A Statistical Intensity Prediction Scheme for Tropical Cyclones in the Western North Pacific

A Western North Pacific Tropical cyclone (TC) Intensity Prediction Scheme (WIPS) is developed based on samples from 1996 to 2002 using the stepwise regression technique, with the western North Pacific being divided into three forecast regions: the region near the coast of East China (ECR), the South China Sea region (SCR), and the far ocean region (FOR). Only the data over the ocean of TCs with tropical storm intensity or higher are used to set up the prediction equations (12-72 h at 12 h interval). Potential predictors include the climatology and persistence factors, synoptic environment conditions, potential intensity of the TC and proximity of the TC to land. The mean absolute errors (MAEs) of all samples are 2.6, 4.2, 5.4, 6.1, 6.1, and 6.3 ms-1 for the 12, 24, 36, 48, 60 and 72 h predictions respectively.

Analyses on the variances explained by the selected predictors suggest that significant predictors are generally different for different forecast regions, except the potential intensity and the proximity of the TC to land which are significant in almost all the equations. Other significant predictors for different region include the vertical wind shear for ECR, the 500-hPa geopotential height anomaly at the TC center and the zonal component of TC speed in the next 6 h of initial time for SCR, and the TC intensity change 12 or 24 h prior to the initial time and the longitude of TC center at initial time for FOR.

To verify the scheme, independent tests are carried out for TCs in 2004 and 2005. The MAEs are 3.1 (460), 4.7 (423), 6.2 (380), 6.9 (353), 7.6 (315), and 7.8 ms-1 (279) (the number in bracket is sample size) for 12, 24, 36, 48, 60 and 72 h predictions respectively, which one only slightly larger than those of the dependent samples. The SCR has the highest MAEs, implying greater difficulty in the intensity prediction of TCs in this region although the scheme performs relatively well for TCs with severe tropical storm intensity or at intensifying stage. The performance of the scheme is also compared to a current operated climatology and persistence prediction scheme on the Shanghai Typhoon Institute. Positive skills are obtained at all lead times with a skill score near or over 20 % after 36 h, implying that the WIPS scheme performs especially well at longer lead times.