Evolution of tropical cyclone forecasts of dynamical - Statistical Cyclone Prediction System (CPS) over the North Indian Ocean during the decade (2010-2019)

Abstract

Tropical cyclone (TC) forecasting over the North Indian Ocean (NIO) has been improved significantly during past decade (2010-2019). The improvement is largely attributed to the numerical weather prediction (NWP) models and dynamical-statistical Cyclone Prediction System (CPS). The CPS has five components namely,          (i) Genesis Potential Parameter (GPP), (ii) Multi-Model Ensemble (MME) technique for track prediction, (iii) Statistical Cyclone Intensity Prediction (SCIP), (iv) Rapid intensification (RI) and (v) Decay model for intensity forecast after landfall. Performance of CPS and NWP models is assessed to review the improvement during the decade. Forecast analysis revealed low false alarm ratio (0.13), high probability of detection (0.95) and high critical success index (0.84) for GPP. Mean track error of MME was ranged from 67 km at 24h to 246 km at 120 h and about 30% less than NWP models. The MME error has also reduced by 52% to 55% for 24 h to 48 h and 41% to 24% for 72 h to 96 h forecast during 2010-2019. Mean landfall point error was ranged from 31 km at 24 h to 127 km at 120 h and landfall time error was ranged from 2.2 h at 24 h to 8.1 h at 120 h. Mean intensity error of SCIP model was ranged from 8.3 kt at 24 h to 12.6 kt at 120 h. Probabilistic rapid intensification index (RII) achieved a good Brier score (BS) 0.079 for RI forecasting. Mean decaying intensity error (after landfall) was ranged from 6.6 kt at 6 h to 3.3 kt at 24 h. There has been improvement of forecasts for NWP models also but MME outperformed all models. Results demonstrate the role of CPS for improvement of cyclone forecast over the NIO in past decade.