ECMWF is augmenting the tropical cyclone track products that it makes available for dissemination in BUFR format. They now include the tracks of tropical cyclones that develop (i.e. undergo genesis) during the forecast integrations (the 'Genesis category'), as a supplement to the tracks that we had already been providing for tropical cyclones that exist in the analysis, i.e. where there is an official observation report for the analysis time (the 'Pre-existing category'). These products are provided for both high-resolution (HRES) and ensemble (ENS) forecast systems for the period T+0 to T+240h.
ECMWF's TC track products contain the following variables: time, position (latitude/longitude in degrees), central pressure (hPa), intensity (i.e. maximum 10m wind speed in m/s in a 750 x 750km box centred on the TC location) and position of that 10m wind speed maximum (latitude/longitude). These are now provided for every tropical cyclone identified and tracked in every ENS forecast, and in HRES.
Naming convention for the 'Genesis category' of tropical cyclones
A "Storm Identifier" label is assigned to each separate TC feature identified in a forecast. This ID is formed by a number (between 70 and 99) and a letter that indicates which basin the cyclone belongs to (among eight basins). Strictly the letter shows the basin in which genesis occurred, so if a TC transfers from one basin into another, the identifier of the original basin will be retained. The range of numbers used was chosen to distinguish from real TC IDs (official reports) at any time. The basins are identified by the following suffixes (letters) which are appended to the end of each TC number:
W - North West Pacific
C - North Central Pacific
E - North East Pacific
L - North Atlantic
A/B - North Indian Ocean (A for Arabian Sea and B for Bay of Bengal)
U - Australia
P - South Pacific
For example, the "Storm Identifier" 71W represents the 2nd new TC feature identified in the North West Pacific basin (W) in forecasts from a particular ENS member or HRES. In general there is no guarantee that '71W' (for example) in one ENS or HRES run is 'the same' feature as '71W' in another ENS or HRES run from the same data time, though for lower numbers they are a bit more likely, perhaps, to be 'the same'. Strictly speaking however one can never say that a certain newly developing cyclone in one member is the same as a certain newly developing cyclone in another member. One could in principle approximate this type of association using a cluster technique though we do not do that. Similarly any apparent association between a cyclone in a certain member from a certain data time and another cyclone in that same member from a later data time that has the same ID will be purely coincidental. This is because we do not attempt to cluster across different data times either.