GGR278 Lecture 09

Completed Notes Status

Central objective: To understand raster data models, including cell values, conversion from vector data, spatial resolution, and fundamental raster operations.
Key concepts:
- Raster Cell Values: Can be floating-point (continuous surfaces, unique values) or integer (classified data, lower memory footprint). Non-classed pixels are assigned 0, 9999, or NoData.
- Vector to Raster Conversion: Rasterization depends on assignment rules (e.g., "Any cell" vs. "Near cell centre"). The chosen rule and grid orientation heavily influence the final output.
- Raster Cell Depth: Determines the number of bits used to store values (e.g., 8-bit stores 256 values). Depth is chosen based on the required analysis resolution versus memory limits.
- Minimum Mapping Unit (MMU): The smallest feature mapped. Cell size must be $< \frac{1}{2}$ of the MMU to avoid severe inaccuracies.
- Mixed Pixel Conflict: Arises when a cell contains multiple classes. Resolution schemes include most common value, centre value, average, or use-case specific rules (e.g., prioritizing water for temperature analysis).
- Raster Operations: Categorized by spatial scope: Local (single pixel), Focal/Neighbourhood (surrounding pixels), Zonal (defined geographic areas), and Global (entire dataset).
Connections:
- Builds on the vector vs. raster comparison from GGR278 Lecture 08 Raw, detailing how vector features are translated into raster grids and the necessary spatial conflict resolutions that follow.

Remember/Understand:
- What is the difference between integer and floating-point cell values in a raster dataset?
- How is the Minimum Mapping Unit (MMU) related to raster cell size?
- What are the four main types of raster operations?
Apply/Analyze:
- If you are converting a vector river network to a raster grid, how would using the "Any cell" rule differ from the "Near cell centre" rule in the final output?
- Why might you choose a "winner takes all" scheme over an "average" scheme when resolving a mixed pixel conflict for categorical land cover data?
Evaluate/Create:
- Propose a scenario where a high raster cell depth (e.g., 32-bit float) is absolutely necessary over an 8-bit integer depth, justifying the trade-off in memory usage.

Minimum Mapping Unit:
- Why it's challenging: It can be unintuitive why cell size must be strictly less than half the MMU rather than equal to it.
- Study strategy: Draw a grid and overlay a small object exactly the size of one cell. Notice how its placement across cell boundaries can cause it to be lost or misrepresented depending on the conversion rule.
Mixed Pixel Conflict:
- Why it's challenging: There is no single "correct" way to resolve a mixed pixel; it relies heavily on the specific analytical goal.
- Study strategy: Memorize the common schemes (majority, centre, average) and practice assigning a scheme to a specific hypothetical use case (e.g., thermal mapping vs. agricultural area calculation).

Immediate review actions:
- Review all [COMPLETED] insertions and verify with course materials
- Re-read Lecture Summary and confirm key links
Practice and application:
- Answer Remember/Understand questions
- Answer Apply/Analyze questions
- Attempt Evaluate/Create questions
Deep dive study:
- Focus on Challenging Concepts (above)
- Extend atomic notes only where the new lecture adds content
Verification and integration:
- Cross-reference notes with textbook/slides
- Identify remaining gaps
- Link this lecture to prior/future lecture notes (lecture notes only; no MOCs)