Studio Matrx Monthly · Volume 1 · Issue 3 · August 2026
Amogh N P
 In loving memory of Amogh N P — Architect · Designer · Visionary 
Vector, Raster, Features & LayersLesson 0.2
GIS for Architecture, Planning & Urban Design/Module 0 · Ground Rules — Thinking Spatially

Lesson 0.2 · Ground Rules — Thinking Spatially

Vector, Raster, Features & Layers

The two ways the world becomes data

12 min Interactive lessonFree · open lessonByAmogh N P· Architect & interior designer
The hook

A road is a line. A satellite photo is a grid. Both are “the map.”

Zoom into any GIS project and you will find two completely different kinds of thing pretending to be the same map. Some things are crisp, editable shapes with data behind them — a plot you can click to read its owner. Other things are photographs made of tiny coloured squares — a satellite image, an elevation surface. These are the two data models of GIS: vector and raster. Almost everything you will ever load is one or the other, and choosing wrong makes easy work hard.

Vector has edges. Raster has resolution. Ask which one your question needs.

Vector: the world as shapes with data

The vector model stores reality as three kinds of geometry:

Points — a single location: a tree, a lamp post, a bore-well, a bus stop. • Lines (polylines) — a length: a road centreline, a drain, a pipe, a contour. • Polygons — an enclosed area: a plot, a building footprint, a ward, a lake.

What makes a vector shape a GIS feature rather than a mere drawing is the row of data attached to it. Every feature sits in an attribute table: one row per feature, one column per fact. A polygon of a plot carries its area, owner, land use and FSI. Click the shape, read the row; sort the row, find the shape. Vector is the model for anything with sharp edges and things you want to count, label or query one-by-one.

One feature the shape (geometry) One row id use area_m2 14 Resi 312 15 Comm 540 16 Park 980 one feature = one shape + one row
Zoom
One feature is one shape plus one row of facts - they are the same thing, seen two ways.

One feature = one shape + one row of data. Lose the row and you’ve just got a doodle.

Raster: the world as a grid of values

The raster model throws a fine grid over the world and stores one value in every cell (pixel). A satellite image is a raster where each cell holds a colour (really, brightness in several bands). An elevation model is a raster where each cell holds a height in metres. A temperature surface, a slope map, a rainfall map — all rasters.

Raster's defining number is its resolution — the ground size of one cell. Sentinel-2 imagery is 10 m; Landsat is 30 m; the ISRO CartoDEM elevation model is ~30 m. A 10 m cell means the smallest thing the data can “see” is roughly 10 m across; you cannot recover a footpath from 30 m data no matter how far you zoom. Raster is the model for continuous phenomena — things that have a value everywhere, not just at sharp edges: height, heat, greenness, wetness.

A raster, zoomed in 812 815 817 818 820 821 811 816 819 822 824 825 810 814 820 826 828 829 809 813 821 828 831 833 808 812 820 829 834 836 one cell = one value value here = 816 (metres) resolution = ground size of one cell
Zoom
A raster is a grid of cells, each holding one number; the cell size is the resolution.

Vector or raster? Match the model to the phenomenon

The rule of thumb: discrete, countable, edged things → vector; continuous, everywhere things → raster.

A plot boundary is discrete — vector. Elevation is continuous — raster. But the same phenomenon can flip models depending on the question. Population is really a set of people (points), yet we often map it as a raster “density surface,” or as vector wards each carrying a count. Roads are vector lines for routing, but “distance to the nearest road” is naturally a raster surface. A large part of GIS craft is converting between the two — rasterising vectors, or vectorising rasters — to suit the analysis. You will do both in later modules.

VECTOR point line polygon exact shapes, any scale RASTER polygon line point same scene, snapped to a grid
Zoom
Vector stores exact shapes; raster approximates the same scene as a blocky grid of cells.

Layers stack both models together

A layer is one dataset of one model shown as one sheet in the stack. A project mixes them freely: a raster satellite base at the bottom, vector roads and plots above, a raster slope surface half-transparent on top. Each layer has its own attribute table (vector) or value range (raster), its own style, and its own coordinate system that the GIS reconciles on the fly.

The discipline is to keep one kind of thing per layer — all the plots in one layer, all the trees in another — so each layer has a clean, single-subject attribute table you can query. Sloppy layering (roads and rivers in one file) is the commonest beginner mistake and it poisons every analysis downstream.

imagery plots roads buildings one subject per layer - toggle, reorder, overlay
Zoom
A GIS holds a place as a stack of transparent layers you can toggle, reorder and analyse across.

One subject per layer. A layer that holds “everything” can be queried for nothing.

Data & formats you will meet in this lesson

Sentinel-2 (Copernicus)

Free 10 m optical satellite raster

The default free imagery raster for the built environment; 10 m cells. dataspace.copernicus.eu

CartoDEM (NRSC)

India's ~30 m elevation raster

Free at 30 m posting via Bhoonidhi; a raster where each cell is a height in metres.

