eduKate Singapore

VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Next useful routeCanonical hubHELP

Why Study Science | Understand How Satellites Measure Earth from Space

—

by

eduKate Asia
NASA Aqua MODIS 14 September 2019 satellite image of fire smoke over Indonesian Borneo, showing its proximity to Malaysia

Why study Science? Because satellites can tell us important things about forests, cities, water and atmosphere without ever touching the ground below. Their measurements become reliable knowledge only when people understand the light, instruments, scales and assumptions involved.

Quick answer: Remote sensing makes it possible to investigate Earth from a distance. Science teaches how passive and active sensors work, why pixels and wavelengths matter, how vegetation indices are calculated and why a colourful image is not automatically proof of a particular cause.

This guide is for Singapore Science learners and families. The calculations and classification tables marked illustrative are invented teaching data, not measurements of any named property, live haze event or satellite pass. Official source and further-reading links appear at the end.

Contents

A satellite image is a measurement before it becomes a picture

From a classroom in Singapore, a student can examine an image showing smoke across Borneo, changing coastlines or the pattern of greenery inside a city. The view appears almost magical: we can investigate large parts of Earth without standing in those locations. Science explains how that view becomes possible and how much confidence to place in it.

An Earth-observing satellite carries instruments that detect electromagnetic radiation or transmit signals and measure the return. Those measurements are processed into data, sometimes shown as recognisable images. A picture with colours and labels may be the final display of a much longer chain involving the instrument, its calibration and the physical surface being observed.

This is a strong reason to study Science. It trains us to distinguish an object from a measurement of that object, and an attractive visualisation from the interpretation based on it. Satellites help people ask questions about the environment, but their outputs do not make the need for evidence disappear.

The central learning question is straightforward: what did the instrument really detect, under which conditions, and what alternative explanations could produce a similar result?

What remote sensing means

Remote sensing is the science of obtaining information about an object or area from a distance. The platform could be an aircraft, a drone or a satellite. The defining feature is that measurements are taken without direct physical contact with the target in the ordinary sense.

NOAA’s National Ocean Service distinguishes passive sensors, which detect naturally reflected or emitted radiation, from active sensors, which send out a signal and measure its return. Both can be useful, but neither is an unlimited window into everything beneath Earth’s surface.

A conventional photograph taken from space is one example, not the entirety of remote sensing. Thermal infrared measurements, microwave radar and other instruments can answer questions that a familiar visible-light picture cannot.

Learning the definition changes how pupils read a news article. They ask which kind of sensor was used before treating the phrase ‘satellite data’ as a complete description.

The Sun helps many passive sensors observe Earth’s surface

Much visible and near-infrared Earth observation depends on sunlight striking a surface and some of that energy travelling back toward the sensor. Water, vegetation, soil and built surfaces can reflect incoming radiation differently across wavelengths.

A passive optical satellite therefore observes interactions between solar radiation, atmosphere and surface. The recorded value is affected not only by the material below, but by illumination, viewing geometry and what happened to the light along its path.

The key misconception to correct is that the satellite simply photographs the true colour of everything it sees under all conditions. A sensor may record bands that the human eye cannot see, and its measurements may require processing to become interpretable.

Science combines Physics and Earth studies here. Reflection, absorption and wavelength—themes from a school light chapter—are used to investigate landscapes that extend over enormous distances.

Why clouds hide some measurements

Clouds can obstruct the view of the ground for many optical sensors. A bright cloud top may dominate a pixel even when the question concerns the plants or buildings underneath it. Thin haze, water vapour and atmospheric scattering can also influence recorded signals.

One satellite pass might be unusable for assessing a small forest patch because it is cloudy. The correct response is not to interpret the white cloud as bare ground or to assume no vegetation exists. Scientists can use cloud masks, other dates or a different sensor suited to the task.

Students should examine whether an image shows the land surface, the cloud layer or a processed estimate produced with additional assumptions. Those are not interchangeable kinds of evidence.

