Seismic cyclostratigraphy: hypothesis testing for orbital cyclicity at ODP Site 1084 using seismic reflection data

Orbital forcing may be preserved as cyclical variation in acoustic impedance in marine sediments due to paleoclimate-related changes in grain size, sorting and lithology. If seismic images of such deposits have the relevant bandwidth, this cyclicity may be imaged as distinct peaks in the power spectra of the seismic traces. In principle this could allow the application of cyclostratigraphic techniques to seismic data. It is still unclear, however, if in practice the statistical power is high enough to reliably discriminate orbital cyclicity from seismic data alone, and how the false detection rate compares to directly sampled data such as outcrop, drill core or borehole logs.

In this study we compare the discriminatory power for cyclostratigraphic analyses between seismic data and an equivalent borehole log. We develop a method for spectral background estimation that accounts for some of the amplitude and frequency filtering effects inherent to seismic data. We forward model the seismic response using 1-D visco-acoustic full-wavefield seismic modelling that includes the contribution of multiples and seismic absorption, which we combine with Monte Carlo ensemble modelling using sedimentary noise models to quantify the discriminatory power of both seismic and borehole significance testing approaches.

We demonstrate this on two examples: i) a simplified model with constant background velocity, sedimentation rate and known seismic source wavelet, and ii) a real-world example based on ODP Site 1084 (Cape Basin, ODP Leg 175). We observe in both cases that the sensitivity and specificity (related to the true and false detection rates) for the seismic case are strongly dependent on the spectral frequency, compared to the largely frequency-independent results for the borehole cyclostratigraphy. For the ODP Site 1084 example we observe a seismic spectral peak corresponding to 95 kyr eccentricity with an uncalibrated confidence level of >95%. Our Monte Carlo ensemble modelling, however, shows that the false positive rate at this frequency and confidence level is around 25%, compared to around 5% for the equivalent borehole cyclostratigraphy. We also demonstrate eccentricity modulation and bundling analysis (TimeOpt) applied to the seismic data, which can successfully invert for the sedimentation rate for the simplified seismic synthetic example.

Our results suggest that reliably identifying Milanković cyclicity from seismic data is possible but is strongly dependent on the sedimentation rate, the geophysical properties of the subsurface and the spectral frequency in question. Where the age model is known (i.e., from a co-located borehole) and an orbital signal is well-preserved in the acoustic impedance, for typical airgun seismic bandwidths, sedimentation rates around 20 cm ka^-1 and seismic velocities around 1600 ms^-1 it should be generally possible to identify eccentricity and obliquity cyclicity in seismic data. This opens the door to widespread use of seismic cyclostratigraphy to identify the preservation of cyclicity directly from seismic data, to extrapolate astronomically-tuned age models away from (and below) boreholes and to screen for the preservation of cyclicity prior to drilling. Similar principles could be applied to other methods such as sub-bottom profilers to identify, for example, higher frequency precessional cyclicity.