What the Map Must Not Say: Density, Honesty, and the Layers You Do Not Draw

The Layer That is True and Wrong

In the Norwegian mining group there is a layer called exploration reports. Two thousand four hundred and thirty-two points. Historical survey records, every place someone has ever filed a report about having looked at the ground. The sources file describes it in four words: "where someone has looked."

It is real data. It is accurate data. Every one of those two thousand four hundred and thirty-two points corresponds to a genuine record. And it is off by default, and that is the correct decision, and understanding why is the whole subject of this episode.

Because here is what happens if you turn it on. Two thousand four hundred points scattered across the Norwegian side of a centerfold do not read as two thousand four hundred individual facts. They read as a grey haze. A texture. The reader's eye cannot resolve them as points, so it resolves them as a fog, and the fog sits on top of the actual story, and now the map says "Norway is a blur" when what you wanted it to say was something specific about mines and permits and the cross-border picture.

The layer is true. Drawing it, on this map, at this scale, would make the map worse and the reader less informed. True and wrong at the same time. A map is not obligated to draw everything it has, and the layers you decide not to draw are as much a part of the map as the layers you do.

Density Has a Breaking Point

Every point layer has a density at which it stops being a set of points and becomes a field. Below that threshold, the reader sees individuals, and can count them, locate them, reason about them. Above it, the reader sees a smear, and the individual data is gone, replaced by a vague impression of "a lot."

The gruvor stack has layers on both sides of that line, and the counts tell you which is which.

Fifteen historical mines. Thirty-two unmined ore bodies. Three Norwegian mines. These are small numbers. Each point is an individual the reader can actually hold. These layers can and should be drawn as distinct symbols, because at that density, the symbol is doing what a symbol is for.

Then the boreholes, one hundred and twenty-seven of them. That is near the edge but still workable, especially across a wide print frame. A hundred and twenty-seven points can still read as points if they are styled with a bit of care.

Then the Norwegian prospecting rights, one hundred and sixty polygons. The Norwegian quarries, eighty-three points. Getting denser. And then the exploration reports, two thousand four hundred and thirty-two. Well over the line. That one is a field, not a set of points, and no amount of clever symbology changes the arithmetic.

The first question for any dense layer is not "how do I style it." It is "which side of the breaking point is this on." If it is over the line, styling it as individual points is not an option that exists. You have three real choices, and we will come to them, but drawing it raw is not one of them.

Nearly Three Hundred Thousand Polygons

The quaternary soils layer, the SGU jordarter, is the extreme case. Nearly three hundred thousand polygons. Close to six hundred megabytes. The sources file is blunt about it: heavy, drop it if a given project does not need surface geology.

That is a density problem and a weight problem at once, and the two are related. A layer that heavy slows every pan, every zoom, every redraw, every export. It makes the project sluggish to work in, which means you iterate slower, which means the map gets less attention per hour, which means it ends up worse. The weight is not just a technical annoyance. It is a tax on your own iteration speed, and iteration speed is most of what makes a map good.

And visually, nearly three hundred thousand polygons of surface geology across a centerfold is, again, a field. At print scale the individual soil polygons are mostly smaller than the resolution of the eye. The reader does not see three hundred thousand soil units. They see a busy coloured texture.

So jordarter follows the bedrock logic from the last episode, only more so. It is not a basemap layer. It is not a quiet editorial layer. It is an entire alternative map, and a heavy one, and it belongs in exactly one place: a specific figure where surface geology is genuinely the question, loaded only for that figure, dropped from every project that does not need it. The sources file already says this. The discipline is to actually do it, and not let it creep into the main project "just in case," because just in case costs you six hundred megabytes of iteration tax for a layer you are not even showing.

Technically Present, Journalistically False

Now the part that matters most, because it is where visualization stops being a design problem and becomes a journalism problem.

The sources file describes a decision made at data prep time. The Norwegian uttak dataset, the extraction sites, was split into two layers: mines, and quarries. Three mines. Eighty-three quarries. And the file says, plainly, why. Without that split, eighty-six Norwegian mines would be technically present in the data and journalistically false.

