ACCOUNTING FOR THE “KNOWN UNKNOWNS” IN A SERIES OF LATE PERMIAN PLANT PALEOCOMMUNITIES FROM THE KAROO BASIN OF SOUTH AFRICA

Ecological theory predicts how community richness and relative abundance should vary relative to factors such as environmental perturbations, invasive species, and ecological succession. The fossil record provides a window into the past but is plagued by sampling issues, most critically time-averaging, spatial mixing, preservational fidelity, and fossilization quality. The Upper Permian (Changhsingian) Clouston Farm locality in KwaZulu-Natal, South Africa contains abundant, well-preserved plant macrofossils in massive, fluvially-deposited shale assigned to the Normandien Formation. Local sedimentology and taphonomy indicate litter accumulated in slack water conditions of an oxbow, a depositional environment with low time-averaging and spatial mixing potential and relatively high preservational fidelity. Bulk collections from 10 beds (C1 to C10) within a 0.7m vertical section were censused. Although taxonomy of glossopterids is generally problematic, Prevec et al. (2009) described the Clouston Farm flora using a morphotype approach. 893 specimens were censused (67% identified to genus or morphotype group) and rarefaction used to assess sampling. Neither richness (Pearson’s ρ = +0.256, df = 7, p = 0.5) nor Shannon’s evenness (ρ = +0.367, df = 7, p = 0.3) were correlated with stratigraphic height of the bed and this does not appear to be a result of variation in fossilization quality (% of unidentifiable specimens per census). There was a slight tendency for younger beds to have better quality but the association with stratigraphic position was not significant (ρ = -0.398, df = 7, p = 0.3). Different relative abundance distribution (RAD) models were fitted to each bed’s observed RAD using likelihood and AIC weights. All but two best fit a geometric model; beds C6 and C9 best fit a Zipf model. An artificially time-averaged collection (combining beds C1 to C10) best fit a Zipf-Mandelbrot model. However, there was a significant, negative correlation between fossilization quality and each bed’s AICc to a lognormal RAD (ρ = -0.006, df = 7, p = 0.988), suggesting that the worse the preservation, the closer the RAD is to lognormal. These results suggest that the Clouston Farm flora generally conformed to a relatively simple model of ecological succession interrupted by brief intervals of more complex processes.