What is GIS? A UK Guide for Flood Risk and Drainage

GIS stands for geographic information system. It is software that holds map features and the data attached to them together, so that questions about a place can be answered by analysis instead of by eye. Anything with a location can become a layer, and the GIS works out how the layers relate to one another.

That is the whole idea, and it is a good deal simpler than the acronym suggests. What follows sets out how a GIS actually works, what the five things it does are, what people across a dozen industries use it for, which software is worth starting with and what it costs, and — the part most introductions leave out — where it stops being reliable and something else has to take over.

A paper map shows you where something is. A GIS can tell you how much of it there is, what it sits next to, what has changed since last year, and what happens if you move it.

What is GIS?

A geographic information system stores two things about every feature on a map: its geometry, meaning where it sits and what shape it is, and its attributes, meaning everything else that is known about it. Holding the two together is what separates a GIS from a picture of a map.

A river is the plainest example. On screen it is a line. In the attribute table behind that line it also has a name, a classification, a catchment area and a gauging record. Ask a paper map which rivers in a county drain more than fifty square kilometres and it cannot answer. Ask a GIS and the answer is a query, because it is those columns, not the line, that the analysis runs on.

Ordnance Survey's definition is the one worth keeping, because it survives contact with practice:

The word to notice is system. GIS is often used as a synonym for the software, and in casual use that is harmless, but the software is only one part of it. A working system is the data, the software, the coordinate framework that makes the data align, and a person who knows which layer to trust. Take away any one of those and the map still draws. It is just wrong in a way nobody looking at it can see. That is why a flood risk assessment can state how much of a site sits in Flood Zone 3, where a screenshot of the same map can only show it.

One piece of vocabulary, because it trips people up. Geospatial is the broader word, covering any data tied to a location and any technique for handling it, from satellite positioning to geospatial analysis of floods, disease or crime. GIS is the part of that world concerned with holding the layers and analysing them together. In everyday use the two are swapped freely and little harm comes of it.

GIS answers questions a map cannot: not “where is the boundary” but “how much falls inside it, what changed, and what happens next”.

How does GIS work? Layers, vector and raster

GIS works by stacking independent datasets in a shared coordinate space so they can be compared. Each dataset is a layer. Turn a layer on and it appears in its correct position relative to every other layer; turn it off and the rest are unaffected. Analysis is then a matter of asking how the layers intersect, and the answer comes back as a number or as a new layer.

Layers come in two data models, vector and raster, and the difference between them governs what can be done with a layer once it is loaded.

Vector data represents discrete things with defined edges. In the QGIS documentation's terms, “a vector feature has its shape represented using geometry”, built from one or more vertices, and carries attributes describing it. Points, lines and polygons: a postbox, a road, a field, a county boundary.

Raster data represents something that varies continuously and has no natural edges. A raster is “a matrix of pixels (also called cells), each containing a value that represents the conditions for the area covered by that cell”. Ground height, rainfall, temperature, an aerial photograph.

The distinction matters because the operations differ. A polygon can be clipped, buffered and measured; a raster is sampled, reclassified and differenced. Area comes from vector geometry. Ground height comes from a raster. Work out which way water runs across a hillside and you start with a raster and finish with a vector polygon.

Every layer in that stack is usually published by a different organisation, at a different scale, on a different update cycle. Making them agree is most of the work.

What is GIS data, and what is an attribute table?

GIS data is spatial data with a coordinate reference attached, and an attribute table is the spreadsheet behind the map. Click one polygon and the attribute table gives its type, its source, its version date and its area. Select two hundred of them and the same table supports a query: show only these, inside this boundary, totalling this much. The map is the interface; the table is where the answer comes from.

Formats matter more than they look. Each one carries or loses different information, and the losses are silent.

• Shapefile. Still the common currency of data exchange, and the oldest. The Library of Congress format registry records its real constraints: a 2 GB limit per component file, attribute field names capped at ten characters, a maximum of 255 fields, no support for null values, and numeric values stored as characters, “thus leading to rounding errors”. It is also not one file but at least three, which must travel together.
• GeoPackage. The OGC's open, single-file replacement, built on SQLite. One file holds vector features, raster tiles and attributes, with no name-length limit and no companion files to lose.
• GeoTIFF. The standard raster container: an image with its coordinate system written into the file's own tags. Opening one in ordinary graphics software and saving it back is a reliable way to destroy that georeferencing.
• GeoJSON. A lightweight text format defined by IETF RFC 7946, used heavily for web mapping. It defaults to latitude and longitude in decimal degrees, which is not what British mapping uses, and any third coordinate is a height above the WGS 84 ellipsoid rather than a height above sea level.

A shapefile will silently shorten the field name “infiltration_rate” to “infiltrat”. A good deal of the friction in data exchange traces back to a format designed in 1998.

What are the main functions of GIS?

GIS is usually described as doing five things, and the sequence is worth knowing because most confusion about the subject comes from collapsing the middle three into “making a map”.

