GIS Fundamentals — A Pre-Course Primer for QGIS

This guide covers the concepts you need before starting the QGIS course: the spatial data model, storage formats, and coordinate reference systems (CRS). CRS and projections are the main source of confusion for new users, so most of the guide is spent there. No software steps are included; the course covers those.

No prior GIS experience is assumed. Basic familiarity with files and folders is enough.

Spatial Data Model

Spatial data combines geometry (location) with attributes (information about that location). A dataset is spatial only when both are present and linked.

• Geometry — where the object is (point, line, polygon; or a pixel grid).
• Attributes — what the object is (name, type, population, measurement, date).

In QGIS, attributes are held in the attribute table — one row per feature, one column (field) per property. Selecting a feature on the map highlights its row, and vice versa. Analysis works by relating geometry to attributes (e.g. “select all polygons where landuse = residential within 500 m of a river”).

Attributes: Field Types and NULL

Every field has a type, fixed when the layer is created. The type controls what the field can store and how expressions and statistics treat it. Most beginner expression errors come from a type mismatch.

Type	Stores	Example	Notes
Integer	Whole numbers	42, -3	Counts, codes, year
Decimal/Real	Numbers with fractions	3.14	Measurements, coordinates
Text/String	Characters	residential	Has a length limit; codes with leading zeros must be text
Boolean	True / false	true	Some formats store this as 0/1
Date/DateTime	Calendar dates/timestamps	2026-07-11	Not the same as text; enables date arithmetic

NULL means “no value recorded.” It is not 0, not empty text, and not false. Expressions treat NULL specially: it usually propagates (arithmetic on NULL returns NULL), and summaries like count or sum skip it. Test with field IS NULL, not field = '' or field = 0.

Feature ID (FID / ObjectID) is an internal row identifier, not analytical data. It can change when you export or re-save a layer, so do not treat it as a stable key unless you have deliberately made one.

A number stored as text will not sort or calculate as a number, and a date stored as text will not support date math. Check the field type before writing expressions.

Spatial Data Types

Vector data

Discrete features defined by coordinates.

Type	Models	Examples
Point	A precise location	City, sensor, tree, address
Line	A linear feature	Road, river, pipeline, track
Polygon	An enclosed area	Building, parcel, lake, boundary

Point clouds

Dense X, Y, Z observations from lidar or photogrammetry. A small area can hold millions to billions of points, requiring specialized formats and processing separate from ordinary vector points.

Raster data

A grid of pixels (cells), each holding a value.

• Imagery — georeferenced photos from drones, aircraft, or satellites (often multi-band: red, green, blue, near-infrared).
• Continuous grids — one value per cell representing a continuous surface (elevation, temperature, population density).
• Mesh — time-varying values at continuous locations; common for weather/ocean data. A modern alternative to stacked rasters.

Tiles

Data pre-cut into small tiles for fast web delivery. Both raster tiles and vector tiles exist; the client fetches only the tiles it currently needs. This is how web basemaps work.

Vector vs. raster — how to choose

Use vector when	Use raster when
Features are discrete with clear edges	Data is continuous across space
You need attributes per feature	You need per-cell values
Precise geometry and topology matter	Imagery, surfaces, or model output
Example: parcels, roads, boundaries	Example: satellite image, DEM, rainfall

Raster resolution is the ground size of one cell (e.g. 10 m). Vector data is captured at a scale it is fit for; data digitized for a national map should not be trusted for street-level measurement.

Spatial Data Formats

Category	Non-spatial analog	Spatial format(s)
Text (vector)	CSV, JSON	CSV with lat/long columns, GeoJSON, KML/GPX
Binary (vector)	Excel, PDF	Shapefile, GeoPackage
Point cloud	—	LAS, LAZ (compressed)
Imagery/raster	PNG, JPEG	GeoTIFF, JPEG 2000
Database	PostgreSQL	PostGIS (vector + raster in-database)

Recommended default: GeoPackage, not Shapefile

Shapefile is common but dated, and its limits cause real problems:

• It is several files (.shp, .shx, .dbf, .prj, and more) that must travel together.
• 2 GB size limit per file.
• Field names capped at 10 characters (names get silently truncated).
• Weak field types and text-encoding issues (accented characters can corrupt).
• One geometry type, one layer per file.

