Appendix E Macroinvertebrate Sample Processing Error Report –

Transcription

Appendix E Macroinvertebrate Sample Processing Error Report –
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Appendix E Macroinvertebrate Sample
Processing Error Report
Appendix E
Page 26
National River Health Program
AusRivAS Quality Assurance and
Quality Control Project
Macroinvertebrate Sample
Processing
Error Report
for
The Department of the Environment
and Heritage
August 2004
Project Manager:
Ross Bannister
telephone
+61 3 9550 1000
email
rbannister@wes.com.au
WATER ECOscience Pty Ltd
ACN 064 477 989
Head Office
68 Ricketts Road Mt Waverley Victoria 3149 Australia
Private Bag 1 Mt Waverley Victoria 3149 Australia
telephone
+61 3 9550 1000
facsimile
+61 3 9543 7372
Wangaratta
1st floor NETC House
90-100 Ovens Street
Wangaratta, Victoria, 3676
Carrum
Eastern Treatment Plant
Thompson Road
Bangholme Victoria 3175
Gippsland
71a Argyle Street
Traralgon Victoria 3844
Geelong
49 Carr Street
Geelong, Victoria 3220
Hobart
20 St Johns Avenue
New Town Tasmania 7008
Werribee
Western Treatment Plant
New Farm Road Werribee Victoria 3030
WATER ECOscience Report Number: 767
August 2004
Cover photo: RBA sweep sampling, Queensland Tasmania. © Commonwealth of Australia
(courtesy of WATER ECOscience 2000)
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Foreword
WATER ECOscience (formerly AWT Victoria) was engaged by the Department of
the Environment and Heritage to undertake a study of Quality Assurance / Quality
Control practices associated with AusRivAS data and collection methods.
The study involved:
•
A national and international literature review of rapid biological assessment
QA/QC techniques and procedures;
•
An independent external audit of the collection of habitat, physico-chemical
and other related state / territory data used in the AusRivAS modelling and
reporting processes;
•
an independent external audit of biological and other related state / territory
data used in AusRivAS modelling and reporting processes, particularly for the
development of AusRivAS models and the First National Assessment of River
Health (FNARH), conducted under the National River Health Program
(NRHP); and
•
an assessment of lead agency QA/QC procedures for data collection and
data management.
The macroinvertebrate sampling and processing component of the study was to
assess and report on state agency performance for live-sort macroinvertebrate
sampling, sample processing, and macroinvertebrate identification against existing
criteria; as well as the incidence of data entry errors, and the effectiveness of
existing QA/QC procedures.
The audit covered data collected and sampling residues retained from randomly
selected sites from the First National Assessment of River Health in each of the
states / territories.
Appendix E
Page i
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Contents
Foreword
i
1 Introduction
1
1.1 Rapid Bioassessment
1
1.2 AusRivAS- National Context
2
1.3 Project Outline
3
1.4 Report Structure
4
2 Methods
5
2.1 Data Entry Audit
5
2.2 Live-sort Audit
6
Sample Collection
Residue Processing
Data manipulations
WISE assessment
1. Live Sort / Whole Sample Estimate ratio
2. Bray-Curtis dissimilarity
2.3 QA/QC Procedures
3 Results
6
6
7
7
8
8
8
9
3.1 QA/QC procedures
9
3.2 Data Entry Audit
10
3.3 Live-sort Audit
10
4 Discussion
12
4.1 QA/QC Procedures
12
4.2 Data Entry Audit
12
Sampling records
4.3 Live-sort Audit
5 Conclusions and Recommendations
13
14
16
5.1 Key Findings and Conclusions
16
5.2 General Recommendations
16
6 References
19
Appendix E
Page ii
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
List of Figures in Appendix E
Figure 1 Location of the 24 sites assessed for the biological data entry
audit and the 55 sites for the live-pick audit for Queensland in
relation to all sites sampled during the MRHI program.
1
Figure 2 Location of the 52 sites assessed for the biological data entry
audit and the 58 sites for the live-pick audit in New South Wales in
relation to all sites sampled during the MRHI program.
2
Figure 3 Location of the 30 sites assessed for the biological data entry
audit in the Australian Capital Territory in relation to all sites
sampled during the MRHI program.
2
Figure 8 Location of the 28 sites assessed for the biological data entry
audit for the Northern Territory in relation to all sites sampled
during the MRHI program.
4
List of Tables
Table 1
Table 2
Table 3
Number and proportion of macroinvertebrate identification sheets
in each state / territory audited for data entry errors
5
Number of samples from each State audited for live-sort
efficiency.
6
The lead agency of each state / territory audited and the relevant
field sampling methods/guidelines employed by each (as was
current during the field audit).
8
Table 4
Assessment of level of detail of field sheet and QA/QC procedures
in audited states / territories.
9
Table 5
Errors associated with data entry of taxonomic information and
abundances.
10
The number of samples and percentages of samples that failed
QA/QC criteria for each State and habitat.
11
Table 6
Appendix E
Page iii
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Appendices
Appendix E:1
Sites selected for audits
1
Appendix E:2
Details of taxonomic data entry errors
5
Appendix E:3
Problems encountered in Data Entry Error and WISE
Analysis
7
Appendix E:4
Whole Individual Sample Estimate (WISE) Analysis
9
Appendix E:5
Description of Habitats Sampled During State and Territory
Field Audits.
11
Lead Agency Internal QA/QC Methods
Live Picking Methods
20
35
Appendix E:6
Appendix E:7
Appendix E
Page iv
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Summary
This component of the AusRivAS Quality Assurance and Quality Control Project
examined the performance of State and Territory lead agencies in performing livesort protocols and taxonomic data entry for AusRivAS models. The general
findings were as follows:
•
documented QA/QC procedures are poorly developed and, although informal
QA/QC procedures were often applied, they did not always show in agency
performance. The level of QA/QC was difficult to assess as the level of
documentation and sophistication varied across states and territories. Also,
whilst some agencies had little or no documentation for QA/QC, their
practices at the time of assessment may have included undocumented
activities which conferred effective QA/QC on the data produced;
•
Approximately 75% of the live-sorted samples audited for this report passed
the QA/QC criteria. This figure is a significant improvement on the first
external QA/QC assessment of live-sorting procedures by Humphrey and
Thurtell (1997) where only 47% of the samples passed the criteria used in
the present report;
•
Approximately 30% of edge assessments and 21% of riffle habitat
assessments using the current AusRivAS procedures may be erroneous.
The remaining habitats assessed had approximately one tenth of samples
with inadequate live-sorted fractions. The percentage of samples passing
the QA/QC criteria in the present study are better than those reported by
Humphrey and Thurtell (1997) who had up to 50% failure rates for edge
samples and up to 31% of riffle habitats. It remains unclear why edge
habitats had a greater failure rate than any other habitat type; and
•
basic record keeping, such as sample labelling and identification sheets, was
found to be inadequate for some live-sort samples.
The following suggestions are for the consideration of the Department of the
Environment and Heritage and the state / territory agencies.
General recommendations:
•
consistent documentation should be established for procedures associated with
field and laboratory work, and QA/QC;
•
standard formats should be developed and applied for entry of field data, desk
assessments, and laboratory work to ensure that all required data is obtained
and calculated correctly;
•
a nationally agreed labelling standard should be developed to ensure that the
labelling of samples is adequate for correct identification, processing, storage,
and auditing;
•
nationally agreed QA/QC standards should be established embracing QA/QC
processes as an integral component of AusRivAS assessment processes;
provide guidance on field and laboratory procedures, equipment, sample
Appendix E
Page v
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
preservation for QA/QC assessment; target error rates for data entry and
taxonomic misidentification; as well as live-sort performance;
•
the performance standards and targets should form part of coordinated national
QA/QC program against which external auditing can be undertaken to review to
assess its efficacy; and
•
QA/QC training, and assessment of operator competencies, should be
undertaken in addition to existing AusRivAS training.
Live-sort Procedures
The substantial improvements in live-sort performance observed in this study
underscore the value of external auditing, nevertheless some areas for
improvement were highlighted, including:
•
improvements to live-sorting of edge samples;
•
extensive study of the relatively poor performance of the live-sort edge results in
turbid habitats.
Whilst it may be concluded that live-sort sampling bias is probably inherent to the
method there is a need to better understand the biases and their quantify their
implications for AusRivAS assessments.
Appendix E
Page vi
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
1
Introduction
1.1
Rapid Bioassessment
Rapid biological assessment (ie. rapid bioassessment - RBA) can be used to
describe two very different types of biological monitoring (Norris and Norris 1995).
The first is a continual monitoring situation to detect trigger or alarm levels of
organisms or toxicants. The second, and the subject of this review, refers to
expeditious sampling of biota with rapid delivery of assessment results (Norris and
Norris 1995). Benthic macroinvertebrate rapid bioassessment techniques and
procedures were developed in 1977 in conjunction with the commencement of the
RIVPACS program in October of the same year (Wright 1997, Davies 2001 pers
comm). The United States Environmental Protection Authority further developed
and expanded the RBA techniques to include fish and the work of Plafkin et al.
(1989) was later used by several other countries (eg. Australia and Canada) to
develop their own rapid bioassessment procedures (Norris and Norris 1995).
Rapid bioassessment offers several advantages over the more traditional
macroinvertebrate sampling methods, which involve a relatively large expenditure
of time to collect, process and identify biological samples (Lenat & Eaton 1991).
Rapid bioassessment reduces sampling effort, and therefore cost, by taking a
relatively large sample instead of several individual replicates and reduces the
number of organisms that must be processed by using a standardised subsampling procedure. In addition, rapid bioassessment programs often employ more
efficient methods of data analysis than traditional biological assessment programs
and produce results that are presented and summarised in a manner readily
understood by non-specialists (Resh et al. 1995).
Although rapid bioassessment of freshwater systems is now used in a number of
different countries, only the United States, the United Kingdom, Canada and
Australia conduct integrated, large-scale programs using comprehensive models
that integrate macroinvertebrate and physico-chemical data to compare test sites to
a benchmark or reference condition.
The bioassessment programs used by these four countries vary in the extent to
which they are applied and in their base method. However, all are based on similar
theory and all require established and documented quality assurance and quality
control systems to ensure that the integrity and veracity of the models used, and the
results they produce, are maintained.
Appendix E
Page 1
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
1.2
AusRivAS- National Context
The Australian River Assessment System (known as “AusRivAS”) was developed
by the Cooperative Research Centre for Freshwater Ecology, in partnership with
state and territory river management agencies, under the auspices of the National
River Health Program (NRHP) funded by the Commonwealth Government. The
NRHP was established in the Prime Minister’s Environment Statement in 1992
(O’Connor et al. 1996).
The objectives of National River Health Program are to:
•
provide a sound information base on which to establish environmental flows;
•
undertake a comprehensive assessment of the health of inland waters,
identify key areas for the maintenance of aquatic and riparian health and
biodiversity, and identify stressed inland waters;
•
consolidate and apply techniques for improving the health of inland waters,
particularly those identified as stressed;
•
develop community, industry and management expertise in sustainable water
resources management and raise awareness of environmental health issues
and the needs of our rivers.
The NRHP, initially called the National River Processes and Management Program
commenced in December 1992 (Davies 1994; O’Connor et al. 1996). The
Monitoring River Health Initiative (MHRI) – a key component of the NRHP – used
aquatic invertebrates to assess on a national level the ecological condition of
Australian rivers (Smith & Kay 1998). As part of the MHRI more than 1500
reference sites were sampled across all states and territories during 1994/96 to
establish the predictive AusRivAS models.
The second phase of the NRHP utilised the AusRivAS models to undertake the
First National Assessment of River Health (FNARH) (Smith & Kay 1998), later
referred to as the Australia-wide Assessment of River Health (AWARH). The
FNARH - AWARH commenced in 1997 and nearly 6000 sites have been assessed
nation-wide. Australia is the first country in the world to undertake such a
continental-scale assessment of the ecological health of its rivers (PIE 1998).
The bioassessment component of AusRivAS uses a series of models to predict the
composition of the aquatic macroinvertebrate community expected at a specific site
in the absence of environmental stress (expected taxa (E)). This is compared with
the macroinvertebrate community composition actually found at the site (observed
(O)). AusRivAS assessments are reported as the ratio of observed to expected
(O/E) taxa for the site, which are then assigned to a band indicating the extent to
which a site has been impacted.
Importantly, the veracity and national consistency of the AusRivAS–based river
health assessments are reliant on the collection and entry of accurate and precise
data.
The current project, the National River Health Program - AusRivAS Quality
Assurance and Quality Control Project, provides a national, external audit of data
collected by the various state and territory government agencies using AusRivAS to
assess river health in Australia. Information from this project augments previous
quality assurance and quality control (QA/QC) work conducted under the NRHP
Appendix E
Page 2
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
and focuses on QA/QC during field operations, during subsequent laboratory
sample processing and data entry, for both biological and environmental data
associated with AusRivAS bioassessment for the NRHP.
The audit encompasses all Australian states / territories and assesses their QA/QC
procedures, field and analytical techniques and methodologies, personnel training,
and data collection, validation and transcription. The purpose is to identify
deficiencies and areas for improvement within state and territory agencies to ensure
accuracy and consistency in the application of the AusRivAS model.
WATER ECOscience (formerly AWT Victoria) was commissioned to undertake the
National River Health Program - AusRivAS Quality Assurance and Quality Control
Project.
1.3
Project Outline
The National River Health Program - AusRivAS Quality Assurance and Quality
Control Project was part of the Toolbox component of the Australia-Wide
Assessment of River Health and involved a national, external audit of data collected
by the various state and territory agencies using AusRivAS to assess river health.
