Discussion of “Synthetic Tropical Cyclone Database” by M. K. James and L. B. Mason

Abstract

The authors present a Monte Carlo technique for generating a database of synthetic tropical cyclone tracks over the Great Barrier Reef Marine Park, situated off the coast of Queensland, Australia. They validate their synthetic tracks by comparing the marginal probability density functions PDFS of various track parameters against the Australian tropical cyclone best-track database Dare and Davidson 2004. They conclude that their synthetic database has statistical properties that are similar to the observational record and that can therefore be used confidently for simulating the climatological probability of occurrence of various aspects of tropical cyclone risk, such as storm surge, waves, and wind. The authors independently model the cyclone east-west x and north-south y components of motion and the intensity change. That is, they assume these variables to be uncorrelated. The discussers will show that this assumption is not supported by the data and that it causes a significant class of storms to be underrepresented in the authors’ synthetic database. The error was not detected by their statistical tests, since they were all tests of the marginal PDFS. The motion components are correlated, and so tests of the joint PDFS would have been more appropriate. The motion of storms within the paper’s three zones, over the same period 1969–2003 according to the Australian best-track database are summarized in Fig 1. It is apparent that the variation of the motion of individual storms about the all-storm mean has a systematic structure. This structure can be summarized by the covariance matrix, plotted in the diagram by an ellipse: here the coordinate rotation that diagonalizes the matrix has been found and the ellipse oriented in this direction, with semimajor and semiminor axes equal to twice the square root of the variances in the rotated covariance matrix. In the northern zone, the mean motion is slow and almost due southward, whereas the covariance ellipse is markedly elongated and oriented almost in the east-west direction, showing that most of the variability in storm motion is in the x-direction and that the x- and y-components of storm motion are not strongly correlated. In the central zone, the mean motion is slightly west of southward and a little faster than in the northern zone. The covariance ellipse is oriented in a northwest-southeast direction, with the ratio of axis lengths equal to 2. The x- and y-components of storm motion are strongly correlated correlation coefficient –0.47 and