Deforestation Monitoring with Satellite Data: How We Track Forest Loss Globally

In 2019, satellite data showed that Brazil lost 1.2 million hectares of primary forest — an area roughly the size of Connecticut — in a single year. That number didn't come from ground surveys or aircraft reconnaissance. It came from algorithms processing billions of Landsat pixels, detecting where forest canopy disappeared between one year and the next.

Satellite-based deforestation monitoring has become one of the most impactful applications of remote sensing, directly influencing policy, enforcement, and international climate agreements.

How Satellites Detect Forest Loss

Optical Detection (Sentinel-2, Landsat)

Forest clearance produces dramatic spectral changes:

• NDVI drops from 0.7-0.9 (dense canopy) to 0.1-0.3 (bare soil or sparse regrowth)
• NIR reflectance decreases as leaf canopy is removed
• SWIR reflectance increases as soil is exposed (previously shaded by canopy)
• Red reflectance increases as chlorophyll absorption disappears

A single cloud-free image pair (before and after) is often sufficient to map clearance events. The spectral contrast between forest and cleared land is among the strongest change signals in remote sensing.

SAR Detection (Sentinel-1)

Forest clearance changes the radar scattering mechanism:

• Standing forest: Strong volume scattering from canopy → moderate-high backscatter in VH
• Cleared land: Surface scattering from soil → lower VH backscatter (typically 3-6 dB decrease)
• Coherence change: Forest always has low coherence at C-band; after clearance, bare soil maintains higher coherence

SAR detection works through clouds — critical in the tropics where persistent cloud cover can obscure optical views for weeks during the wet season, which is often when illegal deforestation accelerates (less chance of being spotted from aircraft).

Combined Optical + SAR

The most robust monitoring systems use both:

1. SAR provides continuous monitoring regardless of weather
2. Optical provides spectral confirmation when cloud-free imagery becomes available
3. Combined detection reduces false alarms from either sensor alone

Global Monitoring Systems

Hansen Global Forest Change (University of Maryland)

The foundational dataset. Updated annually using Landsat imagery:

• Coverage: Global, 2000-present
• Resolution: 30 meters
• Definition: Tree cover loss — any reduction in canopy density from an initial state
• Accuracy: ~87% overall accuracy; higher in tropical humid forests

The dataset maps both gain and loss, enabling net change calculations. It's freely available through Google Earth Engine and Global Forest Watch.

GLAD Alerts

Near-real-time deforestation alerts from the University of Maryland:

• Latency: ~1 week from satellite acquisition to alert publication
• Resolution: 30 meters (Landsat-based)
• Coverage: Pan-tropical
• Method: Time-series analysis detecting persistent NDVI drops below the historical forest range

RADD Alerts (Radar Alerts for Detecting Deforestation)

Developed by Wageningen University using Sentinel-1 SAR:

• Latency: 6-12 days
• Coverage: Tropical humid forests
• Key advantage: Works through clouds, providing detections when optical systems can't
• Method: Detects backscatter drops in forested areas exceeding historical variability

DETER (Brazil - INPE)

Brazil's national near-real-time system:

• Purpose: Operational law enforcement support
• Resolution: MODIS (250m) for rapid detection, Landsat/Sentinel for confirmation
• Latency: Days to 2 weeks
• Impact: Detection data triggers enforcement actions by environmental police

Challenges

Cloud Cover in the Tropics

The Amazon, Congo Basin, and Southeast Asian forests experience 60-80% cloud cover during wet seasons. Optical sensors like Landsat and Sentinel-2 may not acquire a usable image for 2-3 months. During this window, significant deforestation can occur undetected by optical systems alone.

This is why SAR has become indispensable for tropical deforestation monitoring. Sentinel-1's 6-12 day revisit through all weather conditions fills the gap.

Degradation vs. Clear-Cut

Complete forest removal (clear-cutting) is relatively easy to detect — the spectral change is dramatic. Selective logging and degradation — where individual trees are removed but the canopy partially remains — is much harder:

• NDVI may drop only 10-20% (within natural variability)
• SAR backscatter change is subtle
• The spectral signal recovers quickly as remaining canopy fills gaps

Detecting degradation requires:

• Higher resolution data (commercial satellites at 1-5m)
• Multi-temporal analysis tracking gradual decline
• Texture analysis (canopy gaps become visible at high resolution)
• LiDAR or SAR-derived canopy height data (detecting height loss)

False Positives

Not every forest change is deforestation:

• Seasonal deciduous forests lose leaves annually — this is natural, not deforestation
• Fire damage can kill trees without human agency
• Flooding temporarily reduces canopy reflectance
• Phenological variation causes year-to-year NDVI fluctuation

Robust systems use persistence criteria — the change must last through multiple satellite passes to be classified as deforestation rather than a temporary anomaly.

Definition Matters

"Forest" and "deforestation" have multiple definitions:

• FAO defines forest as >10% tree canopy cover, >0.5 hectares, >5 meters height
• Hansen dataset uses >50% canopy cover threshold by default
• National definitions vary widely

The same satellite data can show different deforestation rates depending on which definition is applied. When comparing deforestation statistics, always check which definition was used.

The Impact of Monitoring

Satellite deforestation monitoring has demonstrably reduced forest loss in some regions:

Brazil's DETER system: Studies showed that municipalities covered by DETER alerts experienced 50% less deforestation than similar municipalities without monitoring. The combination of detection and enforcement creates a deterrent effect.

Norway's REDD+ programs: Satellite-verified deforestation data underpins billions of dollars in results-based payments for forest conservation. Countries receive payment for reducing deforestation below a satellite-verified baseline.

Supply chain monitoring: Companies committed to "zero-deforestation" supply chains use satellite data to verify that their commodity suppliers (palm oil, soy, beef, cocoa) aren't clearing forest for production.

The technology for detecting deforestation exists and works. The bottleneck is usually institutional — having the governance structures, enforcement capacity, and political will to act on what the satellites reveal. But without the monitoring data, there's nothing to act on. Satellite systems provide the transparency that makes forest protection possible at scale.