Fundamentals of Rasters and Imagery
The seemingly disparate accumulation of images above all have a common element. Can you guess what common thread binds them together?
The world of GIS is filled with all different types of data. There is so much data, that it can be quite the task to thoroughly understand and categorize all of it. We can start understanding our data framework by considering the two most common data types used in GIS: 1) Vector and 2) Raster.
Vector data is a geographic data type where data is stored as a collection of points, lines, or polygons, along with attribute data.
Raster data is a geographic data type where data is stored as a grid of regularly sized pixels along with attribute data.
In this article, we’ll cover some fundamentals of rasters and imagery, including the different types of rasters and imagery, and some examples of real-world applications and use cases.
What is a Raster?
A raster is a data model made up of a matrix of cells, or pixels, organized as rows and columns in a grid-like fashion. Each cell in the grid has a specific location based on its row and column and contains information. The most common familiar form that rasters take is as images, where the resolution dictates the pixel size, and therefore the density of the information in the raster.
In geographical images, the stored information is unique to its geographic location as represented by its row and column placement in the image.
As stated in Esri’s Introduction to image and raster data, there are four geographic components stored in geographical raster images. These are:
- A coordinate system
- A reference coordinate or x,y location
- A pixel size
- The count of rows and columns
Raster images spanning large amounts of area are collected by satellites, or high-altitude aircraft. The information stored in these small scale images is often less detailed, and broader in scope. This is perfectly fine for many applications of GIS raster analysis, like measuring forest cover in the Amazon, where you would not expect large variety in land cover types over only several meters.
Raster images covering smaller amounts of area are often more detailed and have a specific application in mind when the image is collected. These large scale images are most commonly collected using drones or aircraft during low altitude flights. This scale of imagery is great for monitoring frequent change over a smaller area.
Now that we have a brief understanding of what raster imagery is and how it is procured, let’s move on to discuss more specific types of raster imagery.
Discrete raster imagery has a distinct theme or category, such as land cover class or soil type data. An alternative name for discrete raster imagery is thematic imagery. You can think of it as having qualitative attributes, where the numeric code embedded in the pixel represents the category it belongs to.
Recalling back to our cover image, items A and E are discrete rasters representing thematic data.
In this dataset, Iowa is largely covered in dark brown. As indicated by the legend, the dark brown coloration represents cultivated crops. This makes sense as Iowa is largely an agricultural state of the United States, and is known for growing corn (maize) and soybeans.
In this dataset, several coastal areas display as red, while the northwest portion of the country displays as mostly dark green. As indicated by the legend, red represents built areas while dark green represents trees. This makes sense as Spain has several large port towns and the northwestern portion of the country is largely forested.
Unlike discrete rasters, where items in the imagery are clearly delineated from one another, pixels in continuous raster imagery have less clearly defined boundaries. In this type of imagery, every cell in the map has data that flows in a gradient from one cell to another.
Continuous raster imagery is often used for datasets that represent a certain type of data that changes gradually as the geographic location changes. Typical datasets include elevation, slope, temperature, or precipitation values. These can be integers, but more commonly are decimal values or floating types (ie, 3.14 Celsius, 112.8 degrees, 8.4 inches of rain) to better represent the detail of the subject.
Recalling back to our cover image, item D is a continuous raster representing elevation data.
In Figure 7, the peak of Mt. Fuji can clearly be seen in white after applying an elevation shaded relief symbology to the digital elevation model (DEM). In this example, we simply see a visualization of height. However, this image and other continuous raster images can be used to extract several useful geospatial insights.
According to this excellent article, imagery is a type of data that is useful for many GIS applications, and is defined as any type of photograph. This includes images collected from satellites, aircraft, drones, and other sensors, including thermal and infrared. Additionally, scanned maps and floor plans also count as images.
Back to our cover image, items A-F are imagery. Items B, C, and F are more miscellaneous imagery that have not been processed with any classification methods like a discrete raster has.
Figure 8 is a scanned image of a floor plan. Scanned imagery like this is useful for applications such as engineering and construction where CAD files are used in conjunction with GIS data. The scanned imagery can be georeferenced in GIS software for further analysis.
Figure 9 is basemap imagery of London from OpenStreetMap Vector Basemap . As we are likely aware already, basemap imagery like this is useful for navigation and reference purposes. Very few maps are made without a basemap such as this to add geographical context.
Figure 10 is satellite imagery of Barstow-Daggett Airport in California taken from GoogleEarth. Imagery such as this has many applications in geospatial analysis. In its current state, it can be used as a reference layer like the basemap in the example above. Furthermore, this image would be useful for further processing as it has both crops and developed land. Geospatial analysis such as supervised or unsupervised classification could be performed when packaged into the discrete raster imagery format mentioned earlier in the article.