Cartographic Projection
every map is wrong, and by exactly how much
Concept

Bias — where it appears

A systematic error whose expectation is not zero, so that averaging over more measurements does not reduce it. It is the distinction that decides whether an aggregate is safer than one of its terms: a noise divides by the square root of the count and a bias divides by nothing.

Named by 15 essays across 4 fields — each of them below, with the objects they name alongside it.

10,000 observations of one place, averaged 120 times. Left: 400 single observations of the same place, with a 60-kilometre standard deviation east and north, drawn across ±400 km. They are scattered about the truth and their average is unbiased on the ground. Right: 120 independent averages of 10,000 such observations each, computed on the page and taken back to the ground, drawn across ±1.8 km. The cloud is tight, as averaging ten thousand things should make it, and it is not centred on the cross: it sits 489 m away, against 403 m predicted by the projection's second derivative alone.

The average of noisy positions moves

Average sixty thousand scattered observations of one place on a Mercator map and the answer is 404 metres too far north — at every sample size, because it is a bias and not noise. The same average on the Lambert cylindrical equal-area is 404 metres too far south, the two being ½ (σ²/R) tan φ and its exact negative, and on the plate carrée it is not displaced at all.

distortion · Precision Twenty-four versions of one shape, and not one of them gains area. The same closed boundary rotated twenty-four times and simplified at the same tolerance. If the area error were noise the values would straddle zero and their mean would fall towards it; they do not. Every one is negative, the mean is -0.4644 per cent, and the mean is 71 standard errors from zero. A bias of that size cannot be removed by averaging over more boundaries, which is the only defence anybody has against a rounding error.

A thousand features are wrong in the same direction

The area a simplification costs is unpredictable in sign for one feature. Over a population it is not: twenty-four presentations of one shape all lose area, the mean is seventy standard errors below zero, and no amount of aggregation removes it.

applied · Generalise Three selection rules, and what each one keeps. Keeping one feature in ten from a stated population whose size distribution has a Pareto exponent of a half — the exponent Töpfer's law is a theorem about. Keeping the largest carries 99.99 per cent of the total size and inflates the median feature by a factor of 95. A random sample keeps the median to 1.068 and carries 5.0 per cent of the total. The two rules are right about different things and there is no rule that is right about both, because the total lives in the tail and the median does not.

Which features survive is not a sample

How many features a scale can carry is answered and treats the population as a number. Which ones survive is a different question: keeping one feature in ten carries 99.99 per cent of the total length and inflates the median feature by a factor of 95, and the shape of the size distribution survives both exactly.

applied · Generalise A season of crossings pins the stretch to a tenth only if it learns from where the streak was. The share of 400 seasons at each length whose learned stretch lies within a tenth of 1.25. From the floors' rank histogram: 25 crossings 19 per cent, 50 crossings 27 per cent, 100 crossings 32 per cent, 200 crossings 43 per cent, 400 crossings 55 per cent, 800 crossings 67 per cent. From where the streak turned out to be: 25 crossings 56, 50 crossings 72, 100 crossings 85, 200 crossings 98, 400 crossings 99, 800 crossings 100. At a hundred crossings, 85 against 32 per cent.

A season learns the stretch from where the streak was, not from the floors

An ensemble whose members spread four fifths as widely as the forecast's real error covers the diversion floor in 94 crossings of a hundred instead of 95, and the repair is to stretch every member away from the centre by a factor learned from past crossings. Learned the way the rank histogram teaches it — from where the truth's floor fell among the members' floors — a season of a hundred crossings lands within a tenth of the right stretch one time in three, and a season of eight hundred two times in three. Learned from where the streak turned out to be, which the analysis after each crossing supplies whether the streak reached the gap or not, a hundred crossings land within a tenth 85 times in a hundred and two hundred almost always. The floor is the quantity the plan is about; it is the wrong quantity to learn the spread from.

paths · Paths Averaging divides the hand's wander and leaves its habit: a shared part sets a floor no stack goes below. The drift left in the mean of N tracings, as a share of one tracing's, when a stated part of the hand's variance is a habit shared by every tracing: none (thick), a quarter, and half. With none it falls as one over the square root of N, to 25 per cent at sixteen; a quarter shared stops it at 50 per cent, half at 71. Dotted: the drift at which an independent hand's stacks read the strong seam's easing (two tracings, 71 per cent) and the weak seam's (sixteen, 25).

