UNCERTAINTY INTERVALS FOR UNBINNED SPECIES RICHNESS CURVES: AN EXAMPLE USING ORDOVICIAN CONODONTS

Species richness for the Paleozoic is typically estimated for relatively coarse time intervals of unequal length (e.g. binned into chrons, biozones, or stages) or at boundary instants separated by the same coarse time intervals. Neither method maximizes biostratigraphic resolving power.

Graphic correlation and automated sequencing algorithms have the potential to build an ordered composite sequence of all available first and last appearance events, by seeking the best fit with field observations from all available stratigraphic sections. We have used a constrained optimization algorithm (CONOP9) to sequence 690 Ordovician conodont taxa from 196 sections (Laurentia: 557 taxa, 149 sections; Baltica: 131 taxa, 31 sections; other regions: 189 taxa, 16 sections). Data for Laurentia and Baltica are sufficient to examine their biodiversity curves separately, explore their individual contributions to global species richness, and investigate the effects of provinciality on global correlation. The method was validated by comparison with Walter Sweet's (1995) graphic correlation.

Compared with other conodont biodiversity curves for the Ordovician that include between 6 and 80 sequential estimates of species richness, our composite permits up to 1380 individual estimates. The best-fit composite is not a unique solution because there are clusters of events whose internal order cannot be resolved with the current stratigraphic information, so in practice approximately 1000 events and event clusters can be resolved. The full range of equally well-fit solutions provides one means of placing uncertainty intervals around the species richness curves; another option for establishing uncertainty bounds compares composite sequences generated by different measures of fit with the raw field data. The same composite sequence is also used to explore mean taxon longevity through time.