Thematic Maps, Choropleths and the MAUP Trap

A reference map shows where things are - a topographic sheet, a street map. A thematic map shows how a single theme varies across space - density, income, risk, land use. The whole of this lesson lives in the thematic world.

There is a small family of thematic types, and choosing the right one is half the battle. A choropleth shades predefined areas (wards, districts) by a value - the workhorse, and the trap. Graduated or proportional symbols place a dot or circle sized by a value, good for totals that should not be smeared across an area. A dot-density map scatters dots so that count shows as visible density, keeping a feel for where things actually are inside a zone. An isarithmic map draws smooth contours over a continuous surface (rainfall, elevation, a heat surface). Each answers a slightly different question; reaching for a choropleth by reflex is where most bad maps begin.

This is the single most common error in the discipline, so it gets its own section. A choropleth must show a rate, ratio or density - never a raw count.

The reason is geometric. If you shade wards by total population, the biggest wards will almost always look darkest - not because they are crowded, but because a bigger area holds more people. You have drawn a map of ward size wearing the costume of a map of population. The fix is to normalise: divide the count by something that removes the size effect. Population becomes persons per square kilometre (divide by area). Cases become a rate per 1,000 people (divide by population). Sales become per household. Only after normalising does the colour mean what the reader assumes it means.

The honest exception: if you genuinely want to show totals - where the people are, not how dense - do not shade areas at all. Use proportional symbols, where a circle's size shows the total and the area's size is irrelevant to the reader's judgement.

Put the two ideas together into a simple decision. Do you have a value that is already a rate or can be sensibly divided into one (density, percentage, average)? Then a choropleth is fair - shade away, with a stated classification. Do you have a raw total you must not normalise (number of clinics, number of trees, jobs)? Then use graduated or proportional symbols so size, not area shading, carries the count. Do you want the reader to keep a feel for where inside the zone things sit? Use a dot-density map. Is the phenomenon genuinely continuous, with a value everywhere? Then it was never areal to begin with - map it as an isarithmic surface or a raster, not a choropleth of administrative boxes.

The mistake to unlearn is treating the choropleth as the universal thematic map. It is superb for rates over meaningful areas and misleading for almost everything else.

Now the deep one. Whenever you aggregate point-level reality into areas, your results depend on how you drew the areas - and there is no single correct way to draw them. This is the Modifiable Areal Unit Problem (MAUP), named and dissected by Stan Openshaw, and it never fully goes away.

MAUP has two faces. The scale effect: analyse the same data at ward level, then at district level, and the pattern - even the correlation between two variables - can change or reverse, simply because larger units average away local variation. The zoning effect: keep the same number of zones but redraw their boundaries, and the map changes again. Political redistricting (gerrymandering) is the zoning effect used on purpose.

You cannot eliminate MAUP, but you can be honest about it. Use the smallest, most meaningful units your data supports. Report the unit you used and why. Where a finding matters, test whether it survives a change of scale or boundary - if it evaporates when you switch from wards to a grid, it was an artefact of the zones, not a fact about the city.

In Indian practice the areal units are usually handed to you - Census enumeration blocks and wards, or the ward and zone boundaries inside an AMRUT GIS master plan. That convenience hides a MAUP decision you did not make: someone drew those wards, often for administration rather than analysis, and their shapes now steer your maps.

So when you present a choropleth to a client or an authority, do three plain things. Name the unit (which ward set, which year). Name the normalisation (per square kilometre, per 1,000). Name the classification and source. And where a claim is load-bearing - this ward is under-served - show that it holds at more than one scale or you have merely mapped the boundaries. That discipline is what separates evidence from decoration in a plan-approval room.