THE EFFICACY OF CARBON ISOTOPE STRATIGRAPHY FOR IDENTIFYING AND CORRELATING THE PERMIAN-TRIASSIC BOUNDARY

Stable isotope chemostratigraphy holds great promise for the correlation of critical intervals in earth history, such as mass extinctions. In the case of the Permian-Triassic (P-Tr) boundary, a pronounced negative shift in carbon isotopes has traditionally been used to identify and correlate boundary sections. More refined curves, however, have suggested that there are up to seven large negative excursions (one of which is the boundary excursion) superimposed upon a long-term negative trend. Difficulties with chemostratigraphic correlation for the P-Tr boundary exist at variety of scales. At coarse scales, any P-Tr contact concealed within an unconformity (often cryptic) will show a negative excursion from the juxtaposition of rocks of two different isotopic values. In addition, the high number of negative excursions during this period will lead to aliasing problems when precise biostratigraphic control is lacking. At fine scales, sampling interval will strongly affect the stratigraphic placement of an isotopically-defined boundary.

To address the effect of sampling interval on isotope stratigraphy, we used a computer simulation to test the amount of error produced by iteratively sub-sampling high-resolution carbon isotope curves at specified sampling intervals. The stratigraphic differences between the most-negative excursion in the sub-sampled curves and the most-negative excursion in the true curve approximates the ‘error' associated with a given sampling interval. For a perfectly symmetrical excursion, the maximum possible error and the average error will increase linearly as 1/2 and 1/4, respectively, of the sampling interval. However, for non-symmetrical curves, or curves with multiple excursions, error increases much faster, and extremely high sampling intervals are necessary to recover the true geometry and stratigraphic position of the most-negative excursion. For example, a 1m sampling interval for the ~330m Gartnerkofel-1 core will result in ~1.5m of average error and more than 5m of maximum possible error. Based on these simulations, to avoid large possible errors, stratigraphic sections with sedimentation rates similar to Gartnerkofel-1 should be sampled at intervals of 75cm or less, while condensed sections such as the GSSP at Meishan should be sampled at intervals of 5cm or less.