The same habit applies to experiments: when the measurement channel is blocked or contaminated, the data must be handled honestly rather than quietly turned into a result.

Active radar asks a different physical question

An active radar sensor sends microwave energy toward Earth and measures the radiation returned to the instrument. The response depends on surface properties, geometry, roughness, moisture and the wavelength and polarisation used.

Synthetic aperture radar can obtain useful observations day or night and, for suitable microwave bands, through many cloud conditions that block visible light. That does not mean radar can see all objects through everything, or that a bright radar signal always means a wet surface.

Radar brightness and optical brightness are different measurements. A flat water surface may be dark in some radar images because energy reflects away from the instrument, while rough water can produce a different response. Interpretation requires the observing geometry and physical context.

This comparison teaches students an important rule: the best measurement method depends on the question, not on which image looks most like an ordinary photograph.

Why satellites measure more than one wavelength

A multispectral sensor records data in several wavelength bands. Some are in the visible range; others may detect near-infrared or shortwave-infrared radiation. Different materials and surfaces can show distinct combinations of responses across these bands.

A human eye merges a limited part of the spectrum into perceived colour. A multispectral instrument preserves separate measurements that can be compared quantitatively. The purpose is not necessarily to make a realistic image, but to gather evidence about land cover or material condition.

When three measured bands are assigned to red, green and blue display channels, the resulting image can be called a false-colour composite. That does not mean the image is fraudulent; it means the display colours have a defined mapping that the caption should explain.

Understanding how colour is assigned makes Science students better readers of maps, microscope images and medical visualisations as well as satellite pictures.

What a pixel really represents

A digital satellite image is built from picture elements called pixels. Each value summarises radiation or a derived measurement over a ground area related to the sensor’s spatial sampling and processing.

A 10-metre pixel is not a tiny photograph of one single leaf or object. A ground square of roughly 10 metres by 10 metres may contain tree canopy, paths, shadows and other materials. The recorded value can combine their contributions.

Students may zoom in until the image fills a large screen, but enlarging a pixel does not create previously unmeasured detail. Interpolated or sharpened displays can look clearer without providing a new independent observation.

Science gives a useful distinction between the apparent detail of a display and the actual ground sampling distance and resolving power of the instrument.

Worked problem: area represented by image pixels

Illustrative arithmetic, not a specific satellite product: imagine a hypothetical image with square pixels representing 10 metres by 10 metres on the ground. Each pixel then corresponds to 100 square metres in this simplified geometry.

One hectare is 10,000 square metres. Therefore, 100 non-overlapping full pixels of 100 square metres would cover one hectare in the idealised grid, because 100 × 100 = 10,000.

Now consider a forest patch extending across irregular boundaries. Counting every touched pixel as completely forested would overstate area; omitting boundary pixels could understate it. Mixed pixels and classification uncertainty matter.

This calculation teaches two lessons. Unit conversion is necessary, and numerical precision does not eliminate uncertainty about what the pixels contain. A count of pixels is not automatically a verified land-cover survey.

  • Pixel side length: 10 metres.
  • Simplified area per pixel: 10 × 10 = 100 square metres.
  • One hectare: 10,000 square metres.
  • Idealised whole pixels per hectare: 10,000 ÷ 100 = 100.
  • Boundary and mixed pixels need interpretation, not blind counting.

Spatial resolution does not equal positional accuracy

Spatial resolution describes aspects of the detail a sensor can distinguish or its ground sampling scale. Positional accuracy concerns how closely image features align with their actual geographic locations.

An image may have small pixels but still be misregistered relative to a map. Two clear images taken at different dates can appear to show movement or land-cover change when the alignment was imperfect.

Scientists use georeferencing and geometric correction to place data within a consistent coordinate system. They also inspect control points and uncertainty rather than assuming that two image grids overlap perfectly.

The lesson extends to comparing maps and photographs: a more detailed image does not automatically have better geographic accuracy.

Temporal resolution asks how often we can observe

Temporal resolution concerns the timing and frequency with which a sensor observes an area. Some systems return frequently with wide coverage, while others offer finer detail but visit the same location less often.

