![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Statistics of regional surface temperatures post year 1900. Long-range versus short-range dependence, and significance of warming trends. |
VerfasserIn |
Ola Løvsletten, Martin Rypdal, Kristoffer Rypdal, Hege-Beate Fredriksen |
Konferenz |
EGU General Assembly 2015
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 17 (2015) |
Datensatznummer |
250110225
|
Publikation (Nr.) |
EGU/EGU2015-10203.pdf |
|
|
|
Zusammenfassung |
We explore the statistics of instrumental surface temperature records on 5°x 5°, 2°x 2°, and
equal-area grids. In particular, we compute the significance of determinstic trends against two
parsimonious null models; auto-regressive processes of order 1, AR(1), and fractional
Gaussian noises (fGn’s). Both of these two null models contain a memory parameter which
quantifies the temporal climate variability, with white noise nested in both classes of models.
Estimates of the persistence parameters show significant positive serial correlation for most
grid cells, with higher persistence over occeans compared to land areas. This shows that, in a
trend detection framework, we need to take into account larger spurious trends than what
follows from the frequently used white noise assumption. Tested against the fGn null
hypothesis, we find that ~ 68% (~ 47%) of the time series have significant trends at the
5% (1%) significance level. If we assume an AR(1) null hypothesis instead, then
the result is that ~ 94% (~ 88%) of the time series have significant trends at the
5% (1%) significance level. For both null models, the locations where we do not find
significant trends are mostly the ENSO regions and the North-Atlantic. We try to
discriminate between the two null models by means of likelihood-ratios. If we at each
grid point choose the null model preferred by the model selection test, we find
that ~ 82% (~ 73%) of the time series have significant trends at the 5% (1%).
We conclude that there is emerging evidence of significant warming trends also
at regional scales, although with a much lower signal-to-noise ratio compared to
global mean temperatures. Another finding is that many temperature records are
consistent with error models for internal variability that exhibit long-range dependence,
whereas the temperature fluctuations of the tropical oceans are strongly influenced by
the ENSO, and therefore seemingly more consistent with random processes with
short-range dependence. Four different data products, HADCRUT4, NOAA mlost,
GISS and Berkely Earth, are analyzed in this project, with similar results in all
cases. |
|
|
|
|
|