.gpkg (GeoPackage) is a single SQLite file that holds multiple layers (vector, raster, and saved styles), supports long field names and larger sizes, and is an open OGC standard.

Use GeoPackage as your working default unless a task specifically requires another format.

Moving a .shp without its sidecar files loses data. Losing the .prj file loses the layer’s CRS entirely.

Modern/cloud formats

Cloud-Optimized GeoTIFF (COG), GeoParquet, FlatGeobuf, and PMTiles are designed for large datasets and cloud/streaming access. You will meet them as data volumes grow. Not required for the course.

Under the hood, QGIS reads and writes these formats through GDAL/OGR, the same libraries used by most GIS software. Format compatibility issues are usually GDAL issues.

Data Quality: Metadata, Accuracy, Precision, Scale

Every operation you run is only as trustworthy as the data going in. Evaluate a dataset before using it.

Metadata

Metadata is the data about the data. Good metadata records at least: CRS, source/creator, date, capture scale, positional and attribute accuracy, attribute (field) definitions, and licensing. Without it you cannot judge whether a dataset fits your task, and you may not even know its CRS.

Before trusting any dataset, inspect its metadata — CRS and capture scale first.

Accuracy vs precision

These are different, and beginners routinely conflate them.

• Accuracy — how close a value is to the true value.
• Precision — how repeatable a measurement is, or how many digits are reported. High precision does not imply high accuracy: coordinates printed to six decimals can still be in the wrong place.
• Positional accuracy (is the location right?) is separate from attribute accuracy (is the recorded value right?). A perfectly placed point can carry a wrong population figure, and vice versa.

Storage precision is not accuracy. A GPS reading of Accuracy: ±3 m is an error estimate, not a guarantee of exactness. A 30 m DEM stores one elevation per 30 m cell — that is its resolution, not proof that any elevation is correct to 30 m.

Scale

Vector data has a capture scale it was digitized for. 1:250,000 and 1:10,000 datasets of the same area can produce very different results; the coarser one omits detail the finer one captures. Zooming in within QGIS enlarges vertices and pixels but adds no accuracy — the data is exactly as detailed as when it was captured.

What GIS Is (and what QGIS is)

GIS = Geographic Information System — software that stores, displays, and analyzes spatial data. This is the everyday meaning.

A related but distinct term, Geographic Information Science (GIScience), is the academic discipline. The system is the tooling used within the science.

QGIS specifically

• Free and open source (GPL); no license cost, cross-platform (Windows, macOS, Linux).
• Built on established engines: GDAL/OGR (data I/O), PROJ (projections and datum transforms), GEOS (geometry operations). The same math underpins many commercial tools.
• Extensible through plugins (Python).
• Connects directly to PostGIS, SpatiaLite, and web services (WMS/WMTS/WFS, XYZ tiles).

What GIS Does — and where it lives in QGIS

Capability	In QGIS you will use
View — display, select, zoom	Layers panel, map canvas
Browse/connect to files and DBs	Browser panel
Create/edit geometry and attributes	Editing toolbar, digitizing tools
Analyze spatial relationships	Processing Toolbox
Make maps for print	Print Layout
Transform formats (ETL)	Export / Processing / GDAL tools

The Layers panel controls draw order and visibility, the Browser panel is how you add data, and the Processing Toolbox is where nearly all analysis lives.

Modeling the Earth’s Surface

Analysis needs a mathematical model of the Earth. Three levels, simple to accurate:

1. Sphere — a decent approximation. Mean radius ≈ 6,371 km. Simple math, limited accuracy.
2. Ellipsoid — flattened at the poles, so two radii. WGS84: semi-major 6,378.137 km, semi-minor ≈ 6,356.752 km. Accurate enough for calculation, and the surface all math is done on.
3. Geoid — the true, lumpy surface of equal gravity (historically approximated by mean sea level). Most accurate, but unusable for direct calculation.

You cannot compute on the geoid, so you compute on an ellipsoid chosen to fit it well. Which ellipsoid, and how it is positioned and oriented, is defined by the datum.

Datums

A datum ties an ellipsoid to the real Earth.

