Seasonal tropical cyclone predictions using optimized combinations of ENSO regions: Application to the coral sea basin

Publication Type:
Journal Article
Citation:
Journal of Climate, 2014, 27 (22), pp. 8527 - 8542
Issue Date:
2014-01-01
Full metadata record
Files in This Item:
Filename Description Size
10.1007%252Fs11186-014-9212-9.pdfPublished Version372.25 kB
Adobe PDF
© 2014 American Meteorological Society. This study examines combining ENSO sea surface temperature (SST) regions for seasonal prediction of Coral Sea tropical cyclone (TC) frequency. The Coral Sea averages ~4 TCs per season, but is characterized by strong interannual variability, with 1-9 TCs per season, over the period 1977-2012. A wavelet analysis confirms that ENSO is a key contributor to Coral Sea TC count (TCC) variability. Motivated by the impact of El Niño Modoki on regional climate anomalies, a suite of 38 linear models is constructed and assessed on its ability to predict Coral Sea seasonal TCC. Seasonal predictions of TCC are generated by a leave-one-out cross validation (LOOCV). An important finding is that models made up of multiple tropical Pacific SST regions, such as those that comprise the El Niño Modoki Index (EMI) or the Trans-Niño Index (TNI), perform considerably better than models comprising only single regions, such as Niño-3.4 or Niño-4. Moreover, enhanced (suppressed) TC activity is expected in the Coral Sea when the central Pacific is anomalously cool (warm) and the eastern and western Pacific are anomalously warm (cool) during austral winter. The best crossvalidated model has persistent and statistically significantly high correlations with TCC (r > 0.5) at lead times of ~6 months prior to the mean onset of the Coral Sea TC season, whereas correlations based heavily on the widely used Niño-3.4 region are not statistically significant or meaningful (r = 0.09) for the same lead times. Of the 38 models assessed, several optimized forms of the EMI and of the TNI perform best.
Please use this identifier to cite or link to this item: