Dynamic Cone of Uncertainty Using Ensemble Forecasting of Tropical Cyclones

Abstract

Many operational forecasting centres of tropical cyclones (TCs) issue a static Cone of Uncertainty (COU) to convey the uncertainty associated with the forecast track. This COU is based on the climatological distribution of forecast position errors. The uncertainty information from an ensemble prediction system can help in producing a dynamic COU. The objective of the present work is to build a dynamic COU using multiple member forecasts from India’s National Centre for Medium Range Weather Forecasting (NCMRWF) Global Ensemble Prediction System (NEPS-G), where the radius at each forecast is determined such that it includes 67% of the ensemble members. This dynamic COU is then compared against a static COU constructed using the fixed radii prescribed by the India Meteorological Department (IMD), which are derived from climatological forecast position errors of previous years. All TCs for the period 2019–2021 over the North Indian Ocean (NIO) have been considered in the present study. At shorter lead times (till 18 h), static probability circles are too small to capture most of the best tracks. The dynamic circles show higher detection rate than the static circles till at least 72 h forecast lead time. The static circles outperform the dynamic circles at longer lead times due to large errors in ensemble mean and inadequate ensemble spread. The dynamic circles, over both Bay of Bengal (BoB) and Arabian Sea (AS), perform better till 72-h lead time for the straight and recurving TCs. At longer lead times (84-h onwards), for BOB cyclones, static circles perform better but for AS cyclones, dynamic circles are slightly more effective. For storms with severe cyclonic and higher intensity, dynamic circles are more effective during 18 to 72-h forecasts. During the post-monsoon season for all lead times (except 120-h) best tracks lie within dynamic circles more often than static circles.