Impact of evolution of initial conditions in tropical cyclone forecasting

Accurate prediction of tropical cyclones from very early stages of cyclogenesis has been a great challenge to the operational forecasters. This study examines the impact of evolution of initial conditions with time on improvements of track and intensity forecasts for a typical Bay of Bengal super cyclonic storm AMPHAN. For this purpose, the forecasts of various operational Numerical Weather Prediction (NWP) models have been analyzed from very early stage of the system. The analysis shows that there was wide variation of track forecasts among NWP models when there was very weak signature of the system in the initial condition. The track forecasts of NWP models gradually became closer to the eventual observed track, with the strengthening of the system in the analysis field. The spread of initial locations were 131.6 km, 49.5 km, 55.4 km, 91 km, and 22 km at the stages depression (D), cyclonic storm (CS), very severe cyclonic storm (VSCS), extremely severe cyclonic storm (ESCS), and super cyclonic storm (SuCS), respectively. The spread of forecast locations at 24 h forecast reduced from 128 km at the stage CS to 51 km at the stage SuCS; at 48 h, from 289 to 42 km from low (L) to SuCS; at 72 h, from 496 to 127 km from L to ESCS; at 96 h, from 736 to 250 km from L to CS, and at 120 h, from 890 to 309 km from the stage L to D as initial conditions. The spread of landfall point(time) forecast errors of NWP models were 445 km (23 h), 520 km (18 h), 177 km (13 h), 118 km (8 h), and 79 km (5 h) for the initial stages D, CS, VSCS, ESCS, and SuCS, respectively. Result shows that small errors in the weaker initial conditions contributed large errors in subsequent track forecasts and impact of initial error minimized at stronger stages. No such trend in error is found in intensity forecast of the models.