For a rapidly developing flood, repeated observations can matter more than the ability to see a very small object. For mapping a fixed infrastructure feature, higher spatial detail may be worth the slower revisit time.

A satellite’s stated repeat interval does not guarantee a usable cloud-free image on every pass. Optical coverage depends on weather and processing, while viewing arrangements and sensor modes affect actual revisit opportunities.

Students learn that the best dataset is chosen according to the change rate of the process being studied.

Spectral and radiometric resolution are different again

NASA Earthdata identifies spatial, temporal, spectral and radiometric resolution as four useful considerations for Earth-observation datasets. Spectral resolution concerns the number and widths of wavelength bands; radiometric resolution concerns how finely a sensor distinguishes differences in measured signal.

A camera with many narrow bands is not automatically more informative for every task. Some applications benefit from carefully chosen broad bands, while others need fine spectral distinctions. Signal-to-noise limitations and calibration also influence what can be inferred.

More numerical levels in a detector do not necessarily mean a physically accurate result. A 16-bit data file can still contain uncertainty or require conversion from stored numbers to scientifically meaningful units.

This is a practical reason to study Science: data quality is multidimensional, and the word ‘high resolution’ is incomplete unless we know which resolution is meant.

The difference between stored digital numbers and reflectance

A satellite instrument often records digital values that pass through calibration and processing. An analyst may convert those numbers into radiance or reflectance estimates before comparing materials or dates.

Reflectance describes the proportion of incoming radiation reflected under a defined measurement model. It is not equivalent to the raw brightness displayed on a monitor. Atmospheric correction and calibration can be important when comparing observations taken under different conditions.

If one image looks brighter because the display contrast was stretched, that is not proof that the real landscape became more reflective. The pixel’s physical measurement and the visual display transformation must be separated.

This distinction helps learners resist a common mistake: treating changes caused by image editing as changes in the Earth.

How vegetation indices turn bands into a new quantity

Green plants commonly absorb substantial red light for photosynthetic processes and reflect strongly in parts of the near-infrared. Remote sensing uses this contrast as one way to describe vegetation.

The Normalized Difference Vegetation Index, or NDVI, is calculated as (NIR − Red) divided by (NIR + Red), using appropriate reflectance values. Under conventional inputs its values lie between −1 and +1. It can help indicate vegetation greenness, but does not uniquely identify plant health or species.

A high value may be associated with dense green vegetation, yet soil background, atmospheric conditions, season, species and sensor processing can also affect the result. An index should be interpreted together with the question and data quality.

Science here joins plant biology, algebra and electromagnetic radiation. A familiar school topic becomes a quantitative tool for investigating land.

Worked example: calculate two fictional NDVI values

Invented reflectance values: Pixel A has near-infrared reflectance 0.48 and red reflectance 0.12. Its NDVI is (0.48 − 0.12) ÷ (0.48 + 0.12) = 0.36 ÷ 0.60 = 0.60.

Pixel B has NIR reflectance 0.30 and red reflectance 0.20. Its NDVI is 0.10 ÷ 0.50 = 0.20. Under these assumptions, A shows the stronger relative near-infrared versus red contrast typically associated with green vegetation.

The valid conclusion is about the spectral index. The values do not prove that A is a healthier tree, that B is diseased or that either pixel contains just one plant. The surfaces might differ in species, cover fraction or other factors.

A student who can calculate both values and then state the boundary of the interpretation has made real progress beyond memorising an equation.

Why a cloudy pixel is not automatically a dead tree

If a cloud obstructs the surface, the sensor may record light influenced strongly by the cloud rather than vegetation. Computing a vegetation index for that pixel without masking or quality checks can produce a number with little relevance to the plant below.

The same problem can arise from shadows, haze or unsuitable atmospheric correction. A quantitative index is only as reliable as the measurements and assumptions used to produce it.

