PSLE Science Reality Lab Vol No.527 | “The Change Map Shows a Red Edge” — Did the Ground Really Change, or Were the Images Slightly Misaligned?

PSLE-SCI-REALITY-0527

Wait, What? The Landscape Did Not Change — but the Difference Map Says It Did

Imagine two scientific images of the same hillside. One was collected in March and the other in June. A computer subtracts the first image from the second and produces a bright change map. Along one ridge, a narrow red strip sits beside a narrow blue strip. The caption says, “Surface increased here and decreased there.” It looks decisive. Red means gain. Blue means loss. Surely the ground changed.

Not necessarily. A change map can show a difference even when the physical surface did not change. If the two source images are shifted sideways by even a small amount, a steep edge, river bank, road, roof, cliff or ridge can land in slightly different pixel positions. Subtract one image from the other and the mismatch can manufacture a paired positive-and-negative pattern. The representation has changed because the images do not line up perfectly; the world may not have changed at all.

This is exactly the kind of real-world evidence problem that PSLE Science inquiry prepares you to think about. The goal is not to memorise the word misregistration. The goal is to ask a scientific question: Does the displayed difference come from a real change in the object, or from the way the observations were aligned, processed and compared?

Quick Answer

A coloured difference on a before-and-after scientific map is evidence of a difference between two datasets. It is not automatically evidence of a physical change in the world. Before accepting the claim, check whether the datasets describe the same place, on the same spatial reference, at comparable resolution, and with enough alignment accuracy for the size of change being claimed. If unchanged landmarks also develop opposite-colour edges, misalignment becomes a strong alternative explanation.

The Exact Learner Job This Reality Lab Owns

This article owns one narrow evidence-transfer job: evaluating a change or difference image when small positional misalignment could create false change. It does not replace the core PSLE Science owners for observation versus inference, fair comparison, measurement, maps, variables, image scale or general evidence quality. Those skills are already taught elsewhere. Here, you apply them to one real scientific communication object: a map created by comparing two spatial datasets.

For the broader skill of comparing photographs without confusing viewpoint with physical change, use How to Compare PSLE Science Photographs Without Mistaking Camera Perspective for Scientific Change. For distinguishing what is directly observed from what is inferred, use How to Tell Observation, Inference, Prediction and Explanation Apart in PSLE Science. This Reality Lab assumes those foundations and moves them into a change-detection problem.

Build the Evidence Object Yourself

Consider an original composite case. Scientists map a quarry wall in January and again in April. Both maps use square cells. The January ridge runs through column 20. In April, because the second dataset is shifted one cell east, the same unchanged ridge runs through column 21. A difference map subtracts January from April.

At column 20, the April map now contains lower ground than the January map, so subtraction produces a negative difference.
At column 21, the April map now contains the ridge where the January map contained lower ground, so subtraction produces a positive difference.
The output therefore shows one negative strip beside one positive strip.
Yet in our composite case, the ridge itself never moved and never changed height.

This pattern matters because it gives you a diagnostic clue. Real erosion or deposition can certainly produce neighbouring gains and losses, but a thin, repeated red-blue outline hugging many sharp features should make you ask whether the datasets are slightly offset.

Observed, Claimed and Inferred Are Three Different Things

Observed in the communication object: the difference map contains red and blue values in particular cells.

Claimed: the ground rose in red places and fell in blue places.

Inferred: the change is physical rather than an artefact caused by alignment, resolution, processing, different sensors or other measurement effects.

The first statement can be true even when the second is not yet justified. A disciplined science learner does not erase the map. The learner simply asks what extra evidence is needed to move from “the datasets differ” to “the world changed.”

Why Alignment Is Part of the Method, Not a Cosmetic Detail

When two spatial datasets are compared cell by cell, each cell in the first dataset is supposed to refer to the same physical place as the matching cell in the second. If cell A in January refers to one patch of ground while cell A in April refers to a patch two metres east, subtraction is no longer a clean before-and-after comparison. The method has changed the comparison object.

This is why the U.S. Geological Survey describes precise alignment as critical in elevation-difference work. USGS explains that even a horizontal registration error can appear as a vertical elevation difference, especially in steep terrain, and it documents examples where apparent topographic change was produced by residual misregistration rather than real change. That is a powerful real-world demonstration of a familiar inquiry principle: before comparing two results, make sure you are actually comparing like with like.

Authoritative reference: USGS — SRTM NED Vertical Differencing.

The Representation Check: What Does Each Colour Actually Mean?

Before you interpret any coloured patch, read the legend. A red value may mean “later minus earlier is positive,” “elevation increased,” “reflectance increased,” “probability of change increased,” or simply “difference exceeds a chosen threshold.” Blue may mean the opposite. The colours are assigned representations, not physical substances.

Then ask four questions. First, are the two source datasets measuring the same quantity? Second, do they use the same units and reference system? Third, do corresponding cells refer to the same physical locations? Fourth, is the displayed change larger than plausible measurement and registration uncertainty? A colour cannot answer those questions for you.