• Global (geocentric) datums — ellipsoid centered at Earth’s center of mass, best global fit. Used for global work; made practical by GPS.
• Local datums — ellipsoid oriented to fit one region tightly (minimum local distortion), with its center offset from Earth’s center. Accurate locally, poor elsewhere. Common in older data, still in use.

Worked example — geoid separation

At Nandi Hills, India:

• Signposted elevation (mean sea level / geoid): 1,478 m.
• GPS reading (WGS84 ellipsoid): 1,393 m.

Both are correct against different references. The ~85 m gap is the geoid separation, and it varies by location. Practical consequence: latitude/longitude values are ambiguous without their datum — the same coordinates on different datums can point to noticeably different places.

Example datums

Datum	Type	Notes
WGS84	Global	Used by GPS; the default assumption for most coordinates
NAD83 / ETRS89 / GDA2020	Regional-global	Continental datums (N. America, Europe, Australia)
Everest 1956	Local	Older Indian datum; required for old Indian maps

Moving data between incompatible datums requires a datum transformation. QGIS will prompt you to choose one; grid-based transforms (NTv2) are more accurate than simple parameter shifts. For high-precision work, “WGS84” alone is imprecise — it has multiple realizations/epochs, and plate motion means the specific datum and date can matter.

Coordinate Reference Systems (CRS)

A CRS = a datum + a coordinate grid + units. Two broad kinds.

Geographic CRS

Uses latitude/longitude (angular units, degrees) on the ellipsoid.

• Requires a datum, an origin (equator = 0° latitude; central meridian = 0° longitude), and a unit (degrees).
• Most common: WGS84 — EPSG:4326.
• Good for locating, poor for measuring. One degree of longitude is a different ground distance at the equator than near the poles, so distances and areas are unreliable in a geographic CRS.

EPSG:4326 formally orders coordinates latitude, longitude, but many tools and files use longitude, latitude. Swapped coordinates put your data in the wrong hemisphere. Watch for this when importing CSVs or GeoJSON.

The same point may be written in decimal degrees (40.7128, -74.0060) or degrees-minutes-seconds (40°42'46"N, 74°00'21"W). Both describe one location; know which format your source uses before importing.

Projected CRS

A flat plane with a uniform X/Y grid, an origin, and linear units (meters, feet). Required for reliable distance, angle, and area, and for most analysis.

EPSG codes are short numeric identifiers for coordinate systems (e.g. 4326), letting you select and share a CRS unambiguously. EPSG originated with the European Petroleum Survey Group and is now maintained by IOGP. Look codes up at epsg.io or spatialreference.org.

Map Projections

Flattening an ellipsoid onto a plane is always imperfect. Projections are built by projecting onto a cylinder, cone, or plane and unwrapping; the family hints at the distortion pattern, but the practical rule is what matters.

The trade-off — you cannot keep all three

No projection preserves shape, distance, and area at once.

Preserves	Type	Examples	Use for
Shape / angles	Conformal	Mercator, Lambert Conformal Conic	Navigation, angle-critical work
Distance (from a point)	Equidistant	Azimuthal Equidistant	Range/distance from a center
Area	Equal-area	Albers Equal Area, Equal Earth	Comparing sizes, density maps

Over a small enough area, all three can be held within tolerance. Choosing the right projection for the task is the analyst’s responsibility. Tissot’s indicatrix — small circles drawn on a map — is the standard way to visualize where a projection distorts.

Every common web basemap (OSM, Google, Bing) uses Web Mercator (EPSG:3857). It is fine for display and navigation but grossly inflates area toward the poles (Greenland appears near the size of Africa). Do not measure area or run area-based analysis in EPSG:3857 — reproject to an equal-area or local projected CRS first.

Which Projection to Use

Scale	Recommended	EPSG guidance
Global map	Equal Earth (equal-area, adopted by NASA); do not accept the GIS default	e.g. EPSG:8857
Country / region	Albers Equal Area (area) or Lambert Conformal Conic (shape); prefer a national grid	Varies
City / regional	UTM (Universal Transverse Mercator)	See below

National grids

Country	Grid
USA	Albers Equal Area Conic
India	Lambert Conformal Conic (Survey of India)
UK	British National Grid (Ordnance Survey)
Australia	Map Grid of Australia (MGA), by zone

UTM in practice

UTM is a set of 60 zones, each 6° of longitude wide, split into North and South (120 zones total). Pick the zone your area falls in, then use its projected CRS to minimize local distortion.

