Skip to content
Code by Sergio
PTENES

GeoCarbo

From satellite to a carbon estimate that shows how it was calculated.

Working prototypeMVP stage

Back to category
GeoCarbo

The problem

“The market needs to trust the number before it can trust the credit.”

Measuring carbon in the field is expensive and slow.

In the Caatinga, seasonality, leaf drop and exposed soil throw off generic satellite estimates. GeoCarbo aims for a first reading, automated and transparent, ahead of inventory and certification.

How it works

Sentinel-2 imagery and equations published for the Caatinga estimate the vegetation carbon of each property.

  1. 01PropertyRegistering the area
  2. 02PolygonKML or GeoJSON
  3. 03Sentinel-2Recent scenes, clouds removed
  4. 04Vegetation indicesOver the scene composite
  5. 05BiomassA regression published for the Caatinga
  6. 06Carbon / CO₂eDeclared coefficients
  7. 07ReportResult and PDF

  • Sentinel-2

    Open imagery via Copernicus

  • 10 m

    Resolution of the main bands used

  • up to5 scenes

    Median temporal composite

Published method.Declared limits.

  1. Biomass
  2. Carbon
  3. CO₂e

A regression published for the Caatinga estimates biomass; declared coefficients convert it into carbon and CO₂e.

The system doesn’t hide it when the data goes beyond the model.

  1. Calibrated range

    The model holds for the NDVI range it was calibrated on.

  2. Extrapolation warning

    Outside that range, the estimate is flagged.

  3. Above ground only

    The calculation covers above-ground biomass only.

  4. No field validation

    A preliminary estimate, not yet compared against field measurements.

Technology

Architecture

  1. User
  2. Frontend
  3. FastAPI
  4. Celery / Redis
  5. Copernicus
  6. Processing
  7. Supabase
  8. Result / PDF
Frontend
ReactViteLeaflet
Backend
PythonFastAPI
Processing
CeleryRedisrasterionumpy
Satellite
Sentinel-2Copernicus
Data
SupabasePostgreSQL
Reports
ReportLab
Infrastructure
VPS LinuxNginx

Working prototype.MVP stage.

The pipeline already

  • Takes in the property
  • Processes Sentinel-2
  • Computes the estimate
  • Stores the result
  • Generates the PDF

It isn’t yet

  • A certification platform
  • A product validated by certifiers
  • A complete dMRV
  • A field-validated system
  • A mature commercial solution
Registering the property
Completed reports, with the PDF to download

The architecture already has a slot for models trained on field data.

My role

Co-Founder & CPO at SECCO, working directly on the backend and on GeoCarbo’s evolution.

  1. Backend architecture
  2. API
  3. Processing
  4. Polygon integration
  5. Satellite pipeline
  6. Persistence
  7. Reports
  8. Deployment

Context: SECCO’s incubation at Porto Digital and its participation in Inova Caatinga.