Loading
Loading
Convective outlook
· Live · prismHow likely is a severe thunderstorm in southern UK over the next five days. We combine instability and wind shear from public forecast models into a single category per day, refreshed every few hours. Chasers — full numbers (CAPE, shear, LPI) sit on the day cards below.
Updated · 15 Jun, 02:16
Auto-generated each morning from the convective outlook. Guidance only — not a substitute for official Met Office warnings.
Friday looks like the chase day, with caveats.
UKV ENH, IFS HIGH, ICON SLGT — full split. Ensemble shows 78% probability ≥ SLGT. Peak instability around 13:00 UTC.
Skill caveat · Heads up: IFS has been running almost a full category high at 4-day lead over the last 21 verified days. Discount its HIGH call until the day shows itself — lean SLGT as the working assumption.
Headline level is the areal 5-day outlook — the same call shown on the day cards. The dots show how individual cells vary across the domain; a single hotter cell doesn’t lift the areal call, so the chips above (e.g. 3× SLGT) read as the spatial spread, not the headline.
Isolated showers / weak storms possible (CAPE ~110 J/kg).
No significant convection expected.
Five independent forecast cores on identical CAPE × shear thresholds. Cards flagged diverge are days where models disagree by a category or more — split means three or more give different answers. Both are signals to widen the uncertainty in your planning.
Levels follow the convention used by storm-prediction agencies worldwide, so they line up with what you may see on social media during an event.
Isolated showers / weak storms possible (CAPE ~100 J/kg).
Organised storms possible — CAPE 370, 0-6km shear 25 m/s.
Enhanced risk — discrete/clustered storms, hail and wind. CAPE 700, shear 21 m/s.
UKV is the Met Office’s 1.5 km regional model. IFS is ECMWF’s flagship global model at ~9 km. ICON-EU is the German Weather Service at ~6.5 km. GFS is NOAA’s deterministic 0.25° model. AROME is Météo-France’s 1.3 km HD mesoscale model (southern UK only). Five independent physics cores — when they agree you have high confidence, when they split you don’t.
Where the deterministic models give one answer, an ensemble runs many slightly-perturbed versions and tells you how much they agree. Bars below show the fraction of members at each category per day. A narrow, single-colour bar means high confidence; a spread across categories means low confidence. The number on the right is the probability of slgt or worse — the rough “chance of a chase day”.
GFS-ENS is NOAA’s 30-member ensemble, served as open data. Each member is one plausible evolution of the atmosphere — so reading the bar is reading agreement among 30 slightly-different worlds. ECMWF’s ensemble would be the natural complement here; it isn’t exposed with per-member CAPE through our current source, so for now we lean on GFS.
A single best-estimate level per day, blending all six models weighted by their rolling 30-day categorical hit-rate. Bias arrows (↑↓) show systematic over/under-prediction corrected before blending. A ⚡ live badge means active Blitzortung lightning has boosted day-0 above the model forecast.
The consensus blends UKV, ECMWF IFS, DWD ICON-EU, GFS, Météo-France AROME, and our own CIN-filtered composite. Weights are calibrated from the rolling verification database — as more observations come in, better-performing models gain influence.
Pressure-level stability diagnostics — the other half of what CAPE×shear alone misses. K-Index and Total Totals are the UK forecasting community's standard first-pass thunderstorm filters; 850 hPa temperature drives the instability base.
K-Index ≥ 25 → scattered storms likely · ≥ 30 → widespread. Total Totals ≥ 44 → isolated · ≥ 50 → organised possible. T850 ≥ 12°C → warm base · ≥ 15°C → exceptional UK setup. Lapse rate ≥ 6.5 °C/km → conditional instability.
The three variables that tell you whether instability will actually fire. PWAT sets the moisture reservoir — low PWAT limits storm depth and rainfall. Lifted Index measures how buoyant a parcel becomes — negative is unstable, below −4 is very unstable. CIN is the cap: a strong negative CIN needs a trigger (front, orography) to overcome. High CAPE + strong CIN = nothing fires unless something forces it.
PWAT > 35 mm is exceptional for the UK and associated with Spanish Plume or Azores-ridge moisture surges. LI < −2 with a weak CIN cap is the “storms fire easily” pattern. LI < −4 with CIN > −25 is the most dangerous combination — prolific instability waiting for any trigger. Strong CIN (< −100) with high CAPE is a loaded-gun setup: quiet until a front or convergence line breaks the cap, then explosive.
Two models we build in-house from raw siphon data, targeting the patterns NWP models most often get wrong in the UK. CIN-filtered penalises convective cap strength and boundary layer depth to reduce false alarms. LPI uses the microphysics ice-charge-separation signal to catch prolific lightning days that look modest on CAPE alone.
CIN-filtered: base CAPE × shear level adjusted for convective inhibition, lifted index, and boundary layer depth. Reduces the “high CAPE, no initiation” false-alarm pattern common in UK summer forecasts. LPI: the Open-Meteo lightning potential field derived from mixed-phase microphysics — a prolific lightning day can score ENH even with modest CAPE if the updraft punches through the freezing level efficiently.
The domain-level outlook blends all 10 regions into one number. This shows the individual regional picture — useful when a setup is strongly regionalised (Spanish Plume favouring the South East, or a northward-displaced jet favouring the Midlands).
| Region | Tod | Tom | Wed | Thu | Fri |
|---|---|---|---|---|---|
| South West | slgt | none | none | mrgl | enh |
| South Central | none | none | none | none | mdt |
| South East | none | mrgl | none | slgt | mdt |
| London | none | mrgl | none | mrgl | enh |
| East Anglia | none | none | slgt | enh | mdt |
| East Midlands | none | none | mrgl | mrgl | enh |
| West Midlands | none | none | mrgl | mrgl | slgt |
| Wales | mrgl | none | none | slgt | enh |
| North West | none | none | none | enh | enh |
| Yorkshire | none | none | none | slgt | mdt |
Each region scored independently at its SAMPLE_POINT using the same CAPE × shear thresholds as the domain model. Hover a cell for the raw CAPE and shear values.
Each row is one future date. The dots show how UKV and IFS have called that day across the last 96 hours of model runs — left is oldest, right is latest. Stable dots = high confidence. Jumping dots = the models are still resolving the setup.
Every day in the past month where the system called at least Marginal risk. Verified days show what actually happened — this is the ground truth that drives the rolling skill scores.
Each model’s domain-wide forecast is scored against every chase region’s observed level. A model that over-calls SLGT but only one region saw it earns one hit and partial credit for the others — so the overall rate penalises over-calling honestly.
| Region | Peak precip | Thunder? | Max wind | Min T−Td |
|---|---|---|---|---|
| South West | 48% | no | 11 kt | 0.9°C |
| South Central | 38% | no | 12 kt | 2.4°C |
| South East | 93% | no | 11 kt | 1.3°C |
| London | 60% | no | 9 kt | 3.2°C |
| East Anglia | 16% | no | 10 kt | 0.2°C |
| East Midlands | 33% | no | 10 kt | 1.3°C |
| West Midlands | 48% | no | 9 kt | 3.3°C |
| Wales | 51% | no | 8 kt | 1.5°C |
| North West | 45% | no | 10 kt | 2.2°C |
| Yorkshire | 75% | no | 10 kt | 0.3°C |