Raster and vector data, data input and matching

Detail

1. Two ways to store space: spatial data formats

• A GIS (Geographic Information System) is a computer system that stores, links and analyses data tied to places on the Earth.
• It stores the location part of the data (spatial data) in one of two forms (data models): raster or vector.
• It stores the descriptive part (names, population, soil type) as attribute data, kept in tables.

2. Raster data model

• Raster data model: space is divided into a grid of equal cells, also called pixels (Fig. 4.3, 4.4).
• Each cell has a row-column position, which tells you where it is in the grid.

• Each cell holds one value, such as brightness, land-use class or height.

• Resolution: the cell size sets the resolution.

• Small cells mean fine detail, but more cells to store.

• Large cells mean rough detail, but smaller files.

• Used for: satellite images, aerial photographs and scanned maps. It is also used for low-cost backdrop maps, the background layer that other data is drawn on.

• Raster suits continuous variables, things that change smoothly across space: elevation, temperature, rainfall and land use [6].
• In India: satellite images on ISRO's Bhuvan geoportal (run by NRSC, Hyderabad), including IRS imagery and terrain data, are raster layers [4]. On the Bhuvan platform, raster data is served through UMN MapServer and TileCache [5].

3. Vector data model

• Vector data model: features are stored as points with real coordinates: (x, y), or (x, y, z) when height is added.
• Point: one coordinate pair, e.g. a well or a town shown at small scale.
• Line: an ordered list of points, e.g. a road, river or railway.

• Polygon: a closed chain of ordered points that encloses an area, e.g. a district or lake.

• Vector suits discrete features with clear boundaries: roads, cities, rivers, lake and forest edges [6].

• Topology: the recorded relationships between features.
• Connection (connectivity): which lines join, e.g. which roads meet at a junction.
• Adjacency: which areas share a border, e.g. two neighbouring districts.
• Containment: which features lie inside others, e.g. a village inside a tehsil.

• Topology is built only after errors are corrected. A gap or overshoot would create false links.

• In India: Bhuvan offers vector layers of administrative boundaries, roads and railways, panchayat boundaries, river basins and sub-basins, soils and census data [4]. On the Bhuvan platform, vector data is served through GeoServer and GeoWebCache. Layers are shared as OGC-compliant WMS (Web Map Service, an international standard that lets any GIS software read the map) [5].

4. Raster vs vector (Box 4.2)

Box 4.2	Raster	Vector
Structure	Grid of equal cells	Points, lines and polygons from coordinates
Advantages	Simple data structure; easy overlay; fits satellite images; handles high spatial variability	Compact (less storage); precise map output; good for network analysis (roads, pipelines) and projection change
Disadvantages	Uses much storage; shapes look rough (stair-steps); poor for networks and projection change	Complex data structure; overlay is harder; does not fit satellite images
Best for	Continuous surfaces: images, elevation, rainfall	Discrete features: boundaries, roads, plots

• Exam trap: satellite images are always raster. "Vector is compatible with remote sensing images" is a false statement.

5. Sequence of GIS work

• GIS work follows this order: 1. Spatial data input: entering locations (maps, images). 2. Attribute data entry: typing in the descriptive tables. 3. Data verification and editing: finding and fixing errors. 4. Linking spatial and attribute data: joining each feature to its table row. 5. Spatial analysis: overlay, buffering, network and other queries.

6. Spatial data input: digitisation

• Digitisation: tracing map features into x, y coordinates, by hand on a digitising tablet or on screen over a scanned image.
• It is the best way to input vector data.
• Common digitising errors:

Error	What happens	Result
Omission	A feature is missed	Gaps in the data
Undershoot	A line stops short of the line it should meet	Open polygon, broken network
Overshoot	A line crosses past the line it should meet	Extra small "dangle" bits

• These errors are fixed in the verification and editing step. Topology is built after that.

7. Spatial data input: scanning

• Scanner: a machine that turns a paper map into a raster image.
• How a scanner works:
• A light source shines on the map.
• Reflected light falls on charge-coupled devices (CCDs), tiny light sensors.
• Each CCD turns light (photons) into electron counts.

• Each count is stored as a digital value, giving a raster image.

• The scanned image must then be cleaned (stains, labels, base-map lines removed) or converted to vector.

• Example: in a USGS project (2002), geologic maps were scanned at 300 dpi, 24-bit colour. Base-map details were removed by colour, and leader lines and labels were edited out. The cleaned image was then auto-vectorised: software traced the pixels into lines [7].
• Auto-traced lines had a "stair-step" look with too many vertices. They had to be smoothed [7].
• Using this method, 32 maps of 7.5-minute quadrangles were joined into one 30 × 60-minute map [7].

8. Georeferencing

• Georeferencing: giving a map or image real-world coordinates (latitude-longitude, or a grid such as UTM) so that all layers line up over each other.
• Method: control points with known coordinates are marked, e.g. map corners. The software then stretches the image to fit them [7].
• Errors and distortion:
• Wrong georeferencing: data sits in the wrong place or at the wrong scale.
• Lens faults in cameras.
• Relief: hills and valleys shift positions on aerial photographs.

