Chapter 1 · Maps and coordinatesLesson 1.6Humanities track
Satellite images: remote sensing, spectral bands and false colour
Why does a forest look red in some satellite images when its leaves are green?
Before you read: what do you think?
A forest appears bright red in a false-colour satellite image. What does that colour show?
See the answer
The forest is healthy and dense. The forest is healthy and dense. In false colour, near-infrared is displayed as red. The better plants grow, the more strongly they reflect it. A dried-out forest would look duller.
From 5 am on 14 September 2024, floodwaters struck communities in Galați and Vaslui counties, affecting more than 5,000 households. That day, the Romanian authorities asked the European Copernicus service for maps of the flooded areas made from satellite images.
What the satellite sees and the eye cannot
A leaf looks green, yet it reflects most strongly a kind of radiation the eye cannot see: near-infrared. Chlorophyll in leaves absorbs visible light strongly, while the structure of the leaves reflects near-infrared strongly. Water, by contrast, absorbs near-infrared. So in the near-infrared band, a forest is very bright and a lake is almost black.
An Earth observation satellite measures radiation reflected or emitted by the land surface. Getting information about an object from a distance, without touching it, is called remote sensing. The satellite sensor measures several ranges of wavelengths separately. These are called spectral bands: blue, green and red, which the eye can see, and infrared, which it cannot.
Natural colour and false colour
A screen has three colour channels: red, green and blue. Put the band of the same colour on each channel and the result is a natural-colour image, close to what a person would see from space.
In a false-colour image, at least one channel receives a band the eye cannot see. The most widely used combination puts near-infrared on the red channel. Plants, which strongly reflect infrared, then appear red: the better they grow and the denser they are, the brighter the red. Water appears black or dark blue, and towns silver. False colour is a convention chosen to highlight a feature; the band combination must be stated beside the image.
Three kinds of resolution
Spatial resolution is the length of the side of the ground area covered by one pixel. The first Landsat satellite, launched by the United States on 23 July 1972, had 80 m pixels. Landsat 8 and 9 have 30 m pixels in most bands, and the European Sentinel-2 satellites have 10 m pixels in four bands.
Temporal resolution is the interval before a satellite returns over the same place: 16 days for one Landsat satellite, five days for the pair of Sentinel-2 satellites. Spectral resolution describes how many bands the sensor measures and how narrow they are.
The Sentinel satellites belong to the European Union’s Copernicus programme. Anyone can use Landsat and Sentinel images free of charge. The Landsat archive, which began in 1972, allows the same place to be compared over more than 50 years.
For the curious
Sentinel-2 measures 13 bands and Landsat 9 measures 11. On Sentinel-2, the bands without 10 m pixels have 20 m or 60 m pixels.
What cannot be seen or inferred
Optical sensors, such as those on Landsat and Sentinel-2, cannot see through clouds. Sentinel-1 satellites use radar: they send waves towards the ground and measure the echo, producing images day and night in any weather. For this reason, floods are usually mapped with radar.
A pixel contains a single value for its entire area. A 10 m pixel at the edge of a lake mixes water, reeds and shore. To see changes from one year to the next, compare images from the same season: a deciduous forest looks different in January and July even if no trees have been cut down. The image shows the state of the surface; the cause of a change is established using other data and checks on the ground.
Worked example
A vegetation index and an area
The NDVI vegetation index shows how green a pixel is. It is calculated from the near-infrared (NIR) and red (R) bands: NDVI = (NIR − R) ÷ (NIR + R). Values close to zero indicate no vegetation; values of 0.8–0.9 indicate the densest green vegetation.
In an exercise, one pixel reflects 0.50 of the incoming near-infrared radiation and 0.08 in red. NDVI = (0.50 − 0.08) ÷ (0.50 + 0.08) = 0.42 ÷ 0.58 = 0.72, so the pixel is covered by dense vegetation. The neighbouring pixel reflects 0.25 in infrared and 0.20 in red. NDVI = 0.05 ÷ 0.45 = 0.11, so green vegetation there is very sparse, as on almost bare soil.
Area: a Sentinel-2 pixel with a side of 10 m covers 10 × 10 = 100 m². A patch of 250 pixels of dense vegetation covers 250 × 100 = 25,000 m², or 2.5 hectares.
In Romania
The Danube Delta and the Galați floods from space
In a natural-colour Sentinel-2 image of the Danube Delta from April 2020, with 10 m pixels, expanses of reeds appear brown, as they do in early spring before turning green. In the Black Sea off the river mouths, the water is pale green from sediment brought by the Danube. South of the delta, lakes Razim and Sinoe appear emerald green because of algae.
During the floods of 14 September 2024 in Galați and Vaslui counties, Romania’s request was registered by the Copernicus Emergency Management Service as EMSR758. The service was tasked with mapping the extent of the water from satellite images and assessing damage, so response teams would know which communities were under water.
In brief
- Remote sensing gathers information about the Earth’s surface from a distance by measuring reflected or emitted radiation.
- Sensors measure several spectral bands separately; vegetation reflects near-infrared strongly, while water absorbs it.
