Forestry engineer and carbon project auditor working where remote sensing, GIS and voluntary carbon markets meet.
- I audit AFOLU carbon projects (REDD+, ARR) under VCS, CCB and BioCarbon standards.
- I build open, reproducible tools to check the spatial evidence behind carbon projects: project areas, forest loss, reference regions and monitoring plots.
- I explore how AI agents can run real geospatial analysis through the Model Context Protocol (MCP), with the GIS checks an auditor expects.
Portfolio: albertolozada.github.io · LinkedIn: alberto-lozada-silva
| Project | What it does | Status |
|---|---|---|
| carbon-geo-mcp | MCP server with audit-oriented GIS tools for carbon projects: a CRS plan before measuring, area by stratum, declared vs. calculated areas, overlaps, plot checks, forest loss, satellite imagery availability (STAC) and check maps | v0.2 available |
| deforestation-hotspots-colombia | Where deforestation clusters in the Colombian Amazon and Chocó (Moran's I, Getis-Ord Gi* via gis-mcp), and how past hot spots improve a VT0007-style (VM0048) risk map | Available |
| ndvi-sentinel2-redd | Dry-season Sentinel-2 NDVI 2019-2025 in Tinigua National Natural Park (Colombian Amazon), crossed with ESA WorldCover and checked against Hansen GFC; public COGs read with terrio and cross-checked with GDAL | Available |
| geoqa-precheck | Reproducible pre-checks of the GIS layers, Excel workbooks and PDF reports of carbon projects: CRS, geometry, declared vs. calculated areas, formulas, parameters, monitoring periods and reported reductions, with JSON, Markdown and CSV reports | Available |
| qgis-audit-atlas | Automated QGIS atlas: one evidence map per stratum or plot | Planned |
Every project starts from the geo-project-template, so all of them share the same structure and a "run it yourself" notebook.
- The agent explores, the script confirms. Every result in a report comes from code that anyone can rerun.
- CRS first. Areas and distances are always computed in a metric or equal-area CRS, and every result states which one.
- Say what is missing. Each analysis reports data completeness and limitations, not only the answer.
- Public data only. Case studies use open datasets and public project documents.
Python · GeoPandas · Shapely · PyProj · Rasterio · QGIS · Sentinel-2 / Landsat · STAC · MCP
En español
Soy ingeniero forestal y auditor de proyectos de carbono. Trabajo en la intersección entre teledetección, SIG y mercados voluntarios de carbono.
- Audito proyectos de carbono AFOLU (REDD+, ARR) bajo los estándares VCS, CCB y BioCarbon.
- Desarrollo herramientas abiertas y reproducibles para verificar la evidencia espacial de los proyectos: áreas de proyecto, pérdida de bosque, regiones de referencia y parcelas de monitoreo.
- Exploro cómo los agentes de IA pueden ejecutar análisis geoespaciales reales mediante el Model Context Protocol (MCP), con los controles SIG que exige una auditoría.
Visita el portfolio en albertolozada.github.io, disponible en español e inglés, o contáctame en LinkedIn.