Wait, what? A space-mission website posts a dramatic black-and-white picture only hours after a spacecraft sends it to Earth. The gallery labels it “raw image”. A student downloads the JPEG and says, “Great — this is exactly the original science data that came from the spacecraft. Nothing has happened to it.” The word raw seems to guarantee that conclusion. It does not.
This Reality Lab owns one precise learner job: how to evaluate a scientific image labelled raw when the label might mean “a quickly released, minimally processed public image” rather than “the exact original archival science-data product”. The broader PSLE Science habit is to ask what a label means in that system, what transformations happened between measurement and display, and whether the displayed object is suitable for the claim being made.
Quick answer
Not necessarily. NASA explicitly explains that the “raw images” it provides online for quick public access are generally not the same as truly raw science data, meaning the original image-data format returned to Earth by spacecraft. They are closer to high-quality previews that can be made available rapidly, while science-quality archival data are preserved in systems such as NASA’s Planetary Data System and go through validation and standardised packaging.
The learner job is therefore: do not let a familiar label erase the provenance chain. “Raw” has to be defined by the publisher, not guessed by the reader.
A different kind of case file: follow the image backwards
Instead of starting with a claim and moving forward, we will trace a displayed image backwards.
- You see a picture on a mission website.
- The picture has a filename that ends in .jpg and a “raw” label.
- The website generated or served that public-view file from mission data received on Earth.
- The spacecraft camera originally recorded detector values plus engineering and observational information.
- The science archive later preserves data products, metadata and documentation needed for scientific use.
At each step, information may be arranged, packaged, converted, compressed, labelled or accompanied by metadata. Some changes can be harmless for quick viewing while still making the displayed file different from the original science-data object.
The learner job — and what this page does not own
This page does not become the general owner of image processing, false colour, image enhancement, compression, sensor calibration or graph interpretation. Other eduKateSengkang Science pages retain those jobs. Here, the communication object is the word “raw” attached to a public scientific image and the unsupported inference that the web image must therefore be the exact unaltered spacecraft data.
For separating what is directly visible from what is inferred about its cause or history, route to How to Tell Observation, Inference, Prediction and Explanation Apart in PSLE Science. For checking variables and comparisons when images are used as evidence, use How to Decode Variables and Fair Tests in PSLE Science Questions.
Original composite case: the crater that “grew” overnight
Imagine an invented mission gallery. Yesterday’s “raw” public image of a crater looks pale and low-contrast. Today a science archive offers a calibrated data product from the same observation. A student places the two on a slide and says, “The crater became darker overnight, so the surface changed.”
The conclusion skips a crucial question: are the two files directly comparable representations? A difference in brightness on screen can come from display scaling, conversion, calibration, compression, colour mapping, geometric processing, or other handling rather than a change on the planetary surface.
The first scientific move is not “which image looks more real?” It is “what exactly is each image product?”
Observed, labelled, claimed and inferred
| Layer | Example |
|---|---|
| Observed | A web page displays a monochrome picture and labels it “raw image”. |
| Publisher-defined meaning | The mission may use “raw” for a rapid public-view image that is largely untouched by image-processing software. |
| Unsupported claim | The downloaded JPEG is bit-for-bit the original science data returned by the spacecraft. |
| Scientific inference with provenance | The image can provide an early visual look at the observation, while quantitative analysis may require the archival data product, metadata and calibration information. |
Why a label can be true and still be misunderstood
“Raw” is not necessarily a false label. A publisher can use the term honestly within a defined workflow. The mistake happens when the reader silently replaces that local definition with a stronger one.
NASA’s public explanation says that its raw images are provided in their original, usually monochrome appearance and are largely untouched by image-processing software, but that these public raw images are generally not the same as the truly raw science data originally returned to Earth. See NASA Science: What Are Raw Images?.
