We found the hottest block in each of 34 British cities. 121,819 addresses sit on ground that runs 5°C above their town’s norm.

The league: every city’s hottest block, named

Sorted by anomaly. Each city links to its own heat map. The “hot tenth” is the city’s top 10% of blocks by anomaly — nationally it holds 1,360,194 addresses, 15% of all addresses on measured ground: the heat concentrates where the people do, because dense fabric is exactly what runs hot.

The coolest ground is parks — measured, with the exception named

Crossing every block against OS Open Greenspace: blocks inside a greenspace of ≥2 ha average -2.08°C against their city’s norm while the built fabric averages -0.04°C — parks run 2°C cooler than the streets. The cooling leaks outward: blocks within 150 m of a park average -0.52°C, partway back to the fabric’s heat. 9 cities’ coolest confirmed blocks sit inside a named greenspace — Shenley Wood (Milton Keynes), Parkhall Country Park (Stoke-on-Trent), Cathkin Braes Country Park (Glasgow) among them; several of those parks hold water, which is part of why parks cool. We deliberately do not name a “coolest block” for the other cities — the coldest single pixels in a thermal composite are where artefacts concentrate (shadowed slopes, shorelines, high ground), so the cooling claim rests on averages over thousands of park blocks, never on one cold pixel. The control ran the other way in exactly one city, and we print it: in Cambridge, mapped greens are statistically no cooler than the fabric — the big greens there are unirrigated grazed commons, and browned-out summer grass barely cools. Green on a map is not the same as shade and water on the ground.

Method — and what we caught in our own data

1. Heat: Landsat 8/9 Collection 2 Level-2 land-surface temperature via the Microsoft Planetary Computer, the clearest June–mid-September scenes per city, cloud-masked, median-composited onto a ~150 m grid (clearest summer days, 2021-2024) — the same composite behind the heat row in our paid UK report and the per-city maps at /urban-heat.
2. What we caught: our composite had kept water pixels — cold rivers, docks and sea were sitting in the grids, averaging −11°C, and quietly corrupting any “coolest” claim near a coast. The finding is disclosed rather than patched silently — and it is now fixed at source: the composite masks water via Landsat’s own per-pixel flag and drops any block whose footprint is ≥30% water. The JRC Global Surface Water archive stays on as an independent cross-check; it flags 419 residual blocks in the rebuilt grids, excluded before a single figure on this page was computed. The rebuild also re-centred each city’s median on dry land alone, so anomalies here run slightly lower than first published — the corrected figures are the smaller, truer ones.
3. Artefact guard: a city’s hottest (coolest) block only counts if at least 2 of its 8 neighbouring blocks are also in the city’s top (bottom) 5% — 7 single-block spikes were skipped across the 34 cities rather than published.
4. Names and counts: blocks are named by nearest postcode and ward (postcodes.io); address counts are OS Open UPRN centroids inside each block — 9,346,671 addresses sit on measured ground in total, 1,402 of them on the 34 single hottest blocks.
5. The controls: hottest blocks must land on hard, dry land uses (they do — industrial estates and dense cores, every city); parks must measure cooler than the fabric (they do in 33 of 34 cities; any exceptions are named above, not hidden); and within-city spread must dwarf the sensor’s noise (the narrowest city still spans 5.83°C between its 5th and 95th percentile block; the median city spans 7.6°C).

Full per-city table: cities.csv. Data: USGS/NASA Landsat via Microsoft Planetary Computer · JRC Global Surface Water · OS Open Greenspace & OS Open UPRN (OGL) · postcodes.io. Contains OS data © Crown copyright 2026. This page is dated 2026-08-09. We did not produce the measurements; we produced the join.

Is this air temperature? Would a thermometer read these numbers?

No — this is land-surface (skin) temperature: how hot the ground, roofs and tarmac themselves get, measured by Landsat around 10:50 on clear summer mornings and composited over the clearest days of 2021–2024. Surfaces run much hotter than air on a sunny day, and the gap between a city's hottest and coolest blocks is far larger than any air forecast shows. It is a relative map of which blocks bake, not a forecast and not a heat-health measurement.

Why do industrial estates top every list?

Because that is what the physics says should happen, and it is one of our checks that the measurement works. Dark roofs, tarmac yards and no vegetation absorb the most sun and evaporate the least water. Birmingham's hottest block sits in the Castle Bromwich industrial belt, Middlesbrough's in the dock quarter, Luton's around the Vauxhall works. A surface-heat map whose hottest pixels landed on parks would be broken.

Do parks really cool cities?

Measurably: across 34 cities, blocks inside mapped greenspaces of 2 hectares or more average 2°C cooler than the built-up fabric, and the streets within 150 m of a park sit partway between — they borrow some of the cool. The honest exception: in Cambridge the mapped greens are statistically no cooler than the fabric — the big greens there are unirrigated grazed commons, and browned-out summer grass loses most of its cooling. The finding is that trees and water in a park matter, not green on a map.

My street is not in a covered city — can you check it?

We measure the 34 cities listed here (123,525 blocks) and never guess beyond them. The £19 UK property report includes the same measurement as a row — how many degrees hotter or cooler an address's block runs than its town, as a percentile — for any postcode inside a covered city, and simply omits the row elsewhere.

Check any covered postcode: HeatCheck (free) — or the £19 UK property report, which carries the same measurement as a row. More from the same discipline: Sinking Britain · the clay forecast · the water debt.