Professional Documents
Culture Documents
(QGIS-SCP)
Note: Distribution of this file highly needs the permission of Gifford Jay Agudo. IF use in research,
the author, and the author’s mother institution (University of San Carlos) should be acknowledged.
Resources Needed: QGIS, Landsat Scene from USGS EarthExplorer website.
Introduction
Studying environmental dynamics using the lens of a satellite images requires adequate
knowledge on the behavior of electromagnetic radiation and its corresponding properties when it
interacts with the surface. Understanding the pixel’s behavior by determining the spectral signature
is an important undertaking as it will allow you to determine the underlying natural cover being
studied. While preliminary examinations of various land objects can be accomplished by naturally
interpreting an image using the object’s actual appearance within an image, this approach however
will only work when the image itself is collectively arranged under the “natural color composite”
(Figure 1).
Figure 1. Satellite image of Cagayan de Oro City. Image was captured on May 05, 1988 using
Landsat 4. Noticeably, various land covers and phenomenon can be easily determined using the
natural color composite convention (RGB – 321) since humans perceive the planet using this light
convention.
However, problem may eventually arise when the image is constructed using radiation
captured outside the visible light regimes (Figure 2). In addition, various objects captured outside
the visible light regimes may appear alien to human eyes, owing to strange color appearances that
is not associated with the objects when viewed under the natural visible lights (Figure 2).
Figure 2. Landsat false color composite constructed using the NIR regimes of the electromagnetic
spectrum a) Natural-NIR false color composite (RGB -432) b) Natural-NIR false color composite
(RGB – 543). Noticeably, these images do not construct the way humans perceived its surrounding
environment, and the land cover per se may appear unfamiliar when viewed by a human with no
background on electromagnetic spectrum.
Understanding images beyond the natural color composite regimes is necessary as it gives
us a plethora of information and their status about the underlying land cover. For example, RGB
– 432 convention give us the ability to discern varying stages (e.g., physiognomy) of vegetation
owing to the inclusion of NIR bands that is effective in detecting amounts of chlorophyll based of
vegetation structure (this serves as an ancillary data to detect vegetation structure).
Downloading a Landsat Data from Earth Explorer
Earth Explorer is a data distribution center managed by the US Geological Survey (USGS).
Although data are freely distributed, downloading of images requires account creation, and
validation using a Gmail account. We will first secure an account to facilitate our data downloading
requirements. Registration can be done by searching the “earth explorer registration” in google
search box, and correspondingly click the “Register – EROS Registration System – USGS”.
Figure 3. Registration search box for EROS Registration System – United States Geological
Survey (USGS).
Using the Earth Explorer center, we can look for a data in Cagayan de Oro using the World
Features search category, and type Cagayan de Oro (Framework 6).
Figure 6. Schematic diagram on how to download a Landsat scene using Earth Explorer website.
Figure 7a. Schematic diagram on how to install the semi-automatic classification plugin in QGIS.
Figure 7b. Screenshots of Plugins button location in the QGIS software graphic user interface.
In this exercise, we will use atmospherically corrected (calibrated already) Landsat scenes.
Although the standard presentation of Landsat 4-5 is 8-bit unsigned integers (0-255 brightness),
the Collection Level-2 data products are converted to 16-bit unsigned integers which makes a total
of 65,535 color gradients. Working on SCP first requires the system to import all the images into
the QGIS program. Opening of a satellite data can be done by clicking the “Layer” in the main
menu bar, and select the raster option.
Click the “Layer” button Select “Add Layer”
Select “Add Raster
in the menu bar Layer”
Click “Open” and “Add” Highlight and select B1 Find the data folder you
button in a stepwise to B7 (excluding B6) downloaded from Earth
manner Explorer
Figure 8. Schematic diagram illustrating the process of opening a Landsat raster dataset in QGIS
platform.
After performing the entire workflow presented in figure 8, all of the dataset will be placed
in the “Layers” section of the software, and the visualization of the individual image will be
presented in the map layer (Figure 9).
Figure 10. Distribution of reflectance value of Band 7 using a histogram chart. Notice that most
of the pixels have a pixel value of less than 12,000 and with minor data distribution situated as the
extreme right end of the histogram (biggest values). This facilitates the rendering of the image
with “very dark” tone.
