Resolution &. Scale

1)How does the scale of our analysis dictate our desired resolution?

2)How can data resolution limit the scale of our analysis?

Resolution & Scale

We want to work with high resolution data because: ↑ resolution = ↓ generalization = ↓ uncertainty

  • In reality, this isn't always practical or possible.
  • We must try to strike a balance.

Things to consider

What is the "lowest" acceptable resolution?

  • Smaller scale analysis won't need the same level of detail.
  • Looking at national immigration patterns, maybe you only need provincial level data?

Things to consider

Change the scope of our analysis?

  • If you need a high resolution analysis, work at a larger scale and focus on a small area.
  • If you need to analyze a large area, work at small scales and accept the uncertainty that comes with coarser resolution.

    • Comparing Data Models

    Raster

    Usually continuous fields

    • Grid of cells (pixels) with continuous coverage
    • Each cell has one value per band (layer)
    • One raster image can have many bands

    Vector

    Usually discrete objects

    • Points, Lines, and/or Polygons
    • Each object can have many attributes
    • Objects may overlap, have gaps, or be continuous

    Advantages

    Raster

    • Good for continuous variables: in space and time
    • Simple data structure makes overlay is easy and efficient

    Vector

    • Compact data structure
    • Good for discrete objects
    • Easy to query and select by attributes

    DisAdvantages

    Raster

    • Mixed pixel problem
    • One attribute per cell
    • Large data volumes

    Vector

    • Complex data structure
    • Overlay can be computationally expensive
    • No variability within polygons

    Which data model is “Best”?

    Neither data model is suitable for all types of data or analysis.

    • You will frequently use both the raster and vector models.
    • It is possible to convert back and forth between models.
      • This can introduce error, only do when necessary.