Stable ID: PSLE-SCI-REALITY-0540
PSLE Science asks Primary 5 and Primary 6 learners to do more than read a number correctly. Strong scientific inquiry means asking what that number represents, how it was produced, and whether the representation can support the claim being made. This Reality Lab uses a real-world satellite-data object: a map or table shows a pixel value of 12,000 DN, and someone immediately calls that number “reflectance.”
That shortcut is attractive because satellite images look like measurements of the Earth, and digital numbers look precise. But a stored digital number can be a scaled sensor-data code rather than the physical quantity a learner wants to discuss. In current Landsat products, for example, official USGS documentation explains that stored integer values can require a scale factor and, for some products, an offset before they become physical quantities such as surface reflectance or surface temperature.
This is a useful PSLE Science evidence-transfer problem because it combines interpreting information, evaluating methods, distinguishing observation from inference, and communicating a conclusion with the right scope. The learner job is narrow: before treating a satellite pixel number as a physical measurement, identify the product, processing level, band, units, scaling rule, quality information and valid range. The goal is not to teach remote sensing as a specialist subject. It is to learn how scientific data can be stored in a form that is not yet the meaning you want to claim.
Wait, what? The number on the screen might not be the physical quantity?
Imagine a class downloads an original composite satellite-data table for a fictional wetland. One column says:
| Pixel | Band | Stored value | Product note |
|---|---|---|---|
| P | Red band | 12,000 DN | Scaled integer; use metadata conversion |
| Q | Red band | 18,000 DN | Scaled integer; use metadata conversion |
A student says, “Pixel Q reflects 18,000 units of light, so it reflects 50% more than P.” Another says, “DN means digital number, so the value is only a computer code and tells us nothing.” Both answers go too far. The stored values do contain measurement information, but their physical meaning depends on the calibration and scaling defined for that product.
Quick Answer
A digital number is not automatically the final physical measurement. In many satellite products, the stored integer is a quantised or scaled representation produced for efficient digital storage. To interpret it scientifically, first use the product documentation and metadata to determine what conversion, scale factor, offset, valid range and quality flags apply. Only after that should you describe a physical quantity such as reflectance, radiance or temperature.
For Landsat Collection 2 Level-2 surface reflectance, USGS gives the conversion reflectance = DN × 0.0000275 − 0.2. A DN of 12,000 would therefore correspond to a reflectance of 0.13 under that exact product rule. That does not mean every satellite product uses this formula, and it does not mean every band value of 12,000 everywhere carries the same scientific meaning.
Your owned learner job—and the boundary
This article owns one transfer question: when a scientific image or table gives a raw or stored pixel code, how do you decide whether the displayed value can already be treated as the physical quantity named in the claim?
It does not own satellite orbits, electromagnetic waves, vegetation science, image classification, climate change, geography, computer encoding or radiometric calibration as full subjects. Those belong to other science and technical owners. It also does not replace the site’s existing PSLE Science owners for observation versus inference, variables, graph reading, measurement and checking. Instead, it applies those skills to one communication object.
The six-layer evidence chain
A good way to avoid the DN trap is to read a pixel through six layers. Each layer answers a different question.
- Object: Which satellite product, scene and band is this?
- Stored code: What number is actually stored in the file?
- Conversion: What scale factor, offset or calibration equation applies?
- Physical quantity: Does the converted value represent radiance, reflectance, temperature or something else?
- Quality: Is the pixel valid, saturated, filled, cloudy, shadowed or otherwise flagged?
- Claim: What conclusion is someone trying to make from the pixel?
Most weak interpretations skip from Layer 2 straight to Layer 6. The disciplined learner walks through the middle.
Case File 1: Two numbers that look directly comparable
Suppose a fictional science dashboard shows two red-band values from the same Landsat surface-reflectance product:
| Location | Stored DN | Scale factor | Offset |
|---|---|---|---|
| Marsh A | 12,000 | 0.0000275 | −0.2 |
| Marsh B | 18,000 | 0.0000275 | −0.2 |
Applying the official example rule gives approximately 0.13 and 0.295. The ratio between the stored codes is 18,000 ÷ 12,000 = 1.5, but the ratio between the converted reflectances is much larger than 1.5 because the transformation includes an offset. This means the tempting statement “the DN is 50% higher, therefore reflectance is 50% higher” fails.
The important lesson is not the arithmetic. It is that you cannot assume relationships between stored codes are preserved in the physical quantity until you understand the conversion.
