Case Study · Published applied research
Brazil farm-scale EO soil mapping
Farm-scale Sentinel-2 soil-property mapping with local adaptation, validation, and uncertainty across 50 farms and 7,831 observations.
Problem
Regional soil models often lose reliability when transferred to individual farms. The work examined how satellite covariates and local observations could support farm-scale soil organic carbon and clay mapping while making uncertainty and adaptation needs explicit.
Role
Giannis led and supported technical workflow development for Earth observation covariates, farm-scale soil-property modelling, local adaptation, validation, uncertainty analysis, figures, and publication-ready reporting within the research collaboration.
Data and Methods
- 7,831 georeferenced soil observations across 50 Brazilian farms
- Sentinel-2 bare-soil information and spatial covariates
- Farm-scale soil organic carbon and clay modelling
- Local adaptation and comparison logic
- Validation, spatial error review, and uncertainty communication
Deliverables
- Farm-scale SOC and clay maps
- EO covariate-processing workflow
- Model comparison and local-adaptation evidence
- Validation and uncertainty outputs
- Publication figures, methods, and reproducible analysis components
Evidence and Status
Published applied research based on 50 farms and 7,831 georeferenced observations. The evidence supports technical capability in farm-scale EO soil mapping; it is not described as a commercial client engagement.
Institutional Relevance
The workflow is relevant to soil-monitoring programmes, digital soil mapping, sampling prioritization, land-condition baselines, and MRV design where regional models must be checked against local conditions.
Limits and Next Steps
Transferability depends on soil diversity, observation density, bare-soil visibility, sensor timing, and local calibration. Maps should be interpreted with field evidence and uncertainty rather than as guaranteed property estimates.