GEOS 597e
Spatiotemporal Data Analysis Workshop
Prework 8: Singular
spectrum analysis: Error, stability, truncation and filtering
Last updated 10/25/06. To be
completed prior to class session Weds., Oct. 25th.
Introduction:
This week we'll see how many of
the NINO3 SSA eigenvectors we calculated last week pass a Monte Carlo
test for red noise.
Reading:
- Ghil et al., 2002:
Advanced spectral methods for climatic time series.
Rev. Geophys., 40(1), sections 2.3 (pp. 3-11 to 3-13).
- Vautard and Ghil, 1991:
Interdecadal oscillations and the warming trend in global temperature
time series, Nature, 350, 324-327.
- Elsner and Tsonis, 1991: Do
bidecadal oscillations exist in the global temperature record?, Nature,
353, 551-553.
Reading questions:
- Write down, in words, the Monte-Carlo SSA algorithm proposed by
Allen et al. (1992) and reported in Ghil et al. (2002) to distinguish
'significant' eigenvectors from those expected from SSA on time series
with memory like that of the NINO3 time series we studied last week.
How is this test similar and different than the Rule N test we
applied to the EOF analysis of the SST anomaly field in HW5?
- In Vautard and Ghil's SSA on the global mean temperature time
series, what percent of the variance is accounted for by the trend in
the global mean temperature data set? What percent of the
variance is accounted for by the bidecadal oscillations in global mean
temperature?
- What is Elsner and Tsonis' argument about the nature of the
bidecadal oscillations? What additional information might help to
support the existence of such features in the climate system? Did
Ghil et al. (2002) acknowledge this result?
- Can you add any additional items to your list of
uncertainties for which to look out when performing singular
spectrum analyses?
Products to hand in (keep a
copy for yourself to use in class discussion):
- Answers to the four reading questions listed above.
Back
to Schedule/Syllabus.