera are the acquisition layer, on-device AI performs 3D canopy reconstruction and leaf-image disease diagnosis, and Copernicus Sentinel-2 provides vigour context. The activity also tested the original hypothesis that satellite vegetation indices could be fused with plant-level ground data to produce plot-level maps.
Users and their needs
SAT2LEAF addresses commercial urban and peri-urban growers: market gardens and 'daily-producer' farms supplying restaurants, canteens and citizens, aromatic-herb and leafy-green producers, rooftop and community commercial farms, and the agricultural cooperatives, farm networks and agri-food companies that aggregate them. These growers manage several small, heterogeneous plots, rarely have access to an agronomist, and identify disease only once symptoms are visible. Their priority need, confirmed during the pilot campaign, is early and reliable disease identification with treatment guidance, followed by the reduction of chemical inputs and water. Municipalities, NGOs and public bodies took part as pilot hosts and referral channels rather than as customers.
Service/ system concept
SAT2LEAF is delivered as a dedicated urban module of the iAgro platform, which is already tested and commercially available for conventional agriculture; the module reuses that infrastructure and adds the workflows and models specific to urban and peri-urban micro-plots. The grower registers a plot in the iAgro app by drawing its boundary on a map or by dropping a GNSS point, then walks the plot taking guided canopy scans and leaf photographs with the smartphone. Every acquisition is geotagged by the phone's GNSS receiver. On-device processing reconstructs a 3D digital twin of the plant and extracts canopy metrics — height, volume, Tree Row Volume, leaf area index and above-ground biomass — which drive water and crop-protection prescriptions. An AI model identifies diseases and stress from the leaf images and ranks the most probable diagnoses with treatment advice, while crop-specific infection-risk models combine 1 km gridded weather forecasts with plot conditions to anticipate infection windows and suggest spraying times. Soil properties are retrieved from SoilGrids and a generative-AI engine assembles the whole picture into a plain-language agronomic report. On plots larger than 0.5 ha, Copernicus Sentinel-2 vegetation-index maps add spatial context for zoning.
Space Added Value
SAT2LEAF uses two space assets, with clearly distinct roles. Satellite navigation is the primary one: the multi-constellation GNSS receiver (Galileo/GPS) in a standard smartphone georeferences every canopy scan and leaf image, which is what makes repeatable, plot-level monitoring economically viable on plots of a few hundred square metres — no RTK antenna, no base station, no field hardware, on a device the grower already owns. Satellite Earth Observation plays a secondary, contextual role: Copernicus Sentinel-2 Level-2A imagery, free and open, provides vegetation-index and vigour maps on the larger peri-urban plots. The project established the boundary between the two: on plots below 0.5 ha a single Sentinel-2 pixel covers 100 m², so the satellite signal is mixed with buildings, trees and shadows and cannot drive agronomic decisions on its own; there, the outputs come entirely from the smartphone acquisition chain.
Current Status
The Kick-Start Activity is completed. Five pilot organisations in the Florence metropolitan area hosted nine plots across urban horticulture, olive and vineyard: the Tuscan Horticultural Society, the Municipality of Florence, Villa I Tatti (Harvard University), Fattoria di Maiano, and the intra-urban commercial farm Orti Dipinti, whose two urban horticultural micro-plots are the primary case study of the activity. In total 42 iAgro 3D scans were acquired: 30 olive trees, 10 vines and the 2 urban horticultural micro-plots; the olive and vine sites were retained to verify the super-resolution pipeline and the satellite-to-ground correlation. A dedicated statistical study tested the original hypothesis of fusing satellite indices with plant-level ground data and found no usable relationship on urban micro-plots: the correlation between the super-resolved Sentinel-2 leaf area index and the smartphone ground truth was negligible (Pearson r = -0.115, n = 42, explained variance below 4 %), while the satellite indices were almost perfectly intercorrelated with each other. That result is the central technical finding of the project. The fusion approach was therefore dropped and the architecture was consolidated as smartphone-first, with GNSS as the primary space asset and Sentinel-2 retained as a contextual layer above 0.5 ha. The iAgro application was redesigned around urban micro-plots, with on-device 3D digital-twin reconstruction, canopy metrics and prescriptions, and AI disease recognition — the function the pilot users rated as the most valuable. The business model was narrowed to a single testable B2B segment of urban and peri-urban commercial farms and cooperatives. Next steps are a full-scale demonstration, further urban commercial pilots, investigation of higher-resolution Earth-observation sources, and commercial roll-out of the urban configuration.