Observed, claimed and inferred
| Observed or documented | Claim | Inference that must be justified |
|---|---|---|
| The file stores DN = 12,000 for a particular band and pixel. | “The surface reflectance is 12,000.” | That the stored number is already expressed in physical reflectance units. |
| The product guide provides a scale factor and offset. | “The converted value is a surface-reflectance estimate for this product.” | That the pixel is valid and the correct product rule and band were used. |
If you need a broader refresher on this separation, use How to Tell Observation, Inference, Prediction and Explanation Apart in PSLE Science. Reality Lab 540 does not re-teach that skill; it applies it to encoded satellite data.
Representation check: the screenshot may hide the units
A scientific website can display a number without showing all of the metadata that lives in the original file. A screenshot might show “12,000” beside a cursor. The scale factor could be documented elsewhere. The displayed image might even be a stretched preview whose colours were chosen for visibility rather than quantitative reading.
Ask three representation questions:
- Is this the raw scientific data, a converted layer or only a visual preview?
- Are the units and product level visible?
- Does the image preserve the pixel values, or has it been stretched, rescaled or colour-mapped for display?
For the related question of whether a “raw image” preview is identical to original archive data, route to PSLE Science Reality Lab Vol No.466.
Method check: product level matters
The phrase “satellite pixel value” is too vague to support a quantitative conclusion. Landsat Level-1 data can be delivered as scaled digital numbers that are converted to top-of-atmosphere radiance or reflectance using radiometric scaling factors in scene metadata. Landsat Level-2 products can contain processed surface reflectance or surface temperature stored as scaled integers with their own conversion rules.
Therefore, the same-looking integer can mean different things in different files. A learner should ask: “Which product am I looking at?” before asking “What does the number mean?”
Case File 2: Same DN, different product
Imagine two fictional datasets. Dataset A is a Level-1 product in which DN is converted using band-specific radiance coefficients. Dataset B is a Level-2 surface-reflectance product with the Collection 2 scale factor and offset. Both files happen to contain the integer 15,000 at one pixel.
A learner says, “The physical signal must be the same because the DN is the same.” That conclusion does not follow. The stored code is only meaningful inside the product’s calibration system. Two thermometers can both display the number 20 while using different units; likewise, two digital products can store identical integers while mapping them to different physical quantities.
Band check: one pixel location can have many values
Satellite instruments observe different spectral bands. A single ground location can therefore have one stored value in a red band, another in near-infrared, another in short-wave infrared and so on. The phrase “the pixel value” can hide which band is being discussed.
That matters because a claim such as “this place is highly reflective” is incomplete without saying reflective at which wavelength range, under which product definition and under what correction. A red-band reflectance value and a thermal-band stored code are not interchangeable simply because both are stored as integers.
Valid-range check
Scientific products usually define valid ranges, fill values and special codes. USGS documentation for Collection 2, for example, distinguishes valid data from fill values. A stored zero in one product can be a fill value rather than a physical zero. The correct interpretation comes from the product specification, not from the number’s appearance.
This connects directly to PSLE Science Reality Lab Vol No.453, which owns the separate problem of treating a NoData code as though it were a scientific measurement.
Saturation check
A very large stored value can also tempt a learner to say “very strong signal.” But sensors and data formats have upper limits. If a detector saturates, the stored maximum can mean “at or above the measurable range” rather than an exact physical value. Always inspect the quality information before treating an extreme code as an exact measurement.
For that distinct evidence job, see PSLE Science Reality Lab Vol No.459.
Quality-flag check
A DN can be correctly scaled and still be a poor basis for the intended claim because the pixel is cloud-contaminated, shadowed, snow-covered, saturated or otherwise flagged. Scientific meaning is therefore not just “convert the number.” It is “convert the right number, from the right product, under acceptable quality conditions.”
This is why a pixel value should often be read together with a QA layer or other product-quality information. For the broader skill of reading a data-quality flag beside a scientific value, route to PSLE Science Reality Lab Vol No.040.
Case File 3: A colourful image versus a quantitative layer
A class compares a JPEG browse image with the underlying scientific raster. The JPEG looks bright green in one area, while the scientific red-band layer stores DN = 11,300 and the near-infrared layer stores DN = 26,100.
A student says, “The place must naturally be bright green because the satellite measured green.” That is a representation error. The browse image may combine selected bands and apply a visual stretch. The colours help humans inspect the scene but do not necessarily reproduce ordinary human vision or directly encode one physical measurement.
Another student says, “The green preview is useless.” That also goes too far. A browse image can be valuable for visual selection and interpretation. The correct scientific move is to match the representation to the job: use the preview for visual context and the calibrated/scaled science product for quantitative analysis.
Comparison and baseline check
Suppose two satellite scenes were produced by different sensors or processing collections. A headline claims, “Pixel values increased from 9,000 to 11,000, so reflectance increased by 22%.” Before accepting that comparison, check:
- Are both values from the same physical quantity?
