Machine learning in calibrating tropical cyclone intensity forecast of <scp>ECMWF EPS</scp>

Intensity prediction of tropical cyclones (TC) has been one the major challenges for operational forecast and warning service, as well consequential assessment impacts including high winds, storm surge heavy rainfall caused by TC. With advances in global numerical weather (NWP) modelling systems, TC track intensity forecasts medium range are available every 6 or 12 h, ensemble system (EPS) outputs provide various scenarios producing probabilistic forecasts. The from EPS European Centre Medium-Range Weather Forecasts (ECMWF) shown systematic negative biases, although performance is better than other models general. A machine learning model based on XGBoost, a decision-tree-based algorithm, introduced this paper to post-process ECMWF generate an improved intensity. predictors such selected percentiles members maximum wind minimum pressure previous cases were applied XGBoost calibrated TCs. Verification was made using TCs over western North Pacific during 2016–2019. It found that biases HRES can be reduced with improvement overall accuracy.