The objectives of the project were:
13. Assess and report on state agency performance for macroinvertebrate
sampling, processing and identification procedures against existing criteria.
14. Develop criteria for assessing agency performance in the collection of
environmental and habitat field data.
15. Assess and report on agency performance in the collection of environmental
and habitat data and macroinvertebrate data.
16. Provide feedback and advice on the problems of staff performance in
AusRivAS methods to state / territory agencies and to the new Training and
Accreditation Project of AusRivAS.
The project includes liaison with two other toolbox projects:
•
AusRivAS error analysis project - provide advice on actual error magnitudes
in environmental data to enable evaluation of the consequences of errors
associated with this type of data.
•
Training and Accreditation Project - liaison with the Principal Investigator to
ensure deficiencies detected in the implementation of AusRivAS methods by
lead agency staff can be addressed in any proposed training program.
The AusRivAS Quality Assurance and Quality Control Project involves two types of
audit, broadly categorised as: i) Veracity of macroinvertebrate sample processing
and taxonomic identification, and ii) Collection and entry of field environmental (ie.
habitat) data.
This report deals with the first audit process and includes the macroinvertebrate
sample processing, taxonomic identification, and entry of the biological data. In this
audit an initial analysis of data entry errors was conducted for all states and
territories, followed by an assessment of sorting accuracy for agencies us ing livesort procedures. Data entry errors were assessed by comparing the taxa and their
abundance recorded on the macroinvertebrate identification sheets at the time of
identification against that entered onto the agency databases.
Appendix E
Page 3
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
The accuracy of macroinvertebrate live-sort procedures was assessed for those
states / territories using this method. These include Queensland, New South
Wales, Tasmania, Victoria and Western Australia. The similarity of taxonomic
composition from the live-sort fraction was compared with later identifications and
counts on the residues retained from each sample. This comparison was
performed using the Whole of Individual Sample Estimate procedure (WISE)
(Humphrey & Thurtell 1997). The audit also included an assessment of quality
assurance and quality control (QA/QC) procedures for the validation and
transcription of biological data.
1.4
Report Structure
This report is associated with objectives three and four of the project (see Section
1.3 above), and covers Phase 6: Assessment of Lead Agency Macroinvertebrate
Sample Processing It includes auditing conducted over both the first and second
stage of the project.
The report reviews and reports on state / territory lead agency data entry for
macroinvertebrate taxa and abundance, as well as live-sort performance. The
effectiveness of QA/QC methods in controlling data entry errors and maintaining
acceptable approaches for field collection of site environmental data, with particular
emphasis given to data used as predictor variables in AusRivAS models.
The report addresses the following three areas:
4. Audit of data entry errors associated with a random selection of sites sampled
during the 1997 – 1999 FNARH program. Errors associated with missing or
extra taxa and incorrect abundances were identified.
5. Audit of live-sort residues from the five States using live pick protocols. This
included sub-sampling and picking of sample residues, and analysis of the
performance of live-sorting using the WISE procedure.
6. Discussion of the impact of the various data entry error types and live-sorting
performance on the outcome of AusRivAS river health assessments; and the
adequacy of agency QA/QC practices. Possible improvements and further
investigations are recommended.
Appendix E
Page 4
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
2
Methods
The audit of state / territory lead agency macroinvertebrate sample processing
commenced with an initial analysis of biological data entry errors. It was
considered important that the data entry audit precede the live-sort audit to correct
any data entry or taxonomic anomalies which may have confounded the outcome of
the live-sort audit.
The methods used by individual states and territories are listed in Table 16 below.
More detail on the methods can be found in Appendix E:5 and E:6.
2.1
Data Entry Audit
This audit assessed the accuracy with which lead agencies entered biological data
from macroinvertebrate identification sheets into their database. This was
performed by comparing the taxa and their abundance as recorded on the
macroinvertebrate identification sheets (at the time of identification) against entries
in the agency database.
The lead agencies for each state / territory were requested to forward a list of all the
sites sampled during the 1997 – 1999 FNARH program. From this list ten percent
of sites were randomly selected for assessment. The locations for these sites are
shown in the figures in Appendix E:1 together with all sites from the 1997 – 1999
FNARH program. The list of selected sites was then returned to the lead agencies
with a further request for photocopies of the original macroinvertebrate identification
sheets for these sites, together with an electronic copy of the data held in their
database. Due to time constraints and the volume of data collected from each state
/ territory not all of the randomly selected sites were assessed. The number and
proportion of data sheets audited for each state and territory are listed in Table 14.
Table 14 Number and proportion of macroinvertebrate identification sheets in each state /
territory audited for data entry errors
Number of
Audited
Sheets
Percentage of
Audited Sheets
Queensland
24
5%
New South Wales
52
Australian Capital
Territory
Victoria
State/Territory
Number of
Audited
Sheets
Percentage of
Audited Sheets
Tasmania
59
5%
5%
South Australia
25
3%
30
5%
Western
Australia
27
7%
59
4%
Northern
Territory
28
6%
State/Territory
The electronic data was printed in a format to match the identification sheets. The
paper copy of the data was marked with any errors or discrepancies found in
comparison with the taxa and abundances recorded on the sheets.
Data entry errors were assessed as follows:
•
Taxa missing TM) - taxa recorded on the identification sheet but not on the
database.
Appendix E
Page 5
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
Extra taxa (ET) – taxa recorded on the database but not on the identification
sheet.
•
Incorrect abundance (IA) – incorrect abundance entered on the database,
when compared to the identification sheet.
It should be noted that there is a possibility of recording two errors (TM and ET)
arising from one incorrect entry. For the case where one wrong taxa is entered
there would be two errors generated, one corresponding to the extra taxa and the
other to the missing taxa. To assess the situations where such error duplication
may have occurred would require detailed examination of agency records and was
therefore beyond the scope of this study.
Differences between the identification sheets and the database were individually
assessed to determine whether they represented genuine errors.
Some
discrepancies may have arisen from re-identification or QA/QC assessment by the
lead agency and these were excluded from the analysis.
2.2
Live-sort Audit
Sample Collection
A total of 250 residue samples were assessed for the live-sort audit from the five
States which conduct live picking (Table 15); the location of these sites are
displayed in Appendix E:1 together with all sites from the 1997 – 1999 FNARH
program. The collection of samples by each of these States is described in the
manuals listed in Table 16, extracts of which can be found in Appendix E:7.
Table 15
Number of samples from each State audited for live-sort efficiency.
State
Number of Audited Samples
Percentage of Sites Audited
Queensland
55
21%
New South Wales
58
5%
Victoria
51
12%
Tasmania
59
12%
Western Australia
27
5%
Residue Processing
The live-sort residues were sub-sampled in the laboratory at WATER ECOscience
using equipment and procedures described by Marchant (1989). Subsampling
involved placing the whole sample in a watertight box (35cm x 35cm), the bottom of
which was divided into 100 cells. The box was sealed, gently shaken and then
animals and other material removed from individual cells selected using random
number tables.
A total of 200 individual macroinvertebrates were subsampled in this manner
according to the following criteria. The first 50 animals were identified, picked into
vials and recorded. The following 150 animals were identified, but only new taxa
not represented in the first 50 animals collected were picked and recorded. Finally,
the remaining residue was placed in a large white sorting tray and scanned for 15
minutes with the aid of a magnification lamp. Any new taxa observed were picked
Appendix E
Page 6
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
and recorded. Recorded information includes a breakdown of what was collected
during each component and the total overall percentage of residue that was
subsampled. Pick results for each residue were later used to calculate the WSE for
the sample from which the residue originated.
Macroinvertebrate identification was to family level, except for Chironomidae
(Diptera), which was identified to subfamily, and Collembola, Arachnida, and
Oligochaeta which were identified to class or order level in accordance with
accepted convention. Data was entered, validated and stored on the Zoological
Database (MS Access) maintained by WATER ECOscience.
Data manipulations
The data was converted into the form necessary for input to the WISE database for
assessment of the live-sort performance. Initially the data was checked to ensure
that site, date and habitat details matched on both the Live Sort (LS) data and the
Residue (RES) data for each sample. The WISE database will not run if there are
any discrepancies in these details.
Significant difficulties were encountered during the collation and manipulation of
datasets sent to WATER ECOscience (see Appendix E:3). Generally these
difficulties were associated with inadequate or erroneous detail on sample labels,
identification sheets or database records; or taxa and site coding.
All the data was then checked for duplicate records. The source of the duplication,
identification sheets or the agency database or both, was then established and the
appropriate corrections made.
The WISE procedure was then run through one iteration to generate a printed
output of the combined LS/RES data. These summary output tables were then
assessed for taxonomic anomalies, generally coding errors, and appropriate
corrections made to the database.
As mentioned previously, most macroinvertebrate identifications were taken to
family level and this is the taxonomic level at which data was inputted into the WISE
database. However, if an animal was small or damaged and could not be identified
to family level, the identification was taken as far as possible (usually order level).
In the situation where a particular group of animals was represented in the dataset
by both family level identifications and identifications at a lower taxonomic level
(e.g. order), then all the animals within this group were combined back to the lowest
common level. For example, if Zygoptera (unidentified) was present in either the
LS or RES data then all families present within the order Zygoptera were summed
under the heading of Zygoptera (unidentified). This was performed to avoid the
possibility that taxa belonging to a representative of the family within the order
(unidentified) was already represented in either of the datasets (i.e. a taxa would
not be represented ‘twice’ in the dataset).
The WISE procedure was then re-run through one hundred iterations.
WISE assessment
The WISE procedure uses two criteria to assess the performance of live-sorting the Live-Sort / Whole of Sample Estimate (LS/WSE) taxa number ratio and the
Bray-Curtis dissimilarity measure (BC). These are described below and in more
detail in Appendix E:4.
Appendix E
Page 7
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
1. Live Sort / Whole Sample Estimate ratio
The LS/WSE ratio has been used previously to evaluate the performance of
different sampling methods for the AusRivAS (Humphrey and Thurtell,
1997). The criterion for acceptable QA/QC of samples was that the number
of taxa found in the LS must not be lower than 20% of those found in the
WSE. That is, the LS/WSE must be greater than or equal to 0.8 for the
number of live-sorted taxa to be considered representative of the whole
sample.
2. Bray-Curtis dissimilarity
Bray-Curtis (BC) dissimilarity values are derived by comparing the
macroinvertebrate structure of each live-sorted sample with that remaining
in the sample residue. Adjustment and normalisation of the dissimilarity
values was performed as described in Appendix E:4.
The BC value shows the similarity in the range of taxa collected during the
live-sort compared to taxa present in the whole sample. The criterion for
acceptable QA/QC performance based on presence-absence data is a
maximum Bray-Curtis dissimilarity level of 0.35 (Humphrey and Thurtell,
1997). This criterion was used to assess the QA/QC of samples from the
different States, values greater than 0.35 were taken to indicate live-sorted
fractions unrepresentative of taxa in the whole sample.
2.3
QA/QC Procedures
The methods (including QA/QC procedures) employed by each state / territory are
summarised in Table 16, an assessment of the level of QA/QC is provided in Table
17 and further details can be found in Appendix E:6.
Table 16 The lead agency of each state / territory audited and the relevant field sampling
methods/guidelines employed by each (as was current during the field audit).
State or Territory
Lead Agency
Relevant Field Sampling
Methods/Guidelines
Queensland:
Department of Natural
Resources & Mines
Biological Monitoring and Assessment
of Freshwaters using
Macroinvertebrates, 1997
New South Wales:
Environment Protection
Authority
NSW AusRivAS Sampling and
Processing Manual, 2001
Australian Capital
Territory:
Environment ACT (CRC for
Freshwater Ecology)
ACT AusRivAS Sampling and
Processing Manual, 2000
Victoria:
Environment Protection
Authority
EPA Publication 604, 1998
Tasmania:
Department of Primary Industry,
Water and Environment
Tasmanian AusRivAS Sampling and
Processing Manual, 2000
South Australia:
Environmental Protection
Authority (Australian Water
Quality Centre)
SA AusRivAS Sampling and
Processing Manual, 2000
Western Australia:
Conservation and Land
Management
AusRivAS in Western Australia, 1998
Northern Territory:
Department of Lands, Planning
& Environment
Northern Territory methodology, 2000
Appendix E
Page 8
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
3
Results
3.1
QA/QC procedures
The level of detail and thoroughness of QA/QC procedures varied between lead
agencies. Table 17 provides an assessment of agency QA/QC based on both
documented and accepted but undocumented practices. Most state / territory
agencies were judged to have adequate to good procedures for animal collection,
live-sorting and laboratory QA/QC procedures; NSW, Tasmania (excellent),
Victoria, Queensland and Western Australia (good) had established procedures for
data entry and validation. Sampling QA/QC procedures were mainly poor in all
States and only adequate in Tasmania. Live sorting and laboratory QA/QC
procedures were judged adequate to excellent. Three States, Victoria, Australian
Capital Territory and South Australia, did not have documented QA/QC procedures
for data entry or validation. However, in Victoria entered data is checked against
data sheets a procedure which was not documented in the Victorian RBA manual.
The documented data entry or validation procedures in the Northern Territory were
poor and the remaining States had good or excellent procedures.
Table 17 Assessment of level of detail of field sheet and QA/QC procedures in audited
states / territories.
