Vector geographic data are digital representations of geographic features using coordinate-based geometries, principally points, lines, and polygons, usually associated with descriptive attributes. They are a fundamental data model in a geographic information system (GIS), used to describe objects such as buildings, roads, rivers, land parcels, and administrative areas. Unlike raster data, which organize space into a grid of cells, vector data describe locations and boundaries explicitly. The term vector here refers to this geometric representation, rather than specifically to quantities having magnitude and direction.
Geometry and feature representation
A geographic feature represents an identifiable real-world object or phenomenon. Its geometry describes where it is and its spatial form; its attributes describe properties such as its name, category, population, or observation date.
The principal geometry types are:
| Geometry | Representation | Typical examples |
|---|---|---|
| Point | A single coordinate position | Survey marker, sampling station, address location |
| Line string | An ordered sequence of positions connected by segments | Road centerline, river course, utility cable |
| Polygon | An area defined by an exterior boundary and, optionally, interior boundaries | Land parcel, building footprint, lake |
| Multipart geometry | Several geometries of the same basic type grouped as one object | An island municipality or a road with disconnected sections |
| Geometry collection | A grouping that may contain different geometry types | A complex feature represented by both points and lines |
A line string is commonly called a polyline. Its vertices define its shape; adding vertices can describe bends more closely but does not necessarily improve positional accuracy. Some data models also support circular arcs and other curved segments.
A polygon has a closed exterior ring and may contain interior rings representing holes. For example, the land surface of an island containing a lake can be represented by a polygon with an interior ring. A multipolygon can represent several separated islands as one feature.
Geometry type depends partly on purpose and map scale. A city may be a point on a national map and a polygon in a regional planning database. A river may be represented as a centerline for network analysis or as an area for floodplain mapping. Neither representation is inherently more correct: each preserves different properties of the feature.
Coordinates and dimensionality
Two-dimensional geometries usually store coordinate pairs . Additional ordinates may include elevation or depth, conventionally denoted , and a measure value , such as distance along a route.
A geometry containing elevation values is not automatically a complete three-dimensional model. A building footprint with heights attached to its vertices does not necessarily describe walls, internal spaces, or a closed solid. Surface meshes and solid geometries require more specialized structures.
Vector representations can also describe surfaces. A triangulated irregular network models a surface through connected triangles and is commonly used for terrain and engineering applications.
Coordinate reference systems
Coordinates become geographically meaningful through a coordinate reference system (CRS). This specifies how coordinate values relate to positions on Earth, including relevant axes, units, and reference definitions.
Geographic coordinate systems commonly use latitude and longitude, expressed as angles. Projected coordinate systems use a map projection to express positions on a plane, usually in linear units such as metres. A geodetic datum provides the reference framework underlying these positions.
Several distinctions are important:
- Assigning a CRS identifies how existing coordinates should be interpreted; it does not change their values.
- Transforming coordinates changes their values to express positions in another reference system.
- Georeferencing establishes the spatial relationship between data and a geographic reference framework, often as part of preparing source material for digitization.
Distance and area calculations must use methods appropriate to the CRS. A planar calculation using longitude and latitude as though they were metres is generally unsuitable. Even projected coordinates introduce distortion, so accurate measurements may require a suitable local projection or calculations on an ellipsoid.
Coordinate order is another source of error: some standards and software interfaces use longitude–latitude order, while others follow latitude–longitude or other axis conventions.
Attributes and data organization
Vector datasets usually organize related features into layers or feature classes. A road layer, for example, may associate each line geometry with an identifier, road name, classification, speed limit, and maintenance authority.
In a common tabular model, each row corresponds to a feature and each column to an attribute. Geometry is stored in, or associated with, that record. Unique identifiers allow records to be related to external tables, such as inspection histories or demographic observations.
Attributes can be numeric, textual, categorical, temporal, or more complex. Their interpretation requires information about units, permissible values, missing-data conventions, and definitions. A population field without an observation date or a height field without units may be misleading even when the geometry is correct.
Feature identity is distinct from geometry. Two objects may share the same location but represent different things, such as a bridge and a road segment. Conversely, one feature may have multiple spatial parts or change geometry over time while retaining its identifier.
Topology and spatial relationships
Geospatial topology concerns relationships such as connectivity, adjacency, containment, and shared boundaries. It distinguishes questions about how features relate from questions about their exact coordinates.
In a simple independent-feature model, often called a spaghetti model, each geometry stores its own coordinates. Adjacent polygons may therefore store separate copies of what is intended to be the same boundary. A topological model can explicitly represent shared edges and nodes, supporting consistent editing and relationship checks.
Common topology rules include:
- Parcels intended to partition an area must not overlap.
- Polygon coverages must not contain unintended gaps.
- Connected road segments must meet at appropriate junctions.
- Specified point features must lie on corresponding lines.
- Shared administrative boundaries must remain coincident.
These rules depend on the application. Overlapping polygons may be errors in a parcel fabric but entirely legitimate in habitat or service-area datasets.
