Turn an image into a located observation

Spatial sampling describes the ground represented by a pixel. Geolocation describes where the observation belongs on Earth. A sharp image can have an offset in its position: adding more pixels does not automatically remove a pointing, timing or reference error. This distinction is essential when a map must align with roads, parcels or earlier observations.

Consider a fictional image with one-metre ground sampling whose mapped position is shifted eight metres east. A road may look detailed while appearing in the wrong parcel. At that sampling, the displacement is eight pixels. Those invented values are arithmetic teaching inputs, not Alsat-3 specifications or measured errors.

Separate pixel detail from geographic placement

Image registration aligns observations to a reference. Stable landmarks can supply control, while independent check locations test the result. Using exactly the same points to fit and evaluate an adjustment can hide weaknesses outside those points. Terrain and reference quality also matter. USGS processing documentation explains these principles for Landsat; its product names and accuracy values should not be transferred to Alsat-3.

In a fictional coastal study, four landmarks determine an adjustment and a fifth, unused landmark checks it. If that check is still displaced, the analyst should investigate before treating a thin band along a coastline as new land or erosion. A beautiful overlay at the fitting points is not a substitute for the independent check.

Radiometric calibration connects a detector’s response with a reference; continued validation checks data quality over time. The USGS Landsat calibration program illustrates this ongoing responsibility. A brighter pixel in a later image might reflect a physical change, but exposure, illumination, atmosphere or processing differences can also influence a comparison.

The accompanying ISS photograph is a nighttime regional view published by NASA in 2022. It is not an Alsat-3 product. Its bright city lights are useful for discussing observation conditions: an unlit area cannot simply be labeled empty or undeveloped. To compare dates responsibly, retain acquisition information, processing version, quality flags and the measurement being compared.