• Folded or shrunk paper maps.

• India's base for accurate positions: under the National Geospatial Policy (NGP) 2022, the Survey of India launched a pan-India CORS network (Continuously Operating Reference Stations: fixed GNSS receivers that give exact positions all the time). It is part of redefining the National Geodetic Framework [2][3].

9. Raster-vector conversion

• Why: all layers in one analysis must be in one format.
• Vector → raster (rasterisation): a grid of chosen cell size is laid over the vector data, and each cell takes the value of the feature under it.

• This is more common, because most spatial analysis is raster-based.

• Raster → vector (vectorisation): cell edges are traced into lines and polygons.

• It is done mainly to save storage.
• The output needs smoothing to remove stair-steps [7].

10. Linking and matching data

• Exact matching: two files are joined on a common key that appears in both, e.g. town names in a map layer and in a Census table.
• Hierarchical matching: small-area data is added up until it fits larger units.
• E.g. village data → tehsil → district → state.

• Used when two datasets are at different levels.

• Fuzzy matching: used when boundaries do not line up.

• E.g. crop-field boundaries and soil-type boundaries.
• The layers are overlaid, and the value for each combination (crop × soil) is computed in every new small polygon.

11. Data compatibility traps

• Data from different sources may not fit together because of differences in:
• Scale: e.g. a 1:50,000 map mixed with a 1:250,000 map.
• Georeferencing system: datum and projection.
• Sampling: how and where the data points were collected.
• Classification: e.g. different land-use classes.

• Size of mapping units.

• These problems are worst across district and state boundaries, where agencies use different standards.

12. Where India is taking this: National Geospatial Policy 2022

• Notified: 28 December 2022. It looks ahead to 2035 [2].
• Main idea: geospatial data made with public funds should be openly available as a public good [2][3].
• Targets by 2025: a single unified digital interface for publicly funded data; a redefined National Geodetic Framework; a high-accuracy geoid model (a model of mean sea level used as the zero for heights) [2].
• Targets by 2030 [2]:
• High-resolution topographical surveys: 5–10 cm for urban and rural areas; 50–100 cm for forests and wastelands.
• High-accuracy DEM (Digital Elevation Model: a raster where each cell stores ground height): 25 cm in plains; 1–3 m in hilly and mountainous areas.

• A Geospatial Knowledge Infrastructure (GKI).

• Targets by 2035: high-resolution bathymetric data (depth maps) of inland waters and the deep sea for the Blue Economy; mapping of sub-surface infrastructure (pipes, cables) in major cities; National Digital Twins (live 3D computer copies) of major cities [2].

• SVAMITVA: the Survey of India has mapped over 2.8 lakh villages with drones in Andhra Pradesh, Haryana and Karnataka (reported February 2025). Drone images are raster; the property boundaries digitised from them are vector polygons [2].

Prelims Hooks

• Satellite images, aerial photographs and scanned maps are stored as raster. Raster is the model that fits remote sensing images. Vector does not.
• Raster resolution depends on cell size: smaller cells give finer resolution and larger files.
• Topology records connectivity, adjacency and containment. It is a feature of the vector model and is built after error correction.
• Vector is better for network analysis and projection change. Raster is better for overlay and high variability.
• Order of GIS work: spatial input → attribute entry → verification and editing → linkage → analysis.
• Undershoot, overshoot and omission are digitising errors.
• A scanner's CCD turns photons into electron counts, which are stored as digital values. Scanning produces raster output.
• Vector → raster is the more common conversion. Raster → vector is done to save storage.
• Exact matching uses a common key. Hierarchical matching aggregates small units. Fuzzy matching handles non-coinciding boundaries (crop vs soil).
• NGP 2022 was notified on 28 December 2022. Targets by 2030: DEM accuracy of 25 cm in plains and 1–3 m in hills. The Survey of India runs the CORS network [2].

Mains Points

• Choosing the data model shapes the analysis (GS-I):
• Raster suits continuous surfaces: flood extent from satellite images, DEMs, rainfall.
• Vector suits networks and boundaries: roads, canals, cadastral plots.

• Disaster mapping and Bhuvan's thematic services use both together [4].

• Data compatibility is the weak point of Indian planning:

• Departments use different scales, classes and datums.
• Data does not line up, especially across district and state lines.

• NGP 2022 responds with a unified data interface, a geodetic framework and standards [2].

• Accurate georeferencing and topology support governance (GS-III):

• The CORS network and drone-based SVAMITVA maps (2.8 lakh+ villages) give precise vector property records.

• These help land titles, loans and fewer disputes [2].

• Open data and the economy (GS-III):

• NGP 2022 treats publicly funded geospatial data as a public good.
• This supports private mapping firms, Atmanirbhar Bharat and future Digital Twins for city planning [2][3].