1. Capture. Getting real-world things into the system, by survey, by satellite or aerial sensing, by digitising an existing map, or by joining a spreadsheet of addresses to a set of coordinates.
2. Store and manage. Keeping the geometry, the attributes and the coordinate reference together, with a record of where each layer came from and when.
3. Query. Asking the attribute table questions without touching the geometry: which features are of this type, above this value, recorded after this date.
4. Analyse. The part that makes it a GIS rather than a drawing package. Overlaying layers to find where they coincide, buffering a feature to find what falls within a set distance, measuring areas and distances, and deriving new information from existing layers.
5. Present. Turning the result into a map, a table, a chart or a web service that somebody else can read and act on.

The most famous map in the history of the subject makes the point about that ordering rather well, though not in the way it is usually told. John Snow's 1854 map of cholera deaths around the Broad Street pump in Soho is routinely described as the map that found the source of the outbreak. It did not. Snow had published his waterborne theory in 1849, five years earlier, and had argued the Broad Street case from interviews and death records; the pump handle was removed on 7 September 1854, and the map was not shown until that December, nor published until 1855. It was made to prove an argument that analysis had already settled.

The analysis is the work. The map is how the work gets argued, and it is the step people mistake for the whole discipline.

What is GIS used for?

Almost any organisation that owns things, serves an area, or has to decide where to put something uses GIS, whether or not anyone there calls it that, and the range runs from the council deciding which gullies to clean first to the epidemiologist mapping an outbreak. A few of the commonest applications:

• Local government. Asset registers for street lighting, gullies, trees and bins; planning constraint maps; school catchments; bin-round routing. Much of what a council publishes on a public map viewer is a GIS layer served straight out of the back office.
• Utilities and infrastructure. Network records for water, power, gas and telecoms. The scale of this is easy to underestimate: the National Underground Asset Register now maps 3.2 million kilometres of buried pipes and cables, against a background of roughly 60,000 accidental strikes a year costing an estimated £2.4 billion.
• Emergency response. Mapping an incident as it develops, identifying who is inside an affected area, and routing crews around a closed road or a flooded one. Live sensor feeds go into the same stack, which is how a map of river levels updating through the night is built.
• Retail and commercial property. Catchment analysis and site selection, which is essentially a buffer, a population layer and a question about travel time.
• Ecology and land management. Habitat mapping, designations, tree canopy cover, and measuring change between two dates.
• Public health and epidemiology. The field Snow started. Mapping cases against population, deprivation and access to services.
• Archaeology. Airborne laser scanning has revealed earthworks under woodland and pasture that ground survey had missed for centuries, because the technique can be processed to strip the vegetation out and leave the bare ground behind.

Engineering and environmental consultancy sits in the same list. A drainage strategy, a contaminated land assessment, a noise model and a landscape and visual appraisal are all built on stacked layers and the same handful of operations. The worked example further down takes one of those and follows it through.

If a question starts with “how many”, “how far”, “how much of” or “which of these is nearest”, and it is about real places, it is a GIS question.

GIS in the UK: British National Grid and Ordnance Datum

Anyone using GIS in Britain runs into two national conventions early, and getting them wrong does not produce an error message. It produces a map that is quietly in the wrong place.

British National Grid is the projected coordinate system used for all Ordnance Survey mapping in Great Britain, catalogued as EPSG:27700. It expresses position as eastings and northings in metres, not as latitude and longitude. Satellite positioning works in WGS84 or ETRS89, which are global systems built on a different ellipsoid, so data arriving from a phone, a web map or a GeoJSON file is almost never in the same system as the base mapping underneath it.

Converting between the two is not a formality. Ordnance Survey's own guidance is blunt about the scale of it: different latitude and longitude systems in common use “can disagree on the coordinates of a point by more than 200 metres”, and a simple single-step transformation from ETRS89 to OSGB36 “will give errors of up to 3m (95%) in plan and 3.5m (95%) plan and height”. The correct transformation, OSTN15, is a grid of shifts at one-kilometre resolution rather than a formula, and it brings agreement to better than 0.1 m.

Ordnance Datum Newlyn is the height equivalent. It is the zero point for every OS contour, spot height and bench mark on the British mainland, and it corresponds to mean sea level measured at the Newlyn tide gauge between 1915 and 1921. That figure has never been revised, which is the point: a height above Ordnance Datum does not drift as sea level does. It is why heights in British professional practice, from a railway platform to a finished floor level, are quoted in metres above Ordnance Datum rather than in whatever a satellite receiver happens to report.

Most of the national data worth having is free. Ordnance Survey publishes eighteen OpenData products under the Open Government Licence, including base mapping, the watercourse network and a terrain model; the British Geological Survey publishes geology and borehole records; MAGIC aggregates environmental designations from six government bodies; the Environment Agency publishes elevation, flood and water-quality data through its data services platform, one of nearly two thousand of its datasets listed on data.gov.uk; and DataMapWales does the equivalent for Wales, including the Welsh flood map for planning. Licences vary by dataset and are worth reading rather than assuming — detailed Ordnance Survey and British Geological Survey products are licensed commercially, not open.

