Parts | Days | Selection | Search | Downloads | Help

UP: Umweltphysik

UP 22: Atmosph
äre und Klima

UP 22.5: Fachvortrag

Wednesday, March 15, 2006, 15:15–15:30, D

A Fourier enhanced Monte-Carlo-Method to extract climate trends from time series — •Dieter Ihrig¹ and Christiane Ihrig² — ¹FH Suedwestfalen, Iserlohn, Germany — ²Spurenstofftechnik GbR, Menden, Germany

Using a special apodization function it is possible to calculate Fourier transformations with a relative small number of data points. (100 points are enough). The less points exist the more it is necessary to extract trend functions because the signal of the box exceeds all other signals. The calculation of a trend function is possible using the method of sliding averages. This method leads to more and more uncertainty and to artefacts at the boundaries of the time series because there are less and less points to average. Using the trend function to extract climate trends from harmonic oscillations this effect is very unpleasant because the last few decades are the point of most interest.

Our method uses a Monte-Carlo-method to fit optimized regressions to the data points at the boundaries. This regressions are used to predict more points for the sliding averaging. The Fourier transformation (power spectrum or cosine transformation) of the original data, the corrected data (This are in fact the temperature variations.) and the trend function is calculated. The result of the Fourier transformation gives the criterion of optimization. The performance of the method will be demonstrated using simulated climate trend functions. The method will be applied at real 120 years climate series (34 stations). For this stations the difference in radiation exchange balance from the middle of last century to actual times is estimated.