QLD
NSW 2
ACT
VIC
TAS3
SA
WA
NT
444
4444
4444
4
444
44
444
4444
– Sampling
4
4
4
4
44
4
4
4
– site assessment
44
444
44
4
444
4
4
4
– live-sorting procedure
44
X
N/A
44
444
N/A
444
N/A
444
4444
4444
44
444
444
44
4444
444
4444
X
4441
4444
X
444
4
4
4
4
4
4
4
4
Documented Procedures
Field sheet procedure
(ie. Detailed instructions
included with field sheets)
Field QA/QC procedures
Lab QA/QC procedures
Data entry/validation
– overall
– independent check
– range check5
4
4
Unless indicated otherwise ratings are based on QA/QC procedures documented in individual state /
territory AusRivAS manuals. Level of detail was subjectively assessed by WATER ECOscience
auditors and included supplementary information on accepted practices within agencies which are not
formally documented.
Number of ticks indicates thoroughness of documented procedures: 4 - poor, needs more detail;
44 - adequate; 444 - good; 4444 - excellent; X - indicates no established procedures; N/A – not
applicable.
1 After data is entered it is checked against data sheets. However, this is not documented in the
Victorian RBA manual.
2 Information based on Waddell (2001). QA/QC procedures are not documented in the NSW
AusRivAS manual.
Appendix E
Page 9
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
3 Some QA/QC procedures are detailed in Tasmanian AusRivAS sampling manual, however, ratings
given in table are primarily based on procedures documented in Krasnicki et al. 2001).
4 Based on supplementary information provided, procedures not documented.
5 Data entry/validation checks discussed in Section 4.1.
3.2
Data Entry Audit
A total of 304 identification sheets and 6362 separate data entries were assessed
for the three error types (Table 18). Overall, 11% of the identification sheets had
errors and less than 1% of all data entered were incorrect. The most common error
type was the incorrect abundance of animals in the sample being entered, nearly
twice the rate of the other error types.
Table 18
Errors associated with data entry of taxonomic information and abundances.
State/Territory
QLD
NSW
ACT
VIC
TAS
SA
WA
NT
Overall
Number of taxa entries made
517
1086
553
1376
1212
739
393
486
6362
Number of id sheets assessed
24
52
30
59
59
25
27
28
304
Category
Number of id sheets with errors
2
11
0
2
7
4
3
5
34
17
15
22
21
21
23
30
18
21
Taxa missing -electronic (TM)
0
6
0
0
3
0
3
2
14
Extra taxa (ET)
1
3
0
0
4
3
0
1
12
Incorrect abundance (IA)
1
4
0
2
3
1
1
10
22
Total entries made
0.4
1.2
0.0
0.1
0.8
0.5
1.0
2.7
0.6
Number of sheets
8.3
21.2
0.0
3.4
11.9
16.0
11.1
17.9
11.2
Average number of taxa per sheet
Error types
Error rates (%)
The total number of entries and number of data sheets with errors varied between
the states / territories. The Northern Territory had the greatest error rate, at 2.7% of
all data entered, and the Australian Capital Territory the lowest with no data entry
errors recorded. The greatest percentage of data sheets with errors was found in
New South Wales with approximately one in five data sheets containing errors.
3.3
Live-sort Audit
A total of 250 samples were assessed for QA/QC according to the L/W and BC
criteria (Table 19). Of the 250 samples, 24% failed either the L/W criteria (less than
80% of taxa recorded in the LS compared with the WSE) or the BC criteria (at least
65% similarity between the two sample fractions) or both. The minimum failure rate
for any habitat in any State was for Queensland riffles where 7% of samples failed.
Appendix E
Page 10
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Table 19 The number of samples and percentages of samples that failed QA/QC criteria
for each State and habitat.
Habitat
State
QLD
NSW
VIC
TAS
WA
Data
Bed
Channel
Edge
Macrophyte
Riffle
Total
Number of samples
14
27
14
55
L/W failed (%)
BC failed (%)
7
14
11
15
0
7
7
13
At least one failure (%)
14
15
7
13
Number of samples
31
27
58
L/W failed (%)
45
41
43
BC failed (%)
At least one failure (%)
16
45
0
41
9
43
Number of samples
L/W failed (%)
30
20
21
14
51
18
BC failed (%)
20
5
14
At least one failure (%)
27
14
22
Number of samples
27
32
59
L/W failed (%)
BC failed (%)
30
4
16
0
22
2
At least one failure (%)
30
16
22
Number of samples
L/W failed (%)
BC failed (%)
At least one failure (%)
Overall Number of samples
L/W failed (%)
20
7
27
5
14
7
15
15
14
14
15
15
14
7
20
5
115
27
7
14
94
20
250
21
BC failed (%)
14
15
14
14
2
10
At least one failure (%)
14
15
30
14
21
24
Edge habitats had the greatest failure rate of the five types, with a 30% failure rate,
followed by riffles (20%), channels (15%) and both macrophytes and bed habitats
with 14% failure. The high failure rate of edge and riffle habitats appeared to be
associated with samples from NSW where 45% and 41% failed respectively.
However, the failure rate of edge samples was still nearly double that for riffles in
Queensland, Victoria and Tasmania.
Overall, the percentage of samples failing the L/W criteria was approximately
double that of the percentage of samples failing the BC criteria and 7% failed both.
The L/W criteria was failed more often than the BC criteria in New South Wales and
Tasmania where the L/W failure rates were approximately 5 times and 10 times the
BC rates respectively. The two criteria had similar failure rates for the other three
States.
Appendix E
Page 11
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
4
Discussion
4.1
QA/QC Procedures
The assessment of QA/QC procedures (Table 17) were based on both the
documented procedures detailed in Appendix E:6 and supplementary information
provided on undocumented standard practices within each agency. In assessing
the level of QA/QC it must be recognised that, whilst some agencies had no
documented procedures, their accepted practices encompass a high degree of
QA/QC. Also, where there is low staff turnover and rigorous internal training a high
level of QA/QC may be sustained without adequate documentation.
Practices for data entry mostly included independent checking of entered data
against the source documents. During the MRHI phase in NSW double entry of
data was practiced, however in 1997 this was replaced by an independent check.
In some States additional ‘range’ checks were also undertaken. These include
identification and checking of outliers or data outside acceptable ranges. The
criteria for outliers or acceptable ranges were obtained by plotting frequency
histograms for the parameter of interest. Also, electronic data entry forms have
been set up to mimic the layout of field and laboratory data sheets, and thereby
minimise errors (NSW).
4.2
Data Entry Audit
The overall percentage of data correctly entered from taxonomic identification
sheets was high, with less than 1% of data entered with errors and approximately
90% of all data sheets with no errors. However, the percentage of sheets with data
entry errors varied among states / territories with up to 20% of sheets with errors
within one State. Data entry errors have the potential to affect the outcome of river
health assessments using AusRivAS.
Two of the three error types evaluated, errors associated with taxa on the
identification sheet but not entered onto a database (TM error), and taxa entered
onto the database but not on the identification sheet (ET error), have the greatest
potential to affect river health assessment. Errors associated with the incorrect
abundance being entered are unlikely to affect river health assessment as only
presence/absence data is used in the AusRivAS predictive models.
The total number of these two types of error (missing or extra taxa) that may affect
river health assessment accounted for approximately half of all data-entry errors
found, with up to 17% of data sheets in one State affected by these errors. This
suggests that as many as 20% of site assessments may be affected by data entry
problems within any one State. In addition, of all the data entered onto databases,
only 0.4% (26 out of the 6362 entries made) were associated with TM and ET error
types and generally only one taxon was affected per sample (see Appendix E:B),
and two states / territories did not record any of these error types.
An AusRivAS site assessment is based upon the ratio of animals found in a sample
(Observed) to that predicted (Expected) to occur, the O/E ratio. Since about 20 taxa
occur on average in samples, a TM or ET error, affecting only one taxon per
Appendix E
Page 12
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
identification sheet entry, may only increase or decrease an O/E ratio by 5%. This
level of bias in the O/E ratio is unlikely to affect the health assessment of most
streams or rivers. However, a single TM or ET data entry error may seriously affect
the assessment of a site with a small number of predicted taxa.
Using information from routine analytical laboratories, the acceptable rates of data
entry or transcription error are probably in the range of 0.1%, or one error in 1000
entries. Using this rate as a target for AusRivAS QA/QC, nearly all of the states /
territories exceeded this error rate. Also, the error rate did not appear to correlate
with the degree of QA/QC practiced by each agency (Table 17).
A number of practices can be implemented to reduce the rate of data entry errors,
including:
•
independent checking of entered data against the source documents;
•
double entry of data and resolution of mismatches;
•
formatting data entry screens to mimic the layout of the field or data sheets;
•
range checks based on expected range for data (at a particular site);
•
use of optical character recognition (OCR) software in combination with redesign of the field or data sheets.
Independent checking by another operator should be the minimum level of QA
practiced. Double entry of data does not appear to offer significant benefits over
and above independent checks and may be difficult to implement. The use of entry
screens which match the data sheets should be achievable with current software
packages. This is a basic measure which should be implemented to reduce error
rates.
OCR has been used successfully at NSW Fisheries for data entry however error
rates are highly dependent on the design of the entry forms and the data being
entered. The setup and operating costs may not justify the improvements in data
accuracy at this stage of development of the technique. It is possible that future
developments will make OCR more efficient and cost-effective.
Other potential sources of error, such as the coding system used for
macroinvertebrate families, may also make significant contributions to the overall
error rates. Some coding systems are numeric or counter-intuitive and provide a
greater opportunity for operators to enter or write the wrong taxa code. Adoption of
a uniform national coding system utilising Vic EPA’s taxa codes (to species level, or
to the level of identification) by all states and territories would minimise coding
errors. Alternatively, consideration should be given to using family or species
names rather than surrogate alpha-numeric codes.
Sampling records
A number of basic problems with data recording by agencies were encountering
during the audit process. These included inadequate and ambiguous information on
residue labels, identification sheets and database entries; as detailed in Appendix
E:C. Because of these inadequacies, a number of residue samples were excluded
from the live-sort audit. A matter of further concern is that the number of
deficiencies appeared to be greater for the more experienced agency teams.
Appendix E
Page 13
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
4.3
Live-sort Audit
Approximately 75% of the live-sorted samples audited for this report passed the
QA/QC criteria. This figure is an improvement on the first external QA/QC
assessment of live-sorting procedures by Humphrey and Thurtell (1997) where only
47% of the samples passed the criteria used in the present report. This suggests
that improvements in AusRivAS procedures implemented after the first external
audit have generally been successful and is an endorsement of agency efforts and
the external audit process. The improvements undertaken following the 1997 audit
included a minimum animal count of 200 or extended counting time, and the
establishment of formal training facilities for AusRivAS assessments. However, the
present level of live-sorted sample failures still means that as many as a quarter of
all river health assessments using AusRivAS are potentially incorrect (generally by
around +/-5% of overall condition) due to inadequate picking procedures generally
by as much as 5% of assessed overall condition) due to inadequacies in picking or
the live-pick protocol itself.
The main habitats affected by inadequate picking in the present report were edge
and riffle habitats with 30% and 21% of samples failing the QA/QC criteria. This
suggests that approximately a third of edge assessments and one fifth of riffle
habitat assessments using the current AusRivAS procedures may be erroneous.
The remaining habitats assessed had approximately one tenth of samples with
inadequate live-sorted fractions. The percentage of samples passing the QA/QC
criteria in the present study are better than those reported by Humphrey and Thurtell
(1997) who had up to 50% failure rates for edge samples and up to 31% of riffle
habitats. It is unclear why edge habitats had a greater failure rate than any other
habitat type but Humphrey and Thurtell (1997) suggested that edge habitats tended
to have greater amounts of silt and detritus and therefore animals were harder to
recover from the sample. However, the results of the present report and those of
Humphrey and Thurtell (1997) suggest that significant improvements need to be
made in the edge sample protocols and live-sort procedures. This may include
washing silt from samples; or sorting through detritus by placing the whole sample in
a bucket and sifting through smaller amounts at a time in the sorting tray; or by
increasing the picking time.
The percentage of samples failing the L/W criteria (21%) was approximately double
the percentage of samples failing the BC criteria (10%). The failure of the L/W
criteria has probably more serious implications for AusRivAS assessments than the
failure of the BC values. This is because the O/E index uses the number of taxa
collected in the live-sort in the numerator of the ratio, ie. the number of observed
taxa. If the number of taxa in the live-sort was less than that collected in the whole
sample the site would be assessed as in poorer condition that it actual was. These
results suggest that as many as 20% of all sites, and 40% of cases in New South
Wales, might be classified as in poorer condition than they really are. However the
same biases may equally apply to the data for the reference sites. Therefore
whether a different health assessment (poorer or better) would arise will depend on
whether these biases may have moved one or other of the reference site or the
actual test site out of band and in which direction. The net result is that there may
not be significant changes to classifications.
Humphrey and Thurtell (1997) indicated that small and cryptic taxa were commonly
missed during the live-sorting process and suggested that extra training should be
implemented to increase staff awareness of these taxa. In addition, changes to the
live-sort protocol, namely the introduction of a target sample size of 200 animals
rather than 30 minutes sampling, were implemented at this time (MRHI meeting
Canberra, February 1997). A limited assessment of the data from this study
Appendix E
Page 14
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
suggests that there was no significant difference between taxa represented in livesort fractions versus the whole sample residue. It is not clear whether this indicates
that the implementation of training, minimum counts and other improvements have
been fully effective. Nevertheless, the extent to which small and cryptic taxa still
contribute to the percentage of samples failing the L/W criteria should be the subject
of on-going audit and assessment, to determine the need for further training and
improvements in this area.
