The tracking scheme dependence of simulated future tropical cyclone activity trends

The nature of possible future changes in tropical cyclone (TC) activity is of great interest not only scientifically, but to all of society. In the absence of a general climate theory of TC formation, climate models are the primary tool available for investigating the problem. However, the identification of TCs in model data at moderate resolution is non-trivial, and numerous different schemes have appeared for the detection and tracking of tropical cyclone-like vortices.

We here examine the influence of different tracking schemes on detected TC activity and trends in the Hurricane Working Group (HWG) experiments. These are idealised atmospheric global climate model (AGCM) experiments aimed at determining and distinguishing the effects of increased sea-surface temperature (SST) and of other increased CO2 effects on TC activity. The four HWG experiments simulate present day climate; altered climate with 2K globally increased SSTs; with doubled carbon dioxide concentration; and with increased SST and carbon dioxide in combination. We analyse results in four different AGCMs with resolutions ranging from 1° to around 0.5°.

We apply two different tracking schemes to these data. An altered version of the CSIRO tracking scheme is used across all four models and the Zhao tracking scheme is used for all but one model. We also analyse the tracks provided by each modelling group. The tracking schemes used by the modellers varies from group to group.

Our results indicate reasonable agreement between the different tracking methods in most cases, with some models and experiments showing better agreement across schemes than others. In the case of the combined SST/CO2 increase experiment, the models and tracking schemes are near-unanimous on a decreasing trend in TC frequency compared to the present day experiments, accompanied in most cases with an increase in intensities.

More divergent results are obtained for the separate SST and CO2 experiments. However, we find that the bulk of the disagreement between tracking schemes is due to differences in duration, windspeed, and formation-latitude thresholds rather than fundamental differences in TC identification methods. Filtering tracks to homogenise these factors removes most disagreement between tracking schemes within models. Disagreement between the models themselves remains, but this is unlikely to be ascribable to the tracking schemes. The good agreement between schemes within models after homogenisation indicates that the selection of objective thresholds is the key factor in obtaining meaningful, reproducible results when tracking TCs in climate model data at these resolutions, and that more basic differences between tracking methods have a less significant impact on the trends in activity detected.