This site uses cookies to offer you a better browsing experience. Read about how we use cookies.
Multispectral satellite images are fundamental in deciphering Earth surface features like vegetation or open water areas. With this activity, students will understand the importance of this data for preserving life on Earth and get a hands-on know-how of how real remote sensing satellite data is being analysed, using an educational software package (LEO Works).
There are 4 attachments: Landsat Jan2002, Landsat Jul2002, Worksheet editable and Worksheet PDF.
Students will:
Remote Sensing
The term ‘remote sensing’ indicates a measurement technique that probes and analyses the Earth from outer space. Alongside classical in-situ methods like weather data analysis, field surveying or sample collection, satellite-based measurements are becoming an increasingly important source of data. Their advantage is the fast and complete coverage of large areas. However, satellite data are not always easy to interpret and need substantial analysis.
The most abundant remote sensing devices are weather satellites. By employing suitable sensors, they provide information about cloud coverage, temperature distributions, wind speed and directions, water levels and snow thickness. Keeping the evolving climate change in mind, these data play an increasingly important role in disaster management during draughts and floods and for climate simulations and atmospheric gas content and vegetation monitoring. In addition, urban and landscape management benefit from satellite data.
The first weather satellites were launched by NASA as early as 1960. In the early 1970s, NASA started their earth observation programme using Landsat satellites (Figure 1). In Europe, France was the first, with their SPOT satellite fleet. They were followed by the remaining European countries in the 1990s after the foundation of ESA, the European Space Agency.
The Copernicus Programme
Since 1997, the USA and NASA have been developing a massive programme to explore the Earth, designated the Earth Observation System, which consists of a large number of different satellites. The European equivalent, the Global Monitoring for Environment and Security (GMES), has been under development since 1998. In 2012, the programme was renamed to Copernicus. Information products for six applications are being derived from the satellite data: ocean, land and atmosphere monitoring; emergency response; security and climate change. The data products are offered to the public free of charge. They are supplied via two branches: space-based remote sensing devices (satellite component) and airborne, ground and marine probing (in-situ component). The core of the satellite component is the fleet of Sentinel satellites built exclusively for the Copernicus projects. They are supplemented by other domestic and commercial partner missions. The first Sentinel satellite (Sentinel 1-A) was launched in 2014. Sentinel-2A and 3-A followed in 2015 and 2016, respectively.
Figure 2: Computer model of the Sentinel-2A satellite launched on 23 June 2015 (ESA/ATG medialab, http://www.esa.int/spaceinimages/Images/2014/07/Sentinel-2_brings_land_into_focus).
Electromagnetic Spectrum
Light is the radiation that the human eye can see and interpret. However, the full range of electromagnetic radiation (the spectrum) is much wider. The part that is invisible to us can be detected using special cameras, such as the ones that are part of astronomical telescopes and satellites. Figure 3 provides a good overview on the different kinds of radiation.
Figure 3: The spectrum of electromagnetic radiation. The visible light is only a very small part inside the full range (Inductiveload, https://commons.wikimedia.org/wiki/File:EM_Spectrum_Properties_reflected.svg, “EM Spectrum Properties reflected“, cropped by Markus Nielbock, https://creativecommons.org/licenses/by-sa/3.0/legalcode).
Multispectral Imaging
One of the core purposes of earth observation and remote sensing is taking and analysing pictures. Similar to modern astronomy, taking images with different spectral filters is very diagnostic when identifying and analysing terrestrial surface features. For this kind of data acquisition, the cameras rely on the sunlight that illuminates the Earth’s surface. Hence, they receive the portion of the sunlight that is reflected by the various surface features. Compared to the incident sunlight, the reflected light is modified by brightness and spectral composition.
Table 1: Spectral bands of the MSI camera of the Sentinel-2A satellite (Sentinel Online, https://sentinels.copernicus.eu/web/sentinel/technical-guides/sentinel-2-msi/msi-instrument)
The spectral bands of the multi-spectral instrument (MSI) camera of the Sentinel-2A satellite is given as an example in Table 1. For example, band 2 covers a wavelength range of 0.490 to 0.065 µm. The smallest feature that can be seen in this band would be 10 m across. These bands cannot be chosen arbitrarily because of the wavelength-dependent transparency of the Earth’s atmosphere (grey area in Figure 4). They are referred to as ‘spectral windows’. The main reason for the wavelength ranges, where the atmosphere blocks external radiation, is water vapour. Therefore, observations with cameras have to be designed such that only wave bands where the radiation is transmitted well enough to receive a good signal are used. Thus, these ranges are the ones the optical filters of the cameras are designed for.