• Find your zone from a UTM reference map or epsg.io.
• EPSG convention (WGS84): North zones = 326 + zone number (Zone 43N → EPSG:32643); South zones = 327 + zone number (Zone 43S → EPSG:32743).
• Example: Bangalore → UTM Zone 43N → EPSG:32643.

CRS in QGIS — the part that trips people up

This section prevents the most common early mistakes.

Layer CRS vs Project CRS

• Layer CRS — the CRS the layer’s coordinates are actually stored in.
• Project CRS — the CRS the map canvas displays in (shown in the status bar, bottom-right).

On-the-fly (OTF) reprojection

QGIS automatically reprojects every layer to the Project CRS for display, so layers in different CRS still overlay correctly. This changes only the display, not the files.

Beginners see layers “line up” and wrongly conclude the files share a CRS. They may not. On-the-fly reprojection is a display convenience only.

Assign CRS vs Reproject — different operations

Operation	What it does	When to use
Assign / Set CRS	Declares the CRS the data is already in. Does not move coordinates	The layer’s CRS is missing, undefined, or wrong
Reproject (Warp)	Transforms coordinates into a new CRS, creating new values/a new file	Converting correctly-defined data into another CRS

Rule of thumb: if the data plots in the wrong place, its CRS is likely mis-declared → assign the correct one. If the data plots correctly but you need it in different units/CRS for analysis → reproject.

Practical CRS habits

• Set a sensible Project CRS (usually a projected CRS suited to your area) before measuring or analyzing.
• Keep analysis inputs in the same projected CRS.
• Never trust distance/area readings taken in a geographic CRS (degrees).

A Typical Geoprocessing Workflow

Most analysis follows the same shape. Having the sequence in mind prevents skipped steps (a mismatched CRS or unclean data invalidates everything downstream).

Acquire → Inspect metadata → Check & standardize CRS → Clean & validate → Reproject if needed → Analyze → Style & visualize → Compose map → Export

1. Acquire — obtain the dataset.
2. Inspect metadata — CRS, capture scale, accuracy, field definitions.
3. Check & standardize CRS — confirm each layer’s CRS is correctly declared; get inputs into a common projected CRS.
4. Clean & validate — fix invalid geometries, remove duplicates, check attributes.
5. Reproject — only if a step needs a different CRS or units.
6. Analyze — run the operations that answer your question.
7. Style & visualize — symbolize outputs to read them.
8. Compose & export — build a map layout or export data/results.

The first three steps happen before any analysis. Most “wrong result” problems trace back to skipping them.

Spatial Relationships and Topology

GIS answers questions by testing how geometries relate in space. These tests (spatial predicates) drive Select by Location, Join Attributes by Location, spatial queries, and overlay tools.

Predicate	True when
intersects	Geometries share any point (the most general test)
within / contains	One geometry lies fully inside the other (the two are inverses)
touches	They share only a boundary, not interior
overlaps	Interiors partly coincide, same dimension
crosses	Interiors meet at a lower dimension (a line crossing a polygon)
equals	Geometries are identical

Topology

Topology describes how features connect and share space, independent of exact coordinates: shared boundaries between adjacent polygons, lines that connect at nodes, no gaps and no overlaps (slivers). Many processing and overlay tools assume valid topology; invalid geometry produces wrong or empty results. Enable snapping while digitizing and validate geometry before analysis.

Core Vector Operations

Definitions you will apply repeatedly (all in the Processing Toolbox):

• Buffer — zone of a set distance around features (e.g. 100 m around rivers).
• Clip — cut one layer to the boundary of another (cookie-cutter).
• Intersection — keep only overlapping areas; combines attributes of both.
• Union — keep all areas from both layers, split where they overlap.
• Dissolve — merge features sharing an attribute into one (e.g. districts → states).
• Spatial join — attach attributes based on spatial relationship (e.g. tag each point with the polygon it falls in).
• Centroid — the geometric center point of a polygon.

All distance/area-based operations assume a projected CRS in meaningful units.

