Abstract
The uncertainties associated with the mapping of cyclone numbers and frequencies are analyzed using the 42-yr winter climatology of cyclone tracks derived from 6-hourly NCEP–NCAR reanalysis. Tracking is performed using an automated procedure, based on computer animation of the sea level pressure fields. Uncertainties in the mapping result from an incomplete catchment of cyclones by the grid cells: the coarse temporal resolution of data causes fast-moving storms to skip grid boxes. This introduces error into estimates of cyclone frequency, with cyclone counts systematically underestimated. To minimize these biases, it is possible to simulate higher temporal resolution of the storm tracks by linear interpolation applied to the original tracks. This simple procedure reduces bias in estimates of both cyclone frequencies and numbers and enables quantitative estimation of errors. Errors in cyclone counts are estimated over the Northern Hemisphere for different time resolutions and different grids. Standard errors in cyclone frequencies increase from 5%–15% to as much as 50% as the original temporal resolution of the storm tracks decreases from 6 to 24 h. Mapping the results on circular grids reduces these errors. Appropriate grid sizes for different temporal resolutions of the storm tracks are recommended to minimize the uncertainties in storm occurrence mapping.
Corresponding author address: Sergey Gulev, P. P. Shirshov Institute of Oceanology, RAS, 36 Nakhimovsky Ave., 117851 Moscow, Russia. Email: gul@gulev.sio.rssi.ru