Deforestation monitoring with satellite imagery in Colombia: early-warning systems and how they work

Colombia is one of the most biodiverse countries on Earth and also one that loses some of the most forest area each year, especially in the Amazon. For an environmental authority, a regional corporation or a company with no-deforestation commitments, the operational question isn't whether deforestation is happening, but where and when it's happening, in time to act. Satellite monitoring is the only tool that covers vast, remote, hard-to-reach territories systematically. This note explains how it works, which sensors are used, and how to stand up a useful early-warning system.

Why satellite and not something else

Deforestation happens far away, fast, and on many fronts at once. Patrolling the Amazon on the ground to detect it would be impossible. Satellite solves four problems at once:

• Coverage. A single pass captures thousands of hectares.
• Frequency. Modern sensors revisit the same spot every few days.
• Consistency. The same measurement, repeated over time, lets you compare and detect change.
• History. Decades of archives exist to establish the baseline against which loss is measured.

The underlying idea is simple: forest has a characteristic spectral "signature" (high reflectance in the near-infrared, from healthy vegetation). When that signature disappears or changes sharply between two dates, there's a signal of cover loss worth reviewing.

The sensors used in Colombia

| Sensor | Resolution | Revisit | Strength | |---|---|---|---| | Sentinel-2 (optical) | 10 m | ~5 days | Detail and frequency; the workhorse | | Landsat 8/9 (optical) | 30 m | ~16 days | Historical archive since the 1980s | | Sentinel-1 (SAR radar) | ~10 m | ~6-12 days | Sees through clouds | | PlanetScope (commercial) | ~3 m | Daily | Very high detail, verification |

The great enemy of optical monitoring in Colombia is cloud cover. The Amazon and the Pacific coast are clouded over much of the year, and an optical sensor can't see through them. That's why Sentinel-1 radar is so valuable: microwaves penetrate clouds and let you detect change even in the rainy season. The optical + radar combination is what makes a monitoring system robust in the humid tropics.

Early-warning systems: the logic

A deforestation early warning is an automatic alert triggered when the system detects a probable forest loss at a location, ideally days after it occurs. The general logic:

1. Baseline. Define what counts as "forest" at each pixel from recent imagery.
2. Continuous observation. Each new image is compared against the baseline and the recent history.
3. Change detection. When a pixel's signature drops below a threshold —stops behaving like forest— it's flagged as an alert.
4. Filtering. False alarms (clouds, shadows, noise) are discarded by cross-checking multiple dates and, where available, radar.
5. Prioritization. Alerts are grouped, quantified in hectares, and prioritized by size, location (inside a protected area? an Indigenous reserve?) and speed.

In Colombia, IDEAM operates the Forest and Carbon Monitoring System and publishes official deforestation reports and early-warning bulletins (AT-D). Globally, references like GLAD (University of Maryland alerts) and RADD (radar-based alerts) are accessible via Global Forest Watch. An organization doesn't start from scratch: it can lean on these sources and complement them with its own monitoring over its jurisdiction.

How to stand up operational monitoring

A useful system isn't a map made once; it's a process that repeats. Layer by layer:

1. Define the area and baseline. Jurisdiction, protected areas, Indigenous reserves, known active fronts.
2. Choose the sources. Sentinel-2 as the base, Sentinel-1 for the cloudy season, Landsat for historical trend, and high-resolution commercial only to verify critical spots.
3. Automate the processing. Platforms like Google Earth Engine let you process long time series without downloading terabytes, compute vegetation indices and detect change at regional scale.
4. Set thresholds and filters. Calibrated with local ground truth; what works in a dry forest doesn't work in humid jungle.
5. Connect the alert to action. An alert nobody acts on is useless. The system must deliver the alert to whoever can verify and act —rangers, environmental authority, environmental prosecutor— with location, hectares and priority.

The bottleneck is almost never the technology; it's the alert → verification → action loop. Technically flawless systems fail because no one in the field receives the alert in time, or because there's no capacity to respond.

Applications

• Environmental authorities and regional corporations. Surveillance of protected areas, prioritizing operations, technical support for enforcement processes.
• Local governments. Cross-referencing deforestation with cadastre and land-use plans to understand pressure on rural land.
• Agribusiness and supply chains. Verifying no-deforestation commitments (palm, cattle, cocoa), increasingly required by international markets and regulations such as the EU's EUDR.
• Carbon and restoration projects. Baseline and verification of additionality and permanence.

Common mistakes

1. Ignoring clouds. Relying on optical alone in areas that spend half the year clouded over creates seasonal blindness. Without radar, the system "goes blind" exactly when most deforestation happens.
2. Confusing alert with confirmation. An alert is a suspicion that requires verification, not a verdict. Reporting raw alerts as confirmed deforestation destroys credibility.
3. Uncalibrated thresholds. Copying parameters from another region produces either an excess of false alarms or missed real events.
4. Not closing the loop. Detecting without response capacity turns monitoring into a report, not a control tool.

GeoSAT's role

Deforestation monitoring combines exactly what we do: processing multitemporal satellite imagery, analysis on platforms like Google Earth Engine, and change-detection models. Over 30 years of experience we've supported organizations in analyzing land-cover and land-use change across vast territories, and we know the real problems of the Colombian tropics: clouds, local calibration, and closing the loop toward action.

If your organization needs to stand up or improve deforestation monitoring over its jurisdiction or supply chain, the most useful first step is to define the area, the alert frequency you need, and who will act on each alert. Get in touch and we'll design it with you.