Curated Resources
The top 20 sources for building a deforestation detection platform
Free, open-access, 55 chapters. The single most comprehensive resource covering remote sensing fundamentals through deforestation detection using GEE. Covers classification, accuracy assessment, forest degradation, and SAR.
The authoritative protocol for accuracy assessment and area estimation of land-change maps. Stratified random sampling + area-weighted estimators. Non-negotiable for credibility with government and scientific clients.
The foundational paper behind the Global Forest Change dataset — 2.3M km² of forest loss mapped at 30m. Your baseline benchmark and label source.
Spatial validation methodology mapping where your model can be trusted vs where it's extrapolating. The CAST R package implements spatial CV, forward feature selection, and AOA. Concepts transfer directly to Python/sklearn.
How GLAD-L, GLAD-S2, RADD, and DIST-ALERT are integrated into a single monitoring platform. Documents resolutions, latency, limitations, and coverage. Your competitive landscape reference.
Spatial indexing (GiST), ST_ geometry functions, raster support, and performance tuning. The standard spatial database for production geospatial systems.
JavaScript and Python API guides, data catalog (800+ datasets), and commercial licensing terms. Your current platform — understand both capabilities and business constraints.
The open standard for searchable satellite imagery catalogs. Foundation of cloud-native geospatial architecture and your alternative to GEE lock-in.
Authoritative reference on Sentinel-2's 13 bands, L1C vs L2A processing levels, spatial resolution, and revisit frequency. Official source on your primary optical sensor.
Landsat Collection 2 surface reflectance documentation plus authoritative formulas for NDVI, EVI, NBR, NDMI. Historical baseline sensor spanning 50+ years.
Free, self-paced, certificate-bearing course on satellite remote sensing fundamentals. Available in English and Spanish. Government-grade, no SEO fluff.
Free, open, university-grade course using Python, GeoPandas, and Shapely. Project-oriented with hands-on exercises. Best structured GIS course for Python developers.
Operational forest monitoring training for tropical developing countries. National inventories, change detection, REDD+ reporting. Exactly your customer's context.
Correct usage of cross-validation, metrics, and permutation_importance (use instead of impurity-based feature_importances_, which is biased). The pitfalls section is essential reading.
Python library for reading/writing GeoTIFFs including COGs. The engine under your raster processing pipeline.
Structured intro to GIS concepts: vector/raster, CRS, projections, spatial analysis. Use QGIS as a debug/visualization tool, not the primary workflow.
The clearest explanation of CRS, spatial operations, and geometric transformations in any book. Even as a Python developer, the conceptual chapters are unmatched.
Global 30m forest loss/gain 2000-present, updated annually. Primary benchmark dataset and label source for training data.
Annual land cover maps for South America including Ecuador, with vegetation/deforestation classes. Regional ground-truth and label source for Andean forest.
Dynamic raster tile server for Cloud-Optimized GeoTIFFs. Serves detection results as map tiles without pre-rendering. The modern approach to geospatial visualization.
Comprehensive 8-unit reference covering project scaffolding, raster vs vector, CRS, data formats, data acquisition, raster anatomy, data quality decisions, and network topology. Built around real Cuenca data layers. Contains testable facts for each unit.