Amazon Rainforest Anomaly Detection

Hackathon Submission: Automated detection of deforestation and land-use changes in the Amazon rainforest using OpenAI GPT-4.1 vision models and satellite imagery.

Current Problem

Modis dataset is very coarse, it has a 500m accuracy. I need to get a much sharper dataset.

What It Does

This system automatically detects environmental changes in the Amazon rainforest by:

1. Fetching current satellite imagery from Google Earth Engine (Sentinel-2)
2. Comparing with historical data (MODIS 2020 baseline)
3. Using AI vision models to identify deforestation, construction, and land-use changes
4. Generating detailed reports with coordinates and analysis

Quick Demo

Key Features

• AI-Powered Analysis: GPT-4.1 vision model analyzes satellite imagery
• Real Satellite Data: Live Sentinel-2 imagery via Google Earth Engine
• Automated Detection: Identifies deforestation, roads, and urban expansion
• Geographic Validation: Analysis constrained to Amazon rainforest boundaries
• Batch Processing: Analyzes multiple locations automatically

Quick Start (5 Minutes)

Prerequisites

• Python 3.12+
• GitHub Token (free)
• Google Earth Engine account (free)

1. Clone & Setup

git clone https://github.com/ShaafPlayz/OpenAI-to-Z-challenge.git
cd OpenAI-to-Z-challenge
pip install -r requirements-minimal.txt

2. Configure APIs

Create .env file:

GITHUB_TOKEN=your_github_token_here

3. Google Earth Engine Setup

earthengine authenticate
# Follow the browser authentication

4. Run the Analysis

jupyter notebook "GEE and OpenAI Research.ipynb"
# Open the notebook and run all cells

That's it! The system will automatically:

• Generate random points in the Amazon rainforest
• Fetch current satellite imagery
• Compare with historical land cover data
• Use AI to detect and report anomalies

📁 Project Structure

OpenAI-to-Z-challenge/
├── 📄 .env                                   # Environment variables (local)
├── � .gitignore                             # Git ignore rules
├── � README.md                              # Project documentation
├── � environment.yml                        # Conda environment configuration
├── � package-lock.json                      # Node.js dependencies lock file
├── �📝 requirements.txt                       # Full Python dependencies
├── 📝 requirements-minimal.txt               # Essential Python dependencies
├── 📓 GEE and OpenAI Research.ipynb          # Main research notebook
├── 📁 .git/                                 # Git repository data
├── 📁 .ipynb_checkpoints/                    # Jupyter notebook checkpoints
├── 📁 amazonia_boundary_proposal/              # Amazon boundary shapefiles
│   ├── 🗺️ amazonia_polygons.dbf              # Shapefile attribute data
│   ├── 🗺️ amazonia_polygons.prj              # Projection information
│   ├── 🗺️ amazonia_polygons.sbn              # Spatial index binary
│   ├── 🗺️ amazonia_polygons.sbx              # Spatial index
│   ├── 🗺️ amazonia_polygons.shp              # Main shapefile geometry
│   ├── 🗺️ amazonia_polygons.shp.xml          # Metadata
│   └── 🗺️ amazonia_polygons.shx              # Shapefile index
├── 📁 AppEEARS/                             # NASA AppEEARS MODIS datasets
│   ├── 🗺️ MCD12Q1.061_LC_Prop2_doy2020001_aid0001.tif  # MODIS 2020 land cover
│   ├── 🗺️ MCD12Q1.061_LC_Prop2_doy2021001_aid0001.tif  # MODIS 2021 land cover
│   ├── 🗺️ MCD12Q1.061_LC_Prop2_doy2022001_aid0001.tif  # MODIS 2022 land cover
│   ├── 🗺️ MCD12Q1.061_LC_Prop2_doy2023001_aid0001.tif  # MODIS 2023 land cover
│   ├── 🗺️ MCD12Q1.061_QC_doy2020001_aid0001.tif        # MODIS 2020 quality control
│   ├── 🗺️ MCD12Q1.061_QC_doy2021001_aid0001.tif        # MODIS 2021 quality control
│   ├── 🗺️ MCD12Q1.061_QC_doy2022001_aid0001.tif        # MODIS 2022 quality control
│   └── 🗺️ MCD12Q1.061_QC_doy2023001_aid0001.tif        # MODIS 2023 quality control
├── 📁 AppEEARSjpg/                          # Processed MODIS imagery (empty)
├── 📁 Resources/                            # Additional resources
│   ├── 🗂️ amazon_biome_border.zip          # Original biome shapefile data
│   └── 🗂️ amazonia_boundary_proposal_Eva_2005 (1).zip  # Boundary proposal data
├── 📁 satimagery/                           # Satellite imagery data
│   ├── 🖼️ sentinel_rgb.jpg                  # Processed RGB satellite image
│   └── 🗺️ sentinel_rgb.tif                  # Raw satellite data (GeoTIFF)
├── 📁 Technology Testing (Scripts)/          # Python scripts for testing
│   ├── 📄 .env                              # Local environment variables
│   ├── 🐍 GoogleEarthEngine.py              # Google Earth Engine integration
│   ├── 🔐 Offical_OpenAI_Key.py             # API key management
│   ├── 🤖 OpenAI-o3.py                      # OpenAI o3 model experiments
│   └── 🤖 OpenAI.py                         # Main OpenAI API implementation
└── 📁 WorkFlow Testing (Jupyter Notebooks)/ # Jupyter notebooks
    └── 📓 openai-research.ipynb              # Additional research notebook

