Multifractal Analysis of Greenland Ice Core Data
Abstract
Recent deep drilling of Greenland ice sheet (European GRIP and American GISP2 projects) yielded new rich information about climate of past 140.000 years. The most detailed data were provided by the electrical conductivity measurements (ECM) and dielectric profiling (DEP). It is known that during warm (postglacial) period both direct current conductivity, measured in ECM, and high frequency conductivity, measured in DEP, responds only to the acidity of atmospheric precipitation forming ice cover. Available data rows of electrical properties measurements exceed 100.000 points. Besides obvious large-scale variations, there is a high-frequency component in the rows, usually considered just as a noise. Meanwhile, the amplitude of this component is sufficiently smaller than the instrumental noise level and it contains useful information we lose when smoothing and averaging data rows. It can be, however, extracted with the use of statistical methods. One of the ways to explore the "thin structure" of climatic rows is treating them in the frameworks of multifractal approach. There have been considered the rows of ECM from the hole GISP2 (down to the depth of 1370 m) and DEP from the hole GRIP (down to the depth of 1320 m), corresponding to the time interval of 8000 years -- approximately from the moment of ice cover melting in Europe to the present time. There was no any prominent climatic event in this period; the mean values and dispersions of the rows (after trend removal) were constant. The considered rows were first transformed by the interpolation to the sets of equidistant values with the time step of one month. The analysis has shown that they can be considered as functions close to fractional Brownian noise. The investigation of scaling properties allowed us to choose time interval from 10 to 80 years for the calculation of multifractal spectra f(H). The "running window" of 160 years width with the step of 8 years was applied to the rows. The changes of the position of the ultimate points (Hmin and Hmax) of f(H) spectra were followed. Obtained sequences of Hmin and Hmax were smoothed by the running average for the presentation of the results. The investigation has shown that the rightmost limit of multifractal spectra (Hmax) is changing independently for two considered data rows and apparently reflects the statistics of instrumental errors of the measurements. At the same time the changes of the leftmost limit of multifractal spectrum (Hmin) for the ECM row correlate with the other ones for the DEP row. The comparison with the results of reconstruction of Holocene temperatures from the pollen spectra [Andreev, Klimanov, 2000] has demonstrated that low values of Hmin correspond to the cold periods, when the variability of the curves of electrical properties rises. During warm periods, on the contrary, the values of Hmin increase and the rows become smoother.
- Publication:
-
AGU Spring Meeting Abstracts
- Pub Date:
- May 2005
- Bibcode:
- 2005AGUSM.C43A..12B
- Keywords:
-
- 1620 Climate dynamics (3309);
- 1827 Glaciology (1863);
- 1863 Snow and ice (1827);
- 3250 Fractals and multifractals;
- 3344 Paleoclimatology