- A natural-colour image uses the red, green and blue bands; a false-colour image also uses invisible bands, and vegetation usually appears red.
- Spatial resolution is pixel size (Landsat 30 m, Sentinel-2 10 m); temporal resolution is the return interval (16 days and five days respectively); spectral resolution concerns the number and width of bands.
- Optical sensors cannot see through clouds; radar on Sentinel-1 works day and night in any weather.
- For changes from one year to the next, compare images from the same season and check the cause of a change on the ground.
Common mistakes
WrongA red forest in a satellite image means the image is wrong or the forest has dried out.
RightIn false colour, near-infrared is displayed as red. Healthy vegetation reflects infrared strongly, so it appears bright red; a dried-out forest would be duller.
WrongSatellites can always see, even through clouds.
RightOptical sensors see only by day and under clear skies. Radar satellites such as Sentinel-1 work through clouds and at night.
WrongIf a pixel is 10 m across, the image shows every object larger than 10 m.
RightA pixel has a single value that mixes everything within its 100 m². An object can be identified with confidence only if it covers several pixels.
Test yourself
Six questions. For the first five you see the explanation at once, with the paragraph of the lesson that supports it; for the ones with a written answer you compare with a good answer.
Why does vegetation appear red in a false-colour image?
Answer
b) Because infrared reflected by plants is displayed as red. A screen colour shows the band assigned to that channel. Plants strongly reflect near-infrared, which has been assigned to red.
An image has 20 m × 20 m pixels. A lake covers 25 pixels. What is its area in square metres?
m²Answer
10,000 m². One pixel covers 20 × 20 = 400 m². 25 × 400 = 10,000 m², or one hectare. Multiply by the pixel area, rather than its side.
What does the spatial resolution of a satellite image show?
Answer
d) The length of the side of the ground area covered by one pixel. The number of bands concerns spectral resolution; the return interval is temporal resolution.
For three days, clouds cover a flooded area. What kind of satellite can still show the extent of the water?
Answer
a) A radar satellite, such as Sentinel-1. Radar sends its own waves and measures their echo, so it does not depend on sunlight and is not stopped by clouds.
How does a lake appear in an image made using only the near-infrared band?
Answer
c) Almost black. Water absorbs near-infrared, so almost none returns to the sensor. The boundary between water and land is very clear in this band.
Someone puts two images of the same wooded hill side by side, one from January and one from July, and claims the colour difference proves deforestation. What would you say, and which images should be compared?
A good answer
The difference could come from the season alone: in winter, deciduous trees have no leaves and reflect light differently from in summer. To see deforestation, compare two cloud-free images from the same month in different years, from the same type of satellite. If the difference remains, check its cause on the ground or with other data.
A complete answer names seasonal changes in vegetation, asks for images from the same season and separates detecting a change from establishing its cause.
Terms
- remote sensing
- Obtaining information about the Earth’s surface from a distance by measuring reflected or emitted radiation.
- spectral band
- A range of wavelengths measured separately by a satellite sensor, such as red or near-infrared.
- false colour
- Display of an image in which at least one colour channel receives a band invisible to the eye.
- spatial resolution
- The length of the side of the ground area covered by one image pixel.
- temporal resolution
- The interval before a satellite returns over the same place.
- radar
- A sensor that sends waves towards the ground and measures their echo; it works day and night in any weather.
Sources and link to the programme
- NASA Earth Observatory – Why is that forest red and that cloud blue? How to interpret a false-color satellite image – Natural and false colour; near-infrared on the red channel; plants red and brighter when growing well; water black and blue, towns silver; water absorbs near-infrared
- NASA Earth Observatory – Measuring vegetation (NDVI & EVI) – Chlorophyll absorbs visible light, leaves reflect near-infrared; NDVI formula; meaning of values (zero, 0.8–0.9)
- NASA – Landsat 1 – Launch on 23 July 1972; spatial resolution of 80 m
- NASA – Landsat 9 – 30 m resolution, 11 bands, 16-day return
- NASA – Landsat 8 – 30 m resolution, 16-day return
- ESA – Sentinel-2: Facts and figures – 13 spectral bands; 10 m in four bands, 20 m and 60 m in the others; five-day return with two satellites
- Copernicus SentiWiki – Sentinel-1 mission – Radar working day and night in any weather
- Copernicus – About Copernicus – Copernicus, the Earth observation component of the European Union space programme
- ESA – Earth from Space: Danube Delta – Natural-colour Sentinel-2 image from April 2020 at 10 m: brown reeds, pale green sediment in the sea, lakes Razim and Sinoe emerald green because of algae
- Copernicus EMS – Flood in Galati and Vaslui counties, Romania (EMSR758) – Floods from 5 am on 14 September 2024; more than 5,000 households affected; request to map the extent of water and damage
- NASA Science – Landsat 9, instrumentele – OLI-2: 30 m in optical bands, 15 m in the panchromatic band; TIRS-2: 100 m in thermal bands; 11 bands in total
From the programme: “Using satellite images to analyse terrestrial space” (specialisation curriculum, humanities track). Programme competences covered: 2.1, 2.5, 5.1.