Core Raster Concepts

• Bands — layers of values in one raster (single-band DEM vs multi-band imagery like RGB or RGB+NIR).
• Resolution — ground size of one cell (smaller = more detail, larger files).
• NoData — a reserved value marking “no measurement”; excluded from calculations.
• Resampling — changing cell size/alignment (nearest neighbor for categorical data; bilinear/cubic for continuous).
• Raster calculator — per-cell math across bands/rasters (e.g. NDVI from red and NIR bands).
• Reclassify — group continuous values into classes (e.g. elevation → low/medium/high).
• Zonal statistics — summarize raster values within vector polygons (e.g. mean rainfall per district).
• Terrain products — slope, aspect, and hillshade derived from a DEM.

Common Beginner Pitfalls

• Mixing CRS in analysis — inputs in different CRS give wrong or empty results. Standardize first.
• Assign vs reproject confusion — see the QGIS CRS section.
• Measuring in a geographic CRS — degrees are not distances; reproject first.
• Using Web Mercator for area — inflates area toward the poles.
• Editing Shapefile schemas — 10-character field limit truncates names; prefer GeoPackage.
• Broken Shapefiles — moving a .shp without its sidecar files (especially .prj) loses data or CRS.
• Digitizing without snapping — creates gaps/overlaps (slivers); enable snapping and check topology.
• Comparing NULL as if it were a value — NULL is not 0 or empty text; test with IS NULL.
• Forgetting to toggle editing / save edits — changes are not written until you save the edit session.

Where to Get Data

Source	Content
Natural Earth	Free global vector + raster (small scale)
OpenStreetMap	Global roads/buildings/POIs (QuickOSM plugin, Geofabrik)
Copernicus Data Space	Sentinel satellite imagery
USGS EarthExplorer	Landsat imagery, elevation
OpenTopography	Elevation and lidar
National / open-data portals	Authoritative local data (boundaries, addresses)

Glossary

• Feature — one vector object with its attributes.
• Attribute table — the table of feature properties; one row per feature.
• Field type — the data type of a column (integer, decimal, text, boolean, date).
• NULL — “no value recorded”; distinct from 0, empty text, or false.
• Metadata — data describing a dataset (CRS, source, date, scale, accuracy, licensing).
• Accuracy — closeness of a value to the true value.
• Precision — repeatability of a measurement, or number of digits reported.
• Datum — definition tying an ellipsoid to the real Earth.
• Ellipsoid — the smooth mathematical Earth model used for calculation.
• Geoid — the true equal-gravity surface; basis for elevation.
• Geoid separation — height difference between geoid and ellipsoid at a location.
• CRS — coordinate reference system (datum + grid + units).
• Geographic CRS — lat/long in degrees (e.g. EPSG:4326).
• Projected CRS — flat X/Y in linear units (meters/feet).
• EPSG code — numeric identifier for a CRS.
• Projection — method of flattening the ellipsoid onto a plane.
• UTM — 120 zoned projected systems for local mapping.
• On-the-fly reprojection — QGIS reprojecting layers to the project CRS for display only.
• Spatial predicate — a relationship test between geometries (intersects, within, etc.).
• Topology — rules about how features connect and share space (no gaps/overlaps).
• Band — a value layer within a raster.
• NoData — reserved “no value” cell marker.
• Georeferencing — assigning real-world coordinates to an image.
• Geocoding — converting addresses to coordinates.
• Spatial index — a lookup structure for fast location-based queries.

Appendix: EPSG Codes You Will Use Often

Code	System	Use
4326	WGS84 geographic (lat/long)	Storage, GPS, data exchange
3857	Web Mercator	Web basemaps/display only
8857	Equal Earth	Global equal-area maps
326xx / 327xx	UTM North / South (WGS84)	Local mapping (xx = zone)
27700	British National Grid	UK mapping

Appendix: Terms You’ll Meet Early in QGIS

• Georeferencing — assigning real-world coordinates to an image (a scanned map or aerial photo) so it overlays other layers. Tool: Georeferencer.
• Geocoding — converting addresses or place names into coordinates; reverse geocoding does the opposite. Performed via plugins.
• Spatial index — a lookup structure that lets QGIS find features by location without scanning the whole layer, speeding up joins, Select by Location, and rendering. Built with Create Spatial Index; GeoPackage and PostGIS can store one.