Further, the Protocol Development Project (WATER ECOscience, 2003) found that
whilst live-sort protocols were acceptable for clean sites, the presence of turbidity at
some habitat sites led to poor live pick performance. Four protocol improvements
were studied:
1. Use of visual aids during sorting;
2. Sorting fine and coarse fractions separately;
3. Use of a minimum time/number of animals; and
4. Selection of a minimum number of chironomids
The study concluded that the proposed variants to the live-sorting protocol provided
no increase in efficiency or veracity of results and that the standard method
appeared as effective at retrieving an accurate representation of taxa at any one
particular site. Furthermore, that poor live pick performance appeared to be related
to a small number of animals in the live-sorted fraction and not the live-sort protocols
themselves. Other factors, including lack of homogeneity between amalgamated
samples and the efficacy of the live-sort protocols in some sample types, were also
thought be contributors to poor live-sort performance. A more extensive study,
involving a much larger number of sites and encompassing all live-pick Australian
states and territories, was recommended to establish the validity and generality of
turbid site live-sort QA/QC protocols. A more extensive study could include
assessment of the types of taxa found in whole sample residues but not in the livesort fraction to ascertain if there is any bias for different habitat types in the different
States.
An alternative approach to this issue of probable bias, is to accept that live-sorting in
the field and laboratory examination under a microscope are essentially two different
methods which produce different types of data (Victorian EPA, 2001). Victoria
considered that, like all sampling methods, there will be some inherent bias with livesorting with some taxa consistently over or under-estimated. The EPA concluded
that samples should be collected in a consistent manner and the biases well
understood.
Appendix E
Page 15
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
5
Conclusions and Recommendations
The following key findings, conclusions and recommendations arise from this
project, which assessed the QA/QC performance of live-sort protocols and
taxonomic data entry. Whilst there are some deficiencies in data entry procedures
noted, improvements in the live-sort QA/QC since the previous audit in 1997 are a
positive finding.
5.1
Key Findings and Conclusions
Documented QA/QC procedures are poorly developed and, although informal
QA/QC procedures were often applied, they did not always show in agency
performance; The level of QA/QC was difficult to assess as the level of
documentation and sophistication varied across states and territories. Also, whilst
some agencies had little or no documentation for QA/QC, their practices at the time
of assessment may have included undocumented activities which conferred
effective QA/QC on the data produced;
Rates for data entry errors are higher than desirable, and vary significantly between
agencies with an overall error rate of 0.6%. The most common error was incorrect
entry of the abundance data, which does not influence the AusRivAS assessment.
For the other error types (missing or extra taxa), since there are about 20 taxa per
sample, a single such error would only increase or decrease the Observed /
Expected animal ratio by 5%. This level of bias in an AusRivAS assessment is
unlikely to affect the health assessment of most streams or rivers. Nevertheless, a
single TM or ET data entry error may significantly affect the assessment of a site
with a small number of predicted taxa. There appeared to be no correlation
between the assessed level of QA/QC practices in each agency and the percentage
of error rates.
Inadequacies in basic record keeping, such as sample labelling and identification
sheets, found during live-sort audits are unacceptable.
Accordingly, the following recommendations are suggested for the consideration of
the Department of the Environment and Heritage and the state / territory agencies.
For completeness the recommendations include aspects of QA/QC not directly
related to observations arising from the audit.
5.2
General Recommendations
A QA/QC officer should be nominated for each state / territory agency, to coordinate
all such activities and maintain consistency amongst operators, determine whether
reproducible results are being obtained and performance target met.
Documentation and data entry
Procedures for field work, desk-based assessments and laboratory work should be
adequately documented and made available for staff to follow in both the laboratory
and the field. Standard forms should be developed for entry of field assessments
and laboratory work to ensure that all required data is obtained and that calculations
Appendix E
Page 16
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
are carried out correctly. Similarly, nationally standardised labels should be used for
field samples to ensure that they can be correctly identified and processed. All field
samples should be either logged on field sheets or logged on return to the laboratory
so that they can be tracked through processing and to provide an inventory of stored
samples.
Data entry and validation should be included in documented QA/QC procedures.
These processes should include data validation by printing out database records for
independent checking by another operator against the original data sheets.
In addition, these documented procedures should include scanning for erroneous,
extreme, outlier or ‘out of range’ values based on historical data or collective
experience.
Performance targets
Targets for acceptable rates of error should be nominated immediately on an interim
basis and adopted nationally as a matter of priority. Based on experiences gained
in both this and the previous 1996 audit, the following are recommended as interim
targets for adoption by state and territory lead agencies. They are not meant to be
definitive but are designed to ensure adequate QA/QC targets and benchmarks are
incorporated into AusRivAS processes while more considered standards are
developed nationally. The objective of these interim targets is to improve the quality
of AusRivAS outputs and provide end users of AusRivAS outputs with data of known
levels of error.
It is recommended that, at a minimum, state / territory agencies and those
undertaking AusRivAS (such as the Murray-Darling Basin Commission and Natural
Resource Management regional organisations) seek to achieve the following interim
national QA/QC standards:
40. QA/QC processes are seen as an integral component of AusRivAS
assessment processes and are be incorporated into AusRivAS protocols;
41. A suitably qualified and experienced QA/QC officer should be nominated for
each state / territory agency to coordinate all such activities and maintain
consistency amongst operators, determine whether reproducible results are
being obtained, and performance target are being met;
42. Samples and accompanying field sheets from 10% of all AusRivAS sites
being sampled should be preserved for QA/QC assessment (5% for internal
QA/QC monitoring and 5% for external auditing). Where fewer than 10 sites
are involved, samples from at least one site should be preserved;
Performance targets should require that:
43. Fewer than 10% of field and taxonomic identification sheets have errors;
44. Less than 0.1% error rate for AusRivAS data transcription (data entered into
state / territory data bases);
45. Less than 10% taxonomic misidentification at family level, and less than 10%
taxonomic misidentification at species level;
46. Less than 20% of samples fail either LS/WSE or Bray-Curtis criteria; namely
(a)
(b)
Appendix E
LS/WSE ratio equal to or greater than 0.8,
BC dissimilarity level equal to or less than 0.35.
Page 17
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Also, QA/QC training should be undertaken, in addition to existing AusRivAS
training, for all operators involved in field and laboratory procedures and data entry.
This training should include ongoing assessment of competencies similar to that
incorporated into the base AusRivAS training program.
It is recommended that all such targets and procedures be integrated into a
nationally coordinated QA/QC programs, and that this program then be reviewed to
assess its efficacy through a process of external auditing.
Live-sort procedures
It is recommended that the external audit process undertaken in 1996 (Humphrey
and Thurtell, 1997) be continued on a regular basis, to support an on-going
improvement in the practices and QA/QC associated with AusRivAS assessments.
The substantial improvements in live-sort performance since 1996 are validation of
the external audit process, which includes feedback of results to participants and
actions to improve performance.
As edge samples were the main habitat type failing live-sort QA/QC significant
improvements need to be made in the way that edge samples are live-sorted.
A more extensive study is recommended to determine the poor performance of the
live-sort protocol in particular turbid habitats. This study should include a larger
number of sites for the States using live-sort protocols; and assess the types of taxa
found in whole sample residues but not live-sort fractions to determine whether
there is any bias for different habitat types in the different States. This more detailed
study would provide an understanding of the biases present and whether their
elimination or further reduction is warranted.
It is probable that live-sort sampling bias is inherent to the method and that possible
future improvements will not further reduce this bias. A recent assessment of
potential improvements to live-sort protocols were judged to offer no improvement to
the efficiency of the technique or the validity of the results obtained. Nevertheless,
should this inherent bias be accepted, there is a need to better quantifying the livesorting bias and its implications for AusRivAS assessments.
Appendix E
Page 18
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
6
References
Davies, P. (1994) River Bioassessment Manual Version 1.0: National River
Processes and Management Program, Monitoring River Health Initiative.
LWRRDC, Canberra.
Humphrey, C. and Thurtell, L. (1996) External QA/QC of MRHI agency
subsampling and sorting procedures. Milestone Report to LWRRDC,
December 1996.
Lenat, D.R and Eaton, L.E. (1991) Comparison of a rapid bioassessment method
with North Carolina’s macroinvertebrate collection method. Journal of the
North American Benthological Society, 10: 335-338.
Marchant, R. (1989). A subsampler for samples of benthic invertebrates. Bulletin
of the Australian Society for Limnology 12: 49-52.
Norris, R.H. and Norris, K.R. (1995). The need for biological assessment of water
quality: Australian Perspective. Australian Journal of Ecology, 20: 1-6.
O, Connor, N.A., Lloyd, L.N. and Moore, S.J. (1996). Evaluation of the National
River Health Program. WATER ECOscience Report No. 56/96.
PIE-Newsletter of Australia’s International and National Primary Industries R & D
Organisations. International Acclaim for Aussie River Program August 98
edition. http://www.affa.gov.au/pie/98autumn/
Plafkin, J.L., Barbour, M.T., Porter, K.D., Gross, S.K. and Hughes, R.M. (1989).
Rapid bioassessment protocols for use in streams and rivers. Benthic
macroinvertebrates
and
fish,
EPA/440/4-89/001,
United
States
Environmental Protection Authority, Office of Water Regulations and
Standards, Washington, DC.
Resh, V.H., Norris, R.H. and Barbour, M.T. (1995). Design and implementation of
rapid assessment approaches for water resource monitoring using benthic
macroinvertebrates. Australian Journal of Ecology, 20: 108-121.
Smith, M. and Kay, W. (1998). AusRivAS in Western Australia. An overview of the
development and use of AusRivAS models for assessing river health in
Western Australia. Department of Conservation and Land Management, WA
Wildlife Research Centre, Wanneroo.
Victorian EPA (2001). The Australia Wide Assessment of River Health; Final Report
of the National River Health Program from Victoria. Environment Protection
Authority, Victoria.
Waddell, N. (2001) Australia Wide Assessment of River Health: New South Wales
Program. Supporting Document to Final Report: Internal Quality Control and
Quality Assurance Programs for the NRHP in NSW. New South Wales
Environment Protection Authority.
WATER ECOscience (2003). AusRivAS Protocol Development and Testing Project:
Extended Analysis. WATER ECOscience Report No. 3055/03.
Wright, J.F. (2000). An introduction to RIVPACS. In: Assessing the biological
quality of fresh waters: RIVPACS and other techniques. Wright, J.F.,
Appendix E
Page 19
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Sutcliffe, D.W. & Furse, M.T (Eds.) Freshwater Biological Association, UK.
pp 1-24.
Appendix E
Page 20
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Appendix E:1
Sites selected for audits
The MRHI sites in each state / territory and the sites selected for auditing are shown
in Figures 1-9.
N
MRHI Sites
Audited Sites
Figure 1 Location of the 24 sites assessed for the biological data entry audit and the 55
sites for the live-pick audit for Queensland in relation to all sites sampled during the MRHI
program
MRHI Sites
Audited Sites
Figure 2 Location of the 52 sites assessed for the biological data entry audit and the 58
sites for the live-pick audit in New South Wales in relation to all sites sampled during the
MRHI program.
Appendix E - appendices
Page 1
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
MRHI Sites
N
Audited Sites
Figure 3 Location of the 30 sites assessed for the biological data entry audit in the
Australian Capital Territory in relation to all sites sampled during the MRHI program.
MRHI Sites
Audited Sites
N
Figure 4 Location of the 59 sites assessed for the biological data entry audit and the 59
sites for the live-pick audit for Victoria in relation to all sites sampled for the MRHI program.
Appendix E - appendices
Page 2
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
MRHI Sites
N
Audited Sites
Figure 5 Location of the 59 sites assessed for the biological data entry audit and the 59
sites for the live-pick audit for Tasmania in relation to all sites sampled during the MRHI
program.
N
MRHI Sites
Audited Sites
Figure 6 Location of the 25 sites assessed for the biological data entry audit for South
Australia in relation to all sites sampled during the MRHI program.
Appendix E - appendices
Page 3
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
MRHI Sites
N
Audited Sites
Figure 7 Location of the 27 sites assessed for the biological data entry audit and the 27
sites for the live-pick audit for Western Australia in relation to sites sampled for the MRHI
program.
N
MRHI Sites
Audited Sites
Figure 8 Location of the 28 sites assessed for the biological data entry audit for the
Northern Territory in relation to all sites sampled during the MRHI program.
