A multi-model consensus forecast technique for tropical cyclone intensity based on model output calibration

Forecast errors of the tropical cyclone (TC) intensity in the western North Pacific region are analyzed for seven operational numerical weather prediction models during 2010-2012. It is found that the intensity forecast error is significantly related not only to initial error, but also to initial TC intensity, size, and translation speed. Other factors highly related to the forecast errors include the environment pressure, vertical wind shear and maximum potential intensity. The stepwise regression technique is applied to set up model forecast error estimation equations, which can be used to calibrate the model outputs. The independent experiments exhibit that the calibrated model forecasts have significant skill over the original model outputs. A multi-model consensus forecast technique for TC intensity is then developed based on the calibrated model outputs and it shows a 28% (15%-20%) skill at 12 h (24-72 h) over the climatology and persistency technique forecasts for TC intensity. Such a consensus technique is much more skillful than the consensus based on the original model outputs and has the potential to be applied in real time operation.