Evaluation of Tropical Cyclone Forecasts from Operational Global Models Over the Western North Pacific in 2013

Abstract

ABSTRACT Tropical cyclones (TCs) forecasts from seven global models in 2013 were assessed to study the current capability of track and intensity forecast guidance over the western North Pacific. Analysis of along- and cross-track error revealed stepped decreases in the values of each quantile at each lead time level by showing the annual track error distribution from 2010 to 2013, particularly in the ECMWF-IFS, NCEP-GFS and UKMO-MetUM models. The TC propagation direction was much easier to handle for most of the global models; however, the propagation speed seemed to be more closely linked to the inner-core dynamics and thus processes that take place at smaller spatial scales. A new model evaluation tool, 'track error rose', was used to analyze the models' systematic error in the track forecast using the same concepts as the 'wind rose'. The results showed that as the lead time increased, most of the global models forecast a TC moving speed that was slower than observations and the largest track error often appeared around the rear direction of the observation position. Another new model evaluation tool, the Taylor diagram, was used to evaluate the intensity predictions from the global models. A Taylor diagram provides a way of plotting standard deviation, centered root mean square, and the correlation coefficient on a two-dimensional graph, indicating how closely a predicted TC intensity matches observations. This made it easy to distinguish the intensity forecast performance of the seven global models and determine which models were in relatively good agreement with observations. Furthermore, it also provided a statistical measure of the correlation between modeled and observed TC intensity, offering a practical way of assessing and summarizing model capability.