A raw satellite position pushed through a simple transformation can be 3.5 m out in plan and in height. For most purposes that is invisible. For a floor level or a pipe invert it is the whole job.

A worked example: GIS in flood risk and drainage

Everything above is general. It is worth seeing the same five functions run through one real application end to end, because the abstract version can make GIS sound like colouring in. Flood risk and drainage is the example used here; the shape would be much the same for contaminated land, ecology or highways.

1. Fix the boundary. The site outline is digitised or imported and reprojected to British National Grid. Its area in hectares is then a property of the polygon, not a figure anyone measures by hand.
2. Load the constraints. Flood zones, surface water risk, historic flood outlines, geology, groundwater source protection zones, and watercourses with their main river or ordinary watercourse status. What matters is not that each one is visible but that each one is measurable against the boundary.
3. Bring in the terrain. A national elevation model is clipped to a buffer around the site and read for fall, low spots and flow paths. This is also where a measured topographical survey is ordered or ruled out.
4. Analyse. This is where GIS meets hydrology. Flow direction and accumulation give the catchment draining onto the site, which sets the contributing area for every drainage calculation that follows; intersecting the boundary with each constraint gives the numbers the assessment turns on. Infiltration potential, taken first from published ground conditions and later confirmed by on-site testing, together with groundwater levels and slope, then decides where a soakaway, swale or basin can physically go, which is the spatial half of a compliant surface water drainage strategy.
5. Present the evidence. Constraint plans, flood extent plans, exceedance routing and level drawings, all referenced to Ordnance Datum because the Environment Agency's standing advice requires it. That is what the planning authority and the lead local flood authority actually read.

Planning policy never uses the word GIS, but it assumes the output. Planning Practice Guidance asks a site-specific flood risk assessment for a location map, a national grid reference, block plans and “maps and level surveys of the site”, and states that “if possible, flood levels should be presented in metres above Ordnance Datum”. The 2025 National Standards for Sustainable Drainage Systems, covered in full in Unda's own explainer, require exceedance routes to be “identified across the development” and let the approving body use “appropriate industry mapping” as a first infiltration evidence base, naming British Geological Survey datasets in the footnote.

Georeferenced boundaries, levels to Ordnance Datum, mapped exceedance routes, named geological datasets. That is a GIS specification written by people who never say the word.

What GIS is not: Google Maps, CAD and modelling software

GIS is routinely confused with three other things that also produce maps. The distinctions are practical, not pedantic, because each of the three does something GIS does not.

The prediction point is the one worth holding on to. GIS does not forecast anything. It holds and analyses the evidence a forecast is built from, and it maps the result once the forecast has been made elsewhere. Google Maps is the opposite confusion: it is a beautifully finished map and almost no GIS at all, because the layers behind it are not yours to interrogate.

In flood work that second tool is a hydraulic model. GIS prepares the terrain, the boundaries and the roughness; the model solves the flow equations; GIS takes the results back and turns them into depth and extent mapping. The two are a pipeline, and choosing between 1D, 2D and integrated modelling is a separate decision with its own criteria, compared in detail elsewhere.

GIS software, and how to start using it

Two platforms account for most professional work, and one of them is free. The choice between them turns on ecosystem more than on capability, which is not what most comparisons suggest.

• QGIS. Free and open source, and capable of everything described in this article: vector overlay, terrain analysis, georeferencing, map production, and Python for automation. It reads every format above natively. There is no trial period, no seat limit and no feature held back.
• ArcGIS Pro. Esri's commercial platform, licensed per seat, dominant in local government and large infrastructure. Its advantage is less the software than the estate around it: shared services, organisational data portals and the web maps many public bodies already publish through.
• Learning it. Self-teaching works, and most practitioners have done it. Install QGIS, pick a real place you know, download a few open datasets and try to answer one specific question about it. The learning comes from getting the layers to line up, which is the part no tutorial can hurry.
• What takes longer. Software is the easy part. Knowing which dataset is authoritative for which purpose, how old it is and what it was built to do takes far longer, and it is the difference between a map that looks right and an answer that is right.

Nothing in this article requires paid software. The binding constraint is knowing which dataset answers which question, not the licence fee.

What GIS cannot tell you

A GIS is only ever as good as the data in it, and national datasets are generalisations, assembled at a scale and a date chosen for national coverage rather than for the place you happen to care about. Reading the output as fact instead of as evidence of a particular vintage and resolution is the commonest failure in spatial work, and it is entirely avoidable.

None of that makes the data less useful. It makes the interpretation the valuable part: knowing what each layer was built for, what it cannot carry, and when the honest answer is that somebody needs to go and look.

The skill is not operating the software. It is knowing which of twenty layers decides the answer, and how far that layer can be trusted.

GIS has become the substrate under a great deal of professional work, from asset management to sustainable drainage design, and in Britain most of the national data feeding it is now free to anyone who wants it. What stayed scarce is the judgement to read it correctly. Unda's flood risk assessments, drainage strategies and hydraulic modelling are all built on that foundation, and anyone weighing up a site can talk to Unda's flood risk and drainage consultants about what the data already shows.

Frequently asked questions about GIS