The subjects of my diploma thesis are track forecasts of tropical cyclones, particularly of tropical storms.
If you realize the misery of these catastrophical events, you understand the necessary of a track forecast
as good as possible. A hurricane of category 5 in
Saffir-Simpson - scale reaches velocities in wind
of more than 155 MPH! The ocean is been towered to storm surges up to 20 feet and more! This combination
(strong wind and storm surge) accounts for the tropical storms' great destructions.
Actually all tropical easterly coasts in the world are strucked, so the Indian Ocean or Asia, where are
colossal losts in human lifes (Bangladesh at Nov.,13. of 1970: 300 000 deaths), or middle america and
south of United States, where is a prevailing of material loss (Hurricane Andrew 1992 with a damage of
roughly $30 billion!).
The greatest problem is the forecast of the "land-fall", i.e. of that position the tropical storm hits
the coast. In the moment of the thesis, there was an error of about 200 km in 24 hours.
The thesis desribes a so called self adapting analog model which has been developed in an elder
work tp predict the dynamics of theoretical models. With my work the usage was expanded to atmospheric data.
The program learns stand-alone to find that storm tracks out of a historical dataset which are optimal
to describe the behaviour of a current storm.
The thesis was supervised by Dr. Richard Blender and
Prof. Klaus Fraedrich, department for
University of Hamburg. Meanwhile the method
is running there in operational mode.
An online version of my diploma thesis is available (only german version - sorry!). On the
content page you'll find a reference for downloading a PDF-version of the whole paper.
A few links about tropical storms:
Not enough? Google et al give pages for months: