Test land-change evidence

Amazonia-1’s published five-day revisit describes an observing opportunity, not a promise of a cloud-free image every five days. Clouds and their shadows can hide or alter the surface signal. A regional image may be mostly clear while the particular field being studied remains obscured. Start with usable evidence over the target, not simply the number of files in a catalogue.

In an original schedule, the target is observed on days 0, 5, 10 and 15. Only days 0 and 15 are clear over the target. The usable comparison spans fifteen days despite the five-day observing interval. If change is visible between those two clear observations, the images bracket its timing; they do not identify the exact day on which it happened.

A processed Amazonia-1 view of Sinop shows urban patterns, fields and waterways. It is useful for observing spatial relationships at one time. By itself it cannot establish a rate of deforestation or the cause of a particular land-use pattern. Those claims require additional observations and a clearly defined method.

Suppose a fictional study area contains 100 cells. Only 60 are usable in both of two images, and twelve of those 60 are classified as changed. The observed fraction is 12 ÷ 60 = 20 percent of the jointly usable area. Reporting twelve percent as though all 100 cells were observed would hide missing evidence. Extending the twenty-percent rate to the other forty cells would also require an assumption: clouded locations may differ systematically from clear ones.

Keep the comparison denominator visible

Before treating a difference as land change, align the images to the same places and inspect cloud, shadow, season and processing choices. A shifted image can make a stationary field edge look displaced. Different display stretches can make the same response look brighter. A crop cycle can change a field’s colour without creating a new field. These are hypotheses to test, not reasons to discard all change detection.

Build a defensible claim in stages: identify the observed difference, check data comparability, test alternative explanations and seek independent evidence where needed. A map of possible change can guide further investigation without proving responsibility or legality. Retain dates, masks, product identifiers and analysis settings so another learner can reproduce the comparison and understand its limits.