🌍 Threshold - Urban Development & Service Gap Analysis Platform

Demo Video = https://www.youtube.com/watch?v=4w99ah-wVQk

Threshold is a comprehensive geospatial analysis platform that combines satellite remote sensing, machine learning, and real-time data processing to identify optimal locations for urban development and analyze service accessibility gaps. Built for the NASA Space Apps Challenge, this platform empowers urban planners with data-driven insights for sustainable city development.

🎯 Project Overview

Threshold addresses critical urban planning challenges by providing:

• 🏠 Housing Development Hotspots: Identifies prime vacant land suitable for residential development
• 🏥 Service Gap Analysis: Analyzes accessibility to essential services (healthcare, parks, food, transport)
• 🌡️ Environmental Monitoring: Real-time satellite data visualization for LST, NDVI, and air quality
• 🤖 ML-Powered Scoring: Custom machine learning model for hotspot prediction
• 📊 Real-time Analytics: Live data processing and visualization

---

AQI Calculation

PM 2.5 + PM 10+ NO2 + SO2 + CO = AQI

AQI = [(Ihigh − Ilow) / (Chigh − Clow)] × (C − Clow) + Ilow

🏗️ System Architecture

Architecture Diagrams/ Process Flow

Application Screenshots

---

🛰️ Satellite Data & Remote Sensing

ESA WorldCover Dataset

• Dataset: ESA WorldCover v200 (2021)
• Resolution: 10 meters
• Coverage: Global land cover mapping
• Platform: Sentinel-1 and Sentinel-2 satellites
• Processing: Google Earth Engine integration
• Land Cover Classes: 11 classes (Tree cover, Shrubland, Grassland, Cropland, Built-up, Bare/sparse vegetation, Snow/ice, Water bodies, Wetland, Mangroves, Moss/lichen)
• Primary Focus: Class 60 (Bare/sparse vegetation) for vacant land detection

NASA GIBS Satellite Layers

🌡️ Land Surface Temperature (LST)

• Satellite: MODIS Terra
• Dataset: MODIS_Terra_Land_Surface_Temp_Day
• Resolution: 1 km (GoogleMapsCompatible_Level7)
• Temporal: Daily
• Usage: Heat island analysis, climate monitoring

🌱 Normalized Difference Vegetation Index (NDVI)

• Satellite: MODIS Terra
• Dataset: MODIS_Terra_NDVI_8Day
• Resolution: 250m (GoogleMapsCompatible_Level9)
• Temporal: 8-day composite
• Usage: Vegetation health, urban green space analysis

🌫️ Air Quality Layers

• Carbon Monoxide: AIRS L3 (500hPa, 2km resolution)
• Aerosol Optical Depth: MODIS Aqua (3km resolution)
• Ozone: OMI Total Column (Level 6)
• Nitrogen Dioxide: OMI Tropospheric Column (Level 6)
• Sulfur Dioxide: OMI Planetary Boundary Layer (Level 6)
• PM2.5/PM10: MERRA2 Dust Surface Mass Concentration (Monthly)

---

🤖 Machine Learning & Custom Dataset

Custom Hotspot Scoring Dataset

Our platform features a unique, synthetic dataset specifically designed for urban hotspot analysis:

Dataset Specifications

• Size: 15,000 samples
• Features: 8 input variables
• Target: Hotspot Score (0.0 - 1.0)
• Generation: Sophisticated correlation modeling based on real urban planning principles

Input Features

1. AQI (Air Quality Index): 0-500 scale
2. Population Density: People per km²
3. Distance to Hospital: km
4. Distance to School: km
5. Distance to Airport: km
6. Distance to Bus Station: km
7. Distance to Railway: km
8. Distance to Mall: km

Advanced Scoring Algorithm

# Sophisticated scoring with area type correlation
def calculate_advanced_hotspot_score(aqi, pop_density, distances, area_type):
    # Environmental factor (40% weight)
    env_score = calculate_environmental_score(aqi)
    
    # Accessibility factor (35% weight)  
    access_score = calculate_accessibility_score(distances)
    
    # Population factor (25% weight)
    pop_score = calculate_population_score(pop_density, area_type)
    
    # Area-specific adjustments
    area_multiplier = get_area_multiplier(area_type)
    
    return weighted_combination(env_score, access_score, pop_score, area_multiplier)

Machine Learning Model

• Algorithm: Random Forest Regressor (optimized via GridSearchCV)
• Performance: R² > 0.95, RMSE < 0.05
• Features: StandardScaler preprocessing
• Deployment: Real-time inference with fallback scoring
• Files: hotspot_model.pkl, scaler.pkl