Figure 4: Graphical representation of the spectral bands of MSI/Sentinel-2A compared to the cameras of the Landsat 7 and 8 satellites. The axes depict the wavelength in nanometres (1 nm = 10-3 µm = 10-9 m) and the terrestrial atmospheric transmission (grey) in percent (NASA, https://landsat.gsfc.nasa.gov/sentinel-2a-launches-our-compliments-our-complements/).
A proper choice of optical filters not only permits differentiation between water and landscape but also allows the state of vegetation or surface conditions to be deciphered. For instance, Figure 5 indicates a noticeable difference between the reflective spectra of fresh (green curve) and dry (brown curve) grass. The main reason for this is the absorption power of chlorophyll. In particular, the transition between the red (band 4) and the infrared ranges (bands 7 to 9) sees a sudden jump in the spectrum of fresh, green grass, while the spectrum of dry grass remains rather constant. By subtracting the signals of the bands, one can distinguish between the two states.
Figure 5: Reflective spectra of fresh (green curve) and dry (brown curve) grass in a wavelength range covered by the MSI/Sentinel-2 bands (yellow curves labelled at the top of the graph). There is a strong jump in the green grass spectrum between bands 4 and 7 (USGS Spectral Viewer, NASA, http://landsat.usgs.gov/tools_viewer.php).
Satellite images contain pixel values that represent the brightness or intensity of the light reflected by the surface and detected in an optical band. They are usually displayed in greyscale. Combining these images following the rules of the additive mixture of colour stimuli enables the construction of colour images. When selecting the images of the spectral bands representing the colours red, green and blue, the resulting RGB image shows the colours in a realistic way (Figure 6, left).
Figure 6: Images obtained with MSI/Sentinel-2A. Left: Realistic RGB coloured image of the city of Milan; right: false colour visualisation of the area around the river Po, Italy. The colour red represents the near infrared band which is sensitive to green vegetation (Copernicus data 2015/ESA, https://directory.eoportal.org/web/eoportal/satellite-missions/c-missions/copernicus-sentinel-2).
Spectral Index
By merging data from different optical bands, a lot can be learnt about vegetation or construction areas in a qualitative way (Figure 6, right). If quantitative information is required, however, a more detailed analysis is needed. An established tool for this is the spectral index. This is a number that is calculated from data obtained at different wavelengths and allows a comparison of the relative brightness of different wavelengths of light that is reflected by the Earth’s surface.
Normalised Differenced Vegetation Index (NDVI) An important spectral index used to identify healthy vegetation is the Normalised Differenced Vegetation Index (NDVI). It is calculated from the measured intensities obtained in the red (R) and near infrared (NIR) spectral regimes. As mentioned, the transition between these bands helps distinguished between green vegetation and other features (Figure 5). It is calculated as follows:
NDVI = (NIR - R) / (NIR + R)
With: R: Intensity/brightness of reflected light in the red filter (ca. 0.6–0.7 µm) NIR: Intensity/brightness of reflected light in the near infrared filter (ca. 0.8–0.9 µm)
Figure 7: NDVI world map based on November 2007 data of the ‘Resolution Imaging Spectroradiometer (MODIS)’ of the NASA Terra satellite (NASA, http://earthobservatory.nasa.gov/IOTD/view.php?id=8622).
NDVI is measured using bands 4 and 8 of the Sentinel-2 MSI camera (Table 1). The difference between NIR and R is normalised by their sum, resulting in a range of values between -1 and +1. Negative values indicate water areas. A value between 0 and 0.2 represents nearly vegetation-free surfaces, while a value close to +1 indicates high coverage of green vegetation.
Normalised Differenced Moisture Index (NDMI) Another spectral index is the Normalised Differenced Moisture Index (NDMI) or Normalised Differenced Water Index (NDWI). It is sensitive to humid vegetation and open wetlands. It supplements the NDVI.
