Normalized Convective Characteristics of Tropical Cyclone Rapid Intensification Events in the North Atlantic and Eastern North Pacific

Despite recent increases in operational forecast skill of tropical cyclone (TC) intensity, there still exists room for improvement. This is especially true for those TCs that undergo rapid intensification (RI). Although both statistical and dynamical numerical modeling have demonstrated skill in the prediction of RI episodes, operational forecasts of RI suffer from a relatively low probability of detection and an undesirably high false alarm rate.

In hopes of improving these shortcomings, this study assesses the relationship between TC convective characteristics and TC intensity change, using infrared and passive microwave satellite imagery of TCs in the North Atlantic and eastern North Pacific basins from 1989 to 2016. TC intensity change episodes were placed into one of four groups: RI, slow intensification (SI), neutral (N), and weakening (W). To account for differences in the distributions of TC intensity among the intensity change groups, a novel convective normalization technique is introduced. The normalization technique allows for the analysis of anomalous TC convective characteristics and their possible role in TC intensity change.

A composite analysis of normalized convective parameters shows anomalously vigorous convection in the upshear quadrants is associated with increased rates of intensification, especially RI. For RI episodes in the North Atlantic basin, the emergence of anomalously vigorous shallow convection precedes vigorous deep convection, suggesting deep convection is a symptom of, rather than a precursor to, RI. In the eastern North Pacific basin, the anomalous vigor of both shallow and deep convection increases in tandem near the onset of RI. Normalized convective parameters can be utilized to skillfully predict episodes of RI, as the forecast skill of RI episodes using solely normalized convective parameters is comparable to the forecast skill of RI episodes by current operational statistical models.