Fulltext

Generalized sampling interpolation of noisy gravity/gravity gradient data

While, James ; Biegert, Ed ; Jackson, Andrew

In: Geophysical Journal International, 2009, vol. 178, no. 2, p. 638-650

Ajouter à la liste personnelle
    Titre
    • Generalized sampling interpolation of noisy gravity/gravity gradient data
    Auteur
    • While, James. School of Earth and Environment, University of Leeds, Leeds LS2 9JT, UK. E-mail: jwh@mail.nerc-essc.ac.uk
    • Biegert, Ed. Shell International Exploration and Production Inc, Bellaire Technology Center, Houston, Texas, USA
    • Jackson, Andrew. Institut für Geophysik, ETH, Zürich, Switzerland
    Type de document
    • Postprint
    Identifiant OAI-PMH
    • oai:doc.rero.ch:297362
    Summary
    The generalized sampling expansion (GSE) has been shown as a method for successfully interpolating combined gravity and gravity gradient data sets when the data are undersampled. The presence of noise on data sets renders such interpolation more difficult and many applications (known as expansions) of the GSE can be shown to intolerably amplify noise. However, many key expansions can be shown to successfully interpolate noisy data and even, given limited gradient error and sufficiently narrow line-spacing, reduce noise. These results can be shown to hold for both random noise and along-line correlated (levelling error type) noise. Unfortunately, the only expansion capable of interpolating a data set sampled at 3× conventional line-spacing, the Three-rectangle expansion, has a poor noise response and always acts to amplify data error. The GSE method bares up well against other methods of gradient enhanced interpolation; in numerical tests several expansions for the gravity field produce less noisy output than any of the pseudo-line, gradient enhanced minimum curvature or gradient enhanced Akima spline methods. Despite edge effects and using only gradient data with no gravity component, the GSE applied to real undersampled survey data bares up well against conventional interpolation, reducing noise where the data are clearly undersampled