Federated Farms

Interactive dashboards that translate Earth observation layers into field-scale insights.

Federated Farms image 1
Federated Farms image 2
Federated Farms image 3
Conceptual illustration of privacy-preserving federated learning where local farm models share updates with a central server, enabling collaborative model improvement without raw data exchange.
Flowchart of the stratified data split, client assignment, and processing hub used to aggregate model weights (e.g., FedAvg, WgtAvg) across contributors in large-scale soil spectroscopy experiments.
End-to-end workflow showing bare-soil composite generation in Google Earth Engine, storage via GeoServer, centralized model training, and decentralized fine-tuning across farms for clay and SOC mapping.
Privacy-PreserveSentinel-2Farm ManagementDecentralized AICNN
Earth Observation · Agriculture
2024–2026
University of São Paulo

Federated Farms

Federated Farms is a privacy-preserving federated learning pipeline for soil property prediction across farms, combining in-field and laboratory measurements with Earth observation covariates. It enables collaborative model training without centralizing raw data, supporting scalable deployment for growers, researchers, and partners.

Federated Farms trains shared soil prediction models across multiple farms while keeping raw data local. The approach supports collaboration, improves generalization, and aligns with real-world data ownership constraints.

Local training happens on each farm or partner node, with only model updates shared for aggregation. This enables iterative improvement without central data pooling and supports robust evaluation across sites.

Supports soil spectroscopy and EO covariates (e.g., Sentinel-2 bare-soil composites) to generate field-scale prediction maps and uncertainty layers suitable for interpretation and reporting.

Designed to plug into decision-support dashboards and farm boundary workflows, with export-ready outputs for stakeholders and reproducible reporting for research communication.

Key Features

  • Interactive maps for EO-derived soil/land indicators
  • Farm/field exploration with layer controls
  • Time-series context for monitoring change
  • Built for communication and decision support