NDMI = (NIR - SWIR) / (NIR + SWIR)
With: NIR: Intensity/brightness of reflected light in the near infrared filter (ca. 0.8–0.9 µm) SWIR: Intensity/brightness of reflected light in the shortwave infrared filter (ca. 1.5–1.8 µm)
The NDMI helps distinguish between dry and wet areas.
Modified Normalised Differenced Water Index (MNDWI) The Modified Normalised Differenced Water Index (MNDWI) is regarded as an advanced version of the NDMI. It helps identify open wetland and excludes artificial buildings, vegetation and agricultural areas.
MNDWI = (G - SWIR) / (G + SWIR)
With: G: Intensity/brightness of reflected light in the green filter (ca. 0.5–0.6 µm) SWIR: Intensity/brightness of reflected light in the shortwave infrared filter (ca. 1.5–1.8 µm)
Open wetland shows higher positive values with the MNDWI than with the NDWI, while other landmarks like buildings, vegetation and crop land show negative values.
The LEO Works 4.0 software ESA has developed an educational tool for teaching and learning the basic steps of analysing satellite data. The latest version, 4.0, is being developed by Terrasigna in Romania. Since it is based on Java, it is independent of the operating system. It will be used for the present activity.
Figure 8: Launch window of LEO Works 4.0, a software for treating and analysing satellite data for educational purposes. It can be downloaded at http://leoworks.terrasigna.com and runs on a wide variety of operating systems.
Provide the students printed or digital copies of the worksheet. This includes information from the background information, which is needed to successfully analyse the data.
Install the LEO Works 4.0 software from http://leoworks.terrasigna.com and ensure students can access it on their computers. It is required in order for them to perform this activity.
Introduce the topic by asking the students what they know about observing the Earth. How can we observe the Earth, and what is remote sensing? What information can we collect through remote sensing and what are its uses? The most obvious answers should include weather satellites.
Ask the students if they know where the images in Google Maps or Google Earth come from. The source of the images is mentioned at the bottom of the screen. They might find names like SPOT or Landsat. Ask students to choose one of these satellite campaigns to research. Let them compile information on satellite launch dates, their orbits and countries of origin.
The activity is set up as a step-by-step process for analysing real satellite data. The exercise is interspersed with questions to evaluate the students’ understanding as well as to point to the relevance of the satellite data. Some tasks involve very similar and repetitive procedures that are used to reinforce the steps used in the analysis.
Analysis of satellite imagery data using LEO Works 4.0 This activity introduces the basic tasks for processing and analysing remote sensing satellite data.
The installed version already contains some example datasets that can be used for the purpose of the exercise. They are stored in the leoworks.data folder. When using MS Windows, it can be found in the user directory. From the existing datasets, the one labelled Venice will be used.
Reading the data After it is launched, the software presents its workspace as shown in Figure 9. Open the file Venice_Landsat_ETM_multispectral.tif by clicking on the first icon in the menu bar or via the menu File\Open\Single File Dataset(s). A window appears from which the file is selected (Figure 10).
Figure 9: LEO Works 4.0 workspace. The menu bar contains procedures and tools for displaying and analysing the data. There are three windows below that provide a list of the loaded data sets and image displays.
The file contains seven individual images obtained in seven bands of the camera “Enhanced Thematic Mapper Plus (ETM+)” of NASA’s Landsat 7 satellite (Table 2) covering the vicinity around the city of Venice in Italy. When the window Specify Subset appears, acknowledge by clicking OK.
Figure 10: Window for file selection.
The data automatically appear in the window to the upper left. The element Bands can be expanded by clicking on it to show the list of the seven images (Figure 11). They are labelled band_1 to band_7 and correspond to the spectral bands of Table 2.
Table 2: List of the seven spectral bands of the “Enhanced Thematic Mapper Plus (ETM+)” camera of the Landsat 7 satellite (Source: NASA; column with colours is not revealed to students).
Landsat 7 | Wavelength (µm) | Resolution (m) | Colour --- | --- | --- Band 1 | 0.450 – 0.515 | 30 | Blue Band 2 | 0.525 – 0.605 | 30 | Green Band 3 | 0.630 – 0.690 | 30 | Red Band 4 | 0.750 – 0.900 | 30 | NIR Band 5 | 1.550 – 1.750 | 30 | SWIR Band 6 | 10.400 – 12.500 | 60* (30) | Thermal IR Band 7 | 2.090 – 2.350 | 30 | IR
Action: Fill in the column labelled “Colour” of Table 2 for bands 1 to 5. Use the information provided with the introduction of the spectral indices.
