Extended Prediction of North Indian Ocean Tropical Cyclones Using the ECMWF Variable Ensemble Prediction System - Monitoring and Prediction of Tropical Cyclones in the Indian Ocean and Climate Change - page 117

ensemble mean's 24-hr, 48-hr and 72-hr track forecast error is on an average
10%, 19% and 27% smaller than NCEP's GFS. Since the linear track error
growth per day for the VarEPS ensemble mean (41 n mi/day) is considerably
smaller than the GFS (66 n mi/day), greater track forecast utility is obtained at
longer lead-times with the VarEPS in comparison to other model forecasts.
Figure 4b compares the VarEPS control and ensemble mean absolute wind
error to other forecast models and the JTWC. Generally, for the 2007-2010
period, the VarEPS begins with much higher initial error than any other forecast
model but also shows the lowest intensity error growth through the first 72 hrs.
In addition, the interannual variation of post-genesis intensity forecasts reflects
a more substantial improvement for the 2008-2010 period compared to 2007.
4. Conclusions
Based on this evaluation of the VarEPS TC forecasts, it appears feasible for
warning agencies in the NIO to begin providing a probabilistic TC formation
outlook that assesses the potential for TC development through a lead-time of
seven days. When the probability of formation is within moderate (30-60%)
levels, the VarEPS's probability of detection will average around 60% with a
false alarm rate of about 30% for a lead-time of seven days. In addition, since
the distribution of the VarEPS forecasts provides a dynamical measure of the
forecast uncertainty in the atmosphere's future state (Dupont et al., 2011), some
TCs will be more predictable than others. Therefore, operational forecasts could
include a probabilistic outlook including both TC track and maximum intensity
derived from the VarEPS. Although the VarEPS tends to be slightly
underdispersive at longer forecast lead-times (Majumdar et al., 2010), some
additional statistical post-processing steps including bias-correction and
probability calibration could be incorporated to ensure that the final forecast
track and intensity probabilities are well-conditioned relative to observations.
REFERENCES
Barkmeijer, J., Buizza, R., Palmer, T.N., Puri, K. and Mahfouf, J.-F. (2001). Tropical
singular vectors computed with linearized diabatic physics. Quart. J. Roy. Meteor.
Soc. , 127: 685-708.
Belanger, J.I., Webster, P.J., Curry, J.A. and Jelinek, M.T. (2012). Extended Predictions
of North Indian Ocean Tropical Cyclones. Wea. Forecasting , in review.
Buizza, R. and Palmer, T.N. (1995). The singular vector structure of the atmospheric
global circulation. J. Atmos. Sci. , 52: 1434-1456.
Dupont, T., Plu, M., Caroff, P. and Faure, G. (2011). Verification of ensemble-based
uncertainty circles around tropical cyclone track forecasts. Wea. Forecasting , 26:
664-676, doi: 10.1175/WAF-D-11-00007.1.
Hart, R.E. (2003). A cyclone phase space derived from thermal wind and thermal
asymmetry. Mon. Wea. Rev. , 131: 585-616.