� How It Works

The Analysis Pipeline

1. Boundary Validation: Ensures coordinates are within Amazon rainforest
2. Satellite Data Fetch: Downloads current Sentinel-2 imagery (10m resolution)
3. Historical Lookup: Retrieves MODIS 2020 land cover classification
4. AI Analysis: GPT-4.1 compares current vs. historical imagery
5. Anomaly Detection: Reports deforestation, construction, or land-use changes

Key Functions

• find_and_log_anomalies(): Main detection pipeline
• getAmazoniaAnalysisPoints(): Generates valid Amazon coordinates
• get_modis_class_for_point(): Gets historical land cover data
• promptGPT(): Sends imagery to AI for analysis

Sample Output

Footprint ID: 1 at (-2.8234, -60.1234)
Historical (MODIS) Class: Evergreen Needleleaf Trees
Status: Anomaly Found - Construction activity detected

Setup Instructions

Prerequisites

• Python 3.12+
• Conda (Anaconda/Miniconda)
• GitHub Token (for GitHub Models access)
• Google Earth Engine Account
• OpenAI API Key (optional)

Environment Setup

1. Clone the Repository

   git clone https://github.com/yourusername/OpenAI-to-Z-challenge.git
   cd OpenAI-to-Z-challenge

2. Create and Activate Conda Environment

   # Create the environment from YAML
   conda env create -f environment.yml
   conda activate OpenAI-GoogleEngine
   
   # Or create manually
   conda create -n OpenAI-GoogleEngine python=3.12
   conda activate OpenAI-GoogleEngine

3. Install Dependencies

   # Option 1: Install minimal requirements (recommended)
   pip install -r requirements-minimal.txt
   
   # Option 2: Install full environment
   pip install -r requirements.txt

4. Environment Variables Create a .env file in the project root:

   GITHUB_TOKEN=your_github_token_here
   OPENAI_API_KEY=your_openai_api_key_here  # Optional

5. Google Earth Engine Authentication

   # First time setup
   earthengine authenticate
   
   # Initialize with your project
   # Replace 'your-project-id' with your actual GEE project ID

Usage

Main Research Notebook

# Launch Jupyter and open the main research notebook
jupyter notebook "GEE and OpenAI Research.ipynb"

Individual Components

OpenAI Integration

# Run OpenAI experiments
python "Technology Testing (Scripts)/OpenAI.py"

Google Earth Engine

# Test GEE functionality
python "Technology Testing (Scripts)/GoogleEarthEngine.py"

---

Ready to try it? Just run the Quick Start above! 🚀

Hackathon Submission - June, 2025