 |
| Titel |
Universal Hurst exponent of local and global Earth temperature records? |
| VerfasserIn |
Lene Østvand, Kristoffer Rypdal, Martin Rypdal |
| Konferenz |
EGU General Assembly 2011
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 13 (2011) |
| Datensatznummer |
250054401
|
|
|
|
|
|
| Zusammenfassung |
| Local and global temperature records on time scales from a month up to decades seem to
exhibit Gaussian persistent fluctuations which suggest that they can be completely
characterized as fractional Gaussian noises with Hurst exponents H >0.5. However, standard
tests for long-range memory applied directly to the temperature signals yield relatively low
persistence (H ~0.6) for local records and strong persistence (H ~0.9) for global surface
records. Northern hemisphere sea-surface temperature appears even more persistent on these
time scales. This is true for both standard variogram or R/S-analysis as well as detrended
fluctuation analysis (DFA).
This wide range of Hurst exponents is puzzling if H is interpreted as a quantity
representing a universal characterization of the long-range memory in climatic processes on
monthly to decadal time scales. We argue, however, that the spread is due to spuriously high
Hurst exponents resulting from standard analysis applied to date that exhibits complex trend
patterns like oscillations on different time scales. Even the DFA fails in such cases because
first and second order DFA cannot deal with the complex multi-decadal trend pattern, and
DFA of fourth order and higher suppresses the fluctuations on annual scales and
shorter.
The paradox is resolved by modeling the signals as a fractional Gaussian noise
superposed on a multi-decadal trend found by fitting a fourth-order polynomial to the 150
year instrumental temperature records (1860-2010). Subtraction of this trend from the
temperature records leaves residuals that are fractional Gaussian noises with roughly the
same Hurst exponent (H ~0.6-0.7) for local as well as global records. The relative amplitude
of the stochastic fluctuations compared to the trend signal is much smaller for global
than for local records, and is the reason for the spuriously high Hurst exponents
for the global records. The fourth order polynomial trend is shown to be clearly
statistically significant for global temperature, and involves a sixty-year oscillation
superposed on a monotonically rising signal. We argue that the oscillation is part of
internal dynamics in the climate system, while the monotonic rise is due to external
forcing which most likely is anthropogenic. The residual fluctuations are completely
specified by the exponent Hwhich is characteristic for the present state of global
climate.
At present, however, it is not known whether this value of the Hurst exponent is the same
for other climatic regimes, and it is conceivable that rapid regime shifts associated with
glaciations/deglaciations will involve a change in H. Such changes may be studied in high
time-resolved paleoclimatic records. Some preliminary results from analysis of such data are
presented. |
|
|
|
|
|
|