Averaging tracings reads the easing, until the hand repeats itself

One tracing by a drifting hand cannot show how far a copy was eased, because an easing as long as the drift is a stretch of drift. Average several tracings and the easing stays while the drift shrinks: two tracings read the easing's width at the seam where the copy departs strongly, sixteen read it at the seam where it barely departs, and as the drift falls the rigid count starts buying pieces for the easing again while the eased count stays at two. All of that assumes each tracing drifts afresh. A hand that repeats part of its drift in every tracing leaves that part in the mean, where it sets a floor no number of tracings goes below; and a digitiser who learns the hand from the spread between tracings measures only the part that averages away, and charges the mean too little, so that both counts buy every piece they are offered.

wrong · Identify After a revision narrows the ensemble, a learner that never forgets is still well short of the new stretch three hundred crossings later. The stretch each learner holds before each crossing, the median over 400 winters, from 100 crossings before a forecast revision to 300 after it. Before it the members spread 0.8 of the truth's spread and the right stretch is 1.25; after it they spread 0.6 and the right stretch is 1.667 (dashed). never forgets: 1.28 at 25 crossings, 1.34 at 100, 1.44 at 300; memory 200: 1.31 at 25 crossings, 1.44 at 100, 1.59 at 300; memory 50: 1.42 at 25 crossings, 1.62 at 100, 1.68 at 300; restarts on the date: 1.63 at 25 crossings, 1.67 at 100, 1.67 at 300. Shaded: the middle eight winters in ten for the learner with a memory of 50 crossings, 1.55–1.78 at the end.

A revised forecast is relearned in as many crossings as the learner remembers

An ensemble's stretch is a property of the forecast system, and forecast systems are revised. When a revision narrows the members from four fifths of the real spread to three fifths, a learner that weights every crossing since the start of the winter equally still holds 1.44 for a new stretch of 1.67 three hundred crossings later, and leaves the plan under the floor in 93.8 crossings of a hundred instead of 95. A learner that forgets regains ninety-five per cent coverage after about as many crossings as its memory — 31, 57, 113 and 206 for memories of 25, 50, 100 and 200 — and pays for forgetting only when nothing has changed. A memory of fifty costs nothing measurable over a winter. And a bias the whole ensemble shares, which the stretch alone reads as extra spread, is found to a kilometre by a learner that tracks the mean departure as well.

paths · Paths Sixteen tracings by four hands leave half the drift that sixteen by one hand do, unless the hands share a habit too. The drift left in the mean of sixteen tracings of one sheet, as a share of one tracing's, measured over 120 sheets, for sixteen tracings by one hand, eight each by two, four each by four, two each by eight and one each by sixteen. When half of each hand's drift is its own habit (darker): 1 × 16 74, 2 × 8 54, 4 × 4 40, 8 × 2 31, 16 × 1 25 per cent. When a quarter is the hand's habit and a quarter a habit every hand in the office shares (paler): 1 × 16 73, 2 × 8 63, 4 × 4 58, 8 × 2 56, 16 × 1 54. Dotted: 71 and 25 per cent, the drift at which an independent hand's means read the strong seam's easing and the weak seam's.

The spread between hands charges the mean, and an office's habit hides in the sheet

One hand's tracings cannot see the part of its drift it repeats, so a mean of sixteen by one hand is charged for a third of the error it carries. Sixteen tracings by four hands carry half the error of sixteen by one, and the spread between the hands' means, divided by the number of hands, charges that mean correctly to within a fifth on every sheet from three hands on — without ever estimating how much of a hand is habit. The same spread reads a hand's share of habit from two hands of two. What it cannot read is a habit every hand in the office shares: tracings of a sheet with one are tracings of a different sheet without it, and the charge they give falls further short the more hands and tracings are added.

wrong · Identify An office that overshoots headlands draws every coast wrong by that coast's own curvature. Four coasts at one map size: the compiled coast whose seams are counted, and three control coasts an office might trace with the truth known — an ellipse as thin as the counted coast (axis ratio 2.88), a nearly round one (1.12) and a coast of five headlands. Solid: each coast. Dashed: the same coast as an office would draw it whose every hand pushes a headland out and pulls a bay in by a fixed amount per unit of curvature, exaggerated a hundred and twenty times so that it shows. The curvature's spread round each coast, its variance in units of the coast's size: the sheet being counted 16.9, control, as thin as the target 15.6, control, nearly round 0.1, control, five headlands 29.8. The habit's size on a coast follows that spread, and the habit's slope does not change from coast to coast.

A known coast charges an office's habit, and a bending one takes it out

An office whose hands all share a habit leaves it in every tracing of every sheet, where the spread between hands cannot see it. One control sheet — a coast known independently, traced the office's usual way — reads the habit's size and restores the charge, if the habit is a drift. If the habit answers to the coast's own curvature, a size read on a rounder coast charges a thinner one at little more than half, but the same control read differently does better than charge the habit: its departures, set against its curvature, give a slope that is the office's own on every coast, and subtracting it leaves about one per cent of the habit in the mean. The slope also tells the two kinds of office apart. A coast that hardly bends teaches neither, and corrects worse than nothing.

wrong · Identify The bias is σ² over the length, over two decades. The amount by which a measured baseline is longer than the true one, against its length, for a twenty-millimetre error on each end. The line is the second-order prediction σ²/d. The measured bias times the length is constant to a factor of 1.0014 across the whole range, and it sits 8.2 per cent below the prediction — which is the fourth-order term the expansion drops. The estimate is antithetic, so the first-order scatter cancels exactly and a bias of thirty-seven microns is measured at a t-statistic of 349.

