Impact of steering flow on tropical cyclone predictability

Although tropical cyclone (TC) track forecasts in numerical weather prediction models have improved considerably over the past few decades, there remain cases with large uncertainty. Typhoon Haiyan (2013) and Typhoon Hagupit (2014) are examples of two high-impact storms where, despite similarities in the observed track and intensity, the predictability of the storms differed greatly. Ensemble forecasts showed large uncertainty in the track of Hagupit, whereas the ensemble spread for Haiyan was considerably less. Using the Met Offce's Unifed Model, 5-day global and convection-permitting (CP) ensemble forecasts are analysed for both storms. Global forecasts show Haiyan was located on the southern periphery of the subtropical high and embedded in a strong easterly flow. In contrast, the steering flow of Hagupit was weak as the TC became located between two anticyclones. We show that Hagupit's position between the anticyclones, the strength of the anticyclones, the interactions between the TC outflow and its environment, and the upper-level geopotential height directly to the south of the TC contributed to whether Hagupit would make landfall over the Philippines or turn to the north. The track forecasts in the CP ensembles of both storms produced errors which were not present in the global forecasts. For Haiyan, CP forecasts predicted the motion of the TC to be too slow, whilst for Hagupit the CP forecasts predicted the TC to make a systematic south-west turn, away from the actual storm path. For a third high-impact TC, Hurricane Florence (2018), CP forecasts predicted the storm to move too far to the west before turning north. Analysis of these forecast busts shows differences in how TCs interact with upper-level steering winds, particularly in periods of strong vertical wind shear, can cause differences in the global and CP track forecasts.