AI-generated illustration for Galileo Whispers. Generic spacecraft-instrument and airless-surface concept; not a mission photograph, measured dataset or named sensor design.
Space mission data change meaning as processing turns instrument readings into calibrated measurements and, later, scientific products. Raw data preserve an early form of the observation; calibrated data express measurements in physical units; derived products use those measurements to construct results such as maps or estimates of physical properties.
Those labels describe processing, not an automatic ladder from unreliable to unquestionably correct. A derived result can depend on assumptions that are absent from a raw reading, while a raw reading can be difficult to interpret without calibration.
The processing label answers a limited question
NASA’s Planetary Data System, or PDS, provides an archiving framework for planetary science. Its PDS4 Data Provider’s Handbook, version 1.20.0, dated March 31, 2023, defines broad science-data processing levels in Table 2-1.
| PDS4 level | What it tells the reader | What it does not establish by itself |
|---|---|---|
| Raw | Instrument data have not been processed beyond the allowed handling needed for archiving. | That every stored number is already a physical measurement. |
| Partially Processed | Some processing beyond the raw stage has occurred. | That the product has completed calibration. |
| Calibrated | Data have been converted to physical units under the instrument’s calibration. | That uncertainty, artifacts, or geometry no longer matter. |
| Derived | Results have been distilled from calibrated products or supplied as relevant interpretive products. | That the adopted model or interpretation is unique. |
The categories help readers choose a product. They do not replace its instrument documentation, processing history, or description of limitations.
Raw does not always mean the original radio transmission
A spacecraft may compress, organize, or packetize information for transmission and storage. Reversing that transport formatting can be necessary before data are placed into an archive format.
The PDS distinction allows such handling without treating the resulting raw product as if it had already undergone scientific calibration. “Raw” therefore needs the archive’s definition, not an assumption that the file is a literal recording of every transmitted bit.
Telemetry and a labeled raw science product are related but different things. One concerns the transmitted information stream; the other is an archived representation that a reader can identify and interpret using the accompanying description.
A detector number needs a calibration model
Imagine two exposures of the same scene made with different exposure times. A larger stored signal in the longer exposure does not necessarily mean the source became brighter. Detector response and observing conditions have to be accounted for.
As an illustrative example, subtracting a detector offset from a recorded value can remove an instrumental contribution. Converting the remaining response into radiance requires additional information. The first operation alone does not demonstrate that the second has been completed.
The calibration may include several corrections, and those corrections have uncertainties. A product in physical units is much more useful for many comparisons, but the unit label does not make the detector infinitely accurate.
A map can add geometry and assumptions
Turning calibrated observations into a surface map can involve navigation, projection, resampling, and the combination of multiple measurements. Inferring temperature or composition can add a physical model as well.
That is why a derived product may be easier to use and still require more interpretation. A colored map is not necessarily a direct detector image, and a smooth-looking surface can hide where measurements were sparse or values were interpolated.
The PDS processing-level definition includes examples such as maps and gravity or magnetic-field products. Their usefulness comes from documented transformations, not from the word derived alone.
Choose a product for a question, not the highest label
If the question is what the detector originally recorded, an early product and its documentation can be more useful than a finished map. If the question is how measured brightness varies over a surface, a suitable calibrated product may avoid repeating an entire instrument-calibration pipeline.
For an interpreted physical quantity, a derived product can be the appropriate starting point, provided its model, uncertainties, and input relationships are understood. Requesting the “most processed” file without defining the intended measurement can select an attractive product that cannot answer the question.
Retaining earlier products also allows scientists to revisit a result when calibration improves. A new analysis may transform existing observations rather than require another spacecraft encounter. The archive’s processing history makes that distinction traceable.
Five things to check before comparing files
First, identify the archive and instrument. Processing conventions can differ between missions and disciplines; a numbered “Level 2” elsewhere should not automatically be equated with a PDS4 category.
Second, read the product’s processing level and units together. Third, check the processing and calibration versions. A revised calibration can change a value without the spacecraft having made another observation.
Fourth, check the observing time, target, and geometry. Finally, follow the relationship to input products and relevant documentation. A product identifier and version help make that chain reproducible.
The archive handbook makes metadata and documentation requirements apply across processing levels. Derived products do not make the earlier measurements dispensable: they make the connection between those measurements and a scientific question explicit.
This differs from the broader problem of what observing technology measures. Before using a downloaded result, ask what the instrument supplied, what processing changed, and which assumptions produced the quantity you now see.