A length measured from noisy points is too long

A distance is a square root, a square root is concave, and the average of the distances is not the distance between the averages. The gap is a bias with one sign: 37 microns on a ten-metre baseline with twenty-millimetre marks, following σ² over the length across two decades, and it adds rather than cancelling — so the same boundary is 1.5 parts per million longer when it is measured in more pieces.

distortion · Precision The same field, the same statistic, two pages. A single northern concentration — v = 10 + 90 exp(−d²/(25°)²), d the distance from 20°E 55°N — drawn twice, with the shading rule and the data identical. The ground's own area-weighted mean is 14.009. A reader weighting by the area actually on the page takes 17.490 off Mercator, which is +24.85%, and 14.009 off Gall–Peters, which is +0.00%. Neither map has misdrawn a single cell.

A choropleth is read by area

Every cartography course states the rule — use an equal-area projection for a thematic map — and states it as advice. It is a theorem, and it has a residual: the error a page puts into a reading is exactly the covariance of the value with the areal factor, which is 24.85 per cent for a northern concentration read off Mercator and 0.00 per cent for the same field read off a map that spreads area by 7.7 to one.

distortion · Thematic The same count, scattered two ways. 22 dots in every cell of a 30-cell covering, drawn on Mercator. The upper panel places them uniformly on the GROUND — uniform in longitude and in the sine of latitude, which is what uniform on a sphere means — and the lower places them uniformly on the PAGE, which is what a drawing routine handed a polygon does. Both panels carry exactly the same number of dots in exactly the same regions, so both are honest as totals. They are different pictures, and a reader reads a dot map by density.

A dot map's density is partly the projection's

A dot map carries the right number of dots in every region whichever way it is drawn, so it is honest as a total under both placements. It cannot be honest as a density under both: ground on a uniform field reads 0.099 of its equatorial density at 72° north on Mercator, and scattering inside the polygon on the page moves 64.3 per cent of a cell's dots into its northern half without one of them leaving the cell.

distortion · Thematic One field, one classifier, two sets of breaks. A field that varies with latitude put into 5 classes by area-weighted quantiles, on Mercator. The upper panel weights each region by its ground area and the lower by the area it occupies on this page, which is what a classifier handed projected geometry does. 60 of 150 regions land in a different class, marked in the lower panel. The data has not changed and neither has the number of classes.

The class breaks were computed on the page

The three rungs below price what a reader does with a finished map. A classifier is software, it runs on the geometry it has, and the geometry it has is projected: a five-class quantile classification of one stated field puts half of the three hundred and eighty-four regions in a different colour on Mercator, and 87.5 per cent of them at nine classes.

distortion · Thematic Densifying helps one and hurts the other. The standard deviation of the shoelace area, and the bias of the perimeter, against the number of vertices on the same 100-metre circle with the same 5-centimetre noise on each. The area's spread falls from 9.9 m² to 2.0 — the closed form says it goes as the square root of the vertex count's reciprocal, because a vertex's influence on the area is the vector between its two neighbours and densifying shortens it. The perimeter's bias rises from 0.07 mm to 261, a factor of 3697, because every leg contributes its own σ²/d and shorter legs contribute more.

The area is unbiased and the perimeter is not

A boundary measured from noisy vertices comes out long, always, by σ²/d on every leg. The area enclosed by the same vertices comes out exactly right, because a shoelace is bilinear and the cross terms vanish. So densifying a boundary makes its area five times more precise and its perimeter three thousand times more wrong, and every compactness score computed from it falls short.

distortion · Precision One map, two samples, two answers. The same Miller cylindrical projection carrying two sets of sample points of the same size. On the left the points are uniform over the SPHERE — equal ground area between them, which is what every mean in this collection integrates against. On the right they are uniform over the PAGE, which is what a raster, a pixel loop or any figure that walks its own canvas produces. The right-hand set crowds where the map stretches, and the mean angular deformation it returns is 18.0° against the left-hand set's 7.2°.

The sample was drawn on the page

Every mean in this collection integrates over the sphere, because that is where the ground is. A raster, a pixel loop and any figure that walks its own canvas integrate over the page instead, and the difference is exactly the covariance between the quantity being measured and the map's own area distortion — 7.2° of mean angular deformation on Miller becoming 18.0°.

distortion · Sampling Two places, one written coordinate. The lattice a coordinate written to 5 decimal places lives on, at 45° of latitude, where a cell is 1.11 metres north to south. Two real places a fifth of a cell apart are drawn as open marks and the single value they both round to as a filled one. The distance between them, computed from what was written, is exactly zero — an error a noise model cannot produce, since independent noise never puts two different points in the same place.

Rounding is not noise

Eight rungs treat a coordinate's error as noise that averages down. A published coordinate has a second error that does not: it is deterministic, it is shared between every point in the same cell, and at five centimetres apart ninety-three per cent of pairs come out as the same place — which no amount of independent noise can produce.

distortion · Precision

Named alongside it

The objects these essays reach for when they reach for this one.

EstimatorAveragingVerificationEqual-areaNoiseArea weightingAreal factorGeneralisationPrecisionThematic mappingToleranceAggregation

All concepts