HR: 1340h
AN: GC33B-1260    [Abstracts]
TI: Time Series Analysis of Monthly Precipitation and River Flow Extremes in Germany in the 20^th Century: Evidence of Different Variability Scales as an Indicator of Recent Regional Climatic Changes
AU: Markovic, D
EM: Danijela.Markovic@gmx.de
AF: Department of Geohydraulics and Engineering Hydrology, University of Kassel Kurt-Wolters Str. 3, Kassel, 34109 Germany
AU: * Koch, M
EM: kochm@uni-kassel.de
AF: Department of Geohydraulics and Engineering Hydrology, University of Kassel Kurt-Wolters Str. 3, Kassel, 34109 Germany
AB: We report on an extensive study of time series analysis of monthly precipitation extremes and river stream flows from numerous gauging stations across Germany in the 20^th century. A multitude of spectral and time-scale analysis methods that include a.o. the continuous wavelet tool, principal component (EOF), SSA, and other classical techniqes (R/S,WTMM, DFA) for the estimation of the Hurst coefficient H, as an indicator of possible persistence. The wavelet analysis reveals, beyond the typical annual cycle, dominant oscillations in both precipitation and river flow at the interannual (7-8 yr.) and/or interdecadal (11-14 yr.) low frequency scales, in particular in the second part of the 20^th century, providing evidence for some recent regional climatic changes over Germany. The scaling properties of the time series are determined through the computation of the Hurst parameter. It turns out that the presence of the named dominant periods in the signals may bias the estimated H significantly, depending on the method used, especially when the signal-to-noise ratio and/or the data length are small. Surprisingly, for the case under study the somewhat "outdated" R/S-method appears to fare best. In any case, the estimated H for the precipitation series at the various stations across Germany fluctuate around H=0.5, i.e. do not show a clear picture of systematic persistence. However, data with low frequency components show higher H values, i.e., some amount of persistence, than data without such components, where possible negative autocorrelations and H < 0.5, i.e. anti-persistence, are suspected. As for the river flow, good estimates for H can be obtained after application of an SSA-filter. In any case, as expected, the estimated H are much higher than 0.5, indicating strong persistence of the stream flow. As for the origins of the named interannual (7-8 yr.) and/or interdecadal (11-14 yr.) low frequency scales, using cross-wavelet spectral analysis, we find evidence for a teleconnective influence of the North Atlantic Oscillation on the precipitation and streamflow pattern, primarily, over the last 40 years, and with partly opposite effects in the northern and southern parts of Germany. As such the NAO phase of enhanced decadal variability could be made responsible for the low frequency precipitation components.
DE: 1616 Climate variability (1635, 3305, 3309, 4215, 4513)
DE: 1637 Regional climate change
SC: Global Climate Change [GC]
MN: Fall Meeting 2005