---

🌬️ Air Quality Index (AQI) Calculation

Real-time AQI Processing

• Data Source: Open-Meteo Air Quality API
• Standards: EPA AQI breakpoints
• Pollutants: PM2.5, PM10, O₃, NO₂, SO₂, CO
• Calculation: Real-time conversion from μg/m³ to AQI scale
• Temporal: 8-hour rolling averages for O₃ and CO

AQI Calculation Process

# EPA-compliant AQI calculation
def calculate_aqi(concentration, pollutant):
    # Apply EPA breakpoints
    for c_low, c_high, i_low, i_high in AQI_BREAKPOINTS[pollutant]:
        if c_low <= concentration <= c_high:
            return ((i_high - i_low) / (c_high - c_low)) * (concentration - c_low) + i_low

Unit Conversions

• O₃: μg/m³ → ppb (×0.5)
• NO₂: μg/m³ → ppb (×0.532)
• SO₂: μg/m³ → ppb (×0.382)

---

🏙️ Service Gap Analysis

Service Categories

1. 🌳 Parks & Recreation: Parks, playgrounds, recreational facilities
2. 🛒 Food Access: Supermarkets, grocery stores, restaurants
3. 🏥 Healthcare: Hospitals, clinics, pharmacies
4. 🚌 Public Transport: Bus stations, train stations, airports

Distance Thresholds (Urban Planning Standards)

• Parks: Good (≤0.8km), Fair (≤2.0km), Poor (>5.0km)
• Food: Good (≤1.5km), Fair (≤5.0km), Poor (>10.0km)
• Healthcare: Good (≤3.0km), Fair (≤10.0km), Poor (>20.0km)
• Transport: Good (≤0.5km), Fair (≤1.5km), Poor (>3.0km)

Data Sources

• Primary: Google Places API (20+ place types per service)
• Fallback: OpenStreetMap Overpass API
• Distance: Google Distance Matrix API + Haversine fallback

---

🛠️ Technical Stack

Backend (FastAPI)

# Core Technologies
FastAPI 0.104.1          # High-performance web framework
Motor 3.3.2              # Async MongoDB driver  
Google Earth Engine      # Satellite data processing
Pydantic 2.5.0          # Data validation
Scikit-learn 1.3.0      # Machine learning
Pandas 2.0+             # Data processing
NumPy 1.24+             # Numerical computing

Frontend (React + TypeScript)

// Core Technologies
React 19.1.1            // UI framework
TypeScript 5.8.3        // Type safety
Google Maps API         // Interactive mapping
Google Map React 2.2.5  // React integration
Vite 7.1.6             // Build tool
TailwindCSS 4.1.13     // Styling

Database & Infrastructure

• Database: MongoDB (geospatial indexing)
• Authentication: JWT with bcrypt hashing
• APIs: RESTful with OpenAPI documentation
• Deployment: Docker containerization ready

---

🚀 Features

🎯 Core Functionality

• Area of Interest (AOI) Selection: Interactive polygon drawing on Google Maps
• Vacant Land Detection: Real-time ESA WorldCover processing via Google Earth Engine
• Hotspot Scoring: ML-powered development suitability analysis
• Service Gap Identification: Automated accessibility analysis
• Environmental Monitoring: Multi-layer satellite data visualization

📊 Advanced Analytics

• Real-time Processing: Sub-30 second analysis for most AOIs
• Multi-source Integration: Satellite + API + ML model fusion
• Geospatial Caching: MongoDB-based result caching for performance
• Comprehensive Reporting: Detailed breakdowns with recommendations

🎨 User Interface

• Interactive Mapping: Google Maps with custom overlays
• Layer Control: Toggle between LST, NDVI, air quality layers
• Live Search Results: Real-time API status visualization
• Responsive Design: Mobile-friendly interface
• Dark/Light Themes: Customizable UI themes

