Raster Data Model

A simple, storage intensive format best suited for continuous fields.

Raster Data Model

Represents space continuously:

• Rectangular grid of equally sized cells
• Each cell only has one value

Raster Data Model

Multiple attributes, require multiple bands

• Standard color photos have three bands:
  • Red, green, & blue
• LANDSAT8 has 11 bands:
  • “False Color Images”
  • Spectral Indices (NDVI)

Resolution vs. Extent

Resolution (aka cell size) is the smallest resolvable feature. Extent (aka domain) depends on number of rows and columns along with the cell size.

• 1 m cell size

• 5 rows x 5 columns = 25 cells

• Covers 5 m x 5 m = 25 m²

Resolution vs. Extent

Resolution: 10 m cell size = 100 m² area per cell.

Extent: varies with number of cells and cell size.

Rows	Columns	Extent
5	5	50 m x 50 m = 2,500 m2
5	10	50 m x 100 m = 5,000 m2
100	100	1,000 m x 1,000 m = 1,000,000 m2

TopHat Question 1

Raster resolution is dictated by the number of rows and columns, raster extent is dictated by the cell size.

• True
• False
• Sometimes True, Sometimes False

Mixed Pixel Problem

One cell = one value. What if it covers multiple values?

• A: Winner take all

• B: Cell center

• Other options?

  • Containment: value of interest is present
  • e.g. invasive species presence

Implications

Loss of information during rasterization.

• “Bigger” issue for larger cells
• At some point features become unrecognizable
• Higher resolution = larger file
  • For equivalent areas

TopHat Question 2

This term describes how information is lost when creating a raster image.

• Resolution
• Domain
• Mixed pixel problem
• Rasterization

File Size

The number of cells per image dictates file size.

• # cells = rows x columns x bands
  • Increases exponentially with resolution
    • For a given extent
  • Increases linearly with number of bands

File Size

Imagine four images that all have the same extent:

• 1000 m x 1000 m = 1,000,000 m²

Bands	Cell Size	Rows	Columns	# of Cells
1	100m	10	10	100
1	1m	1000	1000	1,000,000
3	100m	10	10	300
3	1m	1000	1000	3,000,000

Why does this matter?

Downloading, uploading, and processing times will all increase as the file size increases.

• Large raster datasets are computationally “expensive”
  • More cells = more “overhead”
• Practically speaking, how will this impact you?

TopHat Question 3

You downloaded a 3 band LANDSAT8 image covering the UBC campus. The image’s dimensions are 6000m north/south (rows are vertical) by 4500m east/west (columns are horizontal) and the resolution (aka cell size) is 30m. How many cells does the image contain?

• 30,000
• 90,000
• 900,000

• 2,700,000
• 27,000,000
• 81,000,000

Solution

You downloaded a 3 band LANDSAT8 image covering the UBC campus. The image’s dimensions are 6000m north/south (rows are vertical) by 4500m east/west (columns are horizontal) and the resolution (aka cell size) is 30m. How many cells does the image contain?

\(6000 \div 30 = 200\) rows

\(4500 \div 30 = 150\) columns

\(200 \times 150 = 30,000\) cells (per band)

\(30,000 \times 3 = 90,000\) cells in total

Metadata in ArcPro

Raster Overlay

Raster data can be used to evaluate algebraic expressions easily and efficiently.

• A: Winter Temperature
• B: Summer Temperature
• Range: B-A
• Average: (A+B)/2

Raster Calculator

• Perform simple calculations incorporating one or multiple layers
• Ideally, layers should have the same resolution and extent
  • Output will be restricted to “smallest” extent

Raster Calculator

• Perform simple calculations incorporating one or multiple layers
• Ideally, layers should have the same resolution and extent
  • Output will be restricted to “smallest” extent

Raster Calculator

• Perform simple calculations incorporating one or multiple layers
• Ideally, layers should have the same resolution and extent
  • Output will be restricted to “smallest” extent

Raster Calculator

• Perform simple calculations incorporating one or multiple layers
• Ideally, layers should have the same resolution and extent
  • Output will be restricted to “smallest” extent

Key Advantages

Raster model is well suited for continuous phenomena:

• Continuous in space and time
  • Complete coverage within the domain
  • Bands can be stacked to represent multiple attributes or time steps
• Simple data structure overlay easy and efficient
  • Relation between layers can evaluated with expressions
    • Easy to “show” others what has been done

GeoTIFF

One of the most common/functional raster formats, based of the Tag Image File Format (TIFF).

• A TIFF file stores metadata (data about the data) as tags.
  • GeoTIFF adds spatial tags denoting:
    • Extent (minimum x,y and maximum x,y)
    • Resolution (cell size)
    • Projection, Coordinate system, and datum

Other file types

Raster data can come in many different formats. You will likely encounter when working with raster data include:

• IMG: Proprietary format used by ESRI products
• JPEG2000: Geospatial version of the .jpg image
• ASCII: Human readable format (text file)
  • Slower performance than the types listed above

TopHat Question 4

The Raster Calculator tool lets you:

• Evaluate algebraic expressions using raster layers as inputs
• Create raster layers
• Calculate the area of raster layers
• Overlay raster and vector layers

Satellites Data

We’re going to work with LANDSAT8 data in lab.

• LANDSAT8 is one of a large number of satellites that orbit the earth continuously collecting multi-spectral (visible light & other wavelengths) imagery.
  
• We can use multi-spectral imagery for a number of different applications like estimating vegetation health.

Monitoring Vegetation

The lines on the chart below are referred to as a spectral reflectance curves.

\(NDVI = \frac{(NIR-Red)}{(NIR+Red)}\)

• Green vegetation has different reflectance than soil and water

TopHat Question 5

What does NDVI tell us?

• Vegetation Height
• Vegetation Health/Density
• Vegetation Species
• Number of Plants per Square Meter
• Location and Age of Forests