Appendix E - appendices
Page 4
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Appendix E:2
Details
of
entry errors
taxonomic
data
Errors are presented as follows:
Site X - Error type (number of errors): Taxa name (value entered)(correct value)
Queensland
Site 138110A - Incorrect abundance (1) - Caenidae (10 - 4)
Site 136012A - Extra taxa (1) - Ochteridae
Australian Capital Territory
No errors
New South Wales
Abundances of over 26 entered as 99. Sphaeridae changed to Corbiculidae on
database
Site GWYD08 – Taxa missing (1): Chironomidae unidentified (49)
Site MURR126 – Taxa missing (1): Daetidae (159)(99)
Site RICH102 - Taxa missing (1): Psephenidae (6)
Site RICH515 - Taxa missing (1): Corydalidae (2)
Site LACH508 - Taxa missing (1): Diptera unidentified (1)
Site LACH102 - Taxa missing (1): Telephlebiidae (1)
Site TOWA570 - Incorrect abundance (1): Scirtidae (9 - 99)
Site CLYD101 - Incorrect abundance (1): Hydropsychodidae (26 - 99)
Site TURO602 - Incorrect abundance (1): Chironominae(32 - 5)
Site BIDGM1 - Incorrect abundance (1): Gripopterigidae (5 -4)
Site LACH115 - Extra taxa (3): Atyidae, Hydracarina, Parastacidae
Victoria
Site YY1 (15/12/98) - Incorrect abundance (1): Elmidae (A+(L))(14 - 15)
Site EEX (09/04/97) - Incorrect abundance (1): Notonemouridae (5 - 4)
Appendix E - appendices
Page 5
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Tasmania
Site E31 - Taxa missing (1): Tabanidae (2)
Site FT23 - Taxa missing (1): Elmidae (1)
Site G20 - Incorrect abundance (1): Parameletidae (6 - 5)
Site GT18 - Incorrect abundance (1): Turbellaria (1 - 2)
Site A15 - Incorrect abundance (1): Philorheithridae (7 - 8), Extra taxa (1):
Odontoceridae (1)
Site FT19 - Missing taxa (1): Philorheithridae (1), Extra taxa (1): Odontoceridae (1)
possibly wrong taxa entered
Site H10 - Extra taxa (2): Diamesinae, Helicophidae
South Australia
Site G106 - Incorrect abundance (1):
Site J9 - Extra taxa (1): Cherax destructor (1)
Site I51 - Extra taxa (1): Ranatra (1)
Site I123 - Extra taxa (1): Ceratopogonidae pupae (1)
Western Australia
Site EWL01 - Incorrect abundance (1): Dytiscidae (78 - 76)
Site HIL01 - Taxa missing (2): Diptera (unidentified)(1), Lepidoptera (1)
Site SHA07 - Taxa missing (1): Trichoptera (unidentified)(1)
Northern Territory
Site AD08 (17/07/98) - Taxa missing (1): Lindeniidae (1)
Site AD08 (16/10/98) - Incorrect abundance (4): Oligochaeta (46 - 54),Acarina (1 2), Tanypodinae (64 - 63), Chironominae (50 - 51)
Site DA06 (31/10/98) - Taxa missing (1): Odonata (33)
Site DA29 (2/11/98) - Incorrect abundance (1): Atyidae (8 - 7); Extra taxa (1):
Syrphidae (1)
Site DW22 (7/10/98) - Incorrect abundance (5): Nematoda (26 - 31), Planorbidae (1
- 2), Oligochaeta (23 - 21), Acarina (62 - 61), Tanypodinae (16 - 15)
Appendix E - appendices
Page 6
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Appendix E:3
Problems encountered in Data
Entry Error and WISE Analysis
Considerable time was spent converting data supplied by lead agencies to a form
suitable for the assessment of data entry errors and for import into the WISE
database. The issues encountered included:
•
differing formats in which the lead agencies extracted the electronic data from
their database;
•
errors or lack of site/date/habitat detail on sample labels, identification sheets
and database entries;
•
identification of taxa to species level required re-coding and summing to
family level; and the
•
use of taxa codes different from that used by WATER ECOscience (EPA Vic),
necessitating re-coding.
Queensland
Considerable difficulties were encountered with compiling the comparable data sets.
This was mostly due to the incorrect electronic data and missing identification
sheets sent to WATER ECOscience.
•
Several residue labels were incomplete
•
Four samples were removed from the WISE analysis due to unresolvable
uncertainties or incomplete data.
New South Wales
NSW had more problems with site data mismatches than any other State. The
problems were associated with incorrect or incomplete site details entered on either
the residue labels, identification sheets and/or the database. These were further
confounded by the use of different site codes and identification sheets by the two
agencies (EPA and DLWC) responsible for the sampling.
•
Two sites were removed from the WISE analysis due to unresolvable
uncertainties or errors.
Australian Capital Territory
No problems with ACT data, not used in WISE analysis
Victoria
Considerable difficulties were encountered in determining the site locations of the
residue samples supplied to WATER ECOscience due to poor sample labelling.
•
13 residues did not have the site code written on the internal label (only a site
description) and one residue did not have a habitat type/sample method. The
Vic EPA residues internal labels had inadequate sample information. Eg.
“King Parrot Creek - 9/4/97 - Kick Residue – Mike”.
Appendix E - appendices
Page 7
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
Vic EPA electronic data had six date discrepancies when compared with
ours, confirmed to be lead agency transcription errors.
•
Some of the Vic EPA data was identified to species level and some to family
level. All species level identifications were converted to family level for WISE
assessment.
•
The initial electronic data sent to WATER ECOscience was extracted in three
separate sheets, this needed to be compiled and converted into a usable form
for both the data entry error and WISE analysis.
•
During the data entry error assessment significant patterns of errors were
encountered. The assessment was stopped and subsequent investigations
revealed the lead agency had incorrectly extracted the data sent to WATER
ECOscience.
On delivery of the correct data the conversions and
assessment were repeated.
•
Six samples were removed from the WISE analysis due to unresolvable
uncertainties or incomplete data.
Tasmania
No problems with TAS data, no samples removed from the WISE analysis
South Australia
No problems with SA data, not used in WISE analysis
Western Australia
WA data had a different taxa coding system than that used at WATER ECOscience.
Therefore, all taxa entries had to be re-coded before any analysis could be
undertaken.
•
No discrepancies with WA data, therefore no samples removed from the
WISE analysis
Northern Territory
No problems with NT data, not used in WISE analysis
Appendix E - appendices
Page 8
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Appendix E:4
Whole Individual Sample Estimate
(WISE) Analysis
WISE: General overview and Bray-Curtis dissimilarity calculation
The following is a very general overview of the Whole Individual Sample Estimate
(WISE) procedure. For more detailed information refer to Mount and Humphrey
(2001).
The Whole of Individual Sample Estimate (WISE) procedure used in this study was
based on methods initially described by Humphrey and Thurtell (1997) and later
revised by Mount and Humphrey (2001).
The WISE procedure has been incorporated in a Microsoft Access database
developed for use with macroinvertebrate-based river health assessment under the
AusRivAS scheme; it was designed to fulfil two major purposes:
1. Assess the performance of users in recovering small and/or cryptic taxa.
2. Allow a standardised assessment tool of operator performance which enables
live-sort taxa recovery performance to be assessed for the whole NRHP.
The WISE database compares the live-sorted component of a sample (LS) with an
equivalent-sized component representative of the whole sample (prior to sorting)
(WSE, or ‘Whole Sample Estimate’). The WISE database automatically calculates
the Live-Sort/Whole of Sample Estimate (LS/WSE) taxa number ratio and various
Bray-Curtis dissimilarity indices.
The WSE is, in effect, an estimate of the taxa that would have been present in the
original sample (Humphrey and Thurtell, 1997).
The Bray-Curtis dissimilarity measure is calculated in WISE, for two samples, j and
k, based on taxa 1 to N (indexed by i), as:
?∑ |xij - xik |?/ [ ∑ ?
xij + xik?
]
i
i
where xij is the abundance for taxon i in sample j (Mount and Humphrey, 2001).
WISE calculates four different variants of the BC dissimilarity value for each sample:
47. Unadjusted BC dissimilarity for a sample size the same as the LS sample
size.
48. Normalised unadjusted BC dissimilarity for situations where a sorted sample
(ie the LS component) contains fewer than 100 individuals and the sample
size is normalised to 100.
49. Adjusted BC dissimilarity where LS taxa that are unique to LS (ie. not found in
the residue) are eliminated from the ensuing calculations (ie. prior to WSE
calculation).
50. Normalised adjusted BC dissimilarity.
Appendix E - appendices
Page 9
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Adjusted BC values, involving the removal of taxa unique to the LS component, are
provided because often live-sorting has removed all large, rare taxa and,
consequently, such taxa would not be expected to occur in the residue and WSE.
This report used adjusted BC dissimilarity values or, where the LS sample size was
less than 75, normalised adjusted BC dissimilarity values.
Appendix E - appendices
Page 10
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Appendix E:5
Description of Habitats Sampled
During State and Territory Field
Audits.
The level of detail below is indicative of the detail in the manual, guideline or
regulatory publication of each state or territory.
Unless stated otherwise,
descriptions of habitats sampled are derived directly from the AusRivAS methods
manual for each state / territory and, where practical, text has been repeated
verbatim.
Queensland
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling//Qld/)
Two habitat types are sampled, an edge sample and a bed sample. For bed
samples the first choice is a riffle, however if this is absent samples should be taken
from, in decreasing order of preference, a rocky bed or a sandy bed. Bed samples
from the different habitat types are collected using slightly different techniques.
Bed Samples
Riffle habitat
•
Reach of relatively steep, shallow (<0.3 m), fast flowing (>2 m/s) and broken
over stony beds.
•
Net size/type - standard 250 µm mesh, triangular dip net 250mm by 250 mm
by 250 mm opening, 50-75 cm depth and with a 1 –1.5 m handle
•
Technique for sampling - while holding the net downstream with its mouth
facing the sampling area, disturb the substratum by digging the foot well into
the stones and turning them over. Turn and rub stones by hand to dislodge
animals. Continue this process working upstream for 10m, covering both the
fastest and slowest flowing sections.
Stony/rocky bed habitat
•
Stony or rocky bed surface - if this is absent choose a silty area with plant
litter or organic material
•
Net size/type - standard 250 µm mesh, triangular dip net 250mm by 250 mm
by 250 mm opening, 50-75 cm depth and with a 1 –1.5 m handle
•
Technique for sampling - while holding the net downstream with its mouth
facing the sampling area, disturb the substratum by digging the foot well into
the stones and turning them over. If the rocks are too large to kick without
damaging your foot, wash about 10 rocks of a range of sizes, scrubbing
gently with the hands or a light brush into the net. Leave the rocks out of the
water to allow cryptic specimens to emerge. These can then be hand picked.
Appendix E - appendices
Page 11
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Sandy/silty bed habitat
•
Sandy or silty bed with plant litter (not macrophytes) rather than an area of
clean sand
•
Net size/type - standard 250 µm mesh, triangular dip net 250mm by 250 mm
by 250 mm opening, 50-75 cm depth and with a 1 –1.5 m handle
•
Technique for sampling - sample a 10 m length of stream using a short
sweeping action with the net whilst stirring up the bed with your foot. The
suspended benthic animals are captured as the net sweeps through the cloud
of suspended matter.
Bed samples from Pool areas
•
Pools are zones of relatively deep, stationary or very slow flowing water over
silty, sandy, stony or rocky beds. Water velocity is used to determine whether
it is a pool or a run.
•
Net size/type - standard 250 µm mesh, triangular dip net 250mm by 250 mm
by 250 mm opening, 50-75 cm depth and with a 1 –1.5 m handle.
•
Technique for sampling - Disturb the substratum by kicking with feet. If water
is flowing, hold the net downstream with its mouth facing the sampling area.
If there is no discharge, stir up the bed and use the net in a short sweeping
action through the cloud of suspended matter to capture the suspended
benthic organisms.
Edge/Backwater Samples
•
Edges (or banks and underbank areas) occur along the bank where there is
little or no current and extend to approximately 0.5 m from the bank. The
area may be bare or have some terrestrial vegetation or tree roots. A
backwater is a zone where the bank indents and forms a pool away from the
main channel. backwaters may have a circular or back flow and a silty bed
with accumulated plant litter and organic material.
•
Net size/type - standard 250 µm mesh, triangular dip net 250mm by 250 mm
by 250 mm opening, 50-75 cm depth and with a 1 –1.5 m handle
•
Technique for sampling - using short upward sweeping movements at right
angles to the bank, sample a total bank length of 10 m. Stir up the bottom
while doing so, ensuring that benthic samples are suspended and then caught
when sweeping through the cloud of suspended material.
New South Wales
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/NSW/)
For the purposes of AusRivAS a habitat is an instream environment within a
sampling site that supports a distinct macroinvertebrate fauna. In NSW AusRivAS
models were developed for Riffle and Edge habitats. These are defined below.
Riffle habitat
•
The riffle habitat is an area of broken water with rapid current that has some
cobble or boulder substratum.
Appendix E - appendices
Page 12
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Note: In cases where riffles do not have the type of substratum specified in this
definition, the available habitat would generally be regarded as marginal or
unsuitable. AusRivAS assessments from marginal and unsuitable habitats may be
unreliable. For example, samples may be collected from broken waters with only
pebble, gravel, sand or bedrock substratum or a combination of any of these. The
results obtained from such riffles may be assessed as poor by AusRivAS even at
relatively undisturbed sites. There is an advantage to this, however. In riffle zones
where there is evidence to suggest an original cobble/boulder substratum has been
covered by a pebble/gravel substratum as a result of human activities, AusRivAS
assessments may give an indication of the taxa lost due to loss of instream habitat.
In this case, the pre-disturbance substrata should be estimated and used for the
assessment.
Unreliable AusRivAS assessments may also occur as a result of difficult or
dangerous sampling conditions. For example, sampling riffles where the substratum
consists predominantly of large boulders may be difficult and potentially dangerous
especially where flow is deep and fast. In addition, for situations where there are no
riffles deeper than about 0.05m it may not be possible to obtain an adequate sample
because many of the invertebrates released from the substratum when disturbed will
not enter the net. This situation often coincides with slow flows and if riffle samples
are collected under such circumstances, the results should be interpreted with this in
mind. Macroinvertebrate data collected from small streams where the water velocity
is very slow, even though broken water may exist, should also be treated with
caution, as the riffle fauna in such streams may not be distinct from that found in the
edge habitat.
Natural riffle habitats are rare in Western NSW. Consequently the reference sites
used to construct the riffle models in NSW are all from the eastern part of the State.
Therefore using AusRivAS to assess unusual riffles on the Western Plains is not
appropriate.
