Field Inventory Meets Satellite Data: Validating SMFM Map Products
Why Field Reference Data Matters for SMFM Map Products
Satellite data gives SMFM map products their broad spatial coverage, but satellites cannot confirm on their own whether a pixel labeled "flooded" or "bare soil" matches what is actually happening on the ground. That confirmation comes from field-based reference data collected by trained survey teams before any map product is released.

The validation workflow on smfm-project.com pairs every candidate map layer with an independent field dataset. This reference data is collected separately from the model inputs, so it can serve as an honest judge of accuracy rather than simply repeating the same assumptions.
How Field Teams Collect Reference Points
Field crews follow a stratified sampling plan. They visit locations selected to represent the full range of land cover, elevation, and satellite signal conditions found in the project area, not just the easy-to-reach sites.
At each site, teams record coordinates, land cover class, surface condition, and a timestamp. Photographs and short notes are attached to every record so that later reviewers can understand the context behind each observation.
The collection process follows a fixed sequence:
- Select sampling points from the stratified design.
- Navigate to each point and confirm the exact location.
- Record the observed class and supporting evidence.
- Upload the record to the validation database.
- Flag any point that is inaccessible or ambiguous for review.
Checking Accuracy Before Release
Once field data is uploaded, analysts compare each reference point against the corresponding map pixel. This produces a confusion matrix, which shows how often each class is correctly identified and where errors cluster.
Key accuracy measures include overall accuracy, producer accuracy, and user accuracy. If a class falls below the project threshold, the map product is sent back for revision rather than released.
| Metric | What It Measures | Typical Threshold |
|---|---|---|
| Overall accuracy | All correctly classified points | 85 percent |
| User accuracy | Reliability of a mapped class | 80 percent |
| Producer accuracy | How well a real class is captured | 80 percent |
When disagreements appear, the team reviews the field photos and satellite imagery side by side. Some mismatches reveal genuine map errors, while others reflect timing differences between the field visit and the satellite overpass.
Common Challenges in Field Validation
Validation is rarely simple. Access restrictions, cloud cover on the acquisition date, and rapid land cover change can all introduce uncertainty into the comparison.
To manage this, the project uses several safeguards:
- Multiple field visits to the same area when conditions change quickly.
- GPS accuracy checks at every point.
- Independent review of at least ten percent of all records.
- Clear documentation of any point excluded from analysis.
"A map product is only as trustworthy as the field evidence behind it."
From Validation to Publication
Only after field validation is complete does a map product move to publication. Each released layer carries an accuracy statement summarizing the reference data, sample size, and measured performance.
This transparent process lets users on smfm-project.com understand both the strengths and the limits of every map. Field inventory and satellite data are not competitors; together they produce map products that can be defended with real observations.
