How do you calculate the critical success index for rare storms?
The critical success index measures how many forecast or observed storm events were matched hits. A long run of correctly forecast quiet periods leaves the score unchanged.
Start with Probability of detection vs. false alarm ratio: storm exercises.
| Counts | Values |
|---|---|
| Hits | 6 |
| Misses | 2 |
| False alarms | 4 |
| Correct negatives | Excluded from CSI |
Worked example
Let H = hits, M = misses and FA = false alarms. Every case in which a storm was forecast or observed belongs to exactly one of these cells. Their union has H + M + FA cases, so CSI = H/(H+M+FA). Using synthetic counts H = 6, M = 2 and FA = 4 gives 6/12 = 0.50. If neither forecasts nor observations contain events, the denominator vanishes.
Practice questions
- Replace two of the misses with hits and keep false alarms fixed.
- Add 100 quiet windows with correct negatives. How does CSI change?
Worked answers
Answer 1
There are 8 hits and no misses. CSI = 8/(8+0+4) = 2/3.
Answer 2
It stays at 0.50 because none of those windows join the event union.
- There are 8 hits and no misses. CSI = 8/(8+0+4) = 2/3.
- It stays at 0.50 because none of those windows join the event union.