Students should check the quality layer supplied with a scientific dataset. A missing, invalid or cloud-contaminated value should not be silently turned into an ordinary observation.

This is a strong scientific habit: the correct response to unavailable evidence is often to mark uncertainty and obtain better observations, not fill a gap with a confident story.

Why repeated images reveal more than one image

A single satellite image can show the appearance of an area on a particular date. A sequence of comparably processed images can reveal changes in vegetation, coastline shape, built surfaces or water conditions.

Yet not every pixel difference is a real land change. Seasonal leaf growth, different sunlight angles, cloud shadows, instrument calibration or imperfect registration can create apparent differences.

Before claiming deforestation or flood damage from two pictures, analysts examine dates, sensor bands, classifications, viewing geometry and ground evidence. Some land changes may also happen between observation dates and remain unseen.

Science trains students to reconstruct what changed in the world while separating that signal from changes in the observing process.

A change map needs a definition of change

The word ‘changed’ might mean that a pixel became less green, that land cover switched from forest to building, or that water level crossed a threshold. Each definition needs a method and validation.

Suppose a pixel’s NDVI falls from 0.60 to 0.30. That is an index decrease of 0.30 under the stated values. It is not automatically proof that 50 per cent of vegetation disappeared. The relationship between NDVI and vegetation cover is not generally a simple linear identity.

A useful report states the measured indicator first and then evaluates possible causes with context. A numerical change in an index should not be renamed a percentage of forest lost unless a suitable model supports the conversion.

This lesson is transferable to exam graphs, medical markers and economic indicators.

Satellites help scientists study smoke and haze

Earth-observing instruments can record smoke plumes and aerosol-related signals over broad regions. This helps scientists study how particulate pollution is transported by winds and how it changes over time.

An image of smoke over Borneo or Sumatra can illustrate regional transport relevant to Singapore. However, a satellite picture of a plume does not automatically give the concentration someone breathes at street level in Singapore at that moment.

Near-surface air-quality assessment requires suitable atmospheric measurements and modelling. Satellite observations and ground sensors provide related but different evidence, and scientists can combine them cautiously.

This provides a useful bridge to the eduKateSG guide on Singapore PSI and PM2.5: source location, column observation, averaging period and local exposure are not the same variable.

How satellites reveal coastal and marine patterns

Satellites can provide information about sea-surface temperature, ocean colour, shoreline positions and some surface water patterns. Different instruments and processing methods are needed for each question.

A visible change in ocean colour could have multiple explanations involving suspended sediment, phytoplankton, shallow seabed, illumination or processing. A bright streak is not automatically pollution, just as a dark patch is not necessarily deep water.

NOAA identifies coastal mapping and shoreline-change monitoring as important remote-sensing applications. Ground surveys, tides and the timing of the image still matter for interpreting an apparent coastline.

Students learn that a satellite can complement a field observation without replacing every local measurement.

Why flood mapping may use radar

Flooding can coincide with heavy clouds that obstruct ordinary optical views. Certain radar sensors can provide valuable information about surface water during cloudy periods because their microwave measurements are less dependent on sunlight and visible-light transmission.

But radar response changes with surface roughness, vegetation, urban geometry and other factors. Standing water beneath a dense canopy or among complex buildings can be difficult to interpret; radar is not an infallible flood-depth instrument.

A flood-extent map is not the same as a map of floodwater depth or an individual building’s structural condition. Those require additional data and site-specific interpretation.

Science helps students see how the choice of sensor arises from physical conditions and how a useful map still has limits.

Temperature from orbit is not simply air temperature

Thermal infrared instruments detect emitted radiation that can be used to estimate land-surface temperature under appropriate modelling. This is related to but not identical to the air temperature measured by a standard weather station.

A road surface exposed to sunlight can be hotter than the air above it. Trees and shaded surfaces may differ from nearby roofs, but interpretation depends on the observation time, atmospheric correction and materials.

An image labelled ‘surface temperature’ should not be quoted as if it measured the temperature a person feels at head height. The two quantities interact but are not interchangeable.

