LWRRDC Final Report: Development and implementation of QA/QC protocols for sample processing components of the MRHI bioassessment program
Internal Report 299
Humphrey C, Storey A & Thurtell L
Supervising Scientist Division
Department of the Environment
One of the support R&D projects for the first funding round of the Monitoring River Health Initiative comprised external Quality Assurance/Quality Control (QA/QC) audits of MRHI State/Territory agency sample processing procedures (laboratory subsampling and sorting of preserved samples, and field live-sorting procedures), together with research components to refine QA/QCprotocols and acceptance criteria. Sample processing errors have been quantified for the 3rd and 4th sampling rounds of the agency program. These data may be used to assess the degree to which the sensitivity of derived models has been compromised by such errors.
The external QA/QCaudits have confirmed the potential of the live-sort procedure to result in 'high' error rates. For these samples, two main sources of error were identified from the assessments, (i) under-representation of taxa, and (ii) different taxa recovery rates depending upon the efficiency of the operator. Factors contributing to poor taxa recovery in live-sorted samples included low live-sort sample size, operator inexperience, and commonly-occurring taxa missed in samples and across sites. Some preliminary simulations were conducted in the course of the study to evaluate the implications of live sorting errors for model development and sensitivity. Variable findings arose from these simulations, from the inability to derive models from error-ridden data to unexpected consequences for UPGMA classification arising from data sets upon which 'few' and 'many' errors were superimposed. Regardless, all of the simulations indicated the potential for live-sort error to adversely affect the sensitivity of models.
Changes have already been made to live-sorting protocols to reduce sample processing errors whilst additional changes to the protocol will follow as the results of additional R&D come to hand. Improvement in the procedures for taxa recovery in ongoing MRHI studies would only stand to benefit monitoring programs if the quality of data for existing reference sites are also improved through a re-sampling, data replacement and re-modelling program. A sensitivity analysis is required in future to more comprehensively determine the sizes of various sources of error and variation in sample processing, and their effects on the rates of misclassification to quality bands.