Tropical cyclone forecast verification - measures of error
The Met Office verification scheme identifies the centre of a tropical cyclone in the model by a combination of a local maximum of relative vorticity at 850hPa and a local minimum of mean sea-level pressure. The local maximum in relative vorticity is also used in the verification of intensity. A fourth order interpolation is used to locate the centre accurately. Advisory data received in real time from JTWC Pearl Harbour, NHC Miami, CPHC Honolulu, RSMC La Réunion, TCWC Darwin and RSMC Fiji are used as verifying observations of storm location.
The terms used in the statistics tables are explained below.
- Possibly verified - number of forecasts fulfilling requirements 1, 2 and 4 below
- Detection rate - percentage of possibly-verified forecasts which also fulfil requirement 3 below
- Mean DX - mean of positional errors in the east-west direction
- Mean DY - mean of positional errors in the north-south direction
- Mean AT - mean of positional errors in the along-track direction
- Mean CT - mean of positional errors in the cross track direction
- Track skill - percentage skill of model forecast position against CLIPER
- Mean DPE - mean of direct positional errors
- Intensity skill - percentage skill of model in predicting strengthening or weakening (positive if better than chance).
All errors are measured in kilometres except where indicated.
- Observed maximum wind at least 31 knots at the verifying time.
- Forecast tropical cyclone centre equatorwards of 50 degrees.
- Forecast tropical cyclone centre relative vorticity above a critical limit for verification (1.30x10-4 per second).
- Observation within 6 hours of verifying time present for use in verification.
- Track skill is defined as:- (CLIPER DPE - Model DPE) / CLIPER DPE x 100%
- Positive skill indicates the model forecast is better than CLIPER
- Negative skill indicates the CLIPER forecast is better than the model
See below for a diagrammatic explanation of other error statistics and their sign conventions, for full details see the forecast tracks verification method.
Key to terms
|Forecast position: verifying against observation OB2|
|Direct positional error|
|Error in the east-west direction|
|Error in the north-south direction|
|Error in the along-track direction|
|Error in the cross track direction|
- DPE values are always positive
- DX errors are positive if the forecast position lies eastwards of the observed position
- DY errors are positive if the forecast position lies polewards of the observed position
- AT errors are positive if the forecast position lies ahead of the observed position along the tropical cyclone track
- CT errors are positive if the forecast position lies right of the observed track in the northern hemisphere and left of the observed track in the southern hemisphere.
Last updated: 24 August 2011