How Maps Are Made

The first thing to give up about maps is the idea that a good one is an accurate one. A sheet of paper the size of a table cannot hold a county, so something has to go, and the cartographer decides what. Every hedge, every gatepost and every puddle exists on the ground and almost none of them exist on the map. What survives the cut is whatever serves the reader the map was drawn for, which is why a walking map and an aviation chart of the same valley look nothing alike and are both correct.

This selection is not a flaw to be engineered away with better instruments. A perfectly surveyed sheet that showed everything would be unreadable, and a reader who needs to find a footbridge in failing light does not want a complete record, they want the footbridge to be obvious. Good cartography is therefore closer to editing than to photography. The craft is deciding what to delete, what to exaggerate, and what to leave exactly where it is.

Once that is accepted, the rest of the subject makes sense. The famous distortions of world maps stop being embarrassing errors and become declared trade-offs. The symbols stop being quaint and become a compression scheme. Even the blank areas start to carry meaning, because a blank on a map drawn by a careful hand usually means nobody surveyed it rather than that nothing is there.

Measuring a long distance across country directly is miserable and inaccurate. Chains sag, slopes lie, and errors accumulate step after step. The solution, worked out over the eighteenth and nineteenth centuries, was to measure one short line very carefully and then never measure a long one again. That short line is a baseline, and from its two ends a surveyor measures the angles to a distant point. With one length and two angles the triangle is solved, which yields two new lengths without anyone walking them.

Each new line becomes the baseline for the next triangle, and the network propagates outward across a country as a lattice of carefully observed angles. The points sit on hilltops, church towers and purpose-built pillars so that each is visible from its neighbours. Because angles can be measured to a few seconds of arc with a good theodolite, and because every triangle can be checked by the rule that its angles sum correctly, errors are caught locally instead of accumulating silently.

The surviving pillars of these networks are still standing on hills across Europe and beyond, usually ignored by the walkers who pass them. They are the physical skeleton on which the detailed mapping of entire nations was hung, and until satellite positioning arrived they were the most precisely located objects most countries possessed.

A map has to be a map of something, and the something is harder to pin down than it sounds. The planet is not a sphere: rotation makes it wider at the equator than pole to pole by about twenty-one kilometres. It is not a tidy flattened sphere either, because mountains, ocean trenches and variations in rock density pull the local level surface up and down. The surface that gravity actually defines, the one a still ocean would follow, is called the geoid, and it is lumpy in a way no simple formula describes.

Because the geoid is awkward, surveyors fit a smooth mathematical ellipsoid to it and do their arithmetic on that. The choice of ellipsoid, together with how it is positioned relative to the planet, is a datum. Different countries historically fitted different ellipsoids to their own patch of ground, because a shape that fits Europe closely does not fit Australia closely, and each wanted local accuracy more than global consistency.

This is why coordinates can disagree. The same physical corner of a field can be given two sets of numbers that differ by a hundred metres or more, with neither being wrong, because they are quoted on different datums. Modern satellite positioning uses a global datum, which is why old paper maps and new handheld readings sometimes need a stated conversion before they can be compared at all.

A sphere cannot be unrolled onto a plane. Try it with an orange peel and the pieces either tear or overlap, and no amount of care avoids both. This is a result in geometry rather than a limitation of technique, which means every flat map distorts something: area, shape, distance, direction, or some combination. A projection is a rule for doing the flattening, and choosing one is choosing which property to protect.

The best known rule protects direction. On it, a line of constant compass bearing draws as a straight line, which is exactly what a navigator crossing an ocean needs and the reason the projection dominated sea charts for centuries. The cost is area: landmasses far from the equator inflate enormously, so high-latitude countries appear far larger than tropical ones of the same true size. The projection is not lying, it is keeping the promise it made and breaking a different one.