- Do both products use the same scale and offset?
- Are the same spectral bands being compared?
- Were atmospheric corrections and processing versions comparable?
- Are both pixels valid and unsaturated?
- Are the pixels aligned over the same ground area?
A change in stored code can reflect a physical change, a processing change, a calibration difference, a band mismatch or a quality problem. The learner does not need to diagnose the entire remote-sensing pipeline; the job is to recognise that a direct physical comparison needs evidence that the scales are comparable.
Variables and fair-comparison thinking
This is where ordinary PSLE Science fair-test reasoning becomes useful without becoming a new fair-test lesson. If we compare two pixel values while changing sensor, band, processing level, date, illumination, atmospheric conditions and scaling rule, we have many reasons for a difference. A learner should ask which conditions must be comparable for the claim being made.
For the canonical fair-test skill, route to How to Decode Variables and Fair Tests in PSLE Science Questions. Here the skill is applied to metadata and measurement representation.
What evidence strengthens “this is physical reflectance”?
- The product documentation explicitly identifies the band as surface reflectance.
- The correct scale factor and offset have been applied.
- The converted value lies in the valid scientific range.
- Relevant QA flags show the pixel is suitable for the intended use.
- The wavelength band and processing level match the claim.
- The calculation can be reproduced from the metadata.
What evidence weakens the claim?
- The screenshot gives a number but not the product or units.
- The value comes from a stretched browse image rather than the measurement layer.
- The conversion rule belongs to a different product or collection.
- The pixel is a fill value, saturated or quality-flagged.
- The claim compares raw codes from different sensors as though they share one scale.
- The physical quantity is named without showing how the stored code becomes that quantity.
Tempting reasoning—and the repair
| Tempting statement | Why it fails | Repair |
|---|---|---|
| “DN 12,000 means reflectance 12,000.” | The stored integer may require scaling and an offset. | Read product metadata and convert first. |
| “DN 18,000 is 50% more reflective than DN 12,000.” | An offset can break proportionality. | Compare the converted physical values. |
| “Same DN means same physical signal.” | Different products may use different calibrations. | Check sensor, product and conversion. |
| “Large DN means valid strong signal.” | Large codes may include saturation or invalid conditions. | Read quality and saturation flags. |
| “The colourful preview gives the exact measured colour.” | Preview images can combine and stretch bands. | Separate visualization from quantitative layers. |
Worked Case 4: Do the arithmetic, then stop at the correct boundary
A fictional Level-2 surface-reflectance pixel has DN = 20,000. Using the Collection 2 example rule, the learner calculates:
20,000 × 0.0000275 − 0.2 = 0.35
What can we say? We can say the product conversion gives a surface-reflectance value of about 0.35 for that band and pixel, assuming the pixel is valid and the correct rule applies. We cannot automatically say the surface reflects 35% of all incoming light in every direction and at every wavelength, because the value belongs to a particular spectral band and product definition.
This is a good example of how a correct calculation can still support an overwide conclusion if the learner forgets the measurement’s scope.
Worked Case 5: A 30 m pixel is not thirty metres of exact detail
A learner converts a DN correctly and then claims that every object inside the 30 m pixel must have the same reflectance. That conclusion is not supported. A pixel summarises the sensor response over a ground footprint and can contain mixtures of surfaces. Spatial resolution describes the sampling scale, not perfect uniformity inside the pixel.
For the separate problem of resampling a 30 m image to a finer grid without creating new native detail, route to PSLE Science Reality Lab Vol No.395.
How far can the conclusion travel?
A correctly converted, quality-checked pixel can support a claim about that measurement, band, location, acquisition and product. It does not automatically explain why the reflectance changed. It does not prove a biological cause. It does not prove every metre inside the pixel behaves the same way. It does not make one scene representative of all seasons.
Scientific strength comes from saying exactly what the measurement supports and stopping there.
PSLE-style transfer case
This is an original practice case, not a past examination question.
A student receives a satellite-data table. Pixel A has stored value 10,000 and Pixel B has stored value 14,000. The note says, “Values are scaled integers. Physical value = stored value × 0.0001.” A classmate says, “Pixel B has 40% more physical value than Pixel A.”
Question: Is the classmate’s comparison valid under this stated conversion?
Explained answer: Yes, under this particular conversion because both values are multiplied by the same factor and there is no offset. Their converted values are 1.0 and 1.4, so B is 40% higher than A. However, the conclusion would need re-checking if the conversion included an offset, if the products differed, or if either pixel were invalid.
This practice case matters because it shows that the correct rule is not “never compare DNs.” The correct rule is “compare them only when the conversion and quality evidence make the relationship valid.”