•
Net size/type - All macroinvertebrate sampling must be done with a kick net of
0.25 mm mesh size. The preferred net frame is one with a pentagonal
shaped opening with a base of 35cm or greater. The net should be long
enough to not cause backwash (60cm or more) and the net handle should be
long enough (1.2m) to reach animals and microhabitats that are not
immediately near the operator.
•
Technique for sampling - Locate the downstream end of the riffle zone within
the site and begin sampling there. Disturb the substratum with your feet while
holding the net downstream with its mouth facing upstream. Vigorously move
the substratum about by digging your feet well into the cobbles and boulders.
If necessary, turn and rub the boulders and cobbles by hand to dislodge
organisms. Continue this process working upstream over a total distance of
10 metres comprising any number of discrete segments. Sampling should be
conducted in both the fastest and slowest flowing sections of the riffle and at
the maximum possible range of depths. It may be necessary to stop and
rinse the net a couple of times during sampling to remove fine particles that
can block the flow of water through the net which can cause backwash and
loss of captured macroinvertebrates. It is also a good idea to thoroughly rinse
the sample again once sampling is completed. This will assist in the sorting
process by removing fine particles that can cloud water in the tray and
obscure the animals present.
Appendix E - appendices
Page 13
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Edge habitat
•
The edge habitat is an area along the creek bank with little or no current.
Suitable areas for sampling include an alcove or backwater with abundant
benthic leaf-litter, fine organic/silt deposits, macrophyte beds, overhanging
banks and areas with trailing bank vegetation. These areas are often
indicated by the presence of surface-dwelling insects.
Note: In lowland rivers the insides of sharp meander bends (point bars) and outside
of sharp bends (steep eroded banks) should be avoided during sampling. A straight
section is usually best, but suitable areas can sometimes be found just upstream or
downstream of point bars (sandbanks). Ideally, sampling should be done at
sections of the river that include an alcove/small bay between logs or red gums (on
the lowlands of the Murray-Darling system) with logs and tree roots. Some river
reaches have no areas along the edge that meet the requirement of having “little or
no flow”. Such cases should be noted on the field data sheets and the results from
the edge habitats at these sites treated with caution.
•
Net size/type - All macroinvertebrate sampling must be done with a kick net of
0.25 mm mesh size. The preferred net frame is one with a pentagonal
shaped opening with a base of 35 cm or greater. The net should be long
enough to not cause backwash (60 cm or more) and the net handle should be
long enough (1.2 m) to reach animals and microhabitats that are not
immediately near the operator.
•
Technique for sampling - Sweep the net over a total bank length of 10 metres
comprising any number of discrete segments. Use sequential short sweeping
movements at right angles to the bank (Figure 12). Stir up the bottom while
doing so, such that benthic animals are suspended and then caught when
sweeping through the cloud of suspended material. When sampling the edge
habitat, try to sample as many different instream “structures” present in the
reach as possible. Sweep the net in amongst tree roots, trailing bank
vegetation, under overhanging banks and along logs if present. Do not,
however, work into log crevices or use your hands or any means other than
the net to extract animals. Macrophytes can be included in the edge habitat
and should be sampled if abundant, however, small patches of macrophytes
should not be deliberately sought for while sampling.
Australian Capital Territory
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/ACT/)
In the ACT, models have been created for both riffle and edge habitats. Areas of
riffle and edge habitats that are representative of the reach should be chosen for
sampling. The reach is defined as five times the mode bank-full width either side of
the riffle sampling area, unless the bank-full width is less than 10m, then the
minimum reach length is 100m (i.e., 50m either side of the riffle sampling site).
Riffle habitat
•
The riffle habitat is one of flowing broken water over gravel, pebble, cobble or
boulder, with a depth greater than 10 cm.
•
Net size/type - Samples are taken with a 250µm mesh rectangular or Dframed net with a 350mm wide aperture at the base.
Appendix E - appendices
Page 14
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
Technique for sampling - Facing downstream, the operator should place the
net directly on the substratum in front of the feet and vigorously disturb and
dislodge the substratum by kicking and twisting the feet to a depth of
approximately 10cm, slowly moving upstream employing this method. If fine
particles block the flow of water through the net, stop and rinse the net before
continuing. Separate lengths of riffle may be sampled if a continuous 10m
section is not present. Record the collector’s name on the field sampling
sheet along with the total length of riffle sampled, if less than 10m.
Edge habitat
•
The edge habitat consists of slow flowing or still waters adjacent to the bank,
preferably with overhanging or emergent vegetation, undercut banks, root
mats or other suitable habitat providing cover and refuge for
macroinvertebrates.
•
Net size/type - Samples are taken with a 250µm mesh rectangular or Dframed net with a 350mm wide aperture at the base.
•
Technique for sampling - macroinvertebrates are collected by vigorously
sweeping from a distance of approximately one metre from the bank to the
bank edge, disturbing the emergent and overhanging vegetation in the water
if present. The operator should slowly move upstream for a distance of 10
metres employing this method. Separate lengths of edge may be sampled if
a continuous 10m section is not present. Record the collector’s name on the
field sampling sheet along with the total length of edge sampled, if less than
10m.
Victoria
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/Vic/)
Riffle habitat
•
Typically rocky riffles where the flow is rapid and turbulent, but gravel or sand
bars can also be sampled as long as there is flow over the substrate
•
Net size/type - opening 30 by 30 cm, mesh size 250 µm net length
approximately 1 m
•
Technique for sampling - sampler disturbs the streambed by vigorously
kicking while holding a collecting net down current, debris dislodged by the
kicking is collected in the net as the sampler moves upstream, continually
kicking the streambed for 10 metres, typically taking 5 to 10 minutes
Edge habitat
•
Areas of little or no current along the edge of the stream – around
overhanging vegetation, snags and logs in backwaters and through beds of
macrophytes
•
Net size/type - opening 30 by 30 cm, mesh size 250 um, net length
approximately 40 cm
•
Technique for sampling - net is swept along the edge of the stream in areas
of little or no current and around overhanging vegetation, logs and snags in
backwaters and through macrophyte beds for 10 metres, typically taking 5 to
10 minutes
Appendix E - appendices
Page 15
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Tasmania
(AusRivAS Manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/Tas/)
In Tasmania, models have been created for both Riffle and Edge habitats. The
reach is defined as 100m of stream length and within the stream only. Areas of riffle
and edge habitats that are representative of the reach should be chosen for
sampling.
Riffle habitat
•
The riffle habitat is one of flowing broken water over gravel, pebble, cobble or
boulder, with a depth greater than 10 cm.
•
Net size/type - samples are taken with a 250 mm mesh kick net with a
280x340 mm opening.
•
Technique for sampling - facing downstream, the operator should place the
net directly on the substratum in front of the feet and vigorously disturb and
dislodge the substratum by kicking and twisting the feet to a depth of
approximately 10 cm, slowly moving upstream employing this method. Every
2-3 metres the net should be rinsed to remove fine particles which may be
blocking the flow of water through the net. Separate lengths of riffle may be
sampled if a continuous 10 metre section is not present. Note total length of
riffle sampled if less than 10 metres.
Edge habitat
•
Consists of slow flowing or still waters adjacent to the bank, preferably with
overhanging or emergent vegetation, undercut banks, root mats or other
suitable habitat providing cover and refuge for macroinvertebrates.
•
Net size/type - samples are taken with a 250 mm mesh kick net with a
280x340 mm opening.
•
Technique for sampling - macroinvertebrates are collected by vigorously
sweeping from a distance of approximately one metre from the bank to the
bank edge, disturbing the emergent and overhanging vegetation in the water
if present. The operator should slowly move upstream for a distance of 10
metres employing this method. Separate lengths of edge habitat may be
sampled if a continuous 10 metre section is not present. Note total length of
edge sampled if less than 10 metres. The net should be thoroughly rinsed to
remove silt, mud, and fine detritus.
South Australia
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/SA/ )
Areas with riffle and edge habitats that are representative of the conditions found
within the reach of interest along a creek or river should be chosen for sampling.
The site is defined as a 100m section of the stream as stipulated in the
bioassessment manual (Anon 1994). In South Australia, models have been created
for both Riffle and Edge habitats.
Appendix E - appendices
Page 16
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Riffle habitat
•
Areas of shallow turbulent water flowing over a substrate, which is usually
cobble, pebble and/or gravel, but may include sand, detritus, roots, etc.
•
Net size/type - macroinvertebrate samples were taken using a triangular
framed 250 mm mesh pond net (35 x 30 x30cm).
•
Technique for sampling - The substrate was disturbed by vigorous kicking
and rocks were rubbed by hand to dislodge organisms into the net that was
held immediately downstream. Note that sampling proceeded upstream or
was conducted in a manner that prevented the contamination of samples due
to drift and disturbance while working the site. The sampled area was not
necessarily contiguous, and attempted to encompass all the microhabitats
available.
Edge habitat
•
Areas of little to no current, or aquatic vegetation, often in quite deep water. It
may have overhanging or emergent vegetation, undercut banks, root mats or
other suitable habitat providing cover and refuge for macroinvertebrates.
•
Net size/type - macroinvertebrate samples were taken using a triangular
framed 250 mm mesh pond net (35 x 30 x30cm)
•
Technique for sampling - The net was moved in sweeping actions through the
water column as the sampler moved along the bank, and the sediment was
kicked to ensure benthic organisms were collected in the sample. Three
rocks where also rubbed by hand (when present) and the dislodged
acroinvertebrates included in the sample. Note that sampling proceeded
upstream or was conducted in a manner that prevented the contamination of
samples due to drift and disturbance while working the site. The sampled
area was not necessarily contiguous, and attempted to encompass all the
microhabitats available.
Western Australia
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/ Man/Sampling/WA/)
Riffle habitat
•
Areas of steep broken water with rapid current and stony substrate
•
Net size/type - 350 by 250mm opening, 250 µm mesh, 50-75 cm depth and 11.5 m handle
•
Sampling technique - disturb the substratum while holding the net
downstream with its mouth facing the disturbed area. Kick the substratum by
digging the foot well into the stones and turning them over – turn and rub
stones by hand to dislodge animals. Continue this process working upstream
over a distance of 10 metres, covering both the fastest and slowest areas of
the riffle. Do not include material from macrophytes or woody debris located
in the riffle.
Macrophyte habitat
•
An area of dense aquatic vegetation
Appendix E - appendices
Page 17
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
Net size/type - 350 by 250mm opening, 250 µm mesh, 50-75 cm depth and 11.5 m handle
•
Technique for sampling - vigorously sweep the net within the aquatic
vegetation over a length of about 10 m aim to sample the upper middle and
lower portions of the plants. A combination of short lateral sweeps with
vertical lifts will aid in dislodging and catching suspended animals
Channel
•
The central part and margins of the main channel of a stream in areas without
riffles, macrophytes or pool rocks
•
Net size/type - 350 by 250mm opening, 250 um mesh, 50-75 cm depth and 11.5 m handle
•
Sampling technique - vigorously sweep the net through the water column
using short vertical lifts to disturb the substrate and catch the suspended
organisms. Continue this process along the channel for a distance of about
10 m.
Pool rock habitat
•
A pool with little or no current and large numbers of rocks
•
Net size/type - 350 by 250mm opening, 250 µm mesh, 50-75 cm depth and 11.5 m handle
•
Sampling technique - remove about 10 – 20 rocks of a range of sizes from
the stream into a bucket and rub the rocks by hand to dislodge the animals
Northern Territory
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/NT/)
Edge habitat
•
Ideally a vertical or sloping section of the riverbank containing abundant root
matter which is immediately adjacent to a pool. Areas to avoid when
sampling edge habitats are: pandanus roots, adjacent macrophytes (water
plants) and large undercut banks. Pools can be of variable depth from 30cm
to 3cm or more.
•
Net size/type - triangular framed net with 350 mm wide aperture at base, 250
µm mesh, long handled also used is a three pronged rake device used by a
second operator to disturb the sediments and vegetation in front of the net
•
Sampling technique - macroinvertebrates are collected by the rake operator
agitating the root matter in the water and adjacent to the bank. The net
operator sweeps vigorously back and forth following the rake operator and
capturing the material dislodged into the water column. In still waters the net
is kept mobile to ensure water current is generated through the net and the
collected macroinvertebrates do not swim out of the net. Both operators
move upstream for a distance of 10 m.
Appendix E - appendices
Page 18
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Sand/Silt bed habitat
•
Area of uniformly deposited sand or silt. Generally the surface is horizontal
and at a depth of no greater than 40cm. Areas of thick detritus (organic
matter eg. .leaves) and algae are avoided.
•
Net size/type - triangular framed net with 350 mm wide aperture at base, 250
µm mesh, long handled also used is a three pronged rake device used by a
second operator to disturb the sediments and vegetation in front of the net
•
Sampling technique - macroinvertebrates are collected by the rake operator
facing the water and raking in 30 cm scratches across the sand/silt substrate.
The raking pattern is achieved by extending the rake out in front and then
raking towards the body. The net operator faces the water also and employs
a ‘figure of eight’ sweeping motion which collects the dislodged material. The
net operator should endeavour to keep the net sweeping the habitat surface
without collecting large amounts of sediment. Both operators move upstream
for a distance of 10 m.
Appendix E - appendices
Page 19
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Appendix E:6
Lead Agency
Methods
Internal
QA/QC
The internal QA/QC protocols varied between States and this was reflected in each
agencies method manual. The level of detail below is indicative of the detail in the
manual, guideline or regulatory publication of each state or territory. Where
practical, text has been repeated verbatim from the state / territory AusRivAS
methods manual.