Sit with that sentence, because it is the whole job. Eighty-six Norwegian mines. You could put that number on the map. The data would support it. Every one of the eighty-six is a real extraction site in the real dataset. And it would be a lie, because eighty-three of them are gravel pits and stone quarries, and three of them are metal and industrial mineral mines, and a centerfold about mining that shows eighty-six mines has told the reader something false using nothing but true data.

This is the thing about a data-rich map. The data does not arrive pre-sorted into what is editorially true. The dataset has its own categories, drawn up by whoever built the dataset for whatever purpose they had, and those categories are almost never the categories your story needs. The uttak dataset lumps mines and quarries together because to the Norwegian directorate that built it, they are both extraction sites and that distinction did not matter for their purpose. For a newspaper story about mining, that distinction is the story.

So the split is not a data-cleaning step. It is an editorial act. Someone decided, deliberately, that the centerfold will say three, not eighty-six, and that decision is a journalistic judgement that a person has to own. The data cannot make it for you. The software cannot make it for you. A solo journalist, with no media lawyer down the hall, owns every one of these, and the only protection is to make them consciously and be able to explain each one.

The Splits as Editorial Acts

The uttak split is not the only one. The sources file describes a second: the Norwegian mineral rights dataset, bergrettigheter, split into extraction rights and exploration rights. The first means someone has the right to take ore out. The second means someone has the right to go look. Eleven of the first. One hundred and sixty of the second.

Same logic. The source dataset has them in one bucket because to the agency that maintains it, they are both kinds of mineral right. To the story, the difference between "has permission to mine" and "has permission to prospect" is enormous. Conflating them would, again, produce a true number that tells a false story.

And the sources file draws out the editorial line that these splits together create. The Norwegian prospecting rights, one hundred and sixty of them, against the Swedish side, are the basis of a specific claim the map can make: that Norway is prospecting harder than Sweden. That is a real editorial finding, and it is only visible because the rights were split. If extraction and exploration were one undifferentiated layer, the finding would be buried inside it.

So the splits do two things at once. They prevent false statements, and they enable true ones. The same act of dividing a dataset along the line your story cares about is what stops the map lying and what lets the map have something to say. That is not a coincidence. It is what editorial data work is.

The Unlabelled Layer

There is a quieter example of the same principle, and it is one decision in one layer that the sources file flags with a single word in parentheses: intentional.

The unmined ore bodies. Thirty-two points, hollow triangles, and the file notes they are unlabelled, intentionally. Compare that to the historical mines, fifteen points, filled triangles, labelled.

Why label one and not the other. Think about what a label claims. A label says "this thing is individually significant enough that you, the reader, should know its name." For the fifteen historical mines, that is true. A mine that operated is a specific place with a history, and naming it invites the reader to recognise it, to connect it to things they know.

For the thirty-two unmined ore bodies, naming each one would be a different claim. It would suggest thirty-two individually notable named places, when the editorial point of that layer is collective, not individual. The point of the unmined ore bodies is the pattern: here is the ore that is known and sitting in the ground unextracted, here is roughly how much of it there is and where. The reader is meant to take in the constellation, not study thirty-two names. Labelling them would pull the eye into reading mode, name by name, and break the collective read.

So the layer is deliberately unlabelled. That is restraint, and restraint is a real cartographic tool. Not labelling is a decision with the same weight as labelling. The hollow triangle says "known but not extracted." The absence of a label says "read these as a group." Both are design carrying meaning.

The Boreholes Decision

The boreholes layer is the cleanest single example in the whole stack of a density decision and an editorial decision being the same decision.

The raw borehole data has three thousand nine hundred and fifty-three points. The layer in the project has one hundred and twenty-seven. The filter, written right into the sources file, is a subset query: drill year, two thousand and ten or later. Modern boreholes only.

Look at what that filter is doing on both axes at once. On the density axis, three thousand nine hundred and fifty-three boreholes is well over the breaking point. It would be a fog. One hundred and twenty-seven is under the line, drawable as real points. So the filter is a density fix.

But it is also an editorial statement. "Modern boreholes only, two thousand and ten onward" is a claim about what is relevant. It says the story is about recent drilling activity, current interest in the ground, not the full historical record of every hole ever sunk. A different story would pick a different filter. A story about the long history of exploration would want them all, and would then have to solve the density a different way, with a heatmap or with aggregation. The choice of filter is the choice of what the map is about.

