QGIS: The Mining Engineer Who Gave Away the Map

A Living Room in Alaska

It is the winter of two thousand and two. A man named Gary Sherman is sitting at a computer in Alaska, looking out at a yard that may or may not have a moose in it. He has finished his actual job for the day. He has data, geographic data, sitting in something called PostGIS, which is a database that knows about latitude and longitude and shapes on the ground. And he wants to look at it.

He is running Linux on his desktop. Most of the software that draws geographic data into actual maps runs on Microsoft Windows, costs thousands of dollars, and comes with a license that tells you exactly which seat you can install it on and exactly how many people are allowed to look at the screen. None of it will help Gary tonight. So he sits down and starts writing his own viewer. He has no plan to start a movement. He just wants to see his data.

The thing he starts that night will, two decades later, sit on Pär's laptop in Kall with a map of Åre kommun loaded into it.

What the Map World Looked Like Before

To understand why Gary's little project mattered, you have to picture the field he was up against. Geographic information software, the category of tools that takes spatial data and turns it into useful maps and useful analysis, was at that point dominated by one company. That company was, and still is, Esri. Their product was called ArcGIS, and it was excellent. It was also expensive, Windows-only, and licensed in a way that pinned it to specific machines and specific named users at a specific organisation.

[serious] If you were a national mapping agency or a big oil company, that was fine. You had the budget. You had the staff. You had the Windows machines. But if you were a graduate student in Africa, or a small kommun office in Jämtland, or a volunteer mapping the spread of a fire, you were not in the room where the maps got made. The tools simply were not for you.

There were attempts at alternatives. There was GRASS, an older system originally built by the United States Army for environmental work, which was free but had what people generously called a learning curve. There were a few other small projects. None of them gave you a desktop tool where you could just open a file, see the map, change the colours, and print it.

The Geologist Who Wrote Code

Gary Sherman's biography is the kind of thing that explains a lot once you read it. He trained as a geologist. He worked as a mining engineer. He worked as a software engineer. He has, on top of all of that, written novels. He has authored technical books on geospatial software. He ran a small publishing house called Locate Press for about eight years, putting out books about open source maps when almost no one else was publishing them.

He lives in Alaska. He has, by his own account, watched a lot of moose from his living room window. There is something useful about knowing this. The people who build the infrastructure that everyone else later relies on are very often not professional infrastructure builders. They are people with a problem of their own and enough stubbornness to fix it.

In a later interview Gary described how the project began.

QGIS started as a solo project in February two thousand and two. After work, I wanted to view PostGIS data on my Linux machine.

That is the whole origin story. He wanted to look at his data. He was on the wrong operating system to do it with what already existed. So he wrote the thing himself.

Why Is There a Q

The full original name of the project was Quantum GIS. The Q is not for quantum in the physics sense. There are no entangled particles in your map of Åre kommun. The Q is for Qt, which is the cross-platform toolkit Gary chose to build the interface with. Qt was, and is, a piece of software that lets a programmer write one set of code and have it produce a working application on Windows, on macOS, and on Linux, without having to rewrite the buttons and menus three separate times.

Picking Qt was one of those quiet decisions that turned out to matter enormously. It meant that from the very first day, the project could run anywhere. It was not a Linux tool that someone might eventually port to Windows. It was not a Windows tool that might someday work on Mac. It was already, structurally, all three at once. When the project later renamed itself from Quantum GIS to just QGIS, the Q stayed. It is a small monument to the Qt toolkit, baked into the name forever.

From a Solo Project to a Foundation

Version zero point zero point one of Quantum GIS came out in July of two thousand and two. It was a viewer. You could open shapefiles, which is the most common format for vector geographic data, and you could see them on the screen. That was the whole feature set. You could not edit. You could not really analyse. You could just look.

But people noticed. Within a few months Gary had patches arriving from other developers. By the middle of the decade there was a small community. In two thousand and seven, the project entered something called incubation at the Open Source Geospatial Foundation, known as OSGeo, which is a non-profit umbrella organisation that gives open source mapping projects a stable home. In two thousand and eight, QGIS graduated from incubation and became a full OSGeo project, which is the geospatial equivalent of getting your driver's licence.

Version one point zero arrived in two thousand and nine. That was the version that announced to the world that this was a serious tool, not a hobby viewer. It had editing. It had styling. It had the start of a plugin system. From there it grew steadily. New releases every four months. New plugins from volunteers. New translations. By two thousand and fourteen, Gary received the Sol Katz Award, which is the open source geospatial world's most prestigious honour. It is named for Sol Katz, who in the early nineteen eighties had worked on what may have been the very first open source GIS system, called MOSS, for Map Overlay and Statistical System. The line from Sol Katz to Gary Sherman to your laptop is a real line.

