Twenty years ago, learning GIS meant buying ArcView and an expensive textbook. Today, the open-source stack is genuinely better than any commercial product in several categories, and the only barrier to entry is knowing which tool does what. Here is a 2025 tour of the ten that actually matter.
1. QGIS — the desktop workhorse
Get it: qgis.org
QGIS is the open-source equivalent of ArcGIS Pro, except that it doesn't cost €4,000 per seat per year. Version 3.x is mature, fast, and capable of professional cartographic output. It reads every format you'll encounter (Shapefile, GeoJSON, GeoPackage, KML, GML, raster formats, web services), has a built-in Print Layout designer that produces publication-quality maps, and exposes hundreds of geoprocessing tools through a familiar dialog interface.
The Processing Toolbox includes GRASS GIS and SAGA algorithms — between them, there's almost no analytical operation QGIS can't do. Plugins extend the core: QuickOSM for OpenStreetMap downloads, MMQGIS for batch operations, qgis2web for one-click web map export.
If you only learn one tool, learn this one.
2. GDAL/OGR — the universal translator
Get it: brew install gdal, apt install gdal-bin, or conda install -c conda-forge gdal
GDAL (raster) and OGR (vector) are the command-line library that every other tool on this list secretly uses for I/O. QGIS reads a Shapefile through GDAL. PostGIS imports CSV through OGR. Mapbox Studio's upload pipeline calls ogr2ogr under the hood.
The canonical command is ogr2ogr, which converts between any two vector formats:
ogr2ogr -f GeoJSON output.geojson input.shp
ogr2ogr -f GPKG output.gpkg input.geojson
ogr2ogr -f "ESRI Shapefile" output_dir input.kmlFor rasters, gdal_translate and gdalwarp handle conversion and reprojection respectively. Learn ogr2ogr first; it pays for itself within a week.
3. PostGIS — the database brain
Get it: PostgreSQL plus CREATE EXTENSION postgis;
When your dataset outgrows a single file and your workflow outgrows a single analyst, you move to a database. PostGIS turns PostgreSQL into a spatial powerhouse with hundreds of ST_* functions for buffering, intersection, distance, network analysis, and topology.
The killer feature is that SQL composes. "Find all roads within 500 m of a wastewater plant, grouped by watershed, ranked by drainage area" is a single query that runs in milliseconds against millions of rows thanks to GiST spatial indexes.
Cloud-hosted options: AWS RDS, Azure Database for PostgreSQL, Supabase, Crunchy Bridge. For local development, the official Postgres + PostGIS Docker image is the fastest path.
4. Leaflet — the gentle web mapping library
Get it: leafletjs.com
Leaflet is the easiest way to put a slippy map on a web page. The API is small and friendly: create a map, add tile layer, drop in GeoJSON, attach popups. You can build a working map in 20 lines of JavaScript with no build step.
It is also the right tool for most maps. Leaflet doesn't do vector tiles natively or GPU-accelerated rendering, but for raster basemaps with a few thousand GeoJSON features, nothing is simpler. The plugin ecosystem covers clustering, heatmaps, drawing tools, and routing.
5. Mapbox GL JS / MapLibre GL — the GPU-accelerated alternative
Get it: maplibre.org (the open fork of Mapbox GL JS before its licence change)
When Leaflet hits its limits — too many features, slow zoom, no smooth pitched 3D — MapLibre GL takes over. It renders vector tiles on the GPU using WebGL, supports declarative styling via a Mapbox-style JSON spec, and stays smooth at 60 fps with millions of features.
The trade-off is complexity. You'll need to learn the style spec, generate vector tiles (with tippecanoe or martin), and host or pay for a tile server. For most casual maps, Leaflet remains the right choice; for serious web maps with large data, MapLibre is the default.
6. Turf.js — geospatial analysis in JavaScript
Get it: npm install @turf/turf
Turf.js brings GIS analysis into the browser and Node.js. Buffer a point, compute the centroid of a polygon, find the nearest neighbour, calculate distances, perform set operations — all client-side, all on GeoJSON.
It's a natural pairing with Leaflet/MapLibre: load some GeoJSON, run a Turf operation, display the result. For server-side workflows in JavaScript, it removes the need for a database for many simple analyses.
7. tippecanoe — vector tiles from a GeoJSON
Get it: github.com/felt/tippecanoe
tippecanoe (originally Mapbox, now maintained by Felt) generates Mapbox Vector Tile (MVT) sets from large GeoJSON or CSV input. It is the only tool you need to turn a 5 GB road network or a 10 million-point dataset into a smoothly-rendering web map.
The core idea: at low zoom levels, simplify aggressively and drop features; at high zoom levels, keep everything. tippecanoe makes those decisions automatically with sensible defaults, and exposes flags for fine control. Output is .mbtiles (an SQLite-based tile container) or a folder of .pbf tiles for HTTP hosting.
8. Overpass Turbo — OpenStreetMap data in 60 seconds
Get it: overpass-turbo.eu
OpenStreetMap is the world's largest free geographic dataset, and Overpass Turbo is the easiest way to pull pieces of it out. Type a Wizard query like building=hospital in Berlin, hit Run, and a few seconds later you have every hospital polygon in the city as GeoJSON, ready to download.
For anything beyond a city-scale extract, switch to bulk downloads from Geofabrik (regional .osm.pbf files) and process them with osmium or osmosis.
9. Felt — the collaborative web GIS that actually works
Get it: felt.com
Felt is a younger commercial product with a generous free tier, included here because it fills a real gap: collaborative web GIS without infrastructure. Drop in a Shapefile, GeoJSON, or CSV, share a link, and your colleague can edit the map in real time. It is what Google Docs would be if Google Docs were a GIS.
For exploratory analysis with non-GIS colleagues, it removes the install-QGIS-and-figure-out-CRSs barrier entirely.
10. Mapshaper — the underrated topology editor
Get it: mapshaper.org (web app) or npm install -g mapshaper
Mapshaper is a single-purpose tool that does one thing brilliantly: it simplifies polygons while preserving shared boundaries. If you need to take a 500 MB country-level boundary dataset and shrink it to 5 MB for a web map without ending up with gaps and overlaps between adjacent polygons, this is the tool.
It also converts between Shapefile, GeoJSON, TopoJSON, and CSV; performs basic editing; and runs entirely in the browser for files up to a few hundred megabytes.
Putting them together
A realistic 2025 stack for a small GIS project might be:
- Data acquisition: Overpass Turbo for OSM extracts, government open-data portals for everything else.
- Cleaning and analysis: QGIS for inspection and one-off operations, GDAL for batch jobs.
- Storage: GeoPackage for desktop work; PostGIS once you need multi-user editing.
- Simplification: Mapshaper for boundaries; tippecanoe for vector tiles.
- Web publishing: Leaflet for simple maps, MapLibre for large data, Felt for collaborative exploration.
- Server-side analysis: PostGIS SQL queries, or Turf.js for JavaScript pipelines.
Every tool is free. Most have first-class documentation and active communities. The hardest part of learning GIS in 2025 is no longer cost — it's picking the right tool for each step.