---

📁 Project Structure

Threshold/
├── Backend/                 # FastAPI application
│   ├── api/routes/         # API endpoints
│   │   ├── auth.py         # Authentication
│   │   ├── vacant_land.py  # Vacant land analysis
│   │   ├── service_analysis.py # Service gap analysis
│   │   └── aqi.py          # Air quality calculations
│   ├── services/           # Business logic
│   │   ├── gee_service.py  # Google Earth Engine
│   │   ├── esa_worldcover_service.py # Satellite processing
│   │   ├── hotspot_scoring_service.py # ML scoring
│   │   └── service_analysis_service.py # Service analysis
│   ├── models/             # Pydantic models
│   └── database/           # MongoDB connection
├── Frontend/               # React application
│   ├── src/components/     # React components
│   │   ├── Dashboard.tsx   # Main dashboard
│   │   ├── Maps.tsx        # Satellite layer visualization
│   │   └── Hotspots.tsx    # Hotspot analysis interface
│   └── public/             # Static assets
├── Model/                  # ML model training
│   ├── dataset_generator.py # Custom dataset creation
│   ├── train_hotspot_model.py # Model training pipeline
│   ├── hotspot_model.pkl   # Trained model
│   └── scaler.pkl          # Feature scaler
└── README.md               # This file

---

🔧 Installation & Setup

Prerequisites

• Node.js 18+ (Frontend)
• Python 3.9+ (Backend)
• MongoDB 5.0+ (Database)
• Google Earth Engine Account
• Google Maps API Key

Backend Setup

cd Backend
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt

# Configure environment variables
cp config.example.env .env
# Edit .env with your API keys and database credentials

# Run the server
python main.py

Frontend Setup

cd Frontend
npm install
npm run dev

Environment Variables

# Google Earth Engine
GEE_SERVICE_ACCOUNT_KEY_PATH=path/to/service-account.json
GEE_SERVICE_ACCOUNT_EMAIL=your-service-account@project.iam.gserviceaccount.com

# Google Maps API
GOOGLE_MAPS_API_KEY=your-google-maps-api-key

# Database
MONGODB_URI=mongodb://localhost:27017
DATABASE_NAME=threshold

# Security
SECRET_KEY=your-secret-key

---

📊 API Documentation

Core Endpoints

Vacant Land Analysis

POST /api/vacant-land/
Content-Type: application/json

{
  "aoi_geometry": {
    "type": "Polygon",
    "coordinates": [[[lng, lat], ...]]
  },
  "min_area_m2": 5000,
  "max_polygons": 100
}

Service Gap Analysis

POST /api/service-analysis/
Content-Type: application/json

{
  "aoi_bounds": {
    "north": 23.1, "south": 22.9,
    "east": 72.7, "west": 72.3
  },
  "service_types": ["parks", "food", "healthcare", "transport"],
  "grid_resolution": 2.0
}

AQI Calculation

GET /api/aqi/calculate?latitude=23.0225&longitude=72.5714&date=2024-01-15

---

🎯 Use Cases

Urban Planners

• Identify optimal locations for new residential developments
• Assess service accessibility gaps in existing neighborhoods
• Monitor environmental conditions affecting development suitability
• Generate data-driven recommendations for infrastructure investment

Government Agencies

• Strategic urban development planning
• Public service optimization
• Environmental impact assessment
• Resource allocation prioritization

Real Estate Developers

• Site selection for residential projects
• Market analysis based on accessibility metrics
• Environmental risk assessment
• Investment opportunity identification

Researchers & NGOs

• Urban sustainability studies
• Environmental justice analysis
• Public health accessibility research
• Climate change impact assessment

---

🌟 Innovation Highlights

🔬 Technical Innovations

1. Custom ML Dataset: First-of-its-kind synthetic urban hotspot dataset with realistic correlations
2. Multi-source Fusion: Integration of satellite data, real-time APIs, and ML predictions
3. Real-time Processing: Sub-30 second analysis using Google Earth Engine
4. Adaptive Fallbacks: Robust system with multiple data source fallbacks

🎯 Urban Planning Impact

1. Evidence-based Decision Making: Quantitative metrics for development planning
2. Service Equity Analysis: Identifies underserved areas requiring attention
3. Environmental Integration: Considers air quality and climate factors
4. Scalable Solution: Applicable to any global location

---

🔮 Future Enhancements

Short-term (3-6 months)

• Population Prediction: Integrate population growth forecasting
• Economic Indicators: Add economic accessibility metrics
• Mobile Application: Native mobile app development
• Batch Processing: Large-scale regional analysis capabilities

Long-term (6-12 months)

• 3D Visualization: Three-dimensional urban modeling
• Temporal Analysis: Historical trend analysis and future projections
• AI-powered Recommendations: Advanced ML models for planning suggestions
• Integration APIs: Third-party planning software integration

---

🙏 Acknowledgments

• NASA GIBS for providing free satellite data access
• ESA WorldCover for high-resolution land cover datas
• Google Earth Engine for cloud-based geospatial processing
• Open-Meteo for air quality data API
• NASA Space Apps Challenge for the inspiration and platform

Threshold - Where data meets urban planning