Queensland
(AusRivAS manual:http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/Qld/)
Field
The residues of ten percent of all samples taken in the field are retained for analysis.
Half of these samples are put aside for external analysis; the other half are
subsampled and 10% of each sample is analysed by the unit’s staff. The data is
analysed, compared to the sample picked in the field, reports written, and presented
as part of the milestone reports.
Laboratory
•
As the samples arrive in the laboratory from the field, an audit is taken. The
site name and type of habitat sampled is recorded. This list is compared
against the field sheets when they arrive to ensure that no samples have
been lost. The vials are inspected for cracks etc which might allow the
sample to dry out. Low alcohol levels are topped up and the vials stored
upright according to sampling run and district from which they are collected.
The QA/QC residue samples are audited also and checked against the list of
requested samples. The samples are stored in sealed polydrums until
required.
•
As the paperwork arrives in the laboratory from the field, it is stored in ring
binders according to sampling run and district. The field sheets are checked
by the staff member that assisted in the field in each district. Any empty fields
are completed by either the biologist that attended the collection, or the
district staff that undertook the sampling. The photos are stored with the field
sheets as are the major ion and nutrient water quality results from GCL.
Taxonomy
•
Internal QA/QC checks are performed on staff, by staff, on a regular basis. At
each round of QA/QC, a person is assigned to analyse a sample identified by
another. Samples identified during the previous fortnight are selected at
random and re-identified. The resultant taxa lists are compared and
discrepancies in identification checked by other staff in the unit. Any errors
are discussed with the original identifier (both misidentifications and errors of
enumeration) and a report prepared which is read and signed by all members
that underwent the QA/QC check.
Appendix E - appendices
Page 20
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Data Entry
•
Taxonomic data is entered electronically (EXCEL database) as the samples
are completed. At the completion of a district, the electronic database is
verified against the data sheets. This is a 2-person task, where one person
reads aloud the entries from the data sheet and the other person checks
these results against the electronic database. Corrections are made to the
database as the verification proceeds. When a full run is completed and fully
verified, the data is transferred electronically to an ACCESS database.
Random checks are implemented to ensure that the data transferred
successfully.
•
Physical and chemical data is entered directly from the field and GCL sheets
onto an Excel database. Parameters that are measured both in the field and
as part of the GCL analysis (Conductivity, pH, turbidity and alkalinity) are
compared and, if the results are compatible, the field results are recorded on
the database. If there is a major discrepancy, the cause is investigated by
referring to other sampling occasions. A picture of the site is built up and the
more appropriate readings are used in the database. Substrate composition
is checked by adding the components and ensuring all samples measure
100%. Formulae have also been used in the database to calculate mean phi
of the substrate and to rate chemical measurements. On completion of data
extraction to Excel, random checks are done to ensure that data was
transcribed correctly. The data is than transferred electronically to the
ACCESS database. Random checks are implemented to ensure that the data
transferred successfully.
•
Several descriptive variables are also extracted from the field sheets to
describe things such as canopy cover, bank stability, channel condition.
These ratings are entered directly to an Excel spreadsheet. On completion of
the data extraction, the database is transferred electronically to the ACCESS
database. Random checks are implemented to ensure that the data
transferred successfully.
•
Once the data is stored in ACCESS, each parameter is checked for outliers.
These are referred back to the field and laboratory sheets and corrected, if
necessary.
Mapping
•
Several parameters are extracted from maps. Most are done by the district
hydrographers, checked by other hydrographers and sent to us electronically.
We rely upon the data being checked prior being sent. Latitude and longitude
values are mapped to ensure that the sites map to the correct position.
•
Data is also extracted from electronic databases such as BOM, DRF and
vegetation overlays.
These data are entered directly into an Excel
spreadsheet and later matched to the appropriate sampling site. This data is
then transferred electronically to the ACCESS database. Random checks are
implemented to ensure that the data was entered correctly and was
transferred successfully.
New South Wales
(Waddell 2001)
Note: No details of New South Wales QA/QC procedures are provided in the NSW
AusRivAS Sampling and Processing Manual. In reference to such procedures the
Appendix E - appendices
Page 21
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
manual merely indicates that ‘It is also important that in undertaking sampling for
AUSRIVAS, particularly when large-scale monitoring programs are concerned,
appropriate quality control and quality assurance procedures are followed. Details
of quality control and quality assurance programs followed during the MRHI and
AWARH in NSW can be found in Waddell (2001).’
Quality control and assurance programs for the NRHP were undertaken both at the
national level and within each state and territory Program. During MRHI QA/QC
programs were initiated at the national level to address issues of sorting and
macroinvertebrate identification. A second national QA/QC program was also
recently implemented to deal with sorting issues for the AWARH phase. The
individual states and territories assumed responsibility for QA/QC for identifications
during AWARH. The information presented below outlines the QA/QC programs
undertaken out at the state / territory level for NSW.
Field measurements
•
Water quality meters were calibrated prior to each sampling event and
checked daily during sampling.
•
Alkalinity was measured both in the field and in the laboratory from frozen
water samples. Values were compared and suspect data flagged and
excluded from analyses.
•
Field measurements such as stream width and riffle depth were regularly
confirmed using a measuring tape and more subjective measurements such
as disturbance rankings were regularly compared between team leaders to
ensure consistency between sampling teams.
•
Site attributes such as site code, name, position and elevation were checked
using topographic maps and/or GPS on each sampling occasion. Positional
accuracy was reconfirmed in the office using GIS.
Data entry and storage
•
All data collected during the NRHP for NSW were entered and stored in an
Oracle database. To ensure completeness of records in the database, all
samples collected in the field were entered into a field master form within the
database. This was done immediately following all sampling trips and
included information such as site code, date and the habitats from which
biological samples were collected. A complete record of all samples collected
was therefore readily available.
•
To help minimise errors associated with data entry, electronic data entry
forms were set up to mimic the layout of the field and laboratory datasheets.
•
Range checks were also in-built to highlight unusual or incorrect values for
given variables (such as a pH value >10).
•
All entered data was then checked. During the MRHI phase this entailed the
double entry of all field and biological data into a QA/QC table followed by
electronic comparison and subsequent checking of inconsistent results. In
1997 this procedure was replaced by a separate visual check of inconsistent
data between hardcopy records and database records by another operator.
Errors were then rectified and changes noted on the original datasheet. To
provide a measure of NSW EPA QA/QC Document accuracy in data entry, a
second checking procedure was undertaken on the field data from 25
Appendix E - appendices
Page 22
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
randomly selected samples in each season. All errors were recorded and
these data were analysed to provide an assessment data entry accuracy.
Macroinvertebrate identification
•
Five percent of samples were selected from each season on a
stratified/random basis ensuring all identifiers were considered and a range of
habitats and biogeographic regions represented. Selected samples were
then re-identified by an experienced staff member according to the guidelines
presented in Hawking and O’Connor (1997a).
•
Error rates including Percentage of New Taxa and the Bray Curtis
Dissimilarity Index were then calculated for each of these samples as
specified in Hawking and O’Connor (1997b).
•
In addition, Sorrenson’s index (Bennison et al., 1989) was also calculated to
provide an alternative estimate of dissimilarity between original and QA/QC
samples. This index was considered more appropriate than Bray Curtis for
reflecting errors that may affect AUSRIVAS results, which are based on
presence/absence data, as it uses total taxa numbers and not relative
abundance.
•
In accordance with the MRHI QA/QC program (Hawking and O’Connor,
1997c) < 10% error was deemed acceptable and therefore samples with a
‘new taxa’ percentage of 10% or greater and/or a Sorrenson’s index of less
than 0.91 failed the QA/QC test.
•
All identification errors were compiled and appropriate follow-up action
implemented to rectify mistakes and improve identification performance.
Follow-up action was also undertaken to address identification problems
highlighted in the national QA/QC program conducted on samples collected
during the MRHI phase (Hawking and O’Connor, 1997c).
Biological Data Screening
To ensure that only quality-assured samples were included in model development
and performance testing, a formal procedure was undertaken to screen all biological
data collected at reference sites. A set of criteria was developed and biological
samples that failed to pass the quality control and assurance procedure were
flagged in the database and excluded from use in all modeling procedures.
Biological samples were assigned a ‘fail’ for quality control and assurance if they
contained:
•
An unusually low number of taxa for the site (as compared to other samples
collected at the site for the same habitat and season) and/or,
•
Unusually low O/E results and a different fauna composition than expected
for the site in the relevant season and habitat (as compared to other samples
collected at the site) and/or,
•
A lower than expected number of cryptic taxa (as compared to other samples
collected at the site and similar sites for the same habitat and season).
And satisfied one or more of the following criteria:
•
Sorted by an untrained or inexperienced operator,
•
Collected by an untrained or inexperienced operator,
Appendix E - appendices
Page 23
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
Collected from a marginal habitat eg. small bedrock riffle or a fast flowing
edge,
•
Collected from an unusual site eg. acid stream,
•
Sorted under low or artificial light conditions eg. dusk or motel room,
•
Sampled with limited access to available habitats eg. steep, slippery banks,
deep pools or very fast flowing riffles,
•
Sampled under extreme circumstances eg. heavy rain,
•
Sampled under extreme flow conditions ie. during or immediately following a
flood or drought,
•
Collected outside the acceptable date range for a given season.
This screening procedure was also applied to some test site data where replicate
samples were available for comparison. The quality of test data, however, is a lot
harder to assess than reference data.
Whereas data collected in other
years/seasons from the same site and in the same year from similar sites could be
used as a benchmark for assessing reference site data, such benchmarks are not
available for test sites because it is neither possible nor appropriate to attempt to
anticipate results from test sites. Even when replicates were available from a
disturbed test site it was often not safe to assume that there should be consistency
in the results over time because in most cases the degree of disturbance would
change greatly over time. Consequently the screening procedure for the test site
data was probably less reliable and depended on the results of the reference data
screening procedure to identify possible problems such as consistently poor sorters,
flood or drought extremity, difficult sampling conditions etc.
Environmental Data Screening
Map/GIS derived data
All predictor variables derived by topographic maps or GIS including elevation,
distance from source, latitude, longitude, slope and mean annual rainfall, were
checked and screened for unexpected results.
Field data
A rigorous screening procedure was undertaken to ensure only reliable field
recordings were used in all aspects of model development, testing and site
assessment. This followed the finding of a recent internal study, conducted to
assess temporal variation in AUSRIVAS outputs, that one of the major factors
affecting group probabilities and hence O/E values between samples from the same
reference site was variation in environmental data. This was particularly evident for
substrate variables such as clay and silt where large differences were evident
between different sampling occasions. As a consequence, potential predictor
variables including substrate composition, mode stream width, mode riffle depth and
field alkalinity readings were screened for all reference and test samples. Records
that were inconsistent with other samples and/or inconsistent with the checkers
knowledge of the site were recorded as unreliable and eliminated from any analysis.
For modeling purposes substitute values were then derived for all deleted and
previously missing environmental records. The mean value of all quality assured
records for the site was used for this purpose. If no quality assured values were
Appendix E - appendices
Page 24
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
available for a site an estimate was derived from values recorded at similar sites in
combination with the samplers knowledge of the site.
For alkalinity, field records were also compared to values derived in the laboratory
from frozen water samples. If consistent with laboratory-derived values field records
were used preferentially for modelling purposes. If reliable field data were not
available for a sample the lab value was used followed by the mean value of all
quality assured field records.
Australian Capital Territory
(AusRivAS manual http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/ACT/)
Field
•
It is critical that once the field sampling sheet has been completed, it is rechecked to ensure no measurements have been missed. The discovery of
missing data once back in the laboratory can mean returning to the site, a
costly mistake in both time and money, and one that could have been easily
avoided. It is also critical that measurements of zero are recorded as such on
the field sampling sheet and not left blank. A blank value may be interpreted
as missing data when being entered into a database by another person and
can result in the site not being assessed.
Laboratory
Quality control/quality assurance (QA/QC) procedures are designed to establish an
acceptable standard of macroinvertebrate sorting and identification. The quality
control component reduces the level of error in sorting and identification, while the
quality assurance component provides potential users with the assurance that the
accuracy of results is within controlled and acceptable limits.
Sorting
•
For new persons, projects, or sampling runs, quality control staff should
check the residues of the first five samples sorted for missed organisms. In
order to pass the QA/QC process, = 95% of the total number of organisms in
the sub-sample must be recovered. If one of the first five samples fails the
QA/QC protocol, the process begins again until five consecutive samples
have passed.
•
Following the initial five samples, a random selection of two samples in the
following ten, two samples in the following 30 and two samples in the
following 50, will be checked. If any of these samples fail the QA/QC
protocol, the person must again pass five consecutive samples.
•
Staff checking samples will have adequate experience in sorting. Where
possible, QA/QC of sample residues should be conducted by more than one
person to avoid bias and increased workloads. If less than 95% of the
organisms are recovered from the sub-sample, commonly missed taxa should
be shown to the person and suitable instruction given to rectify the problem.
Laboratory sample record sheets are to be fully completed by the person
conducting the QA/QC check. Error/action codes are to be recorded on
laboratory sample record sheets if appropriate.
Appendix E - appendices
Page 25
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
QA/QC error and action codes
Error codes
•
CC Number of organisms recovered from the sub-sample represents less
than 95% of the total number of organisms in the sub-sample.
•
IE Identification error (i.e., “Percent Taxa Error” and/or the “Percent Incorrect
Identifications”) greater than 5%.
•
LE Labelling error
•
SE Sub-sampling error – e.g., if the sample was stored in more than one
sample container and only one container was sorted.