This is a superb evidence lesson because both statements can be true at once:
- the web image can reasonably be called “raw” in the gallery’s public-release sense;
- the web image can still differ from the original archival science-data format.
Scientific literacy often depends on resisting false either/or thinking.
The provenance ladder
When a picture is used as scientific evidence, ask where it came from. A useful provenance ladder is:
- Instrument event: what detector recorded the signal?
- Transmission: how were the data sent and reconstructed on Earth?
- Data product: what format, units and metadata were created?
- Processing: what calibration, correction, geometric handling or combination occurred?
- Display: what scaling, cropping, compression or colour choices were used for human viewing?
- Caption: what claim does the publisher make about the image?
You do not need every technical detail every time. You need enough of the chain to know whether the image can support the intended conclusion.
Representation check: the screen is not the measurement
A camera detector records numerical values. Your screen turns numerical values into visible brightness and colour. That display step is already a representation. Two programs can display the same underlying values differently if they stretch brightness differently. A JPEG can also use lossy compression, which changes some pixel information to reduce file size.
Therefore, “I can see it” and “I have the original measurement” are different statements.
Method check: what happened before the image reached you?
Before using an online image quantitatively, ask questions such as:
- Is this a quick-look image, preview, browse product or archival science product?
- Has it been compressed?
- Has the brightness range been stretched for display?
- Has geometric correction been applied?
- Is the image calibrated into physical units, or does it show detector counts?
- Has colour been assigned from multiple filters or wavelengths?
- What metadata accompany the archival product?
- Is the public image appropriate for measuring what I want to measure?
These are not accusations. They are ordinary scientific provenance questions.
Worked case 1: a “raw” JPEG and a calibrated archive file
An invented rover camera captures a rock. The mission website posts a 1600-pixel JPEG labelled raw. Months later the archive releases a larger calibrated product with metadata and instrument documentation.
A student wants to calculate the rock’s exact reflectance from the JPEG’s displayed grey value. That is too strong. The public image may be excellent for recognising shapes and following mission activity, yet quantitative reflectance analysis could require calibration, the original data values, exposure information, filter information and other metadata.
The key distinction is not “good image versus bad image”. It is fit for purpose.
Worked case 2: one image looks brighter
Two public “raw” images of the same area are taken on different days. The second looks brighter. A social post says, “The ground reflected twice as much light.”
Before accepting that claim, compare exposure settings, illumination geometry, detector conditions, display scaling and whether the web service processed both images identically. A visible brightness difference is an observation about the displayed representations. “The surface reflected twice as much light” is an inference requiring quantitative comparability.
Worked case 3: “untouched” does not mean “context-free”
Suppose a quick-look image has had almost no cosmetic processing. A viral post circles a bright dot and calls it a new object. Even if the image really is minimally processed, the bright dot could be a detector artifact, cosmic-ray hit, saturation effect, hot pixel or another imaging feature. Provenance and instrument knowledge still matter.
Minimal processing does not guarantee that every visible feature is a feature of the scene.
Worked case 4: early release versus validated archive
NASA notes that public raw images can be released quickly, while archival data undergo validation and standardised packaging. This creates a useful trade-off. Rapid images support timely public access. Validated archives support rigorous scientific reuse. Speed and archival completeness are different jobs.
A headline that says “the archive corrected NASA’s earlier picture” may therefore be misleading. The early image and later archive may simply belong to different stages and purposes in the data lifecycle.
Comparison and baseline check
When comparing two scientific images, keep constant what must be constant for the claim:
- same wavelength or filter if colour/brightness is being compared;
- comparable illumination and viewing geometry where relevant;
- same or understood calibration state;
- same display scaling if judging visible intensity;
- same spatial scale if judging size;
- same processing stage if a pixel-by-pixel numerical comparison is intended.
If those conditions differ, the image comparison can mix scene change with representation change.