Transfer case with an offset
Now use the conversion physical value = DN × 0.0000275 − 0.2. Pixel C = 12,000 and Pixel D = 18,000.
Question: Can you say D is 50% greater than C in the physical quantity?
Explained answer: No. The stored values differ by 50%, but after applying the scale factor and offset they become approximately 0.13 and 0.295. The physical values are not related by the same 1.5 ratio.
Practice laboratory
Practice 1
A map cursor says “Value = 0.” The product guide states zero is the fill value. Does the sensor prove the physical quantity was zero?
Answer: No. The code marks missing or fill data, not a measured physical zero.
Practice 2
Two images from different sensors both show DN = 8,000. Can you conclude they measured the same radiance?
Answer: Not without confirming that the two products use compatible calibration and conversion rules.
Practice 3
A pixel is correctly converted to reflectance but the QA layer marks heavy cloud. Can it still be used as direct evidence about the ground surface?
Answer: The cloud flag weakens that ground-surface interpretation because the signal may not represent the surface cleanly.
Practice 4
A graph compares converted reflectance values from the same band, same product and valid pixels over time. Is this stronger than comparing raw screenshot colours?
Answer: Yes. The evidence chain is clearer and the quantities are defined, though other differences such as season, illumination and surface conditions may still matter.
Practice 5
A social post states, “This pixel rose from 10,000 to 20,000, so the physical quantity doubled.” The conversion includes a non-zero offset. What is the first repair?
Answer: Convert both values using the correct scale and offset before comparing the physical quantities.
Delayed independent return
Return later without looking back. A dataset stores surface-temperature values as integers and provides a scale factor plus an offset. A student sees 45,000 and says, “The surface temperature is 45,000 degrees.” What is the exact error?
Return answer: The student has mistaken the stored digital code for the final physical unit. The documented conversion must be applied first, and the resulting temperature must be interpreted with the product’s units and quality information.
Model and measurement limits
This guide uses Landsat because its documentation provides a clear public example of scaled integers, but the reasoning pattern is broader. Different sensors can use different digitisation, calibration, scaling, offsets and quality systems. Some products distribute already-converted floating-point data. Others store physical values multiplied by a constant. Still others use lookup tables or more complex calibration.
Therefore, never memorise one Landsat scale factor as a universal satellite rule. Memorise the evidence habit instead: find the product definition, recover the conversion, check quality, then name the physical quantity.
How this connects to PSLE Science without becoming a trick
The current 2026 PSLE Science assessment objectives include interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning. The 2023 Primary Science syllabus also encourages healthy scepticism and evaluation of how Science is presented through different forms and media.
A learner does not need a magic phrase such as “check the scale factor” in every question. The evidence must earn that move. If the source already gives values in physical units, no extra conversion is required. If the source gives stored codes with a conversion note, ignoring the note is the problem.
Parent and tutor teaching guide
- Round 1: Show a learner three cards: “stored code,” “conversion rule,” and “physical unit.” Ask them to place the cards in order.
- Round 2: Give two proportional conversions with no offset and ask whether ratios are preserved.
- Round 3: Add an offset and ask the learner to test the ratio again.
- Round 4: Add a QA flag and ask whether correct arithmetic is enough.
- Round 5: Swap “surface reflectance” for “surface temperature” and see whether the learner checks the product instead of reusing the old rule.
- Round 6: Finish with a new context such as a sound recorder, digital camera or air-quality sensor whose stored values also require calibration. Ask which parts of the reasoning transfer.
Reward precision. “The number is wrong” is weaker than “the stored number is not yet the physical quantity; the documented scaling and quality checks are needed before the claim.”
Authoritative source trail
The 2026 PSLE Science syllabus from SEAB states that candidates apply scientific inquiry by interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning.
The Ministry of Education’s 2023 Primary Science Teaching and Learning Syllabus includes healthy scepticism, questioning assumptions and uncertainty in evidence, evidence-based model building, and understanding how Science is communicated through different forms and media.
The U.S. Geological Survey’s Landsat Level-2 scaling guidance explains that science products can be distributed as scaled integers and provides the applicable scale factors and offsets. Its Landsat Level-1 data guide likewise explains that Level-1 bands are delivered as digital numbers that can be converted to physical quantities using metadata coefficients.
Quiet return: numbers need a measurement identity
A scientific number is never only a number. It has an object, a method, a unit, a scale, a valid range and a quality status. The satellite-data version of that habit is simple: DN first tells you what was stored. Metadata tells you what it means.
When a learner refuses to jump from “12,000” straight to “12,000 reflectance,” that is not hesitation. It is good scientific reasoning: identify the representation, convert it correctly, check whether the pixel is usable, and only then allow the conclusion to leave the screen.