Other rules make other bargains. Equal-area projections keep relative sizes honest and let shapes shear instead. Conformal projections keep local shapes and angles right while sizes drift. Compromise projections spread small errors everywhere rather than concentrating a large one anywhere, which is usually what a general reference map of the world wants. There is no neutral choice, only a declared one.

Scale is the ratio between a length on the map and the same length on the ground. At one to twenty-five thousand, a centimetre on paper is two hundred and fifty metres of country, and a line drawn a fifth of a millimetre wide is already five metres across on the ground. That single fact drives most of what a cartographer does, because a road drawn to true width at that scale would be invisible, so it is drawn far too wide on purpose and everything near it must shuffle aside to make room.

The process of simplifying as scale shrinks is called generalisation, and it has recognisable moves. Lines are smoothed, so a meandering stream loses its smallest bends. Features are merged, so a cluster of farm buildings becomes one block. Features are displaced, so a road, a railway and a river squeezed into the same gorge are drawn side by side rather than on top of one another. Features are selected away entirely, so the smallest ponds simply stop existing.

Done badly this produces a map that misleads, with a footpath apparently running through a building. Done well it produces a map that is geometrically looser than reality and far more useful, because the reader can see at a glance a relationship that the true geometry would have buried. The skill is keeping the topology honest, so that what connects to what, and what lies on which side of what, survives the simplification intact.

Height is the hardest thing to put on a flat map, because it is the one dimension the sheet does not have. The rigorous answer is the contour: a line joining points of equal height, drawn at a fixed vertical interval. Contours are quantitative, so a reader can measure a slope from them, and they are unambiguous, because a closed loop with higher loops inside it is a hill and the same pattern with lower loops inside it is a hollow. They are also, for many readers, genuinely hard to see as shape.

The alternative is to imitate light. Hill shading assumes illumination from one side, conventionally the upper left, and darkens the slopes facing away from it. The result reads instantly as terrain even to someone who has never been taught to read a map, which is why it appears on almost every modern topographic sheet. Its weakness is that it is qualitative: nobody can measure a gradient from a grey tone, and a shading scheme can flatten or exaggerate a landscape depending on the chosen light.

The usual resolution is to print both, with shading underneath for immediate comprehension and contours over it for measurement, plus scattered spot heights giving exact values at summits and junctions. Each method covers the other method weakness. A reader who wants the gist takes it from the shading, and a reader who needs a gradient counts the lines.

Latitude and longitude name any point on the planet with two angles, and they are the right system for global work. They are awkward for local work, because a degree of longitude shrinks towards the poles, so the same numerical change means a different distance depending on where you are. Anyone computing an area or a distance on the ground would rather work in metres on a flat sheet than in angles on a curved surface.

Hence the projected grid. A country picks a projection, lays a square grid of metres over it, and quotes positions as two distances from an origin. Because the grid is flat and metric, distances and areas can be computed with ordinary arithmetic, which is why national mapping, land registration and engineering all use grids rather than angles. The cost is that the grid only behaves well over a limited belt, so a country or a zone gets its own.

Two details catch people out. The first is that grid north, true north and magnetic north are three different directions, differing by an angle that varies with place and, for magnetic north, with time. The second is the convention for quoting a reference: the distance east is always given before the distance north. Reverse them and you will confidently name a point somewhere else entirely, which is a mistake with a long history in search and rescue.

A symbol lets a map say something that would take a paragraph. A small circle with a line above it means a particular kind of structure; a dashed line of one pattern means a right of way while a dashed line of another means a parish boundary. None of this is guessable, which is why every serious sheet carries a legend. The legend is not decoration, it is the declaration of a private language, and a map read without consulting it is being half guessed.

Good symbol design follows a few quiet rules. Symbols for related things look related, so all water features share a colour and all boundaries share a line grammar. Symbols are distinguishable at the size they will actually print, which rules out fine detail that looks elegant on a designer screen and turns to mud on paper. And symbols are stable across sheets and across editions, because a reader who has learned the vocabulary once should not have to learn it again.