The Comparison and Baseline Check

A change map always contains a baseline, even if the designer does not make it visually obvious. If the map calculates “April minus January,” then January is the comparison reference. Reverse the subtraction and every sign reverses. Red may become blue without any new observation being made. This reminds us that change maps are derived products: they are built from source observations plus a rule for comparing them.

You should therefore ask: Which date is the baseline? Were both datasets collected with the same coordinate reference? Were they resampled onto one grid? Was one dataset shifted, warped or transformed? Are there known position errors? If different sensors were used, were their positions checked against common reference points?

A Simple Landmark Test

Suppose a concrete building was known not to move between two survey dates. In the two source maps, its western wall appears three metres east in the second dataset. That is important evidence. If a supposedly unchanged object shifts, you have found a positional mismatch independent of the quarry or river change you were originally studying.

A strong investigation may use several stable landmarks distributed across the scene, not just one. If every stable landmark shows nearly the same shift, that supports a systematic registration problem. If only one location disagrees, local measurement noise, object change or identification error may be more plausible. The pattern of disagreement matters.

Method and Variable Check

The real-world object may look like “just a map,” but it comes from a method. Treat the method as part of the evidence. Useful questions include:

Were the two observations made by the same sensor or different sensors?
What is the spatial resolution of each dataset?
Were both mapped to the same coordinate system?
How accurately are the images registered to one another?
Were cells resampled, averaged or interpolated?
Were clouds, shadows, vegetation or buildings present on one date but not the other?
Was the threshold for declaring “change” larger than expected noise and alignment error?
Do independent measurements support the same physical change?

Notice the PSLE Science habit underneath all of these. We are checking whether another factor changed besides the factor named in the claim. In a school investigation, that might be temperature or amount of water. In a spatial comparison, it might be image position, sensor geometry or cell size.

Alternative Explanations You Should Consider

Misregistration is one alternative explanation, not a magic answer. A good scientist does not replace one overconfident claim with another. Other possibilities include real surface change, different vegetation height, different water level, shadows, seasonal cover, sensor noise, different viewing geometry, different measurement method, a processing correction, missing values, or a different resolution.

The question is not “Can I invent an alternative?” It is “Which explanation best fits the full pattern of evidence?” If red-blue edge pairs occur mainly along every steep ridge and unchanged road edge, alignment is a strong candidate. If a broad excavated pit appears only in the later dataset and is confirmed by photographs and field measurements, real change is better supported.

Evidence That Would Strengthen a Real-Change Claim

Stable control landmarks align well between dates.
The change is much larger than the known registration and measurement uncertainty.
The same change is visible in an independent sensor, survey or field observation.
The pattern has a physically plausible shape rather than a repeated thin opposite-colour outline.
Multiple observations show the change appearing and persisting through time.
The method documents co-registration and quality-control checks.

Evidence That Would Weaken the Claim

Known fixed structures shift between source datasets.
Paired positive and negative strips hug sharp edges throughout the map.
The apparent change is about the same size as the positional uncertainty.
Different coordinate references or cell-center conventions were not reconciled.
The result disappears after improved co-registration.
No independent observation confirms the physical change.

Worked Case 1: The River Bank

A composite river-monitoring report shows a narrow red strip on the eastern bank and a narrow blue strip immediately beside it. A student says, “The river deposited soil on one side of the bank and removed exactly the same amount from the next strip.”

That is possible, but the pattern also fits a small sideways offset. The learner should inspect stable bridge piers, roads and buildings. If they show the same paired-edge pattern, then the map comparison may be misregistered. A stronger claim would be: “The datasets show a bank-edge difference, but alignment must be checked before concluding that the bank physically migrated.”

Worked Case 2: The Roof That “Moved”

Two height maps of a school are subtracted. Every roof has one positive edge and one negative edge. The school did not rebuild its roofs. A field photograph shows the buildings unchanged.

The repeated edge pattern around many roofs is evidence against a story of many separate building changes. A common positional offset explains more observations with one cause. This is a good example of scientific economy: prefer the explanation that accounts for the evidence without inventing many unrelated changes.

Worked Case 3: Real Change After Alignment

A landslide area is mapped before and after heavy rain. The team first checks stable rock outcrops and road intersections. They line up correctly within the stated uncertainty. After alignment, a broad zone still shows a large decrease in elevation, while a lower zone shows deposited material. Ground photographs and a later survey agree.

Here, the alignment test does not remove the signal. Independent observations agree, and the spatial pattern is physically plausible. The real-change claim is much stronger. Notice that healthy scepticism did not mean rejecting the map. It meant making the evidence survive a better test.

Worked Case 4: When the Shift Is Smaller Than a Pixel

A student argues, “The images cannot be misaligned because they are not shifted by a whole pixel.” This reasoning fails. Positions can differ by fractions of a pixel, and resampling can distribute the effect across neighbouring cells. Whether the shift matters depends on the shape of the surface, resolution and size of the claimed change.

Near a flat field, a small horizontal offset may cause little height difference. Near a steep cliff, the same offset can produce a much larger apparent vertical difference. USGS explicitly notes this slope dependence in elevation differencing. The scientific question is therefore not simply “Is there a shift?” but “Could the plausible shift generate a difference of this size here?”

