Abstract
An objective index was proposed to determine the intensity of typhoons in this study. This was achieved using an image edge processing technique to examine meaningful discontinuity characteristics and thereby calculate the gradient of brightness temperature in satellite infrared images. By taking the typhoon centre as a reference point, the angle between the position vector and the gradient vector was defined as the deviation angle. Following this definition, the probability density and standard deviation of the deviation angle may be derived. After creating a scale from 1 to 0 (0-1) to, respectively, represent the maximum and minimum values of the probability density (standard deviation), this research proposed a non-dimensional typhoon intensity (TI) index. Analysis results reveal a high accuracy when the TI index was used to objectively measure TI. The bias, average error, root mean square error, and R-2 value reached 0.6, 3.5, 4.8 m s(-1), and 0.89, respectively. Meanwhile, various evaluation parameters in assessing the forecasting skill were also employed, where a specific 'yes' and 'no' threshold for each typhoon stage was established. The ratio of the number of correct determination to the number of events for a specific typhoon stage was 0.74 (mild), 0.76 (moderate), and 0.89 (severe), respectively, for 557 infrared images of five validation typhoon cases in 2011. The results demonstrated that the TI index technique had good performance in assessing the TI even during typhoon stage changes.