Map projection distortion
A Socratic walk-through of map projection distortion — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #What must a flat map of a round world distort, and why can no projection escape that cost?
It is tempting to hear "all maps distort" as an apology for imperfect craftsmanship, as though a cleverer cartographer might one day get it right. But the claim is stranger: it is not that no one has managed it, but that no one can, for reasons having nothing to do with technique. So the first question is not which map is best, but what exactly cannot be done?
Reasoning it through
REASONING #Ask what "no distortion" would mean. That the distance between any two points on the map, at the stated scale, matches the distance between the corresponding points on the globe — every pair, everywhere. Achievable? Peel an orange and press the peel flat. It splits; push the edges together and it buckles. Peeling more carefully will not help.
Why not? Gauss answered this, and it is the one piece of real mathematics worth carrying: a surface has a curvature measurable from within it, using only distances along it, without stepping outside to look. A sphere's is positive, a flat sheet's is zero, and any transformation preserving all distances must preserve that quantity too — so none exists. Notice this rules out a possibility, not a technique. And note what it does not rule out: a cylinder and a cone unroll flat with no distortion at all, being curved in only one direction.
So something must give. What can we save instead? Shrink the question: put a tiny circle on the globe and see what the map does to it. It returns as an ellipse, stretched more one way than another — and that ellipse encodes everything, its shape reporting whether directions are skewed and its area whether sizes are.
Now two ambitions become visible. You could insist the ellipse is always a circle, so that at each point the scale is the same in every direction and angles and small shapes come out right — a conformal map. Or that it always has the same area, so sizes are comparable across the sheet — an equal-area map. And here is the question that closes the argument: what would demanding both mean? A circle, always, of the same size, always — the same scale everywhere in every direction, which is exactly the distortion-free map Gauss has just ruled out. The two goals are not merely hard to combine; combining them would be the impossible thing.
The analogy
THE ANALOGY #It is like a household budget with a fixed total. You may spend on rent or on food in whatever mixture you like, and a skilled hand can spread the pain so nothing hurts badly — but you cannot spend more than you have, and no arrangement of the columns creates money.
a budget's total is contingent and a raise could change it, whereas a sphere's curvature is negotiable by no means whatsoever — and unlike money, the "spending" here varies from place to place on the same map, so a projection can be nearly free of cost near one line and ruinous far from it.
Clarifying the model
THE MODEL #One point deserves emphasis, because most confusion lives there: distortion is not a single number attached to a map but a field varying across it. Mercator is the famous case — conformal, essentially faithful along the equator, but the north-south stretch it needs to keep angles honest grows without limit toward the poles, so areas inflate badly at high latitudes. Which is why Greenland can look comparable to Africa when Africa is about fourteen times larger by area. Worth defending it, though: it was built so a line of constant compass bearing draws straight, exactly what a navigator needs, and it remains right for that job and for zoomed-in web maps where the whole-world view is incidental.
Two more corrections. "Distance" is not on the menu the way the other two are: no flat map preserves all distances, and equidistant projections only promise correct distances from one or two chosen points. And most everyday world maps — Robinson, Winkel Tripel — are neither conformal nor equal-area, deliberately taking a little of both errors to look reasonable overall. That is a legitimate answer to the tradeoff, not a failure to grasp it.
A picture of it
THE PICTURE #How to readRead rightward for how well shapes survive and upward for how well areas survive, then look at the top-right quadrant — it is empty, and that emptiness is the point of the picture rather than an oversight. The bottom-right cluster buys faithful shapes by letting polar areas blow up; the top-left buys faithful areas by shearing shapes at the edges. The two points near the middle are the compromises, deliberately below and left of both ideals: not worse, but a different choice about where to spend a fixed budget.
What became clearer
WHAT CLEARED #A flat map is not an imperfect copy of the globe but a surface of different intrinsic curvature, and the mismatch is a theorem rather than a limitation of craft. Once that is accepted, "which projection is correct?" dissolves into the only question that survives: which distortion does this map's job tolerate?
Where to go next
ONWARD #- Why great-circle routes look absurdly curved on Mercator and straight on a gnomonic projection.
- How the choice of projection is made almost invisible in web maps, and what that hides.
Key terms
TERMS #| Term | What it means |
|---|---|
| Theorema Egregium | Gauss's result that a surface's curvature is measurable from within it, and so cannot be changed by bending without stretching. |
| Conformal | a projection preserving angles and local shapes, at the cost of varying scale. |
| Equal-area | a projection preserving relative areas, at the cost of shape. |
| Tissot's indicatrix | a tiny circle on the globe drawn as the map renders it, showing local distortion as an ellipse. |
Every term the collection defines is gathered in the glossary.