Assessing reproducibility in sedimentary macroscopic charcoal count data

Abstract

AbstractCurrent understanding of global late Quaternary fire history is largely drawn from sedimentary charcoal data. Since publication, CharAnalysis increasingly has been relied upon as a robust method for analyzing these data. However, several underlying assumptions of the algorithm have not been tested. This study uses replicated charcoal count data to examine the assumption of Poisson distribution and reproducibility of peak detection. Results show <10% of the replicate counts are Poisson distributed, a maximum peak replication rate of 60%, and, for >90% of the data, intra-level count differences were larger than the threshold used to identify significance in inter-level differences. A pronounced “edge effect” was observed at the beginning and end of the records, cautioning against validation of results based on sections corresponding to the historical period. The proximal cause for low reproducibility is likely a lack of spatial randomness of charcoal particles at the scale of a core diameter. Until and unless decomposition methods can be developed that accommodate the observed limitations inherent in particle count data, best practices for interpreting charcoal records may be to rely on qualitative interpretations based on smoothed influx values and minimum particle count values in the hundreds.