Final Report Summary - SEISSEA (Seismic Inversion and Stochastic Spectral Analysis of Thermohaline Staircases in the Tyrrhenian Sea)
The SEISSEA (SEismic Inversion and Stochastic SpEctral Analysis of thermohaline staircases in the Tyrrhenian Sea) project was conceived to address the limits to which useful physical oceanographic data can be reliably extracted using seismic oceanography. The subject of this study was the Tyrrhenian Sea, which is known to contain well-defined thermohaline staircases, which are stratified layers of varying temperature and salinity that can only form in the most quiescent of waters. The centre of the Tyrrhenian Sea is largely isolated from external turbulence and mass convection. As such, it is a natural laboratory to study a type of mixing which is not well understood, but is thought to be ubiquitous throughout the world’s ocean, namely diapycnal mixing, or mixing across lines of equal density. Due to the study area’s stability there is a constraint on mixing by other means, allowing diapycnal mixing to be studied in isolation. Diapycnal mixing is important because it contributes significantly to global thermohaline circulation, which in turn strongly affects climate. Conversely, excess heat from a warming planet is stored in the ocean, which may in turn affect circulation through its influence on mixing.
Before seismic data could be inverted and analyzed for their spectral content and stochastic parameters, detailed processing needed to be carried out. This included data sorting, the precise application of the acquisition geometry using a custom designed back projection algorithm, low-cut band pass filter, attenuation of the near-surface direct wave noise, minimization of wrap-around coherent noise trains resulting from non-optimal acquisition parameters, normal moveout correction using sound speeds derived from simultaneously and coincidentally acquired in situ temperature probes (calibrated using nearby in situ salinity data), a radial prediction filter to improve signal coherency, high-cut band pass filter, stacking and migration. After this data preparation stage the seismic stacks were inverted to determine the sound speed, temperature and salinity distributions in the zone of interest. These data have been presented at numerous conferences and workshops over the tenure of the postdoc, accumulating in a comprehensive paper that is in the final stages of preparation. This paper presents the final processed seismic data and interprets them in the context of the current knowledge of the Tyrrhenian basin through, among other things, the quantification of sound speed gradients. It is a collaboration between geophysicists and physical oceanographers at GEOMAR, CSIC (Consejo Superior de Investigaciones Científicas) in Barcelona, Spain, Durham University in the UK and ISMAR (Istituto di Scienze Marine) in Italy.
The second aspect of the postdoctoral fellowship involves the extraction of the Hurst number from seismic data. The Hurst number is the exponent in the power law function which controls fractal scaling and therefore characterizes the roughness of surfaces. Through our research we have found that it can be used to estimate internal wavenumber spectra directly from seismic data. This is of intense interest to physical oceanographers because, until now, wavenumber spectra were obtained by measurements from a sparse distribution of oceanographic probes or a horizontal tow mechanism. Previous studies have shown that wavenumber can be estimated from seismic data by picking the peak amplitudes of coherent reflectors, but our analysis shows that by inverting the seismic data for Hurst number, we can calculate the wavenumber spectra globally, across the entire seismic section, even in places where the signal coherency is low. This analysis has been presented at various stages of completion at several conferences during the postdoc and most notably was published in a preliminary stage as a conference proceeding to the International Congress on Acoustics in Montreal, Canada. Contributions for this study come from GEOMAR, Durham University, Memorial University of Newfoundland in Canada and CSIC.
A third paper, also in the final stages of preparation, describes the theory and explores the boundaries of power spectra estimations in seismic oceanography showing that the temporal resolution of the seismic source is traded into lateral resolution through the first Fresnel zone for unmigrated data and λs/2 for migrated data where λs is the dominant wavelength of the seismic source wavelet. By doing so, we show that the loss of short wavelength resolution can be masked on real data by noise, itself possibly the product of interference of seismic energy reflected by heterogeneous turbulent mixing or, more significantly by discretisation of the signal during digitization. The recommendation of this paper is that seismic oceanography surveys actually require a source peak frequency content in excess of 200 Hz with 1) high temporal sampling, 2) adequate spatial sampling (~6.25 m) and 3) a sufficient signal-to-noise ratio so that event picking is not corrupted. This paper is a collaboration between GEOMAR and Durham University.
Before seismic data could be inverted and analyzed for their spectral content and stochastic parameters, detailed processing needed to be carried out. This included data sorting, the precise application of the acquisition geometry using a custom designed back projection algorithm, low-cut band pass filter, attenuation of the near-surface direct wave noise, minimization of wrap-around coherent noise trains resulting from non-optimal acquisition parameters, normal moveout correction using sound speeds derived from simultaneously and coincidentally acquired in situ temperature probes (calibrated using nearby in situ salinity data), a radial prediction filter to improve signal coherency, high-cut band pass filter, stacking and migration. After this data preparation stage the seismic stacks were inverted to determine the sound speed, temperature and salinity distributions in the zone of interest. These data have been presented at numerous conferences and workshops over the tenure of the postdoc, accumulating in a comprehensive paper that is in the final stages of preparation. This paper presents the final processed seismic data and interprets them in the context of the current knowledge of the Tyrrhenian basin through, among other things, the quantification of sound speed gradients. It is a collaboration between geophysicists and physical oceanographers at GEOMAR, CSIC (Consejo Superior de Investigaciones Científicas) in Barcelona, Spain, Durham University in the UK and ISMAR (Istituto di Scienze Marine) in Italy.
The second aspect of the postdoctoral fellowship involves the extraction of the Hurst number from seismic data. The Hurst number is the exponent in the power law function which controls fractal scaling and therefore characterizes the roughness of surfaces. Through our research we have found that it can be used to estimate internal wavenumber spectra directly from seismic data. This is of intense interest to physical oceanographers because, until now, wavenumber spectra were obtained by measurements from a sparse distribution of oceanographic probes or a horizontal tow mechanism. Previous studies have shown that wavenumber can be estimated from seismic data by picking the peak amplitudes of coherent reflectors, but our analysis shows that by inverting the seismic data for Hurst number, we can calculate the wavenumber spectra globally, across the entire seismic section, even in places where the signal coherency is low. This analysis has been presented at various stages of completion at several conferences during the postdoc and most notably was published in a preliminary stage as a conference proceeding to the International Congress on Acoustics in Montreal, Canada. Contributions for this study come from GEOMAR, Durham University, Memorial University of Newfoundland in Canada and CSIC.
A third paper, also in the final stages of preparation, describes the theory and explores the boundaries of power spectra estimations in seismic oceanography showing that the temporal resolution of the seismic source is traded into lateral resolution through the first Fresnel zone for unmigrated data and λs/2 for migrated data where λs is the dominant wavelength of the seismic source wavelet. By doing so, we show that the loss of short wavelength resolution can be masked on real data by noise, itself possibly the product of interference of seismic energy reflected by heterogeneous turbulent mixing or, more significantly by discretisation of the signal during digitization. The recommendation of this paper is that seismic oceanography surveys actually require a source peak frequency content in excess of 200 Hz with 1) high temporal sampling, 2) adequate spatial sampling (~6.25 m) and 3) a sufficient signal-to-noise ratio so that event picking is not corrupted. This paper is a collaboration between GEOMAR and Durham University.