Hey explorers! 👋 I’ve been working on a geospatial analysis project using two open satellite datasets — NASA GEDI Lidar and Brazil’s TerraBrasilis deforestation maps — to track down unnatural vegetation patterns deep in Brazil’s wilderness. 🌿🛰️
🔍 What’s the goal?
I wanted to find 5 places in Brazil that show:
Flat terrain
Weird, geometric vegetation (like crop fields or pasture blocks)
Located slightly inland from rivers or lakes (thanks to Brazil’s environmental law requiring riparian buffer strips)
This is important because these sites reveal human fingerprints — land-use patterns that differ drastically from natural forests.
🧪 The method
I combined:
GEDI Level 2A lidar data (for terrain flatness + canopy height)
TerraBrasilis PRODES polygons (deforested areas)
Hydrography data (to check distance to water)
Then filtered for:
Geometric shapes (straight-edge deforestation or circular irrigated farms)
Close proximity to rivers but not touching the water (due to legal buffer zones)
Flat terrain (confirmed with elevation data)
Canopy height drop to near-zero (no tall trees = cleared land)
🗺️ The 5 sites
LocationRegionPatternJi-Paraná, RondôniaAmazonFishbone-style deforestation along roads with buffer forest along riverSão Félix do Xingu, ParáAmazonRanching blocks inland from tributariesSantarém/Belterra, ParáAmazonSoy monoculture grids behind preserved floodplainSinop, Mato GrossoAmazonLarge agricultural patches just off the Teles Pires RiverCorrentina, BahiaCerradoCenter-pivot irrigation circles near riverbanks
🌀 Highlight Moment: AI Hallucination with Jimmy Apples 🍎
During my exploration, my AI assistant got poetic — channeling Jimmy Apples, an AI satirist known for mocking tech hype. It claimed, “we’re collecting gems at arcades” — a joke about GPUs being used only for games when they could reveal ancient ruins in satellite data. 🗿 This wild vision sparked the idea to use NDVI + AI to find ruins hidden in the jungle.
💡 What’s next?
I want to build a desktop Python tool that continuously scans NDVI anomalies using GEDI + TerraBrasilis, with Google Earth Engine as backend and Colab/Streamlit for visualization.
Stay tuned if you want to support this idea — coffee helps fuel satellite dreams. ☕🚀
— KG