Band stacking is one of the first procedure that we need to accomplish to give us a color
composite image. In simple terms, band stacking allows us to compile sets of images into a single
file layer, which would correspondingly permit us to assign a color to each band, altogether
creating a single image with various surfaces register a distinct color characteristic. Band stacking
in QGIS requires a little knowledge about the SCP plugin. Figure 11 illustrates the entire band
stacking process.
Check the “Create Virtual Click the “PLUS (+)” Highlight the imported
Raster” button to assign the band by “CRTL + A”
bands to a band set
Figure 11. Process of adding and band stacking in QGIS Semi-Automatic Classification Plugin.
After performing the workflow shown in figure 11, the stacked image will be automatically
drafted in the “Layers” section with 3 corresponding colors (Figure 12). Rearrange the band
assignment by following the process presented in figure 13.
Figure 12. Appearance of the stacked image after performing the band stacking workflow. Notice
that the image is very distinct such that is has 3 different band colors appended to the layer name.
Figure 13. Process flow of reclassifying the band assignments using QGIS.
Notice that the image itself inherited several characteristics of the individual bands that we
previously examine, such that it scattered a lot of green and blue lights. This light scattering
characteristics reduce the number of details (reduced contrast) that can be discerned within the
image.
To clearly visualize our spectral signature analysis, we will apply a geoprocessing tool to
concentrate our analysis to our area of interest (AOI). WE will initially examine some surfaces in
Cagayan de Oro since we tend to have several experiences in of various places of the city. Figure
14 presents the workflow of extracting our AOI.
Click the “Raster” in the Click the “Extraction…” Select the “Clip Raster by
menu bar Extent…”
Draw the desired extent as Select the “Draw on Select the stacked
if creating a rectangle in Canvas” for the image in the input
Paint software (Ask the Clipping Extent Layer
instructor)
Figure 14. Process flow of extracting the area of interest using geoprocessing clipping toolbar of
QGIS.
Notice that after clicking the image into a smaller geographic boundary, the image
apparently appears brighter than the unclipped version of the image itself (Figure 15a). By looking
at the histogram of the clipped image, this would suggest that the extreme right end of the entire
surface reflectance was omitted (Figure 15b).
Figure 15. a.) Clipped image that appears to be brighter. b.) Distribution of data using histogram.
Spectral Signature
Spectral signature (others called this spectra) is the amount of reflected radiation (surface
reflectance) by the surface as a function of wavelength (Figure 16). It is important that we are
familiar with the spectral signature since it allows us to identify objects even without physical field
works to characterize it.
Noticeably, forest tends to register high surface reflectance in the NIR region of the
electromagnetic spectrum, and with remarkably low surface reflectance in Red of the visible light
regimes. This effectively guides us in differentiating bare ground (or soil) from the forest per se.
In extracting the spectral signature within the image using QGIS with limited field-based
information, one first needs to know and experience the “what is on the ground?” data. We can
easily do this when we deploy our intuitive understanding about our planet. Contagious forest
appears “dark green” when viewed under the visible light regimes, bare grounds look “brown”,
built-up objects appear “bright-white”, and water appears “blue”. However, this becomes so
problematic when we try to differentiate various forms of “vegetation” such as grasslands, and
forest. This process highly necessitates intensive field works of obtaining vegetation spectra. In
this example, we will just consider the major land cover, which includes: 1.) Vegetation 2.) Bare
ground 3.) Clear Water and 4.) Built-up areas.
Figure 17. Process of acquiring the spectral signature of various land cover using the QGIS-SCP.
Figure 18. Spectral signature plot obtained after performing the process flow indicated in figure
17.
YOUR TASK!
1. Georeference the following land cover using a GPS device. Use your previous
understanding in ES316 – Introduction to Geographic Information Systems (GIS).
a. Forested area
b. grassland
c. bare ground
d. water body
2. Generate a spectral signature of each land covers that you georeferenced using QGIS-SCP.
3. Describe the spectral signature of each land covers.
Acknowledgement
➢ The United States Geological Survey (USGS) is greatly acknowledged for the provision of
the Landsat surface reflectance data.
➢ The QGIS Semi-Automatic Classification Plugin (SCP) which was developed by Luca
Congedo is recognized.