Geology Reference
In-Depth Information
C18n.2n
C18n.1n
C17n.3nC17n.2n
C17n.1n
10
8
C18r C18n.1r
C17r
C17n.2r
C17n.1r
C16
6
4
2
10
8
6
4
2
0
100
200
300 400
Meters upsection
500
600
700
800
Figure 4.2
Examples of continuous and discontinuous stratigraphic series from the Eocene Arguis Formation,
Pyrenees, northern Spain. Discontinuous magnetochrons identified in this section are shown at the top (Kodama et al.,
2010); middle: continuous anhysteretic remanent magnetism (ARM) series; bottom: discontinuous grain size series.
The ARM series will be used throughout Chapters 4 and 5 to demonstrate the time series tools that are typically used in
cyclostratigraphy.
4.3 Time Series Analysis Tools and Eocene
Arguis Rock Magnetic Cyclostratigraphy
4.3.1
Sampling and Interpolation
Sample rate is a basic issue with natural data, especially deep-time Earth
data with uncertain timescales (for which stratigraphic thickness is the
independent variable), or data that cannot be or have not been collected at
strict uniform spacings. This is the case with the Arguis ARM series, which
was sampled in the field at close spacings of Δd = 20 cm at the base of the
formation, where lithologic cycles were relatively thin, then adjusted to
Δd = 75 cm as bedding became thicker up section, and finally Δd = 1.5 m for
the final 200 m (Figure 4.3a). At various points along the formation, there
were gaps (covered outcrop); at other levels, local changes in sampling had
to be made. Thus, even at this early stage in the analysis, the practitioner is
often confronted by the problem of a nonuniformly sampled series.
There are several options available at this point, for example, windows-
based applications SPECTRUM (Schulz & Schaltegger 1997) and for
larger datasets, REDFIT (Schulz & Mudelsee 2002) use Lomb-Scargle
spectral analysis for nonuniformly sampled time series. Alternatively, the
nonuniformly sampled time series can be resampled to a uniform spac-
ing, which allows application of many powerful time series methods that
were developed for uniformly sampled data (Thomson 2009). This is the
approach taken in our analysis of the Arguis ARM data. However, care
needs to be taken to understand the limiting effects of the original
