DETERMINING THE SMALLEST SAMPLE SIZE REQUIRED FOR PALEOCOMMUNITY RESEARCH: A CASE-STUDY FROM THE FINIS SHALE OF TEXAS

Exploring various sampling and analytical methods when conducting paleoecological research is necessary to discover which methods produce the most accurate representations of past ecosystems. Determining the smallest sample size required for examining fossil communities is important to ensure accurate interpretation with no excessive costs or wasted time. Diversity is often used as a metric to determine an appropriate sample size, but this size may differ from that representative of the community signal. Here, we compare the ordination-based community result from a high-resolution (13 bulk-samples in a 4.1 m section), large sample size (13 L of sediment) dataset to that of four sub-sampled datasets to determine the smallest sample size required to produce the same ordination-based result.

Specimens were collected from the Pennsylvanian Finis Shale in Jacksboro, Texas. Brachiopod and mollusk specimens that had a maximum dimension >2.8 mm were tallied for a total of 4921 specimens from 51 genera. One hundred sub-sampled matrices were created for each of four fractions (10 L, 4 L, 2 L, 1 L) of the full 13 L dataset by randomly selecting the appropriate fraction (1/13 for 1 L, 4/13 for 4 L, etc.) of specimens from each of the 13 samples. Q-mode Polar Ordinations were conducted for all 401 matrices using PC-Ord v. 5.10. All 10 L, 4 L, and 2 L richness, evenness, Shannon’s H, Simpson’s diversity, and ordination axis-one scores were significantly correlated (r = 0.69 to 0.98, p < 0.01) with the coeval values for the full 13 L dataset. The 1 L sub-sample had 99 of the 100 matrices’ ecological measures and ordination axis-one scores significantly correlated (r = 0.78 to 0.99, p < 0.002) with the coeval values for the 13 L dataset.

The same ordination-based result was demonstrated whether using a 2 L sub-sample or the 13 L dataset. Furthermore, the sample-size needed for capturing the ordination-based result is less than that needed to capture total richness, which could mean fewer hours are required collecting and processing material. However, these results may be unit-specific because two abundant brachiopod taxa may be driving the signal regardless of manipulation to the dataset. Creation of an ordination sub-sampling program to conduct similar research on various lithologies, scales, and preservation is underway.