The deepest constraint is that the set must stay small. Every new symbol is a demand on the reader memory, and a legend that runs to several hundred entries has defeated its own purpose. So cartographers reuse, group and generalise relentlessly, and accept that two slightly different things on the ground will share one mark on the sheet.

A map is a dense picture that a reader scans rather than studies, so visual hierarchy does most of the work. The features that matter most must be the most visible, which is achieved with contrast rather than size: a dark line on a pale ground reads first, and a pale line on a busy ground reads last. Beginners usually make everything prominent, which is the same as making nothing prominent, and the sheet turns into noise.

Type has its own craft. A label must sit close enough to its feature that ownership is obvious and far enough that it does not bury it. Names of linear features curve along them, names of areas spread across them, and names of points sit beside them in a preferred order of positions. Where labels collide, something must move, shrink or go, and automated placement systems spend most of their effort on exactly this problem.

Colour carries meaning that has to survive reproduction and the reader eye. Conventions help, because blue reading as water and green as vegetation are learned almost everywhere, but they are not universal and they are not free. A significant fraction of readers cannot separate red from green reliably, so a well made sheet never carries critical information in that distinction alone, and keeps a difference in lightness or pattern as well.

For most of the history of mapping, detail was drawn in the field by someone standing in it with a plane table, sighting features and sketching them in place. That method produced superb maps slowly. Photography from aircraft changed the economics completely, because a single flight captures a strip of country in minutes and the detail can be plotted afterwards indoors, in comfort, by people who never visited.

The technique that makes this rigorous is photogrammetry. Two photographs of the same ground taken from different positions give a stereo pair, and the apparent shift of a feature between them yields its height, exactly as two eyes give depth. With known camera positions and a few surveyed control points on the ground, an entire landscape can be measured in three dimensions from the images. Contours stopped being walked and started being plotted.

Laser scanning went further. An airborne scanner times the return of its own pulses and produces a dense cloud of measured points, often several per square metre, including returns from the ground beneath a tree canopy that no camera can see through. That has quietly transformed archaeology and flood modelling, where a bare-earth surface matters more than a photograph. What remains for the field surveyor is the ground control that ties it all to a datum, and the checking of everything the instruments cannot judge.

The ground does not hold still. Roads are built, buildings come down, rivers shift their channels, coastlines erode and administrative boundaries move for reasons that have nothing to do with geography. A sheet is therefore correct as of a date, and the date matters as much as the content. Serious maps print it, and a reader who ignores it can be looking at a junction that was replaced years ago.

Revision is organised by change rather than by calendar. Cheap signals, such as building permissions, aerial imagery comparison and reports from users, flag where something has likely changed, and the expensive field visit is spent only there. This is why a busy urban fringe is resurveyed far more often than a stable upland, and why two parts of the same sheet can legitimately carry different revision dates.

Digital publication changed the rhythm but not the problem. A continuously updated database can absorb a change the week it is reported, which is a real improvement, but it also removes the fixed edition that made two readers certain they were discussing the same document. Careful practice replaces the edition with a timestamp, so that a decision made from a map can be traced to the state of the data when it was made.

Someone who reads a map well does a few unglamorous things consistently. They orient the sheet to the ground before trying to interpret it, so that what is ahead of them is ahead on the paper. They check the scale and the contour interval before estimating any distance or climb, because both vary between series and an assumption carried over from a different sheet is a reliable source of error.

They also read the shape rather than the labels. A valley, a spur and a saddle each have a contour signature, and recognising those patterns means a feature can be identified without any name attached to it. This is what allows a reader to work out where they are from the landscape alone, which is the difference between navigating and following. It takes deliberate practice and very little equipment.

Finally they treat the map as evidence rather than truth. They expect generalisation near crowded detail, they expect the oldest-looking corner of the sheet to be the least current, and they believe the ground over the paper when the two disagree. A map is the best available summary of somebody else fieldwork, which is a great deal, and it is not a substitute for looking up.