COMPARING THE GRID ANALYSIS METHOD OF PLANT BEARING COAL BALLS WITH DIGITAL IMAGE ANALYSIS

Pennsylvanian coal balls are calcareous concretions of permineralized peat. They contain vascular plant tissue preserved at the cellular level, giving us insights into the ecology of peat swamps during this time. Thin sections of coal balls are generally made with the cellulose acetate peel method, where a polished surface is impressed in 5% HCL to dissolve part of the matrix revealing the cellular structure; acetone is then used to transfer material to a cellulose acetate sheet. These so-called peels are sampled by overlaying a transparent cm2 grid and analyzed using a dissecting microscope. For each grid cell, researchers identify and record the plant organ or arthropod trace present, using discretion to determine the defining feature (Phillips et al., 1977). This method provides a quantitative survey of the taxonomic composition of a peel sampled from a coal ball. However, because of the coarse-scale of identification and requisite subjectivity, the grid analysis does not accurately represent the area-based composition of a peel. Pryor (1986) devised an alternative random mm2 method, where random grid cells were chosen on a peel that was overlaid with an mm2 grid. Following Pryor’s criteria, this method displayed equivalent accuracy to that of the cm2 grid analysis. To evaluate the accuracy of the cm2 grid method we use Photoshop to analyze randomly selected peels of coal balls from the Mt. Rorah coal bed in the Tradewater Formation of Illinois (Desmoinesian). We outline each plant organ and arthropod trace and calculate their contribution to the total peel area. This allows us to break down peel composition by specific taxon, tissue, or trace fossil abundance for comparison with the cm2 and mm2 methods. Additionally, we assess the accuracy of the cm2 grid analysis by comparing it to a point count method, where only material marked by the crosshairs of the cm2 grid was identified, partially eliminating the need for a subjective assessment of the representative organ or trace in each cell, while limiting the total assessed area. Finally, we offer a comparison of the time investment and accuracy trade-offs between the Photoshop method and the cm2 grid and point count methods. Our digital image analysis methods can be applied to other assemblages, increasing the quality of quantitative data that can be extracted from a survey.