| description abstract | ropical cyclone (TC) intensity prediction, especially in the warning time frame of 24?48 h and for the prediction of rapid intensification (RI), remains a major operational challenge. Sea surface temperature (SST) based empirical or theoretical maximum potential intensity (MPI) is the most important predictor in statistical intensity prediction schemes and rules derived by data mining techniques. Since the underlying SSTs during TCs usually cannot be observed well by satellites because of rain contamination and cannot be produced on a timely basis for operational statistical prediction, an ocean coupling potential intensity index (OC_PI), which is calculated based on pre-TC averaged ocean temperatures from the surface down to 100 m, is demonstrated to be important in building the decision tree for the classification of 24-h TC intensity change ?V24, that is, RI (?V24 ≥ 25 kt, where 1 kt = 0.51 m s?1) and non-RI (?V24 < 25 kt). Cross validations using 2000?10 data and independent verification using 2011 data are performed. The decision tree with the OC_PI shows a cross-validation accuracy of 83.5% and an independent verification accuracy of 89.6%, which outperforms the decision tree excluding the OC_PI with corresponding accuracies of 83.2% and 83.9%. Specifically for RI classification in independent verification, the former decision tree shows a much higher probability of detection and a lower false alarm ratio than the latter example. This study is of great significance for operational TC RI prediction as pre-TC OC_PI can skillfully reduce the overestimation of storm potential intensity by traditional SST-based MPI, especially for the non-RI TCs. | |
