I led Crop Monitor from the hypothesis to the commercial product. The system combines satellite imagery, weather, and AI to flag emergence, crop type, harvest proximity, and harvest start — reducing field visits, anticipating crop failure, and mitigating fraud. I worked as technical lead, PM, and PO, managing the team and data quality at scale.

From 2017 to 2021, I mapped agriculture and planted forests across Brazil with Landsat imagery from 1984 onward. I am a coauthor, not the first author, of the MapBiomas network paper in Remote Sensing (2020, Souza Jr. et al.). Its roughly 1,800 citations measure adoption of that infrastructure, not a personal citation count. The data left the lab and now guides environmental monitoring, rural credit, and territorial management.

I led the satellite-image processing that fed the greenhouse-gas emission and removal matrices for the land-use sector (1994, 2002, 2005, 2010, and 2016), used in Chapter 2 of Brazil's Fourth National Communication to the UNFCCC — the country's official climate report.

I led the mapping of irrigation pivots and urban infrastructure in NexGenMap. The initiative produced the method MapBiomas later adopted for irrigation, and it supported detection of deforestation and land-use change with daily high-resolution imagery — input for conservation policy.

I published, as first author, the method for automatic detection of center-pivot irrigation systems (Remote Sensing, 2020): 99% precision, 88% recall, and more than 42,000 PlanetScope images. The work is cited as the basis for MapBiomas irrigation mapping and for water-demand studies, from Brazilian public policy to the Zambezi River Basin.

As technical lead, I delivered SIMFaz 2.0: three modules — Environmental and Social, Land Appraisal, and Agricultural Monitoring — so banks, insurers, and agribusiness companies can decide risk, value, and compliance of rural properties with satellite and weather data, on the timeline of the operation.

I delivered the COAMO portal from scratch: integrated databases, a soy-identification model, and a web platform. The cooperative can trace the origin of the grain against sustainability criteria, from the farm to the market.

I implemented the method for identifying irrigated and fertigated sugarcane areas in Brazil. This first-of-its-kind map became a reference for the sugar-energy sector and for water-resource management in the supply chain.

I brought the SSEBop model to the national territory on Google Earth Engine, with local adaptation and validation. The result is country-scale water monitoring — data for water management, irrigation, and policy.

I implemented the method that identified Brazil's main irrigation hubs. The map now supports public policy and strategic water management.

I led the analysis of the 2024 floods in the Jacuí and Guaíba basins, combining satellite imagery, thematic maps, and official data. The result quantified affected crops and storage structures — input for rapid loss assessment, insurance, and public response.

I delivered to ANA, Brazil's National Water and Sanitation Agency, an operational application for actual evapotranspiration at national scale. The data left the paper and entered the routine of the people who manage water in the country.

I integrated satellite imagery and INPE alerts into a near-real-time monitoring application, so the alert leaves the spreadsheet and reaches the screen of the people doing enforcement.

I write GeoDescomplicado to make geographic information systems easier to apply. The posts cover the methods I use: SIRGAS 2000 and WGS 84, what a projection actually conserves, Cloud Optimized GeoTIFF and GDAL, Sentinel-2 cloud masks in Google Earth Engine, elevation models, and satellite imagery.