This helps pupils understand urban heat questions without mistaking one colourful thermal map for a complete account of human heat exposure.

Singapore’s urban landscape needs careful map comparisons

Singapore combines buildings, transport corridors, parks, waterways and coastal edges within a compact area. Remote sensing can help describe broad land-cover patterns or track visible changes over time, provided the data and classifications are suitable.

A tree canopy and a small roof can fall within one coarse pixel; an elevated structure may create shadow; construction activity can alter surface reflectance before land use has definitively changed.

Students who investigate their own neighbourhood should therefore check scale and source instead of relying on a pixelated screenshot. A mapped category can be useful while still being too coarse to settle an individual property-level question.

The value of Science lies in choosing evidence appropriate to the geographic detail of the claim.

Ground truth checks what the satellite interpretation means

Researchers compare remote-sensing results with independent ground observations, reference maps, field measurements or other suitably reliable evidence. This process is often called validation or ground truthing.

If a model labels a pixel as forest, a reference dataset may help determine whether the label is correct. However, reference data have their own dates, definitions and errors. Validation requires aligning the comparison criteria.

An algorithm’s high accuracy on one sample is not a guarantee of equal performance in every season or country. A model trained on one kind of landscape may perform differently elsewhere.

Science encourages a culture of checking rather than accepting a compelling map because it was produced by a computer.

Accuracy percentages need a suitable denominator

Invented teaching confusion matrix: a land-cover model evaluates 100 reference pixels. It correctly identifies 36 of 40 forest pixels and 54 of 60 non-forest pixels. Its total correct count is 90, so overall accuracy is 90 per cent.

However, four forest pixels were missed and six non-forest pixels were falsely labelled as forest. If the task concerns accurately identifying scarce forest patches, these different errors may have different consequences.

The calculation also assumes the reference sample adequately represents the area and categories of interest. Choosing 100 unusually easy pixels could make the score misleading.

Science teaches students to name the class definitions, sampling method and errors before making claims about the reliability of a map.

Machine learning does not remove the need for interpretation

Computer algorithms can classify satellite data, locate patterns and help manage enormous image collections. Their outputs are shaped by training data, labelled examples, model design and the conditions under which new observations are processed.

A model may mistake shadows for water, bright roofs for bare ground or seasonal vegetation changes for permanent clearing. These errors can sometimes be reduced with better data and validation, but not erased by calling the method artificial intelligence.

Students should ask whether the algorithm’s predicted category is supported by independent evidence. A probability score is not the same as a measured physical property or certified truth.

This is a valuable connection between Science and digital literacy: computational sophistication does not replace the logic of evidence.

Why uncertainty belongs on satellite maps

Satellite products can include uncertainty estimates, cloud masks, flags or quality categories. A map rendered in crisp colours may conceal those additional fields unless the analyst makes them visible.

Suppose a measurement is considered low quality because of clouds or sensor conditions. Excluding it may leave a gap, while displaying it as ordinary data may create false confidence. The proper choice depends on the scientific question and quality criteria.

Students learn to ask for the legend, units, dates and quality information. The absence of a visible warning on a map does not prove there are no limitations.

A well-designed scientific image communicates both what the instrument observed and what it could not reliably determine.

A 14-day inquiry plan using existing data

Begin with an official NASA Earthdata overview and identify passive versus active sensing. Ask the student to explain how a surface feature becomes a detector signal. Next compare visible and false-colour images, recording what each displayed colour represents.

During the middle sessions, practise scale conversions, pixel-area estimates and the fictional NDVI calculation. Introduce the four types of resolution and ask which type would matter most for detecting an afternoon flood versus mapping broad forest cover.

In the second week, compare two prepared images from different dates. Require the learner to check cloud cover, alignment and the source before writing that a surface changed. Finish with the fictional classification-accuracy table and explain why 90 per cent overall accuracy can hide important errors.

These are desk-based scientific tasks. Students do not need to launch a drone, enter restricted places or claim access to live remote measurements.

