Getting Your Head Around Bando Gz and the Rio Reforestation Effort
I got pulled into this space about a year ago when a colleague sent me a link to a GitHub repo I'd never heard of. I expected another overhyped open-source tool that delivered nothing. Instead, I found something that actually works, even if the documentation is rough around the edges. Bando Gz is essentially a geospatial data aggregation and visualization platform that was built to track and coordinate reforestation work across the Atlantic Forest biome, particularly in and around Rio de Janeiro. It pulls from satellite imagery, field survey data, and local community reports, then stitches them together into usable maps for NGOs, researchers, and municipal environmental agencies. The name comes from the Portuguese word for "crew" or "gang," which tracks with how grassroots the whole thing started. A handful of developers and ecologists in Rio put the first version together in 2022 after noticing that existing tools were either too expensive or didn't support the local coordinate systems properly. Here's how it actually works when you sit down to use it. The core platform runs on a web interface backed by PostGIS. You upload shapefiles or GeoJSON from field surveys, register GPS coordinates from handheld devices, and the system cross-references everything against a living basemap of deforestation alerts from INPE's DETER system. The output is a set of priority planting zones ranked by soil viability, proximity to existing forest fragments, and water table depth. For someone who's spent time in this domain, the integration with DETER is the part that actually matters. Most tools pull from MODIS or Landsat, which gives you a 16-day revisit window. DETER is near-real-time via PRODES follow-up, and that difference is the reason Bando Gz got traction faster than expected.
I ran into a specific edge case pretty early on that almost made me write it off entirely. Our team was working with field data collected on a garmin GPS unit that used the SIRGAS 2000 datum, but the platform's default projection was WGS 84. When I imported the shapefile, the coordinates were off by roughly 80 meters in the vertical component and about 3 meters horizontally. That doesn't sound like much until you're trying to match a planting site to a specific slope aspect for microclimate modeling. The workaround wasn't documented anywhere obvious. I had to reproject the data server-side using pyproj with the correct EPSG codes (EPSG:4674 for the datum, EPSG:31983 for the UTM zone relevant to the Rio area), then re-upload through the API endpoint rather than the web form. The web form appears to strip or misapply CRS metadata during ingestion. Once I started pushing data through the REST API with explicit CRS headers, the alignment issue disappeared completely. The counter-intuitive part that most people miss is that Bando Gz isn't really about the software. The software is competent but unpolished. The actual value is in the data pipeline and the community of local contributors who feed it. I've seen teams try to deploy it in other biomes without the field network, and the maps come out sterile and unusable because there's no ground truth feeding the model. The platform was designed around the idea that the people doing the planting are also the ones collecting the data. That's why the mobile app, which is still in beta, allows offline data entry with sync-on-connect. It's clunky but it's the only reason the tool works in areas with zero cell coverage, which is most of the terrain you'd be working in around the Serra do Mar corridor. There are real bottlenecks you need to plan around. The platform's rendering engine struggles when you load more than about 50,000 point features on a single map view. It doesn't crash, but it becomes effectively unusable with lag that makes navigation pointless. The workaround is to aggregate your data into tiles using the built-in clustering tool before visualizing, or to slice your work area into municipal-level subsets. Another issue is the API rate limit, which sits at 100 requests per minute for unauthenticated users and something closer to 500 for accounts verified through a partnering organization. If you're processing bulk uploads, batch them. Don't fire off individual POST requests.
The downside nobody talks about is the dependency on Brazilian federal data sources. When INPE's servers go down during budget cycles or political transitions, which happens more often than it should, the DETER feed cuts out and a lot of the priority zoning logic stops functioning. We lost access for about three weeks during a funding lapse in late 2024, and our entire planting schedule had to be reorganized around older satellite composites. If you're building a workflow on top of Bando Gz, you need a fallback data source. Sentinel Hub or even manual field verification can fill the gap, but it slows everything down significantly. If you want to start using it, the main repository and documentation are hosted on GitLab under the Bando Gz organization. The installation guide is adequate but assumes you're comfortable with Docker and have a Linux environment. There's no native Windows client. The mobile app is available through their website for both Android and iOS, though iOS support is more limited in terms of offline functionality. I'd recommend setting up a test instance with sample data from the Rio Metropolitan Region before committing any real fieldwork to the platform. The learning curve is maybe two days for someone who knows geospatial tools, or a week if you're coming from a non-technical background. The ROI kicks in pretty quickly after that, especially if you're coordinating multiple teams across a large area. It's not a perfect solution. No tool in this space is. But it's one of the few that was built by people actually working in the Atlantic Forest rather than designed in a lab and handed off to practitioners. That makes a measurable difference in how usable it ends up being.
Get the Full Details
