[A-AS07] Analyses and Prediction of Extreme Weather and Climate in Southeast Asia

[AAS07-P01] Preliminary Evaluation of Multi-Week Tropical Cyclone Forecasts in the Philippines

*Marcelino Q. Villafuerte II¹, Maria Czarina M. Tierra¹ (1.Philippine Atmospheric, Geophysical and Astronomical Services Administration)

Keywords:Multi-week forecasting of tropical cyclones, western North Pacific, the Philippines

With the aim to improve the tropical cyclone (TC) forecast guidance at the extended-range timescales in the Philippines, this study examines the predictability of TC activity within the Tropical Cyclone Information Domain (bounded by 110°-160°E and 0°-40°N), which is operationally being used by the Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA), up to four (4) weeks in advance. Hindcast data from three ensemble prediction systems, namely: the NCEP Coupled Forecast System version 2 (NCEP-CFSv2), the European Centre for Medium-Range Weather Forecasts Ensemble Prediction System (ECMWF-EPS), and the NCEP Global Ensemble Forecast System version 12 (NCEP-GEFSv12) covering the period from 1 June 2020 to 31 October 2021 were used. TC-like vortices (TCLV) were detected and tracked based on a set of dynamic and thermodynamic conditions in the domain, and weekly information on the forecast TC tracks were extracted and evaluated against the Joint Typhoon Warning Center (JTWC) best track dataset. Verification showed that the ECMWF-EPS model performed best in Week-1 (1-8 days of forecast period) and Week-2 (9-15 days of forecast period), wherein its hit rates were relatively high and false alarm rates were low. Beyond Week-2, ECMWF-EPS obtained lower hit rates but consistently maintained the lowest false alarm rates among the models. In contrast, the CFSv2 model acquired higher hit rates (but with notable high false alarm rates) in Week-3 (16-22 days of forecast period) and Week-4 (23-30 days of forecast period), while the GEFSv2 model performed poorly in Week-4. Comparison of the forecast and observed tracks of westward-moving TCs revealed that the forecast tracks have a significant poleward bias in latitudinal position starting from Week-2 (95% confidence level) and that most of them tend to propagate slower than the observed tracks.