GOES-East channel-3 (water vapor) imagery and NCEP global analyses are used as input to an algorithm to estimate the probability of tropical cyclone formation within the next 24 hours in 5 by 5 degree latitude/longitude areas from 0 to 45 degrees north and 140 to 10 degrees west.
The probability estimates are based upon the following input parameters:
- CLIMATOLOGICAL FORMATION PROBABILITY: This was derived from the NHC Atlantic and east Pacific best track files from 1949-present. Formation was defined as the first point for each storm in the best track after all extratropical cases were excluded.
- PERCENT LAND: The percent of each 5 by 5 degree square over land
- DISTANCE TO PREXISTING STORM: The distance from the center of each 5 by 5 degree square to existing tropical cyclones.
- CLIMATOLOGICAL SST: The maximum SST in each 5 by 5 degree area from the Levitus climatology.
- VERTICAL SHEAR: The 850-200 hPa shear determined by averaging the 850 hPa and 200 hPa wind vectors over each 5 by 5 degree area, and then calculating the magnitude of the shear vector.
- 850 HPA CIRCULATION: The 850 hPa circulation determined from a line integral of the wind component tangent to the boundary of each 5 by 5 degree area.
- VERTICAL INSTABILITY: The vertical average temperature difference between the equivalent potential temperature of a parcel lifted from the surface to 200 hPa, and the saturation equivalent potential temperature of the environment, for each 5 by 5 degree area.
- GOES COLD PIXEL COUNT: The percent of GOES-east channel 3 pixels colder than –40 degree C in each 5 by 5 degree area. All full disk images within 3 hours after and 6 hours before each synoptic time are include, so that this parameter represents the amount of sustained deep convection.
- CLOUD-CLEARED WATER VAPOR BRIGHTNESS TEMPERTATURE: The average channel 3 brightness temperature, after the cold pixels in parameter 8 above have been eliminated. This parameter is a measure of mid- to upper-level moisture.
The formation probability is calculated by a screening step, followed by a discriminant analysis.
In the screening step, the probability of formation is set to zero for 5 by 5 degree areas where formation almost never occurred based upon the best track data (1949-present), NCEP reanalysis fields (1981-present), and GOES climatology (1995-present). The following areas are eliminated:
- Maximum climatological SST < 21 degree C
- Latitude < 5 degrees north
- Areas that are 100% over land
- Areas that already contain a tropical cyclone
- 850 hPa Circulation < -5 kt
- Vertical Shear > 65 kt
- Vertical Instability < -8 degrees C
- GOES cold pixel count = 0%
- GOES brightness temperature > -23 degrees C
The parameters for all 5 by 5 degree areas that passed the screening tests for the May-December, 1995-present sample were used as input to a linear discriminant analysis procedure (about 350,000 inputs). This procedure provides a discriminant function that can be converted to a 24-hour formation probability. Separate discriminant analyses were performed for the East Pacific and Atlantic cases, but the results were very similar for each. In both cases, the probability depends on the following input at the post-screening 5 by 5 degree areas.
DISCRIMINANT ANALYSIS INPUT:
- Climatological formation probability
- 850 hPa Circulation
- GOES Cold Pixel Count
- Distance to Pre-Exisiting Tropical Cyclone
- Vertical Shear
- Percent of Area Over Land
To provide some time continuity of the product, the formation probability and the primary parameters used in the screening and discrimination analysis are summed (probabilities) or averaged (all other parameters) over eight sub-basins. Time series products comparing the current and climatological parameters over the sub-basins are provided on the web site.