Geometric intersection also does not necessarily imply real-world connectivity. Two road lines crossing in plan view may represent an overpass rather than an intersection. Transport networks may need levels, directions, turn restrictions, and explicit junction information beyond line geometry alone.
Acquisition and development
Vector data can originate from field surveys, satellite positioning, manual digitization, image interpretation, automated extraction, engineering drawings, or existing administrative records. Paper maps and imagery may first require georeferencing before features can be traced into geographic coordinates.
Automated extraction can identify roads, buildings, or land-cover boundaries from imagery, but the results still depend on image quality, classification methods, and validation. A visually plausible outline may be displaced, incorrectly classified, or incomplete.
Vector methods developed alongside digital cartography, computer-aided drafting, and GIS from the mid-twentieth century onward. Early geographic systems used coordinate-based boundaries and line networks to support mapping and spatial inventories. Subsequent developments in database technology, standardized geometry models, and web services made vector features easier to store, exchange, edit, and analyze across systems.
Spatial analysis
Vector data support a wide range of spatial analysis operations.
Spatial queries and joins select or combine features according to location. A point-in-polygon operation can associate monitoring stations with administrative areas.
Buffering creates areas within a specified distance of a feature. The result depends on the distance model, coordinate system, and treatment of line ends and corners.
Overlay combines geometries and attributes from different layers. Intersection identifies common portions, union combines their spatial subdivisions, and difference removes areas occupied by another dataset.
Network analysis examines movement or connectivity along linked features. Applications include route finding, utility tracing, and accessibility analysis. Reliable results require appropriate network structure and attributes, not merely visible lines.
Aggregation and generalization combine or simplify features. Dissolving polygons by an attribute can remove internal boundaries; cartographic generalization can reduce geometric detail for smaller-scale display.
Vector data also contribute to spatial interpolation, in which observations at sampled locations are used to estimate values elsewhere. The resulting surface may be stored as raster data or represented through other models.
Storage, standards, and delivery
Vector data can be stored in files, spatial databases, or web-accessible services. Common formats include:
- Shapefile: a widely supported format whose geometry, attributes, and indexing information are distributed across several associated files. It has restrictions affecting field names, attribute types, and some geometry capabilities.
- GeoJSON: a text-based format for features and geometry. Under RFC 7946, coordinates use WGS 84 longitude and latitude in decimal degrees, with an optional third coordinate for height.
- GeoPackage: a standardized SQLite-based container that can hold vector feature tables as well as other geospatial content.
- Spatial databases: database systems with geometry types, spatial indexes, spatial queries, and transaction support.
The Open Geospatial Consortium develops standards relevant to feature models, storage, and exchange. Interoperability nevertheless depends on more than geometry encoding: CRS interpretation, attribute schemas, character encodings, identifiers, and geometry validity must also agree.
Vector tiles divide and encode geographic content for efficient map delivery. They commonly contain clipped and simplified geometries appropriate to particular display scales. They should not automatically be treated as substitutes for the full-resolution source dataset.
A Web Map Service, by contrast, typically delivers rendered map images rather than directly editable feature geometries.
Quality, uncertainty, and limitations
Vector data can appear precise because coordinates contain many decimal places and boundaries look sharp. Numerical precision is not the same as accuracy. A highly detailed geometry can still be systematically displaced or derived from an uncertain source.
Important quality dimensions include positional accuracy, attribute accuracy, completeness, temporal currency, and logical consistency. Measurement uncertainty may arise from surveying, interpretation, coordinate transformations, or ambiguity in the feature itself. A coastline, vegetation edge, or urban boundary may not have a single stable, sharply defined position.
Geometry validation can identify problems such as self-intersecting polygon boundaries, invalid holes, or collapsed components. Dataset-level checks are also needed: individually valid polygons may still overlap or leave unwanted gaps. Overlay operations can expose small mismatches as narrow sliver polygons, while inappropriate snapping tolerances can move or merge legitimate features.
Analytical conclusions may depend on how features are delineated or aggregated. Statistics calculated for administrative polygons, for example, can change when the boundaries or aggregation scale change, an issue associated with the modifiable areal unit problem.
Relationship to raster data
Vector and raster models emphasize different aspects of geography. Vector data are often effective for individually identifiable objects, explicit boundaries, and networks. Raster data are often effective for imagery and regularly sampled fields such as temperature or elevation. A digital elevation model, for example, may be stored as a raster grid or represented using a triangulated surface.
Conversion between models changes the representation. Rasterizing a polygon assigns values to cells and introduces resolution-dependent boundary effects. Vectorizing classified cells can produce complex, stair-stepped boundaries without recovering detail absent from the source.
Neither model is universally smaller, faster, or more accurate. Performance depends on feature complexity, cell resolution, indexing, compression, and the operation being performed. Practical GIS workflows commonly combine them: vector parcels may be used to summarize raster flood depths, while vector observations may provide inputs for a raster prediction surface.