Benthic Invertebrate Sample Sorting Quality Assurance

Sorting and subsampling of aquatic invertebrate samples involve removing a minimum number of individuals from a uniformly distributed benthic macroinvertebrate sample. The product is a collection of individuals separated from both inorganic and organic material, which optimizes taxonomic accuracy and efficiency. Human error can occur if the sample is not split correctly and/or if all organisms are not removed from a sample split. To reduce the potential error associated with subsampling procedures, we take the following standardized actions.

1. Use effective and repeatable methods – Our technicians use a Caton tray subsampling method.
2. Intensively train all new technicians – Our training period lasts 6 months for all technicians. This includes 1:1 training in our standardized sorting procedures using previously processed samples where the type and number of each organism is known. We calculate sorting efficiency (see below for explanation) for all samples during the 6-month training period. Additionally, we review the first 10 subsamples for accuracy in enumeration and morphotyping. Our technicians must achieve a minimum sorting efficiency of 95% to complete the probationary training period.
3. Record accurate processing data – Our technicians record all relevant data in our postgreSQL database. Recorded fields include technician’s name, date processed, time required for sorting, the number of invertebrates removed from the sample and sample split proportion. Our technicians record this information on sample vial labels as well as in our database. On set completion, our staff reviews all technician processing data to identify potential anomalies. Specifically, we look for large differences in the percent of the sample sorted, total processing time, or final invertebrate counts among individual samples within a sample set.
4. Eliminate cross-sample contamination - Our staff makes every attempt to avoid the contamination of samples with individuals from a previous sample. They thoroughly wash and dry equipment after each sample split. They inspect all equipment prior to splitting a new sample. If an invertebrate appears desiccated with respect to other invertebrates in that sample, our technicians discard that organism. These desiccated individuals may be remnants from a previous sample (i.e., they were not removed from the equipment during a previous subsampling event and were inadvertently added to a different sample).
5. Conduct systematic checks of sorting effectiveness – We reprocess one randomly chosen sample from each technician each month. Specifically, a second technician examines the organic and inorganic matter from which organisms have been removed (i.e., remnant material) and removes all remaining organisms for further identification. In cases where the original sort time exceeded 20 hours (organic matter-rich samples), the second technician will re-examine the sorted material for approximately 10% of the time that the sample was originally sorted (2 hours, in this case). We add newly removed organisms to a separate sample vial for identification and enumeration. We compute sorting effectiveness (E_s) using:
   

   E_s = 100 * S/(R + S)

   Where R is the total number of organisms obtained during the re-sort of the remnant material and S is the total number of organisms originally obtained from the sample sorting. Our minimum quality objective for sorting effectiveness is ≥ 95%.
6. Follow remedial action plan - If sorting efficiency is 90% ≤ E_s ≤ 95%, we review additional samples until three consecutive samples achieve 95% sorting efficiency. If a technician fails to achieve mean monthly E_s of 95% twice in a 3-month period, we re-enter the technician into our sorting training process. If sorting efficiency is below 90%, we look for an additional three randomly selected samples that achieve 95% sorting efficiency. Additionally, we place the technician into a 3-month probationary period, during which time, we review all processed samples. We remove technicians from our sorting staff if mean E_s ≤ 75%.
7. Publish all quality control results – Our average sorting effectiveness for 2023 was 95.2% across 129 QC samples.