0–100

Experimental score

How the Forecast Success Score is graded

The score compares archived General risk polygons, the hazard probabilities attached to those polygons, separate tornado, hail and severe-wind risk layers, ESWD reports and archived lightning from the same 06:00–06:00 UTC period.

Important: this is an experimental CESTOF verification guide, not an official scientific verification product. Missing reports, reporting bias and incomplete archives can affect the result.

When a day can be rated

A final numeric score is only produced when at least 100 archived lightning strikes are available in the Central European verification area. With fewer strikes, the day is shown as Not Rated.

Hazard placement may still be calculated internally, but the final number is withheld because a quiet or incomplete lightning archive could produce a misleading result.

Base-score weighting

32%

General risk placement

All reports are checked against the archived General risk category at their exact location.

13%

General report capture

Measures whether severity-weighted reports were inside any General forecast risk.

25%

Polygon hazard probabilities

Checks the hazard chances attached to the exact General polygon covering each report or lightning strike.

10%

Lightning capture

Measures the share of sampled lightning strikes inside any General forecast risk.

12% when used

Separate hazard placement

Checks tornado, hail and severe-wind reports against their separate hazard-risk layers.

8% when used

Separate hazard support

Checks whether the issued hazard percentages and matrix values had matching observed support.

General risk placement, General capture, polygon hazard probabilities and lightning together control 80% of the base score. Separate tornado, hail and wind layers refine the result but do not outweigh the General forecast.

Which forecast checks each observation?

Observed event	Forecast information used
Tornado or waterspout	General Tornado chance and separate Tornado risk
Hail	General Hail chance and separate Hail risk
Hail at least 20 mm	Hail and Large hail chances
Hail at least 50 mm	Hail, Large hail and Giant hail chances
Severe wind	General Wind chance and separate Severe-wind risk
Wind at least 25 m/s	Wind and Damaging wind swath chances
Heavy rain	Heavy rain chance inside the General polygon
Flood / flash flood	Flash flooding chance inside the General polygon
Lightning	Lightning and Frequent lightning chances

Supercells, significant supercells, bow echoes and MCS / squall lines are recognised in the archive, but they are not numerically verified until a dependable observed storm-mode dataset is available.

Polygon-specific hazard probabilities

Hazard chances are stored inside individual General polygons. They are not treated as one percentage for the entire forecast day.

A report or lightning strike is checked against the exact polygon, or overlapping polygons, covering its location. This matters because hazard chances are often attached only to selected polygons, usually the highest-risk part of the forecast.

Issued chance at the observation location	Placement credit
No matching hazard chance	0 / 100
2%	68 / 100
5%	78 / 100
15%	88 / 100
30%	95 / 100
50% or higher	100 / 100

The exact percentage attached to the polygon is used. Any correctly placed percentage earns strong credit, while higher percentages add a calibration bonus. A 2% or 5% forecast is therefore not treated as a failure when the matching event occurs inside it.

Report severity weights

Reports do not all have equal influence. More severe events receive a larger weight, so missing a destructive event damages the score much more than missing a weak event.

Report severity	Approximate weight
Funnel cloud or weak sub-severe report	0.5–0.65×
IF0 tornado, 20–29 mm hail, or 25–32 m/s wind	0.9–1.0×
IF1 tornado, 30–49 mm hail, or 33–41 m/s wind	1.4–1.5×
IF1.5–IF2 tornado, 50–69 mm hail, or 42–49 m/s wind	1.8–2.3×
IF2.5–IF3 tornado, hail at least 70 mm, or wind at least 50 m/s	2.8–3.5×
IF4–IF5 tornado	4.5–5.0×

When severity is missing, the system uses a neutral default weight for that report type.

When no polygon hazard percentage is issued

If no hazard percentage is attached to any General polygon for that archive day, the Polygon hazard probabilities component is completely excluded. It does not receive a zero and does not lower the score.

The remaining available components are automatically rebalanced to 100%.

How the issued percentage is used

The calculator reads the strongest matching percentage at each observation location. For example, a hail report inside a polygon carrying Hail 5% is graded from 5%, while a report inside Hail 15% is graded from 15%.

The system is intentionally forgiving: correctly locating the hazard matters more than issuing a very high number. Higher percentages improve the score, but lower percentages still receive substantial credit.

High-risk and matrix support

The highest General risk and the separate tornado, hail and wind risks are checked for observed support. The system also reads the stored probability and intensity matrix when that information exists in the archived polygon.

Reports inside any issued hazard area earn credit. Reports inside the highest issued percentage earn additional credit. A matching probability–intensity matrix cell adds further credit.

Final score = weighted base score × moderate General support adjustment × moderate hazard support adjustment

Unsupported high risks lower the result, but the adjustment is deliberately moderate so incomplete ESWD reporting does not destroy an otherwise strong forecast.

How the calculation works

Load the archived General, tornado, hail and severe-wind forecast layers.

Load ESWD reports and archived lightning for the matching 06:00–06:00 UTC period.

Determine each report's hazard type and severity weight.

Check each report against the General risk at its exact location.

Read the hazard chances stored in the exact General polygon covering the report or strike.

Check tornado, hail and wind reports against their separate hazard layers when those risks were issued.

Read the probability–intensity matrix when it is available.

Calculate the weighted score, apply moderate support adjustments, round the result and assign a grade.

Examples

Strong polygon verification: a hail report falls inside the highest General polygon where Hail 15% was attached. The General risk, matching polygon hazard chance and separate hail layer all receive positive credit.

Useful lower probability: a severe-wind report occurs inside a Wind 5% polygon. It receives meaningful credit because the correct hazard and location were forecast, even though the issued percentage was low.

Hazard chance in the wrong place: Heavy rain 30% is attached to one General polygon, but the heavy-rain reports occur in another polygon without a Heavy rain chance. The reports still count for General placement, but not for the polygon hazard probability component.

Missed severe event: a destructive hail report occurs outside both the General Hail chance and the separate hail-risk layer. Its larger severity weight lowers the score more than a weak report would.

No separate hazard risk issued: the day is still fully graded from General placement, General capture, polygon hazard probabilities and lightning. The missing separate hazard components are excluded and the remaining weights are rebalanced.

Grade labels

Excellent90–100

Very good80–89

Good70–79

Fair60–69

Limited45–59

Poor0–44

Limitations

ESWD reports are not evenly distributed and some events are never reported. A missing report does not prove that no severe weather occurred. Lightning does not directly measure tornado, hail or wind intensity. The score should therefore be used alongside manual forecast review, radar analysis and other verification data.

Return to Forecast Archive

Marginal Tornado Potential

Marginal Tornado Potential is ignored only in the tornado hazard component when no actual tornado or waterspout is reported. The rest of the forecast day is still graded normally.

Funnel-cloud and gustnado reports alone do not positively verify Marginal Tornado Potential.

Important limitations

ESWD reports are not evenly distributed and some events are never reported. One storm may also produce several nearby reports. Lightning does not directly measure tornado, hail or wind intensity.

The Experimental Forecast Success Score should therefore be used together with manual review, radar analysis and other verification information.

Return to Forecast Archive