Vector vs Raster¶

Goal: Stop second-guessing this. Know exactly when to use vector and when to use raster.

The 10-second version¶

Visual¶

VECTOR ▭ ─── ●           RASTER  ┌─┬─┬─┬─┐
   parcel  road  point           │░│▒│▓│▒│
                                 ├─┼─┼─┼─┤
                                 │░│▒│▓│▓│
                                 └─┴─┴─┴─┘
        Discrete                  Cells with values
        shapes                    (elevation, NDVI, …)

When vector wins¶

✅ Use vector when:

• The thing has a clear boundary — a building, a parcel, a stream.
• You need to attach attributes (population, owner, name).
• You'll measure length / area / perimeter.
• The output goes to print at varying scales (vectors stay crisp).
• You'll edit the data interactively.

When raster wins¶

✅ Use raster when:

• The phenomenon is continuous — elevation, temperature, NDVI, rainfall.
• You're doing map algebra (calculate slope from elevation, NDVI from bands).
• You're working with imagery (satellite, aerial, drone).
• You need a pixel-by-pixel analysis (suitability, change detection).
• You want a density / heatmap surface.

The hybrid¶

Most real workflows mix both:

Conversion¶

You'll convert between them often:

• Polygon → Raster — Polygon to Raster tool. E.g., zoning polygons → input to weighted overlay.
• Raster → Polygon — Raster to Polygon tool. E.g., classified land cover → polygons.
• Points → Density raster — Kernel Density tool.
• Raster cells → Points — Raster to Point tool.

Storage tradeoffs¶

A US-wide vector layer of all roads might be 300 MB. A US-wide DEM at 10 m might be 30 GB.

Why? Vector stores only vertices. Raster stores every cell, even empty ones.

For huge rasters, use Cloud-Optimized GeoTIFF (COG) — you can read just the part you need without downloading the whole file.

Common confusions¶

Performance notes¶

• Vector spatial joins are O(n × m) without an index. Add a Spatial Index.
• Raster operations are O(cells). A 4× resolution increase = 16× more cells = ~16× slower.

Practice¶