Oral Presentation Freshwater Sciences 2023

Derivation and use of results from taxonomic quality control analyses (#528)

James Stribling 1
  1. Tetra Tech, Inc., Owings Mills, MARYLAND, United States

The first principle of any taxonomy, biological or otherwise, is communication. Framing and labelling items in our surroundings provides the foundation for discovery, planning, decision-making, and effective application of nomenclature. The centerpiece of many current biological assessment and monitoring programs is morphology-based taxonomic identification of benthic macroinvertebrates, while simultaneously exploring the feasibility and applicability of metagenomic techniques. The principal goals of this work are to provide objective and rigorous estimates of error at two scales, that of the: 1) sample and 2) taxon. These direct quality control (QC) procedures use independent sample re-identification comparisons to quantify error rates with performance measures, and use results to configure corrective actions for samples, taxa, and taxonomists. Sample-based QC results in taxonomic data of known and acceptable quality for use as input to indicator calculations; this is routinely accomplished by objectively demonstrating error rates (percent taxonomic disagreement) <15 percent.  A secondary goal of these analyses was to develop taxon-specific uncertainty ratings. Comparison data from >900 samples were compiled from monitoring programs throughout the US, and error rates (relative percent difference [RPD]) calculated for 1,003 mostly genus level taxa. Frequency of observation for each taxon (FREQ) in the dataset was calculated as the number of samples found relative to the total number. Plotting RPD and FREQ against each other revealed that most taxa are identified with relatively low error rates, even those that are infrequently observed. Graphical patterns were used to delineate six uncertainty-frequency classes (UFC 1-6) encompassing taxa identified with variable levels of confidence with relative rarity/frequency of observation ranging from rare (a single sample) to common, being observed in 73.7 percent of all samples (n=674). Results of these analyses could be used to differentially weight taxa counts prior to index calculation or model input, target research activities, and mediate site condition assessments.