Improvement of Sub-seasonal Tropical Cyclone Forecasting Using Dynamical Downscaling

Abstract

Improving the sub-seasonal forecast of tropical cyclones (TCs) is challenging for the climate models. The characteristics of TCs in sub-seasonal forecasting with a dynamic model, the Global Seasonal Forecast System 5, 6 (GloSea5, 6), which is an operational Seasonal to Sub-seasonal forecast model of the Korea Meteorological Administration (KMA) were assessed for June–September (JJAS) from 1991 (1993) to 2010 (2016) over the western North Pacific (WNP). The performance of the dynamic model was examined for its ability to reproduce observed TC climatology as well as changes in TC genesis with the El Niño-Southern Oscillation (ENSO) and a 1998/1999 climate regime shift (e.g., frequency, genesis spatial distribution). The dynamic model showed reasonably skillful performance in predicting the frequency and genesis spatial distribution of TCs in climatology and both ENSO phases; this performance was best during periods of La Niña. Environmental fields related to TC genesis (e.g., sea surface temperature [SST], vertical wind shear [VWS], 850-hPa wind and relative vorticity) were also reasonably captured, despite some systematic biases in SST, low-level circulation, relative vorticity, and VWS. However, the Dynamic model underestimated TC frequency, TC track, especially in mid- latitude activities, TC lifetime, and intensity in all months, but the errors are particularly large in August. We aimed to determine whether employing dynamical downscaling with coupling a columnar ocean mixed layer model during August 2016, characterized by the lowest TC forecast skill in GloSea6, would enhance the sub-seasonal TC forecast. First, we developed a sub-seasonal to seasonal (S2S) dynamical downscaling modeling system and subsequently conducted experiments with it. We identified added value in applying dynamical downscaling, which directly improved the simulation of the tropical cyclone itself in terms of frequency, structure, intensity, and lifespan, albeit with some slight overestimation. It also improved the reproduction of the Indian monsoon and circumglobal teleconnection (CGT), which affect the western North Pacific Subtropical High (WNPSH). Improvements in reproducing the Indian monsoon and CGT pattern resulted in enhanced forecasts of the WNPSH, which, in turn, improved forecasts of TC activity, particularly in mid-latitudes and East Asia. In addition to the direct added value in sub-seasonal TC forecast through dynamical downscaling, this has indirectly enhanced TC activity forecasts by improving the reproduction of other variables (WNPSH, Indian monsoon, CGT) associated with TC activity. This developed a dynamical downscaling system for sub-seasonal TC forecast, which was validated in a 24-year experiment. We aimed to assess the forecast skill of the developed modeling system for sub-seasonal TC during the hindcast period spanning 24 years. We found added value in applying dynamical downscaling that the forecast skill of TC activity by enhancing the simulation of the environment field and teleconnection through verification of S2S dynamical downscaling modeling systems.