That is why the filter belongs in the sources file, written down, visible. It is not a technical detail. It is one of the load-bearing editorial decisions of the centerfold, and it is the kind of thing that, six months later, you need to be able to see and re-justify. A filter that changes the count from nearly four thousand to one hundred and twenty-seven is not cleanup. It is the map deciding what it means.

Density Tools: Scale Visibility, Aggregation, Heatmap

For the layers that genuinely are over the breaking point and that you still want to represent somehow, there are three honest tools.

The first is scale-dependent visibility. A dense layer can be set to simply not draw at the centerfold's print scale, and only appear when someone zooms in far enough that the density resolves back into individual points. The exploration reports could be invisible on the printed centerfold and present in an interactive version where the reader can zoom. The layer still exists, it is just shown only at the scale where it can be read honestly.

The second is aggregation. Instead of two thousand four hundred individual points, you compute counts per kommun, or per grid cell, and you draw the aggregate. One number per area. This turns "a fog of points" into "a clear statement about density per region," which is usually what the dense layer was actually evidence for anyway. Nobody needs to see all two thousand four hundred exploration reports. They need to know that the western kommuns have had ten times the survey activity of the eastern ones, and an aggregation says exactly that, cleanly.

The third is the heatmap. A density surface, smooth, where the dense layer becomes a wash of intensity rather than a scatter of marks. This keeps the spatial pattern, the where-is-it-concentrated, while explicitly abandoning the individual points. It is honest precisely because it looks like what it is, a density estimate, and does not pretend the individual points are legible.

All three share one quality. They do not pretend. Drawing two thousand four hundred points at print scale pretends the reader can read two thousand four hundred points. Scale visibility, aggregation, and heatmap each admit, in their form, that this layer is a field and treat it as one. That honesty of form is the point.

What Off By Default Protects

Now the thematic group's "off by default" reads as more than a performance setting. It is a structural protection.

Every layer that is off by default is a layer that cannot accidentally end up in the printed centerfold. It cannot creep in. To put it on the map, someone has to make a deliberate choice to turn it on, for a specific figure, with a specific reason. The default state of the project is the honest state: the editorial story, on a quiet ground, with the fields and the alternative maps held back.

This matters for a solo operation more than it would for a newsroom with layers of editors. There is no second person who will catch it if the bedrock or the soils or the exploration reports drift onto the main map. The project's own defaults have to be the editor. Off by default means the project, left alone, tells the truth, and telling a richer or different story requires a conscious act each time. That is a good architecture for someone who is going to come back to this project tired, at the end of a long day, and needs it to be safe to work in.

The Map's Argument

Pull it together and the shape of the whole thing is clear.

A centerfold is not a container for everything true. It is an argument. And like any argument, it is defined as much by what it leaves out as by what it includes. The forty layers in the gruvor project are the available evidence. The map is the case built from a chosen subset of that evidence, arranged so the reader can follow it.

The exploration reports are left off because two thousand four hundred points is a fog. The soils are left off because three hundred thousand polygons is a different map and a heavy one. The uttak set is split because eighty-six mines would be a true-sounding lie. The rights are split because the prospecting-versus-extraction line is where the story lives. The boreholes are filtered because the story is about recent activity. The ore bodies are unlabelled because they are meant to be read as a pattern.

Every one of those is a decision a person made and has to own. Not the dataset, not QGIS, not the build script. A person, deciding what this map says and what it does not.

The good news is that the sources file is already, quietly, a record of those decisions. The splits are documented. The filter is written down. The intentional unlabelling is flagged. That file is not just an inventory. It is the editorial reasoning of the centerfold, kept where it can be re-read and defended. Keep writing it that way. The map is the argument the reader sees. The sources file is the argument you can still explain six months later, and for a solo journalist, being able to explain every choice is the whole of the protection you get.

That is the trilogy. The stack is the composition. The colour budget is what keeps it readable. And the honesty layer, the deciding of what not to draw, is what keeps it true. Three things to hold in your head tomorrow in Mörkret, while the build script runs and the centerfold comes back together.