Premium
Numerical prediction of tropical cyclogenesis part I: Evaluation of model performance
Author(s) -
Liang Mei,
Chan Johnny C. L.,
Xu Jianjun,
Yamaguchi Munehiko
Publication year - 2021
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.3987
Subject(s) - cyclogenesis , tropical cyclone , climatology , tropical cyclogenesis , monsoon trough , ocean gyre , numerical weather prediction , cyclone (programming language) , african easterly jet , meteorology , tropical wave , environmental science , tropical cyclone forecast model , geology , geography , subtropics , computer science , field programmable gate array , fishery , biology , computer hardware
This is a two‐part study investigating the numerical prediction of tropical cyclogenesis. This paper (Part I) presents a comprehensive statistical assessment of the performance of the high‐resolution European Centre for Medium‐Range Weather Forecasts (ECMWF) deterministic forecast in predicting tropical cyclone (TC) genesis over the western North Pacific (WNP) between 2007 and 2018 using The International Grand Global Ensemble (TIGGE) data. Each genesis forecast at each lead time is classified into one of these categories: well predicted (WP), early formation (EF), late formation (LF), or failed prediction (FP). Based on the synoptic patterns at the actual genesis time, the genesis of a particular TC is grouped into one of five flow patterns: monsoon shear line (SL), monsoon confluence region (CR), monsoon gyre (GY), easterly wave (EW), and preexisting tropical cyclone (PTC). Any case that does not fall in one of these patterns is labeled as unclassified flow pattern (UCF). Overall, the prediction skill of SL cases is the highest, followed by CR, GY, and PTC, while the EW and UCF patterns have the lowest skill. No significant improvement in the model performance is found during the 12 years, probably partly because of the low percentage of cases with the SL pattern in the later years. The prediction skills of TC genesis in May, June, September, and October are the highest, while those in July, August, and November are the lowest. TCs that form further to the south are easier to predict in the SL pattern. In Part II, we will examine the model predictions of the dynamic and thermodynamic variables in the WP cases to identify the physical processes responsible for TC genesis.