Tempting but Invalid Reasoning

“The map is computer-generated, so the change must be objective.” A calculation can faithfully process misaligned inputs.
“The colours are bright, so the change is large and real.” Display colour depends on the chosen scale and threshold.
“Two different dates automatically make a fair before-and-after comparison.” The observations must also refer to comparable positions and methods.
“If alignment is imperfect, the entire dataset is useless.” Not necessarily. The uncertainty may be small enough for large changes but too large for tiny ones.
“A correction proves scientists made a mistake.” Registration is a normal part of building comparable spatial evidence.

How Far Can the Conclusion Travel?

A verified change in one mapped area does not automatically prove the same process happened everywhere. A result measured between January and April does not automatically describe every intermediate day. A change detected at one resolution may not locate the boundary to finer precision than the source data support. A derived difference may show where evidence is strongest without making every cell equally certain.

This is conclusion scope. Good scientific reasoning keeps the conclusion inside the space, time, quantity and uncertainty actually covered by the evidence.

PSLE-Style Transfer Case

A research team compares two height maps of a coastal cliff. The later-minus-earlier map contains a narrow +1.2 m strip beside a −1.1 m strip along several cliff edges. A car park, known to be unchanged, also shows a similar paired pattern. After the team realigns the maps using fixed survey markers, most of the paired strips disappear. One broad −3.8 m patch on the cliff remains and is also seen in field photographs.

Question: Explain why the +1.2 m and −1.1 m edge strips should not immediately be accepted as real ground change, and why the −3.8 m patch has stronger support.

Explained answer: The paired edge strips also appeared at an unchanged car park and mostly disappeared after the maps were realigned, so misregistration is a plausible cause of those differences. The −3.8 m patch remained after alignment and was independently supported by field photographs, so there is stronger evidence that it represents a real physical change in the cliff.

Notice what the answer does not say. It does not rely on a magic keyword. It links observations to an explanation and compares competing causes using evidence.

Delayed Independent Return

Close this page for ten minutes. Then answer this without looking back: A difference map shows thin red and blue outlines around every building. What two checks would you perform before claiming the buildings changed?

A strong return answer would include checking whether stable landmarks are aligned between dates and checking whether the apparent difference is larger than known position or measurement uncertainty. You might also ask whether the pattern survives improved registration or is confirmed independently.

Practice: Five Original Evidence Decisions

1. A wetland boundary shifts two pixels east, but roads shift two pixels east too. Best interpretation: suspect a positional offset before claiming wetland movement.

2. A broad quarry pit deepens by 12 m, fixed benchmarks align within 0.2 m, and a ground survey agrees. Best interpretation: the evidence strongly supports real change.

3. A change map contains values smaller than the reported registration uncertainty. Best interpretation: the map may not support confident cell-by-cell physical-change claims at that scale.

4. The change pattern disappears when the later image is shifted half a cell. Best interpretation: the original pattern was sensitive to alignment, weakening a real-change interpretation.

5. The same changed area appears in two independently processed datasets. Best interpretation: independent agreement strengthens the claim, although you should still check whether the datasets share the same source or systematic error.

Routes Back to Canonical PSLE Science Skills

Parent and Tutor Teaching Guide

Do not begin by teaching the specialist vocabulary. Begin with two transparent sheets containing the same square and shift one sheet slightly sideways. Ask the learner what a subtraction or difference picture might show around the edges. Then reveal the term misregistration only after the learner understands the mechanism.

Next, give the learner three evidence packets: one where the pattern disappears after alignment, one where real change remains, and one where the evidence is ambiguous. Ask for a decision plus one sentence describing what would strengthen or weaken it. This prevents the lesson from becoming “misregistration is always the answer.” The habit you want is comparison of explanations.

Finally, return to ordinary PSLE language. Ask: What was observed? What was inferred? What other variable changed? What evidence makes one explanation stronger? How far can the conclusion travel? The specialist object becomes useful because it exercises familiar scientific inquiry under unfamiliar conditions.

Authoritative Sources and Official Frame

U.S. Geological Survey — SRTM NED Vertical Differencing, including the effect of horizontal registration error on apparent elevation change.
U.S. Geological Survey — example of residual misregistration causing false detection of vertical change.
Singapore Examinations and Assessment Board — PSLE, current official examination information.
Ministry of Education Singapore — 2023 Primary Science Syllabus.

The official Primary Science frame values evidence-based inquiry, interpretation and evaluation rather than unsupported certainty. This article does not claim any examiner-specific wording rule or mark-scheme shortcut. Its purpose is to train a durable scientific habit that transfers to unfamiliar evidence.

Quiet Return: The Map Is Evidence, Not the World Itself

A change map can be extraordinarily useful. It can reveal patterns too large, too slow or too spatially complex to see from the ground. But every map is made from observations, choices and transformations. If two observations are compared, their positions must be comparable before the difference can safely become a physical story.

So when a red edge appears beside a blue edge, do not rush to either belief or disbelief. Ask the better scientific question: What evidence shows that the datasets line up well enough for this difference to represent a real change? That question is small, calm and powerful. It turns a dramatic colour into an investigation.

Small Group Tutorials