ROAD-TESTING TWO LATE ORDOVICIAN-EARLY SILURIAN GSSPS USING HORIZON ANNEALING FOR QUANTITATIVE STRATIGRAPHIC CORRELATION

Global Stratotype Sections and Points (GSSPs) should contain a range of marker horizons or other data points relevant for long-distance correlation, including as wide a diversity of fossil occurrences, chemostratigraphic events and geomagnetic reversals as possible. The best practice is to use the totality of the evidence through the entire continuous section that contains the boundary and correlate with the GSSP based on the sum of this evidence. Although there are several methods available for integrating multiple data sets into a correlation project, such as graphic correlation (GC) and constrained optimization (CONOP), relatively few studies have employed these methods specifically for correlation with GSSPs and fewer still have used these methods as part of the process of selecting GSSPs. Horizon Annealing (HA) is a recently developed method similar to CONOP. Like CONOP, HA simultaneously integrates data from multiple stratigraphic successions and uses a simulated annealing algorithm to find a correlation between all of the sections (a multidimensional line of correlation) that minimizes the misfit between the composite succession and the data from each of the individual sections. CONOP does this by ordering events that occur in each of the sections, such as first and last appearances of taxa or other kinds of marker events, whereas, HA orders sample horizons, each sample horizon being characterized by the all of data marking that level. Since GSSPs are, by definition, sample horizons, HA is well suited to the task of correlation of the GSSPs with correlative levels in other sections. We have examined the correlation of the GSSPs for the bases of the Hirnantian Stage and the Silurian System using 27 relatively graptolite-rich sections on four paleo-continents spanning the late Katian into mid-Rhuddanian using HA. An analysis conducted using only graptolite occurrence data permitted construction of what appears to be a reliable composite graptolite range chart, but graptolite data alone did not provide sufficient constraints for high-resolution correlation of individual levels between sections, such as the GSSP levels. Integration of data from other fossil groups, carbon isotopes and regional marker horizons are required to achieve the highest possible resolution in correlation among the study sections.