Premium
Selective ensemble‐mean technique for tropical cyclone track forecast by using time‐lagged ensemble and multi‐centre ensemble in the western North Pacific
Author(s) -
Du Yugang,
Qi Liangbo,
Cao Xiaogang
Publication year - 2016
Publication title -
quarterly journal of the royal meteorological society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.744
H-Index - 143
eISSN - 1477-870X
pISSN - 0035-9009
DOI - 10.1002/qj.2838
Subject(s) - environmental science , ensemble forecasting , meteorology , tropical cyclone , ensemble average , climatology , track (disk drive) , precipitation , quantitative precipitation forecast , computer science , geography , geology , operating system
This article investigates the impacts of both the time‐lagged ensemble technique and the multi‐centre ensemble prediction system (EPS) on tropical cyclone (TC) track forecasts. Four EPSs from European Centre for Medium‐range Weather Forecasts (ECMWF), Japan Meteorological Agency (JMA), National Centers for Environmental Prediction (NCEP) and China Meteorological Administration (CMA) participating in the TIGGE project are considered, and 60 TCs in the western North Pacific from 2010 to 2012 are investigated. The time‐lagged ensemble uses single‐centre EPS (SCE) initialized at different times to increase ensemble members for specific forecast times. Both the means of selected EPS members (SEAV) method and the ensemble‐mean method are applied to the time‐lagged ensemble, respectively. Verification results show that the time‐lagged ensemble yields marginal changes in forecast position errors compared with its zero‐lag counterparts for all SCEs by applying the SEAV method. The mean forecast position errors by the ensemble‐mean method are greater when adding previous data compared with the SEAV method. A variety of multi‐centre EPSs (MCEs) composed of SEAV‐integrated SCEs are investigated. These MCEs, labelled according to their component SCEs, are EC&NCEP, EC&CMA, EC&JMA, EC&NCEP&CMA, EC&NCEP&JMA and EC&NCEP&CMA&JMA. The EC&NCEP MCE, for example, is a SEAV‐integrated combination of the ECMWF and NCEP ensemble TC forecast tracks. It is found that MCE track forecasts are more accurate than SCE track forecasts. EC&NCEP has the best performance among all MCEs. In addition, MCEs are compared with an equally weighted consensus of deterministic model forecasts (MCDs). The MCEs outperform their corresponding MCDs at 24 h. Sensitivity tests on the weights of component SCEs in SEAV‐derived MCEs are conducted. For lead times within 72 h, if the component SCEs in an MCE have comparable performances, the MCE with equal weights of SCEs will be more accurate than those of non‐equal weights of SCEs. Otherwise, non‐equal weights of SCEs are more suitable for MCEs. However, for lead times longer than 72 h, the above result is not convincing.