A statistical analysis on the predictability of tropical cyclogenesis
Dianna N. Nelson, Univ. of Wisconsin, Madison, WI; and M. C. Morgan
This study examines the feasibility of using a linear statistical model to forecast the probability of tropical cyclogenesis by using a linear discriminant analysis (LDA) technique, which acts to find the greatest separation between the features of two predetermined groups. The focus of this study concerns the development of depressions located in the eastern Pacific basin. All vortices that formed during the peak of the eastern Pacific hurricane season (1 July - 31 September) in 2005, 2006 and 2007 were tracked and grouped into one of two categories: developing or non-developing. LDA was then applied to test various linear combinations of vortex-centered variables such as vorticity, vertical shear, and deformation as well as larger scale factors including the Madden Julian Oscillation (MJO). The combinations that demonstrated the greatest skill in correctly grouping individual cases into the developing and non-developing categories were then used to hindcast the development of vortices into tropical storms during the 2008 hurricane season. Results showed promising predictive skill for this method with only approximately a ten percent error in forecasting (within a 24 to 48 hour window) developing and non-developing vortices during the 2008 season.
Session 17, Mesoscale predictability and data assimilation I
Thursday, 20 August 2009, 1:45 PM-3:15 PM, The Canyons
Browse or search entire meeting
AMS Home Page