  • Days 1–3: sensors, reflected light and radar.
  • Days 4–5: pixels, scale and geographic alignment.
  • Days 6–7: four types of resolution and data quality.
  • Days 8–9: NDVI and responsible index interpretation.
  • Days 10–11: compare images across dates.
  • Days 12–13: false classifications and validation.
  • Day 14: independently interpret a new image with evidence limits.

What parents can ask about a spectacular image

Ask which satellite and instrument produced the image. Then ask whether the colours are intended to resemble normal human vision or show selected wavelengths. Encourage your child to find the legend, date and scale before offering an explanation.

If a student concludes that every red area in a false-colour picture is on fire, ask what the red display channel represents. If a tree seems to have disappeared between dates, ask whether the images are properly aligned and free of clouds.

These questions do not require parents to be remote-sensing specialists. They simply make the learner accountable to what was actually measured.

The result is more than image appreciation: it is the ability to read a scientific visualisation without mistaking its design for direct physical truth.

What satellite Science cannot guarantee

An image cannot automatically tell us whether a specific building is safe, who caused a fire, how much pollution a person inhaled or the exact reason a patch of vegetation changed. Those claims need additional evidence.

Even high-quality satellite datasets vary in resolution, timing, calibration and environmental suitability. A claim about a particular street or person’s exposure may exceed the scope of a regional image.

The responsible approach is to treat remote sensing as a powerful observational method that is strongest when paired with suitable models, independent checks and clear uncertainty.

Understanding both the capability and limitation of the tool is the educational achievement.

The lasting reason to study Science

Satellites show how a simple physical interaction—radiation reflected, emitted or returned to a sensor—can become evidence about Earth’s forests, coasts, atmosphere and cities. The path from signal to conclusion is built from many careful scientific steps.

A learner who can inspect wavelengths, pixels, spatial scale, time windows and uncertainty gains a method for reading far more than satellite imagery.

That is why Science is worth studying: it teaches us to understand what a measurement means, challenge an unwarranted interpretation and use evidence responsibly to learn about a world bigger than our immediate view.

Frequently asked questions

What is remote sensing?

Collecting information about an object or area from a distance through an instrument, often on a satellite or aircraft.

What is the difference between active and passive sensors?

Passive sensors detect naturally available radiation; active sensors transmit a signal and measure its return.

Can optical satellites see through every cloud?

No. Clouds can obstruct surface information in many visible and infrared products. Suitable radar can complement them.

Are false-colour satellite images fake?

Not necessarily. Colours are assigned to measured bands so that features or wavelengths can be interpreted; the mapping must be explained.

What does a 10-metre pixel mean?

In a simplified ground-sampling example it represents roughly a 10 by 10 metre ground area; it does not show individual objects inside that area.

What is NDVI?

An index combining red and near-infrared reflectance, commonly used to study vegetation greenness.

Can NDVI prove a tree is healthy?

No. It can be useful evidence but is affected by cover, species, season, sensor conditions and other factors.

Does a thermal satellite measure ordinary air temperature?

Often it estimates surface temperature, which is not identical to air temperature at a weather station.

Can a satellite map establish an individual’s haze exposure?

No. Ambient observations and exposure depend on different quantities and may require ground data and time-location information.

Why must images from different dates be aligned?

Misregistration can create false differences that look like movement or land-cover change.

What four resolutions are relevant?

Spatial, spectral, radiometric and temporal resolution; their importance depends on the question.

How should I judge a claimed change from space?

Check sensor, date, band choice, cloud quality, geographic alignment, scale and independent validation.

Primary sources and connected Science reading

NASA Earthdata, NOAA and USGS provide the scientific foundations for the remote-sensing definitions, resolution types and vegetation-index calculation described here. The published imagery in eduKateSG’s media library is used as a historical or illustrative reference and is not evidence of current conditions.

eduKateSG’s learning principle is to make understanding explicit: identify the variable, connect the mechanism, test a changed case, check the evidence and practise independently. Properly taught kids shine a bright light into the future.