OpenStreetMap (OSM)

Global vector features (points/lines/polygons)

The reference example of vector data: every road, building and POI is a feature with attributes.

Hands-on workshop

Workshop — hold both models in one project

You will load a vector layer and a raster layer for the same area, read a feature's attribute table, and read a raster cell's value — feeling the difference in your hands.

QGIS 3.44 or ArcGIS Pro; a free USGS EarthExplorer or ISRO Bhoonidhi account for the DEM.

Given & goal
Goal: one project with (a) OSM building polygons and (b) an elevation raster
Data: OSM buildings (free) + SRTM or CartoDEM 30 m DEM (free)
Time: ~40 minutes
  1. 1In QGIS: use QuickOSM to download building polygons for a small area. In ArcGIS Pro: add OSM buildings from the Living Atlas or a downloaded extract.
  2. 2Open the vector layer's attribute table (QGIS: right-click → Open Attribute Table; ArcGIS Pro: right-click → Attribute Table). One row per building — that is vector.
  3. 3Download a DEM raster: SRTM 30 m from USGS EarthExplorer, or CartoDEM 30 m from ISRO Bhoonidhi. Add it to the project.
  4. 4Use the Identify / value tool on the raster (QGIS: Identify Features on the raster; ArcGIS Pro: Explore → click). A single cell reports one number — an elevation. That is raster.
  5. 5Style the DEM with a colour ramp and set it 50% transparent over the buildings. You are now looking at both models, stacked — the everyday reality of a GIS project.

You’ll walk away with
A single project holding a vector layer (with a table) and a raster layer (with per-cell values), and a clear, felt sense of when each model applies.

The worked example

Three altitudes on the same idea

Read the band that fits you — or all three.

For the architectSite, form & environment

You will live mostly in vector, and lean on raster for terrain and imagery. Plot, building, tree, setback line — vector. Contours you derive from a raster DEM; the satellite backdrop for a context plan — raster. Know that a 30 m DEM is too coarse for fine site grading but perfect for reading the plot's overall slope and drainage.

For the plannerLand use, zoning & infrastructure

Statutory plans are vector; the evidence base is often raster. Land-use zones, ward boundaries and road networks are polygons and lines you edit and total. But land-use change, built-up spread and green cover come from raster satellite analysis (Module 6). A planner must be fluent moving counts between the two — e.g. summarising a raster land-cover map by vector ward.

For the urban designerStreets, blocks & public realm

Street and block analysis is vector; experiential surfaces are raster. Block polygons, street centrelines and frontages are vector. But sunlight hours, visibility and “distance to a park” are continuous — raster surfaces you drape over the plan. Choosing the right model keeps a walkability study honest instead of blocky.

Misconception check

Vector is for accurate work and raster is just low-quality background imagery.

Neither model is “better.” They store different kinds of reality. A raster elevation model can be exquisitely precise; a vector layer can be hopelessly wrong. The question is never “which is more accurate” but “which matches the phenomenon”: sharp-edged and countable, or continuous and everywhere. Treating raster as mere wallpaper throws away half of GIS — most environmental analysis is raster.
Try it

Do it yourself

Sort the world into the two models.

  1. 1List five things on a site and tag each vector or raster: bus stop, ground temperature, plot line, tree, rainfall.
  2. 2For a vector plot layer, write three columns its attribute table should carry.
  3. 3Explain in one sentence why a 30 m DEM cannot show a 2 m-wide drain.
  4. 4Name one phenomenon you could store either way, and say what question would decide it.
Take this with you

The grammar, in one line

Vector stores sharp, countable things as points, lines and polygons each carrying a row of data; raster stores continuous things as a grid of valued cells whose resolution sets what it can see. Pick the model that matches the phenomenon, keep one subject per layer, and half of GIS suddenly reads clearly.
Take it further
References & further reading

Peer-reviewed journals & authoritative standards

  1. 01Chang, K.-T. — Introduction to Geographic Information Systems, 9th ed.McGraw-Hill Education, 2019.
  2. 02Burrough, P.A., McDonnell, R.A. & Lloyd, C.D. — Principles of Geographical Information Systems, 3rd ed.Oxford University Press, 2015.
  3. 03Copernicus Data Space Ecosystem (Sentinel imagery)European Space Agency, ongoing.
Related lessons
Recap
Two data models: vector (features + attribute tables) for discrete things, raster (grid of cell values, defined by resolution) for continuous things. Layers stack both.
Carry forward →

Every feature and every cell claims to sit at a spot on Earth — but Earth is round and our screens are flat. How we squash the globe onto a plane, and why two “correct” datasets can still refuse to line up, is the coordinate-systems lesson next.

A

The author

Amogh N P

Architect, interior designer, and creative polymath. Studio Matrx began in his notebooks — his vision of design made honest, useful, and open to everyone. Its Academy is written and taught in his memory, and free, forever.

More about Amogh →