High-resolution time-lagged ensemble prediction for landfall intensity of Super Typhoon Haiyan (2013) using a cloud-resolving model

Abstract

The prediction of tropical cyclone (TC) intensity at landfall is crucial for regions vulnerable to high winds and storm surges, but its accuracy has experienced only limited improvement. At high resolution with a grid size of 2.5 km, the Cloud-Resolving Storm Simulator (CReSS) is applied to Super Typhoon Haiyan (2013), one of the strongest TCs to ever make landfall (at 85 m s−1 in peak wind speed and 900 hPa in central pressure. Predictions are made every 6 h using a time-lagged strategy so that the computational cost is relatively low. Averaging at 64 m s−1 and 925 hPa, our hindcasts during 4–7 November show large improvements in pre-landfall intensity over global models. Furthermore, when the previous CReSS result that best matches the observed intensity is used as the initial field, the predicted intensity consistently reaches 73–76 m s−1 and below 900 hPa. Thus, this approach is shown to produce more accurate TC intensity forecasts for storm-surge simulations for Haiyan.