How UK weather data is gathered

Surface observations

The backbone of UK weather observing is the Automatic Weather Station network. The Met Office runs around 270 official AWS sites across the UK, reporting every hour and, at many sites, every minute internally. Each site bundles the same basic kit: a thermometer and hygrometer in a ventilated screen, a tipping bucket rain gauge, an anemometer on a 10 m mast, a barometer, and increasingly a present-weather sensor and a laser ceilometer for cloud base.

Alongside that, around 50 UK airfields produce METAR reports — the standardised aviation weather code — every 30 minutes, with a fresh hourly one on top, and SPECI reports out of cycle when conditions change quickly. METAR is dense and ugly to read but it is the most timely surface dataset we have. Heathrow, Stansted and Boscombe Down show up constantly in UK weather writing because their METAR stations are reliable, long-running and well-sited. A METAR gives you pressure (in hectopascals and inches of mercury), temperature, dewpoint, wind direction and speed (with gusts), visibility in metres, cloud cover and base in eighths and feet, and a coded present-weather group.

Radar

The UK has around 15 weather radars, run by the Met Office and sited to give overlapping coverage from the Shetland Isles to the Channel. They are all C-band, and since the 2010s upgrade programme they are all dual-polarisation. The radar sends out pulses on two polarisations (horizontal and vertical) and compares the returns — that gives it enough information to classify hydrometeors, not just measure their reflectivity. Rain looks different from hail, which looks different from wet snow, which looks different from ice pellets. Pre-dual-pol radar could only really tell you how much was up there; dual-pol can tell you, with reasonable confidence, what it is.

Each radar scans every five minutes, sweeping through several elevation angles. The individual scans are stitched into a single national composite at 1 km horizontal resolution, which is what you see on most public radar maps. There are limitations baked in. Terrain blocks the beam, so the Highlands have known gaps where the radar simply cannot see past a ridge. Every radar also has a cone of silence directly overhead — the antenna cannot tip vertically, so anything straight above the station is invisible until it drifts into an observable elevation angle. And the beam climbs with range, so showers 200 km from the nearest radar are being sampled several kilometres up rather than at the surface.

Chaseit's live radar page reads from the Met Office composite directly.

Upper-air: radiosondes

Twice a day, at 00Z and 12Z, the UK launches weather balloons from a handful of upper-air sites. Camborne in Cornwall, Castor Bay in Northern Ireland, Herstmonceux in East Sussex, Lerwick in Shetland and Watnall in Nottinghamshire are the current operational sites, with Larkhill having featured historically. Each balloon carries a small instrument package called a radiosonde — a thermometer, hygrometer and pressure sensor, with a GPS receiver so the ground station can track its position and infer winds from drift. The balloons climb at around 5 m/s and reach roughly 30 km before they burst.

A sounding is the single most information-dense product in operational meteorology. From one ascent you get the vertical profile of temperature, dewpoint, pressure and wind, and from that you can derive CAPE (the energy available for convection), wind shear (how the wind changes with height — critical for organised storms), freezing levels, inversion layers and tropopause height. Forecasters live and die by soundings on convective days. The trouble is that two ascents a day is not enough resolution to capture how quickly things change ahead of a thundery setup, which is why aircraft-derived upper-air data has become so valuable.

See the soundings surface for live profiles, and winds for the derived upper-air wind atlas.

Upper-air from aircraft

Every commercial flight in UK airspace is already an upper-air weather instrument. Aircraft transponders broadcast ADS-B (Automatic Dependent Surveillance-Broadcast) and Mode-S EHS (Enhanced Surveillance) messages many times a second. Buried in those messages, especially in Mode-S EHS, are the true airspeed, magnetic heading, Mach number, static air temperature and selected altitude. With a bit of trigonometry and a known ground track, you can recover the wind vector and the temperature at the aircraft's flight level.

One aircraft does not tell you much. Tens of thousands of them, climbing and descending through the troposphere all day, tells you a great deal. Aircraft-derived data fills the gap between radiosonde launches and gives near-continuous upper-air coverage along busy airways. Chaseit operates its own ADS-B and Mode-S receiver network and feeds the derived winds into the upper-air wind atlas — the result is a higher-resolution picture of jet streaks and shear layers than soundings alone can give.

Lightning

Lightning detection lives on two technologies. The first is ground-based radio direction finding. A lightning stroke is a massive broadband radio emitter, and a network of receivers timing the arrival of the same pulse can triangulate its location to within a kilometre or so. The Met Office runs ATDnet, a global very-low-frequency network optimised for cloud-to-ground strokes. Alongside it sits Blitzortung, a volunteer-run citizen network that has put enormous density over Europe by encouraging hobbyists to host receivers in their attics.

The second technology is optical detection from space. The EUMETSAT Meteosat Third Generation Lightning Imager (MTG-LI), launched in 2022, is the European equivalent of the American GLM. It sits in geostationary orbit and watches the full disk continuously, looking for the optical flash of lightning at 777.4 nm — an oxygen emission line that fires when a stroke heats the surrounding air. MTG-LI sees both cloud-to-ground and intra-cloud activity, which makes it especially useful for picking up storm intensification before any rain reaches the ground.