Figure 11: List of loaded data.
Image display A double-click on the band name issues a command that displays the image.
Action: Do this for band 1 first.
You will see an image of the city of Venice and its surroundings. It consists of different shades of grey, a greyscale display, that correspond to the brightness or intensity measured at a given spot (pixel) in the image. The contrast is quite poor and should be adjusted using the tool Interactive Stretching.
Action: Find the corresponding button or menu item.
You can explore the meaning of the different buttons when moving the mouse pointer above them. After clicking, a new window appears as shown in Figure 12.
Figure 12: Windows for adjusting the contrast levels using Interactive Stretching. The window contains two graphs, one showing the distribution of pixel values in the image and the other used for display. Adjustments can be made by moving the flags. The setting is adopted by clicking Apply. Left: Distribution before adjustment; middle: after adjustments were made; right: the same shown in logarithmic scale, displayed by clicking the bottom left icon to the right.
The scaling of the contrast is accomplished by moving the flags. The window provides additional tools like displaying the data in a logarithmic scale.
Figure 13: Image of band 1 before (left) and after (right) adjusting contrast scaling.
Action: Display the seven images and adjust their scaling.
Creating a realistically coloured image After having adjusted the contrast settings, a colour picture can be produced by superposing three images. A bad contrast will lead to pale colours. For a realistic impression, the three bands representing blue, green, and red have to be selected.
Action: Find the corresponding bands in Table 2. If you need help assigning colours to wavelengths, research the missing information on the internet.
Select View\New RGB View. A new window appears (Figure 14). Choose the matching bands for red, green, blue and click on OK.
Figure 14: Window for selecting the bands to be used for constructing an RGB image.
A new colour image appears. If necessary, you can adjust the colours with Interactive Stretching.
Actions: Inspect the result and try to identify landscape elements (buildings, water, soil, vegetation).
Find the airport. Figure 15: Three-colour image (RGB) created from satellite data of Venice.
Creating a false colour image You have just produced an RGB image that corresponds to the natural impression of colours how humans see it. It consists of the colours red, green, and blue. Imagine other species like bees or snakes. They can see other parts of the electromagnetic spectrum like the ultraviolet (UV) or the infrared (IR). We can simulate such kind of vision skills by combining different spectral bands than red, green and blue. The resulting colours do not match the natural ones we can see with our eyes, but they can help making interesting details visible.
Use the knowledge that the chlorophyll in green plants absorbs red light but reflects infrared radiation.
Actions: Produce a three-colour image from the near infrared (ca. 0.8 µm), red (ca. 0.65 µm), and green (ca. 0.5 µm).
What are the corresponding bands?
Put the infrared band in the red channel, the red band in the green channel and the green band in the blue channel of the RGB image.
Compare this image with Figure 15. Where do you find green vegetation?
Can you distinguish between green crops and green water (algae)?
What does uncultivated land look like?
Figure 16: False colour image produced by combining the green, red and infrared bands.
Analysis via NDVI You have already seen in the information section that the NDVI is a colour or spectral index
that is particularly sensitive to green vegetation. The index provides a number that objectively reflects the degree of vegetation. Remember that there is a jump in the spectrum of green vegetation between the red (R) and the infrared (NIR) range (Figure 5). You will now construct a map that contains the NDVI for every image pixel. LEO Works provides a tool for this.
Action: Find the NDVI tool.
After activating that tool, a new window pops up (Figure 17). You select the dataset at the top. The next line contains the name of the image to be constructed and how it appears in the list of data. A name is already suggested. Select the suitable bands in the following rows below.
Action: What are the bands to be selected here? The answer can be found in the section about the NDVI and Table 2.
The formula is shown below. In the beginning, the variables show “null” as long as no band is selected. It is automatically updated as soon as you select the band corresponding to the NIR and the R bands. The NDVI map is created by clicking OK. A suitable false colour representation is chosen automatically, which helps identify green vegetation. However, the scaling of the colour table must be adjusted.
Figure 17: Window of the NDVI computing tool.