Today the QGIS project lives at qgis dot org and is governed from Switzerland, with a board of directors, a treasury, an enhancement proposal process modelled on the way Python is governed, and somewhere between dozens and hundreds of regular contributors depending on how you count. Gary himself stepped back from running it years ago. He now runs a small consultancy called GeoApt and has talked about retiring fully and doing more travelling.

What QGIS Is Actually Doing On Your Screen

Now let us talk about the program itself, because the most interesting thing here is not the history. The most interesting thing is what is happening when you open QGIS, load a map of Åre kommun, and look at the screen.

A geographic information system is, at its core, a stack of layers. Imagine a pile of transparent sheets, one on top of another. The bottom sheet might be a satellite photo of northern Jämtland. The next sheet up might be a digital elevation model, which is a grid of numbers where every pixel tells you the height above sea level at that spot. The next sheet up might be the outlines of every lake in the country. The next sheet might be the road network. The next sheet might be the boundary of Åre kommun, drawn as a single closed polygon. The next sheet might be a list of dots, each one marking a borehole drilled by a mining company between nineteen seventy and last Tuesday.

[calm] All of these are layers. QGIS is the thing that holds the stack, lets you turn each layer on and off, lets you change the colours and the line widths and the symbols, and lets you ask questions across the stack. Which lakes lie inside Åre kommun. Which boreholes are within five hundred metres of a road. Which slope angles are above thirty degrees. The map you see on the screen is just the visual surface of that stack of data.

Now, the data in those layers comes in two big flavours. The first flavour is called vector data. Vector data is points, lines, and polygons, with a table of attributes attached to each one. A dot on a map with the words drilled in two thousand and nine attached to it is vector. The outline of a lake with its name and surface area attached is vector. Roads, kommun boundaries, mining claims, anything that has a definite shape, is vector.

The second flavour is called raster data. Raster data is a grid of pixels, like a photograph, except every pixel carries a number that means something specific. A satellite image is raster, where each pixel carries colour values. A digital elevation model is raster, where each pixel carries the height of the ground at that spot. A land cover map is raster, where each pixel carries a code that says forest or open water or pasture or built-up area.

QGIS handles both. Most of the work you do in QGIS is some combination of pulling layers in, deciding how each one should look, stacking them in the right order, and exporting the result either as a printable map or as a new dataset.

Coordinates, Projections, and a Slightly Squashed Sweden

There is one more concept that every QGIS user eventually has to make peace with, and it is the concept of the coordinate reference system. The Earth is round. Maps are flat. There is no honest way to flatten a sphere onto a rectangle without distorting something, and every map projection is a different choice about which thing to distort.

The data in your Åre kommun map was collected and stored in a particular coordinate reference system. In Sweden, the most common one is called SWEREF ninety-nine, in any of several variations. SWEREF ninety-nine is a system designed specifically to keep Sweden looking the right shape and the right size, at the cost of being slightly wrong about every other country. If you load a layer that was made in a different coordinate reference system, say one of the global systems used for satellite data, QGIS has to be told how to reconcile the two. Usually it does this for you on the fly. Sometimes it gets confused, two layers do not line up, and you spend an afternoon figuring out why. Welcome to GIS.

The thing on the bottom of the QGIS window that says something like EPSG colon three thousand and six, is telling you which coordinate system the current map is using. EPSG is a registry, originally maintained by the European Petroleum Survey Group, that gives every coordinate system in the world a numerical code. SWEREF ninety-nine TM is EPSG three thousand and six. Standard latitude and longitude in the WGS eighty-four datum is EPSG four thousand three hundred and twenty-six. Once you start recognising those numbers, the bottom of the QGIS window stops being mysterious.

Lantmäteriet, Boreholes, and a Pile of Layers

This is where the project you have been working on comes in. The gruvor centerfold, which is a printable spread for the paper showing mines and mineral activity across a region, is being built with three Python scripts that together produce a single QGIS project file.

The first script pulls and clips Topografi fifty, which is the standard Swedish topographic dataset, and exports the part you care about as a GeoParquet file. GeoParquet is a relatively new format that takes the popular Parquet columnar data format and adds spatial awareness. It is one of those quietly important formats that the GIS world has settled on in the last few years.

The second script reaches into Lantmäteriet's open data through a system called STAC, which stands for SpatioTemporal Asset Catalog. STAC is a standard way for an organisation to publish its geographic datasets so that programs can find them, query them, and download just the pieces they need. Through STAC, your script pulls a fifty metre digital elevation model, a derived hillshade layer, the kommun boundary geometry, and the marktäcke land cover layer for your area of interest.

The third script is the one that actually opens QGIS in the background, loads about twenty layers into a single project, applies a coherent set of colours and symbols and labels, and writes out a print layout sized for A3 landscape. The print layout is the centerfold. That is the file that, when you open it on your laptop, looks like a real map made by a real cartographer.