Chaseit reads MTG-LI directly via EUMETSAT's data feed rather than relying on a second-hand network.

Satellite

Satellite weather observation splits cleanly into two orbits. Geostationary satellites sit 36,000 km above the equator and rotate with the Earth, so they appear to hover over a fixed point. Europe's geostationary fleet is Meteosat, now transitioning to the MTG generation, parked at roughly 0° longitude over the Gulf of Guinea — perfectly positioned to watch the Atlantic and Europe. MTG provides full-disk imagery every 10 minutes across multiple spectral channels.

The other family is polar-orbiting satellites — Metop, NOAA, and the various ESA Sentinels. They fly low, around 800 km up, so resolution is much higher than geostationary, but each satellite only passes over a given spot a couple of times a day. Polar orbiters are where high-detail soundings, scatterometer winds, sea-surface temperature and most of the science-grade products come from. Geostationary tells you what is happening now; polar-orbiting tells you what it really looks like.

Rainfall and rivers

Radar gives you the spatial picture of rainfall, but the ground truth comes from rain gauges. The Environment Agency operates around 1,000 telemetered rain gauges across England, most of them tipping-bucket designs that record every 0.2 mm of rainfall as a discrete tip. Pair that with a river-level monitoring network of around 1,500 gauging stations, and you have the input to the flood-forecasting models that drive warnings to Property Flood Resilience teams and the public.

Equivalent agencies cover the rest of the UK. SEPA — the Scottish Environment Protection Agency — runs the network in Scotland, with its own dense coverage of Highland and Borders catchments. NRW, Natural Resources Wales, looks after Welsh rivers and rainfall. All three publish telemetry openly, and it is that combined dataset that drives the Flood Forecasting Centre's national outlook.

Citizen networks

Official networks are accurate but sparse. Citizen networks are noisy but dense, and they fill a lot of gaps. The Met Office runs WOW (the Weather Observations Website), a citizen-science platform where anyone with a backyard weather station can submit observations. Netatmo's commercial personal weather stations expose their data through a public API and produce a remarkably dense temperature and pressure map across the south east. Amateur radio operators run a sprawling network of APRS weather stations, broadcasting on 144.800 MHz and aggregated through aprs.fi.

Data quality varies wildly. A Netatmo on a south-facing wall in full sun is reading something, but it is not air temperature. The right way to use these networks is as a mesoscale supplement — they tell you that a sea breeze has actually crossed the M25, or that a thunderstorm gust front has just rolled through Watford, well before the nearest official AWS catches it. For exact climatology, stick with the official network.

Marine

The sea has its own observing system. The Met Office maintains a network of moored buoys around the UK coast and out into the North Atlantic, reporting wave height, sea-surface temperature, wind and pressure. Drifting buoys, released into ocean currents, plug the gaps between the moorings. On top of that, the ASCAT scatterometer instrument on the Metop satellites measures ocean surface winds globally by bouncing radar off the centimetre-scale roughness of the sea — a calmer sea reflects differently to a windier one, and the instrument retrieves wind vectors across a wide swath.

All of this feeds the marine forecasts you hear on the Shipping Forecast and read through MetOffice DataPoint. HM Coastguard re-broadcasts the same picture on Marine VHF channels for vessels at sea, and the inshore waters forecasts fold in the same coastal AWS data plus tide gauge readings.

Reanalyses and history

All of the above is the present-tense picture. The historical record is its own enormous project. ECMWF's ERA5 reanalysis is the workhorse: it runs a modern forecast model backwards through every weather observation ever collected since 1940, producing a globally consistent hourly history at roughly 30 km resolution. ERA5 is what you reach for when you want to ask "how unusual was that" in a rigorous way, because it normalises away the changes in observing networks over time.

For UK-specific climatology, the Met Office's HadUK-Grid is the canonical product — gridded temperature, rainfall and related variables on a 1 km grid covering the whole of the UK from 1862 to the present. It is a careful reconstruction from the surface station network across more than a century and a half, and it is what underpins almost every UK climate-change statement you read. Our history section draws on the same kinds of records.

How it all gets to a forecast

Observations on their own are not a forecast. What turns them into one is data assimilation: every few hours, an operational model takes its previous forecast and gently nudges it towards what the observations are actually saying, weighting each observation by its expected uncertainty. The result — the analysis — is the starting point for the next forecast cycle. In the UK, the high-resolution model is the UKV, a 1.5 km grid covering the British Isles, run by the Met Office on their supercomputers and refreshed every hour. Behind that sits the Global Model at around 10 km resolution, and around the world other agencies run AROME (Météo-France), the ECMWF IFS (the European medium-range workhorse), and many others. The same observations feed almost all of them.

Model output is what you actually consume. The Shipping Forecast, the BBC weather symbols, the MetOffice text products, the DataPoint API, the ECMWF charts, the convective outlook on this site — every one of them is a view onto a model run that has been initialised from the observation network described above. The Met Office page covers the UK side of that pipeline in more depth, and the wiki collects the chase-relevant definitions in one place.