Satellite evidence · 2026-10-09
A satellite shows a signal. What does it take to interpret it?
A tree cover loss alert can guide a review, but a plot-level decision needs context. Scale, dates, land use and field evidence determine what a conclusion can support.
The question starts after the alert
A cooperative opens a map and finds a tree cover loss signal over a supplier's plot. Should it pause a purchase, request documents or send a technician? The answer depends on what the source detected, where the change occurred and what evidence can explain it.
An alert helps prioritize a review. A decision about a producer or a lot requires a recorded interpretation: what was observed, which hypotheses were considered and what remains unresolved.
Tree cover and land use answer different questions
Tree cover loss data can capture changes in natural forests, timber plantations and tree crops, from human or natural causes. Loss can be temporary or permanent. Identifying conversion from forest to another land use requires additional information.
For shade-grown coffee and cocoa, review site history and tree management. A map layer's label alone does not establish previous land use, the cause of change or an individual's responsibility.
Resolution matters as much as the plot boundary
The Hansen Global Forest Change v1.13 catalog lists a pixel size of approximately 31 meters. This describes the product's observation unit; it does not represent boundary accuracy or identify every tree on a farm.
At a plot edge, a pixel can include land on both sides. Before attributing change to a producer, review the geometry, its provenance and its overlap with the original source. Zooming in on a map does not add detail to the data.
Detection date is not the exact event date
Optical alerts depend on usable observations. Clouds can delay detection, and later observations can increase confidence. The GLAD laboratory describes its alerts as early indicators and cautions against using them to estimate the area of loss.
A review record should distinguish observation date, detection date and the likely period of change. When a decision depends on a cutoff date, the interval between clear images before and after the change can be more informative than the alert's displayed day.
Illustrative example: reviewing shade-grown coffee
Suppose a signal appears near a coffee plot's boundary. The producer says shade trees were renewed; the map includes vegetation on a neighboring farm. Both explanations are hypotheses to examine, rather than reasons to close the case automatically.
A review can compare images from different dates, confirm the boundary and retrieve management records. If material uncertainty remains, a visit with located observations and photographs can add context. A current photograph without site history does not, by itself, establish what was there before the change.
A conclusion may support an explanation, confirm a relevant change or leave a gap open. Each outcome requires a preserved basis.
What to record before closing a review
For another person to review the decision, the record should allow them to reconstruct at least these six elements:
- Plot: identifier, polygon version, survey date and survey source.
- Satellite source: product name, version, geographic coverage, resolution and access date.
- Signal: location, observation period, alert definition and confidence level when supplied by the source.
- Cross-checks: before-and-after imagery, observation quality and field or documentary evidence.
- Interpretation: explanation assessed, alternatives considered and remaining limitations.
- Decision: responsible reviewer, date, agreed action and conditions for reopening the case.
Case status should reflect the available evidence
An organization can distinguish a signal awaiting review, a change interpreted with supporting evidence and a case with insufficient evidence to conclude. These states keep uncertainty from disappearing inside a traffic-light indicator.
The absence of alerts also needs context: covered area, observation period and the product's detection capability. It can support a statement limited to that query; it does not automatically establish the absence of all change.
A universal pixel or confidence threshold is unsuitable for all crops and regions. Action should consider the source methodology, geometry quality and the consequences of an incorrect decision.
From a signal to a decision others can review
The operational value of monitoring lies in guiding work and preserving a verifiable explanation. Satellites help locate and compare changes; agronomic knowledge, records and territorial review help interpret their meaning.
At Nova Silva, this principle guides our approach to evidence: preserve the source, context, interpretation and limits of each conclusion. A useful decision should explain what is known, why it is supported and what information is still missing.