Some of the decisions in that build script are pure QGIS in action. The boreholes were filtered from three thousand nine hundred and fifty-three points down to one hundred and twenty-seven, by keeping only the ones drilled in two thousand and ten or later. That is a vector filter expression, written once, that QGIS applies to the layer at draw time. The mines were re-symbolised as triangles, because triangles read more clearly at print scale than circles when the dots are dense. The place name labels were limited to the largest size class, because at A3 every name fights for room with every other name and you have to pick which ones win.

A Print Layout the Size of a Pizza Box

The print layout system in QGIS is the part of the program that turns the live screen view into a real, printable, exportable page. It has its own window. It has a canvas the size of whatever paper you choose. It lets you place map frames, scale bars, north arrows, legends, titles, and explanatory text wherever you want them.

[happy] An A3 landscape page is roughly the size of two large pizza boxes laid side by side. For a newspaper centerfold, this matters, because the centerfold is the two centre pages of the paper opened flat. It is the largest single canvas the paper offers. People stop at it. People keep it. A good centerfold map in Årebladet is, in some ways, the most permanent thing the paper produces in any given issue.

The fact that all of this, the data pulls, the styling, the layout, can be scripted in Python and run as a command line job is one of QGIS's quietly powerful features. QGIS ships with its own Python binding, called PyQGIS, which exposes basically every menu in the program as a function you can call from code. You can open a project file, change a layer style, swap a data source, and re-export the print layout, all without ever clicking anything. For a paper that needs the same centerfold built fresh every quarter from slightly different data, this is the difference between an afternoon of human work and a thirty second job.

Cathedrals, Bazaars, and a Lot of Moose

There is a famous essay in open source software called The Cathedral and the Bazaar, written by Eric Raymond in nineteen ninety-seven. The cathedral, in Raymond's metaphor, is software built by a small priesthood of architects who plan everything carefully and release polished versions to the public. The bazaar is software built out in the open, by anyone who shows up, with rough edges, loud arguments, and the occasional brilliant patch from a stranger.

Esri builds cathedrals. They are beautiful cathedrals. ArcGIS is a deeply considered, very integrated product, with twenty-five years of consistent engineering behind it. If you work in a national mapping agency or a big city planning department, ArcGIS will probably still be on your desk. The cathedral has its place.

QGIS is a bazaar. There is no single architect. There is a board, a project steering committee, a release schedule, and a process for proposing changes, but the actual work is done by a shifting cast of contributors from different countries working for different clients with different priorities. Some of those contributors are paid by consultancies who make money supporting QGIS deployments. Some are paid by governments who use QGIS internally and contribute features back upstream. Some are volunteers. Some are graduate students. Some, like Gary in the beginning, are people who just want their own particular problem to go away.

The bazaar model has known weaknesses. Features can arrive in odd orders. Two contributors can build similar things without coordinating. Documentation lags behind code. But the bazaar model also has a property that no cathedral can match. It cannot be turned off. There is no single company that can be acquired, no single licensing decision that can be reversed, no single bankruptcy that can erase the work. QGIS belongs to its community, and its community is large enough that it cannot be uninvented.

What He Built

Today QGIS is used by national mapping agencies, by humanitarian organisations responding to disasters, by ecologists tracking species, by archaeologists mapping dig sites, by city planners, by farmers, by journalists. It is taught in universities on every continent. It has been translated into more than forty languages. The download numbers run into the millions.

None of this would have happened if Gary Sherman had decided, in February of two thousand and two, that it was easier to dual-boot into Windows and pay for ArcGIS. None of it would have happened if he had decided, in July of that year, that his little viewer was too embarrassing to release publicly. The whole thing turned on the willingness of one mining engineer in Alaska to put a half-finished tool on the internet and let other people poke at it.

That is how almost every important piece of open infrastructure starts. Someone has a small private problem. They solve it badly. They share the bad solution. Other people improve it. Twenty years later, the bad solution has become a foundation that a tourism kommun in northern Sweden can build a newspaper centerfold on top of.

Your Living Room, Your Alaska

Here is the connection that is worth holding on to. When you open QGIS on your laptop and load the map of Åre kommun, you are sitting at the end of a chain that starts in a living room in Alaska in winter. The same toolkit. The same layer model. The same coordinate system mechanics. The same plugin architecture. The same project file format. All of it routed through every contributor between Gary's first commit and the current release.

[calm] You do not need to understand all of QGIS to use it well. Almost nobody understands all of QGIS. What you need is a working mental model. A QGIS project is a stack of layers, each in some coordinate reference system, each either vector or raster, each with its own symbology. You add layers from files, from databases, from web services, or from scripted pulls like the Lantmäteriet STAC fetch. You style them. You arrange them. When you want a printable result, you open the print layout and compose the page. When you want to repeat the whole thing automatically, you write a Python script that does what you would have done by hand.

The map of Åre kommun on your screen is, in QGIS terms, just one example of that pattern. Once you can name the parts, the program stops being mysterious. The mining engineer made a tool, the tool grew up, and now it knows the shape of your home valley.

That is the whole story.