The tool Color manipulation is used for this. Move the flags of maximum value to the upper end of the distribution histogram. Then move the flag of the minimum value until the first green coloured flag reaches a value of 0.2 (Figure 18). The new setting is adopted after clicking Apply.
Figure 18: Window that allows adjusting
The result should look similar to Figure 19. You see large white zones with alternating yellow and green areas in between.
Action: Compare the NDVI map with the previously produced images. What can you say about the degree of vegetation in the green and yellow areas?
Would you be able to detect a seasonal change, if the images were taken at a monthly rate?
What would be the situation during a draught?
Figure 19: Map of the NDVI in the vicinity of Venice, based on Landsat 7 satellite data.
Analysis via MNDWI You will now use the satellite data to identify open wetland with the MNDWI. MNDWI = (G - SWIR) / (G + SWIR)
Especially small ponds and narrow rivers are not easily found on naturally coloured images. The MNDWI can theoretically be constructed using the NDVI tool. However, the correct assignment of the corresponding bands can be confusing. LEO Works provides a generic tool to does all kinds of mathematical operations with the spectral bands. The procedure is called Band arithmetic.
Action: Find the tool in the tool bar or in the menu and open it.
Similar to the tool for calculating the NDVI, you first select the dataset and the name of the image to construct (Figure 20, left). Then click Edit expression … for opening a new window (Figure 20, right). This is where you enter the formula for calculating the spectral index.
Figure 20: Window for doing mathematical operations on the spectral band images.
Action: Find out what bands are needed to calculate the MNDWI.
From the formula of this index you see that you divide the difference of the intensities of the reflected light measured in two spectral bands by their sum. Be careful with placing operators and brackets according to the formula.
After confirming the formula, it also appears in the first window. The procedure is executed by clicking OK.
The resulting image presents the values of the index in greyscale. To improve the readability of the map, you can change colours to certain values via the Color manipulation tool. A colour table is assigned by clicking on the symbol “Import palette” as shown in Figure 21.
Figure 21: Colours can be assigned to image values to improve the readability of the map.
Action: Select the file gradient_red_white_blue.cpd.
Adjust the flags such that the values are well covered and the central flag represents the value 0.
What is the colour coding of water?
Compare the MNDWI map with the previously produced images. Would you be able to find wetland also on the naturally coloured image?
Can you imagine situations for which the identification of water levels can be important or even life-saving?
What would the image look like, if the water level rises?
If you have time, produce a map of the NDMI. Compare it with the other results.
Figure 22: Map of the MNDWI in the vicinity of Venice based on Landsat 7 satellite data.
For advanced students Two additional datasets are provided that show the same area in January and July 2002. The already analysed dataset is from August 2001.
Action: Load the two additional datasets like the previous one.
Produce naturally coloured RGB images.
Produce images of the NDVI distributions.
Compare the results from the three datasets obtained at different dates during the year. Indicate how the vegetation changes.
In light of the results, describe and explain the advantage of satellite remote sensing.
The major part of this activity is analysing satellite images. The output of this exercise are images generated by combining the satellite images in a certain way. Success can be evaluated by comparing the maps and images with the ones provided with this material. In addition, students will answer questions that will show how well they understand the importance of satellite data for various aspects. These answers can be discussed in class after the activity.
UK KS3 Geography - Geographical skills and fieldwork.UK GCSE (2016) Geography AQA Skills UK GCSE (2016) Geography Edexcel Cartographic skills UK GCSE (2016) Geography OCR A and B Geographical skillsUK GCSE Geography WJEC A and B (2016) Cartographic skillsUK AS/A level Geography AQA (2016) ICT skillsUK A level Geography Edexcel Geographical SkillsUK A level Geography OCR Geographical skillsUK A level Geography WJEC Geographical skills non-numerical
Suitable image material from other areas can be downloaded via the ESA Eduspace image server at: http://www.esa.int/SPECIALS/Eduspace_EN/SEMLK0F1EHH_0.htmlAnother source of suitable satellite data: https://earthexplorer.usgs.gov
The students used the LEO Works software to inspect and analyse real satellite data at a close to professional standard. They combined datasets to produce colour images and maps of spectral indices and learnt how to interpret them. Students should understand the importance of satellite data when dealing with issues like disaster management and climate change.
