Management of ambiguities in magnetostratigraphic correlation

Magnetostratigraphy is a powerful tool to access absolute dating of sediments enabling good and detailed chronostratigraphic correlation. It is based on the correlation of a magnetic polarity column, observed and measured in a given sediment section, to a magnetic polarity reference scale where polarity changes are well dated via other independent methods. However, magnetostratigraphic correlations are loose because only constrained by binary magnetic chrons (i.e normal or reversal) and their thickness, which are both defined from depth variations of the magnetic remanent directions. The thickness of a given magnetic chron is a function of time and sediment accumulation rate, which may not be stationary, leading to ambiguities when performing the correlations. To address these ambiguities, a numerical method based on the Dynamic Time Warping algorithm is proposed. Correlation is computed in order to minimize the local variation of accumulation rate. Non-recorded magnetic polarity inversions (due to non-deposition, erosion, fault or approximate sampling) is managed by allowing multiple gaps in the DTW algorithm. In the case of absolute dating of sections, i.e. correlation between the studied section and the reference scale additional rules are used: (i) gaps are not allowed on the reference scale; (ii) extremities of the studied section are considered as free in the DTW algorithm. The correlation method is designed to compute not only the “best” dating of a given polarity column (i.e. the one minimizing local variation of accumulation rates) but also to output the “n best” correlations offering insight into uncertainties on sediment dating and resulting accumulation rate estimation. This method is applied to the Subei (Gansu Province, western China) and the Yaha sections (Xinjiang province, western China) to discuss about conflicting interpretations proposed in the literature. http://www.gocad.org/www/people/index.xhtml?lallier