Every UNESCO World Heritage Site (WHS), Intangible Cultural Heritage element (ICH), Memory of the World inscription (MoW),
Global Geopark (GEO), and Biosphere Reserve (BR) is assigned to the nearest or most representative city from the UN's list of
519 built-up areas ≥1,000,000 people (World Urbanization Prospects 2025). The map on the left is that same 519-city network
— node radius models each UNESCO site as a unit of volume added to a sphere (so radius grows with the cube root of
1 + site count, tapering more gradually than a flat-area scaling would), while colour still marks the country. The lines are an
independent city-to-city connection graph carried over from the source map, unrelated to UNESCO data; only outgoing connections
are shown when a city is selected. Click any node to see what's assigned there, or use the connections panel to hop to a
connected city.
- Coordinate match: where the site's own coordinates are known (via Wikidata), assigned to the geographically nearest UN city, restricted to the same country when that country has one.
- Country tiers (used for ICH/MoW, which usually have no coordinates of their own, and as a fallback elsewhere):
Tier A — country has exactly one UN city, used directly.
Tier B — country has none, assigned to the true geographic nearest UN city across the border (via the country's centroid).
Tier C — country has several; a place/region is extracted and geocoded where one can be identified, otherwise it falls back to the country's capital (or largest city if the capital isn't on the list).
- Intangible Cultural Heritage place signals were extracted, in order of precedence, from: (1) the element's own title (e.g. "Sango Festival, Oyo"), (2) its Wikipedia article text, and (3) — for anything still unresolved — the element's own official description on ich.unesco.org, which is often far more specific than either. About half of all ICH elements carry a real place signal this way; the rest are genuinely national or region-wide practices with no specific location to name, and fall to the country default.
- Memory of the World items were matched to their actual holding institution's city (e.g. an archive or library) via Wikidata's "location" property, not just the nominating country.
- Transboundary sites (spanning multiple countries) are assigned to multiple cities, one per country, where that is the more honest representation.
- A modest fraction of ICH/BR assignments rely on manual geographic judgment where automated data was unavailable or ambiguous.
- Balancing pass 1 (no-cost): a genuinely national ICH element (e.g. "Yoga" in India) has no real connection to its country's capital over any other city in that country — the capital was only ever a neutral default. Such elements, in countries where the capital had accumulated many diffuse items while other cities sat empty, were freely reassigned to a different city in the same country to spread coverage — never to a city that would itself end up empty, and never at any cost to geographic accuracy since none of these had a specific location to begin with.
- Balancing passes 2 & 3 (small-cost): the same near-tied-alternate logic, first for countries with no UN city of their own, then generalized to every country. Both capped by relative distance — at most 50% farther than the original (with a 300km ceiling) — not a flat km budget, since a site 10km from its city and one 500km away don't tolerate the same absolute detour.
- Cross-border reassignment (accuracy fix, not balancing): physical sites (WHS/GEO/BR) had always been restricted to a city within their own nominating country. But "which city would you actually fly into to see this" is a better test, and for many sites the honest answer is a city across the border — sometimes dramatically so (a New Caledonia reef site was assigned to Paris, 16,478km away, when Brisbane is 1,410km). Gated by how far the original pick already was: sites ≥2,000km from their own country's city were moved to any meaningfully closer alternative; sites in the 500–2,000km range needed a much stronger case (the new city under 35% of the original distance, or at least 500km closer).
- Audit correction, round 1: "no automated place signal found" is not the same as "genuinely diffuse." All freely-reassigned ICH elements from balancing pass 1 were re-audited against their official UNESCO description text; about a third turned out to have a real location the earlier passes missed and were corrected — e.g. Hezhen Yimakan storytelling to Harbin (a small Heilongjiang minority), Meshrep to Ürümqi (a Uyghur/Xinjiang tradition), Congolese rumba back to Kinshasa (its well-documented living center, even though the joint DRC/Congo nomination text is diplomatically non-specific).
- Audit correction, round 2 (wider sample): a 74-element random sample was drawn across every ICH matching method, including elements untouched by any correction pass, and checked the same way. Found the earlier passes were mostly sound (94% clean), but surfaced a few more real misses, including two Siberian traditions assigned to Novosibirsk at 2,800–2,100km when the true nearest city (via the same cross-border logic already used for physical sites) was 400–1,800km away in Mongolia or China.
- Audit correction, round 3 (full remaining coverage): the other ~90% of ICH elements were split into batches and independently reviewed against their official description text, the same rules, and the same reference city lists used throughout. Net result: dozens more corrections, roughly a 6% error rate — consistent with round 2, suggesting that rate holds fairly evenly rather than being concentrated anywhere in particular. Every corrected site carries a note explaining the change, visible in its detail entry.