Alternative explanations checklist
When a visible feature or difference is used to support a scientific claim, consider at least four classes of explanation:
- scene: the real world actually changed;
- instrument: detector behaviour or settings changed;
- processing: calibration, correction or combination changed;
- display: scaling, colour, cropping or compression changed what you see.
The existence of alternatives does not prove the headline wrong. It tells you what needs to be ruled out.
What evidence would strengthen an image-based claim?
- The exact product name and processing level.
- Metadata describing the observation and instrument.
- Calibration documentation appropriate to the measurement.
- A comparison using like-for-like data products.
- Independent observations showing the same change.
- A claim limited to what the displayed product can actually support.
What would weaken it?
- Assuming “raw” always means bit-for-bit original spacecraft data.
- Measuring a web preview as though it were a calibrated science product without checking.
- Comparing brightness from images with unknown display scaling.
- Ignoring metadata and instrument context.
- Treating a visible artifact as a scene object without independent evidence.
How far can the conclusion travel?
A public raw image may support statements such as “this feature is visible in the mission’s quick-look image” or “the image provides an early view of the observation.” Stronger quantitative claims — exact brightness, calibrated reflectance, temperature, composition, precise dimensions or change through time — may require science products and metadata designed for those purposes.
The stronger the claim, the stronger the provenance chain must be.
PSLE-style transfer case
This is original practice.
A mission website publishes a “raw” JPEG of a moon. An archive later provides a calibrated science product from the same observation. The JPEG looks darker. A student concludes, “Calibration made the moon brighter.”
Question 1: Why is the conclusion too strong?
Explained answer: The two files may differ in display scaling, format, calibration and other processing. A brighter displayed image does not by itself show that the physical moon or its measured signal became brighter because of calibration.
Question 2: What should be checked before comparing their pixel values?
Explained answer: Check the product type, units, calibration state, metadata, scaling and whether both images represent values in a comparable way.
Question 3: Does the label “raw” prove the JPEG is the exact original science-data format returned by the spacecraft?
Explained answer: No. The publisher’s definition must be checked; NASA, for example, explicitly distinguishes its quick public raw images from truly raw archival science data.
Delayed independent return
- What is the first question to ask when a scientific file is labelled “raw”?
- Why can a JPEG be useful but still not be the original science-data product?
- Name two transformations or representation choices that can change how an image looks.
- Why is a visible difference not automatically a physical difference?
- What determines whether an image is suitable for a scientific claim?
Check: Ask how the publisher defines raw; public-view files may involve conversion or packaging and may lack the archival form and metadata; examples include calibration, scaling, compression, colour assignment and geometric correction; representation changes can create visible differences; and suitability depends on the claim and the image’s provenance, processing and metadata.
Parent and tutor teaching guide
Teach this with ordinary phone photos before talking about spacecraft. Save one photo as a full-resolution file, send a compressed copy through a messaging app, take a screenshot of that copy, and display all three. Ask, “Which one is the photo?” The answer is: all are representations of the same scene, but they are not the same data object.
Then transfer the habit to science. Give the child the labels “quick-look image”, “calibrated product”, “archive file” and “social-media screenshot”. Ask what each might be good for and what evidence would be needed before making a numerical claim. This builds a mature idea: data quality is task-dependent.
Authoritative sources and curriculum frame
- SEAB: 2026 PSLE Science syllabus — includes interpreting and analysing information, evaluating observations, information and methods, and communicating reasoning.
- MOE: 2023 Primary Science Teaching and Learning Syllabus — advocates healthy scepticism and careful handling and communication of data and information.
- NASA Science: What Are Raw Images? — explains that mission “raw images” offered for rapid public access are generally not the same as the truly raw science-data format returned by spacecraft, and describes archival validation in the Planetary Data System.
The quiet habit to keep
Scientific words are not magic seals. “Raw” does not erase a data pipeline. Ask where the image came from, what happened to it, what metadata travel with it, and what job the file was designed to do. Then make a claim no stronger than the evidence object you actually have.
