Variability of tropical error characteristics in the NCEP (MRF) lower tropospheric wind forecasts during two contrasting seasons

Abstract

This study has addressed the issues of the diagnosis and the variability of the tropical error and its growth rate, and the several dynamical processes involved in the error growth of a forecast model during two contrasting seasons. The geographical distributions of the systematic and random error energy, their growth rates and the different dynamical terms of the error energy growth rate equation in the NCEP (MRF) model are presented in this paper. The data used for this purpose are 1–7 day forecasts of wind fields for the consecutive boreal winter (DJF, 2000–01) and boreal summer (JJA, 2001) seasons at the 850 hPa level. The predictability limit in certain geographical regions over the tropical belt (30° S–30° N) and in the global tropics as a whole has been estimated by calculating the predictability score in the wind field. The major findings of this paper are as follows: The variation found in the geographical distribution of the systematic and random error energetics during the two contrasting seasons has been well established by the statistical investigation. The predictability limit of the winds in the Indian region varied from a 2-day to 1-day forecast during the boreal winter and summer seasons, respectively. For the tropical region as a whole, the limit of predictability was found to be 2-day and 1-day forecast in the zonal and meridional wind, respectively, during both seasons. The location of maximum in systematic error energy shifted from the south to the north of the equator in accordance with the seasonal shift of the ITCZ during the boreal winter and summer months, respectively. Random error energy was present over land and oceanic regions at 850 hPa with large errors over the oceans, indicating that insufficient data coverage might be the major cause for the generation of the tropical random error. In terms of the geographical distribution, each dynamical process contributing to the error growth rate budget showed a maximum value at the location where the error was found to be large during both the seasons.