Tracking TCM thunderstorm forecast accuracy from today onward — comparing model-predicted lightning activity against real observed strike counts across UK regions.
A plain-English guide to what every number on this page means
Every lightning bolt detected by the Blitzortung sensor network is saved to our database in real time. At the end of each day we count how many strikes fell inside each UK region (South England, Midlands, North England, Scotland, Wales, Ireland, and UK-wide). This is the "Observed" number.
The TCM (Thunderstorm Convective Model) assigns each grid point across the UK a lightning score from 0 to 100. We average all those scores within a region and scale that average into an estimated strike count. This is the "Predicted" number. Note: the TCM is primarily a risk tool, not a strike-count tool, so the prediction is a rough estimate.
Accuracy = 1 − (|Predicted − Observed| ÷ max(Predicted, Observed)) × 100. In plain English: if the model predicted 80 strikes and 100 were observed, the difference is 20. Divided by the larger number (100) that's 20%, so accuracy = 80%. A score of 100% is a perfect match. If both predicted and observed are zero (no storm, no strikes), that also counts as 100%.
Bias = Predicted ÷ Observed. If bias is 1.0 the model is well-calibrated. A bias above 1 means we predicted more strikes than happened (over-forecast). A bias below 1 means we predicted fewer than happened (under-forecast). Over time, a consistent bias tells us which direction to nudge the model.
This page reset on 23 June 2026. A verification record is written automatically at midnight each day. The charts and tables will fill in as days pass — after a week you'll see a 7-day chart, after a month the trends become meaningful.
Quick reference