Appearance
Choose vector when your historical features are discrete and need attributes, queries or links — parishes, roads, addresses — and choose raster for continuous surfaces and for the scanned map images themselves. Most history projects use both: a georeferenced raster as backdrop, with vector layers digitised on top for the things you measure and analyse. This guide makes the choice concrete and shows where each model earns its keep.
What is the actual difference?
A raster is a grid: every cell holds a value, like pixels in a photo or metres of elevation. A vector is a set of geometric objects — points, lines, polygons — each with coordinates and an attribute record. A scanned tithe map is a raster; the field boundaries you trace from it are vector.
The mental test: is the phenomenon continuous everywhere, or made of distinct named things? Elevation is continuous (raster). A parish is a distinct, named, bounded thing (vector).
When should you use vector?
Reach for vector whenever the data is discrete and you want to do something with it beyond looking:
- Attach attributes (population, owner, date, source).
- Run queries and spatial joins.
- Measure exact lengths and areas.
- Link to external records or a gazetteer.
python
import geopandas as gpd
parishes = gpd.read_file("parishes_1851.geojson")
parishes["area_km2"] = parishes.to_crs(27700).area / 1e6
big = parishes[parishes["area_km2"] > 50] # query by attribute + geometryVector is editable, compact and resolution-independent — it looks crisp at any zoom.
When should you use raster?
Use raster for continuous fields and for images:
- The historical map scan itself — a georeferenced image is raster by definition.
- Surfaces: elevation (DEM), slope, density, cost surfaces.
- Cell-by-cell analyses: least-cost paths, viewsheds, hillshade.
bash
# rasters shine for surface analysis
gdaldem hillshade dem.tif hillshade.tif
gdaldem slope dem.tif slope.tifIf your method works on neighbourhoods of cells, raster is the natural home.
How do raster and vector compare at a glance?
| Aspect | Raster | Vector |
|---|---|---|
| Best for | Surfaces, images | Discrete features |
| Attributes | One value per cell | Rich record per feature |
| File size | Often large, fixed resolution | Usually small, scalable |
| Zoom behaviour | Pixelates | Stays crisp |
| Typical history use | Scanned maps, DEMs | Boundaries, roads, points |
| Editing | Awkward | Natural |
Can you convert between them?
Yes, but convert only when the analysis demands it, because both directions lose something:
- Vectorise (raster to vector): tracing scanned boundaries with
gdal_polygonizeor QGIS produces noisy, stepped shapes that need cleaning. - Rasterise (vector to raster):
gdal_rasterizeburns polygons onto a grid, fixing them to one resolution and discarding sub-cell detail.
A common, justified conversion is rasterising parish polygons so they can feed a cell-based density or cost analysis.
A practical workflow for historians
Most projects converge on the same pattern. Keep the georeferenced scan as raster for visual context. Digitise the features you care about as vector layers, each with a clean attribute table and source notes. Run surface analyses on rasters, feature analyses on vectors, and only convert at the boundary between the two. This keeps each dataset in the model that makes it cheap to store, easy to query and honest about its precision.
Key Takeaways
- Vector for discrete, attributed, queryable features; raster for continuous surfaces and images.
- A georeferenced historical map scan is inherently raster.
- Digitise to vector only when you need to measure, query or link the features.
- Raster powers cell-based analysis: hillshade, slope, least-cost paths, viewsheds.
- Vector is smaller, editable and resolution-independent.
- Convert between models only when the analysis truly needs it — both directions lose detail.
- Real projects mix both: raster backdrop, vector layers on top.
Frequently Asked Questions
What is the difference between raster and vector data?
Raster stores the world as a grid of cells each holding a value (like a scanned map or elevation grid); vector stores discrete points, lines and polygons with coordinates and attributes (like parish boundaries). Rasters suit continuous surfaces and images; vectors suit named, bounded features.
Should a scanned historical map be raster or vector?
The scan itself is raster — that is what a georeferenced image is. You only digitise it into vector when you need to measure, query or join the features drawn on it, such as boundaries or roads.
When is vector the better choice for historical GIS?
When features are discrete and you need attributes, queries, exact boundaries, or to link records — parishes, railways, addresses, ownership parcels. Vector data is editable, light, and scales cleanly at any zoom.
When is raster the better choice?
For continuous phenomena (elevation, cost surfaces, density) and for the historical map images themselves. Raster also suits analyses like least-cost paths and viewsheds that operate cell by cell.
Can I convert between raster and vector?
Yes. Vectorising (raster to vector) traces cells into shapes and is lossy and noisy; rasterising (vector to raster) burns shapes onto a grid. Convert only when the analysis genuinely needs the other model.
Does the choice affect file size and performance?
Greatly. A detailed raster can be huge and fixed in resolution; equivalent vector data is usually far smaller and resolution-independent, but complex polygon overlays can be slow to render and process.