Computer Science – Performance
Scientific paper
May 2008
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2008agusmsm24a..08w&link_type=abstract
American Geophysical Union, Fall Meeting 2007, abstract #SM24A-08
Computer Science
Performance
4430 Complex Systems, 4440 Fractals And Multifractals, 4494 Instruments And Techniques, 7924 Forecasting (2722)
Scientific paper
This paper provides a method to forecast extreme events observed in the physical sciences. In particular, we will demonstrate forecasting of magnetic storm events based on the time series of the Dst index over the period 1981-2002. This method is based on the multiple scaling of the measure representation of the Dst time series, and is applicable to many different geophysical systems. This measure is modeled as a recurrent iterated function system, which leads to a method to predict storm patterns included in its attractor. Numerical results are provided to evaluate the performance of the method in outside-sample forecasts. The method works reasonably well in predicting storm patterns for hourly data when we only pay attention to 2-symbol scenarios such as {storm, no storm}, or {big storm, no big storm}. A further point to note is that this perfomance is achieved from the scaling behavior of the Dst series captured in its measure representation. This distinguishes our approach from the usual approach based on the correlation structure of Dst. As the method does not rely on additional information such as a solar wind driver, it is simple to implement and its prediction can be used as a benchmark to evaluate more elaborate systems.
Anh Vo Van
Wanliss James A.
Yu Zhenbao
No associations
LandOfFree
a New Forecasting Technique for Extreme Events in the Geosciences does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with a New Forecasting Technique for Extreme Events in the Geosciences, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and a New Forecasting Technique for Extreme Events in the Geosciences will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1400542