Module 7: Seeing the Unseen: Fusion and Time Travel

By Gopi Krishna Tummala

Beyond the Visible

If you rely on visible light, clouds are the enemy. They block your view, making satellite data useless. But what if you could use a completely different type of light to see through the clouds? This is where Data Fusion comes in, particularly combining optical data with Synthetic Aperture Radar (SAR). SAR sends its own microwave signal and measures the reflection, which easily penetrates clouds, smoke, and darkness, allowing for 24/7, all-weather mapping.

Furthermore, by analyzing images of the same location over weeks, months, or years (Time-Series Analysis), we can track the rate of deforestation, monitor urban sprawl, or predict crop yields—literally watching the planet breathe.

💡 The Math Hook: The Normalized Difference Vegetation Index (NDVI)

The simplest and most beautiful example of this multi-source power is the NDVI. Healthy plants strongly absorb red light (for photosynthesis) and strongly reflect Near-Infrared (NIR) light. Bare ground or water does neither.

The NDVI is a simple ratio that exploits this difference:

$NDVI = \frac{\text{NIR} - \text{Red}}{\text{NIR} + \text{Red}}$

The resulting value ranges from -1 to +1, with values close to +1 indicating dense, healthy vegetation. This simple formula is the backbone of modern agricultural monitoring and is a classic example of using mathematics to extract profound, unseen information from light.

Interpretation:

• NDVI > 0.6: Dense, healthy vegetation
• NDVI 0.2-0.6: Sparse vegetation or stressed crops
• NDVI 0.0-0.2: Bare soil, rock, or urban areas
• NDVI < 0: Water, clouds, or snow

Key Topics

Multi-Spectral Indices

Using Multiple Color Bands to Highlight Specific Features:

Multi-spectral indices combine different wavelength bands to highlight specific surface properties that aren’t visible in individual bands.

Other Important Indices:

1. NDWI (Normalized Difference Water Index): $NDWI = \frac{Green - NIR}{Green + NIR}$

   • Highlights water bodies

2. NDBI (Normalized Difference Built-up Index): $NDBI = \frac{SWIR - NIR}{SWIR + NIR}$

   • Identifies urban/built-up areas

3. EVI (Enhanced Vegetation Index):

   • Improved version of NDVI
   • Better in high-biomass areas
   • Reduces atmospheric effects

Data Fusion (Optical + SAR)

Combining Optical Imagery with Synthetic Aperture Radar (SAR):

Different sensors provide complementary information. Fusion combines their strengths.

Optical vs. SAR:

Optical (Visible/Infrared):

• ✅ Rich spectral information
• ✅ Intuitive interpretation
• ❌ Blocked by clouds
• ❌ Requires daylight

SAR (Radar):

• ✅ Works day/night
• ✅ Penetrates clouds
• ✅ Sensitive to surface structure and moisture
• ❌ More complex interpretation
• ❌ Speckle noise

Fusion Strategies:

1. Pixel-Level Fusion:

   • Combine at raw pixel level
   • Example: Pan-sharpening (merge high-res pan with multi-spectral)

2. Feature-Level Fusion:

   • Extract features from each sensor
   • Combine features for classification
   • Example: Optical texture + SAR backscatter

3. Decision-Level Fusion:

   • Classify each sensor independently
   • Combine classification results
   • Voting or probability fusion

Applications:

• Cloud-free mapping (SAR fills optical gaps)
• Crop monitoring (optical for species, SAR for structure)
• Flood mapping (SAR detects water, optical provides context)
• Urban mapping (combine building detection from both)

Long-Term Change Detection Using Landsat and Sentinel Time-Series

Time-Series Analysis:

Time-series analysis tracks changes over days, months, or years to understand dynamic processes.

Key Platforms:

1. Landsat (NASA/USGS):

   • 16-day revisit (30m resolution)
   • 50+ year archive (since 1972)
   • 11 spectral bands

2. Sentinel-2 (ESA):

   • 5-day revisit (10-60m resolution)
   • Free and open data
   • 13 spectral bands

3. MODIS (NASA):

   • Daily global coverage
   • 250m-1km resolution
   • 36 spectral bands

Time-Series Applications:

1. Deforestation Monitoring:

   • Track forest loss over time
   • Detect illegal logging
   • Measure carbon sequestration

2. Urban Growth:

   • Monitor city expansion
   • Track infrastructure development
   • Plan for future growth

3. Agricultural Monitoring:

   • Crop growth stages
   • Yield prediction
   • Irrigation management

4. Disaster Response:

   • Before/after comparisons
   • Damage assessment
   • Recovery tracking

Analysis Techniques:

• Change Detection: Compare images from different dates
• Phenology Analysis: Track seasonal vegetation cycles
• Trend Analysis: Identify long-term patterns
• Anomaly Detection: Find unusual events

The Power of Specialized Sensors

Advanced DEM Generation: Interferometric Synthetic Aperture Radar (InSAR)

InSAR uses phase differences between two SAR images to measure very precise vertical changes.

How InSAR Works:

1. Two SAR Images: Captured from slightly different positions
2. Interferogram: Phase difference between images
3. Phase Unwrapping: Convert phase to height
4. DEM Generation: Create elevation model

Key Advantages:

• Very high vertical accuracy (cm to mm level)
• Works through clouds
• Large area coverage
• Can measure ground deformation

Applications:

1. Topographic Mapping:

   • Generate high-accuracy DEMs
   • Fill gaps in optical DEMs
   • Map areas with persistent cloud cover

2. Ground Deformation Monitoring:

   • Subsidence (sinking ground)
   • Landslide detection
   • Volcanic activity
   • Earthquake effects

3. Glacier Monitoring:

   • Ice flow velocity
   • Thickness changes
   • Climate change impacts

Differential InSAR (DInSAR):

• Measures changes between two time periods
• Detects mm-level ground movement
• Critical for infrastructure monitoring

Challenges:

• Phase unwrapping complexity
• Atmospheric effects
• Temporal decorrelation
• Requires stable targets

Combining multiple data sources and time-series analysis unlocks powerful monitoring capabilities. In the final module, we’ll explore real-world applications and future trends.