•
WE Washing error – some sample bypassed sieve when rinsing
•
DE Data entry error on data sheet
•
CE Calculation error – mathematical error on data sheets
Action codes
•
LC Labels corrected – contact person who collected the sample if error is on
the original sample label
•
SC Sample re-subsampled, processed, re-checked and data sheets
corrected
•
WC Material bypassing the sieve caught in washbasin, sample combined and
rewashed
•
WI Material bypassing sieve lost, partial sample processed
•
DC Data entry corrected (strike out incorrect entry with one line and write in
the correct entry, initial).
Identification
Two methods are used to calculate a sample’s identification QA/QC result.
These are the “Percent Taxa Error” and the “Percent Incorrect
Identifications”. A sample must pass both methods to achieve an overall
pass.
Percent Taxa Error
A "Taxa Error" occurs when a mis-identification results in the loss or addition of a
taxon. The “Percent Taxa Error” is the "Number of Taxa Errors" divided by the
"Total Number of Original Taxa", multiplied by one hundred. Samples pass if the
"Percent Taxa Error" is less than or equal to 5% at the family level (10% at species
level). The manual provides examples (not included here) of how to calculate ‘Taxa
Error’.
Percent Incorrect Identifications
The “Percent Incorrect Identifications” is the "Number of Organisms Incorrectly
Identified" divided by the "Total Number of Organisms in the Original Count",
multiplied by one hundred. Samples pass if the "Percent Incorrect Identifications" is
less than or equal to 5% at the family level (10% at species level). The manual
provides examples (not included here) of how to calculate ‘Percent Incorrect
Identifications’.
Appendix E - appendices
Page 26
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
For new persons, projects, or sampling runs, quality control staff should
check the first five samples identified. If one of the first five samples fails the
QA/QC protocol, the process begins again until five consecutive samples
have passed.
•
Following the initial five samples, a random selection of two samples in the
following ten, two samples in the following 30 and two samples in the
following 50, will be checked. If any of these samples fail the QA/QC
protocol, the person must again pass five consecutive samples.
•
Staff checking samples will have adequate experience in identification.
Where possible, QA/QC of sample identifications should be conducted by
more than one person to avoid bias and increased workloads. All misidentifications will be shown to the person and suitable instruction given to
rectify the mis-identification. Other samples containing taxa that were misidentified are then checked for identification errors by the original identifier.
Laboratory sample record sheets are to be fully completed by the person
conducting the QA/QC check. Error/action codes are to be recorded on
laboratory sample record sheets if appropriate.
Victoria
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/Vic/)
•
Prior to a field trip, the number of samples to be taken in the catchment(s)
visited during the trip shall be determined. Coloured marbles equal to this
number, plus additional white marbles greater than or equal to 10% of the
number of expected samples are to be added to a jar. Immediately after
sampling and live-picking has been conducted at each site a marble shall be
randomly selected from the jar for each sample. Marbles are not returned to
the jar after selection. If the marble is other than white, no residue sample
need be collected. If the marble is white, the sample residue remaining after
live-picking shall be collected in a large, sealable container, preserved in 80%
ethanol solution and returned to the lab.
•
EPA will retain five per cent of residue samples for internal quality control of
live-picking efficacy.
Macroinvertebrate samples
•
Each residue sample shall be randomly sub-sampled using a box subsampler and ten per cent recovered for picking and identification of
macroinvertebrates under a dissecting microscope. Macroinvertebrates shall
be identified to family as for the live-picked samples and where taxonomic
resolution allows. Results from the residue samples will be recorded and
entered onto the EPA biological database under method code 31 for sweep
sample residues and method code 32 for kick sample residues.
•
For quality control purposes, and in particular to ensure consistency across
laboratory staff, a random selection of 10% of all sorted macroinvertebrate
samples shall be re-identified by a senior taxonomist or ecologist. Errors at
family level should be less than 1% and errors at species level should be less
than 10%. Sample processors are responsible for ensuring that their
identifications are checked for quality.
•
For each sample processor, quality control checks shall be undertaken on
10% of samples identified to species level and independently on 10% of
Appendix E - appendices
Page 27
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
samples identified to family. Samples from different catchments can be
grouped for QC testing, providing that all samples have been identified by the
same processor at one level of taxonomic resolution.
•
The checker shall select in a random process the sample(s) that will be
quality controlled. Appropriately qualified staff shall undertake quality control:
•
For all discrepancies between processor and checker identifications, an
appropriate third person shall be consulted and a consensus decision made
on the actual identification.
•
The checker will complete the standard QC report form and discuss the
results with the processor. The form shall then be placed into the laboratory
QC folder, and details of the QC check entered onto the front index page of
the folder.
Sample sorting
•
As a minimum, 10% of all sample residues are to be preserved for quality
assurance/quality control assessment, using the same procedures (envelopes
etc) as conducted under MRHI Phase 1 rounds 3 and 4. Sample selection is
organised by the lead agency. Stratified random allocation of samples must
be conducted with respect to operator, catchment and habitat. The person
with knowledge of specific sites/operators for which samples require
preservation must not be involved in sample sorting. Half of these samples
(ie. 5% of the total sample number) must be sorted and an assessment made
of individual operator quality assurance/quality control performance. A
proportion of the remaining samples must be forwarded for external quality
assurance/quality control (to be arranged).
•
Additional internal quality assurance/quality control is to be conducted as
agreed with the Program Coordinator.
Taxonomy
•
5% of all processed samples, accompanied by their data sheets, must be
forwarded for external quality assurance/quality control for taxonomic
identification (to be arranged).
Training
•
All staff involved in live picking must be provide with supplementary basic
training in the above modified sampling techniques proper to conducting any
field sampling for the assessment.
Tasmania (1)
(Krasnicki et a 2001)
NOTE: The following information has been taken from Krasnicki et al. (2001) which
documents the internal QA/QC programs that were run from the start of the MRHI
program through to the end. It is unknown whether indicated QA/QC guidelines and
procedures are/were applied during AusRivAS sampling. To this effect, the QA/QC
procedures documented in the Tasmanian AusRivAS sampling manual, which are
far less comprehensive than those described by Krasnicki et al. (2001), are listed
separately under Tasmania (2).
Appendix E - appendices
Page 28
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
As outlined in Krasnicki et al. (2001), quality assurance/quality control procedures
are designed to establish an acceptable taxonomic standard of macroinvertebrate
sorting and identification. The quality control component is to determine the
variation in the level of identifications, and quality assurance provides potential
users with the assurance that the accuracy of results is within controlled and
acceptable limits.
The aim of the internal QA/QC program was to assess individual performance,
detect problems at an early stage and thereby allow intervention and training to
proceed before the quality of the data gathered in the program was seriously
compromised.
Rapid assessment sampling and site/habitat assessment
•
Ecologists involved in the MRHI program attended a training day that involved
demonstration of the sampling protocol and practice sessions prior to the first
round of sampling (Spring 1994). Any staff involved in the sampling process
since then have also undergone similar training in order to standardise the
MRHI sampling protocol.
•
New staff were trained by experienced staff in the appropriate rapid
assessment protocol and environmental variable assessment techniques
developed for NRHI sampling. This process involved visiting a range of sites
around Hobart, where ecologists sampled riffle and edgewater habitats and
performed live picks and completed all aspects of the habitat data sheets.
This allowed sampling techniques to be standardised according to the River
Bioassessment Manual (CEPA, 1994) and live-picking protocols and also
enabled staff to compare and standardise physical descriptive techniques
according to the Tasmania AUSRIVAS Sampling and Processing Manual.
Upon commencement of fieldwork, the two field teams also assembled at
several sites and cooperatively sampled the sites covering a range of habitats
from upland streams to lowland rivers, giving the staff experience in sampling
from a wide range of river types and was used to identify potential problems
and sources of error associated with the sampling regime.
Identification
•
All staff employed on the MRHI program had prior experience in
macroinvertebrate identification
•
Initial training for ecologists was given in a two-day taxonomic workshop held
by the Murray-Darling Freshwater Research Centre. Identification sessions
were held by the relevant taxonomic experts to train participants in the use of
taxonomic keys and to assist in identification of specimens from difficult
groups.
•
Ecologists associated with the MRHI program have attended these
workshops and have passed on knowledge gained in the use of new
macroinvertebrate keys to all staff involved in identification.
•
DPIWE laboratory uses a reference collection of the most recent keys from
these workshops in macroinvertebrate identification.
•
The number of ecologists involved in identification was kept to a minimum
where possible to help reduce misclassification errors. All staff worked in the
same laboratory and any questions were encouraged and dealt with at the
time by experienced biologists on hand or at the Zoology Department,
University of Tasmania.
Appendix E - appendices
Page 29
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
Approximately 5% of samples collected from all sampling rounds were cross
checked. Samples were systematically selected to cover a range of
biogeographical regions, habitats and staff.
•
Internal process for QA/QC for macroinvertebrate identifications consisted of
staff experienced in taxonomic identification of macroinvertebrates reidentifying and counting the selected samples and then cross checking the
results against those provided by the original staff. A miss-identification error
of <10% of the total number of animals was deemed acceptable at the family
level.
•
Identification issues were resolved with additional training or updating of staff
in taxonomic issues. Past samples were rechecked and where appropriate
the database was updated.
Rapid assessment live picking efficiency
•
For two sampling rounds, a few sites were sampled intensely by two
operators to monitor live picking efficiency in the field. Each operator’s
sampling efficiency was compared both within the riffles at each site and
between the two samplers, with the aim of identifying any taxa that were
consistently missed. As this method was time-consuming and representative
of limited conditions it was not continued through the remainder of the MRHI
program.
•
The QA/QC process involved preserving a minimum of 10% of all sample
residues, both riffle and edgewater, from each sampling round. Samples
were selected to cover a broad range of biogeographical regions, habitats and
staff. Half (5%) were processed internally to assess operator sorting
efficiency and the other halt were sent to be assessed externally. Analysis
was carried out as per previous external audits conducted by ERISS (Bray
Curtis dissimilarity values and Live Sort/WSE ratios).
Site and Habitat Assessment
•
This part of the sampling procedure was performed by two ecologists whose
job was to solely describe and record environmental variables. To minimise
the effects of operator error an attempt was made to keep staff who
performed the habitat assessments constant and this was maintained for the
Autumn 1995 to Spring 1997 sampling rounds. Subsequent rounds usually
then had at least one experienced staff member who was able to train any
additional staff required to perform site and habitat assessment.
•
To assess consistency of data recorded, ecologists from two of the rounds
were required to independently assess site and habitat parameters at the
same site, from a number of NRHI locations. Variables assessed were those
that required some element of subjective judgement (eg. substrate
composition). Ecologists new to the program were trained in all aspects of
habitat assessment incorporating those areas that were highlighted as
potential for operator error.
Data Entry
•
All biological and environmental data were checked for entry errors prior to
performing the final classifications.
•
A person independent of the data entry examined the data and cross checked
it against the data sheets. Any missing or incorrect data was corrected.
Appendix E - appendices
Page 30
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
Additional checks were performed by plotting biological and environmental
variables are frequency histograms. Outliers were examined to establish
whether they were in the correct range and the independent data checker
also had the task of judging whether the data, even though it fell within the
reasonable ranges, was still appropriate or usual for that type of site. This
meant that the checker had a working knowledge of the data set.
Tasmania (2)
(AusRivAS Manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/Tas/)
After picking, 10% of the residues are randomly selected and preserved in
formalin for QAQC analysis of operator sorting efficiency.
Sorting
The QA/QC program aims to assess the effectiveness of individual operator sorting
procedures using as its basis, comparison of the composition of live-picked samples
with associated residues. The MRHI Bioassessment manual states that the aim of
the live pick procedure is to ensure that the broadest range of biota are collected at
a site. This implies that the taxa list derived from a live pick will encompass more
taxa than would be expected if a random sample of animals of equivalent number to
the live sort total were drawn from the sample (ie. ‘whole sample estimate’).
•
Approximately 10% of all riffle and edgewater residues are to be preserved
from each sampling round. Half of these (5%) are to be processed so that
operator sorting efficiency can be assessed
•
Analysis is carried out as per previous external audits conducted by ERISS
Identification
•
Approximately 5% of the samples collected each round are cross-checked by
persons with adequate identification experience.
•
Samples are selected to cover a broad range of biogeographical regions,
habitats and staff.
•
A miss-identification error of < 10 % of the total number of animals is deemed
acceptable at the Family level. this is the error rate used by the Murray
Darling Freshwater Research Centre who conducted external quality control
checks of all State agencies
•
In all cases, identification problems are to be resolved with additional training
of staff. Past samples containing taxa that were found to be misidentified are
to be rechecked and, where appropriate, the database updated.
South Australia
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/SA/)
Field sampling
•
The South Australian program has involved a small group of biologists from
the same laboratory collecting all samples from 1994 to 2000. This has
Appendix E - appendices
Page 31
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
ensured that an experienced operator has been involved with sampling each
site, and over time all members in the group have gained considerable
experience sampling a wide variety of river types throughout the State. This
has ensured a consistent approach to identifying habitats and field sampling
effort during the course of the program.
•
It is recommended that any future use of this protocol should involve the
participation of at least one of the team members involved with this work at
the Australian Water Quality Centre or EPA to minimise the chance of
introducing significant sampling errors.
Laboratory
•
The sub-sampler is subject to an internal check of an additional 5% of
samples to ensure random sub-samples are being produced with the
laboratory sorting method used in S.A.
•
In addition, ERISS carried out external checks of residues from the 1995
surveys to test the performance of the sub-samplers used by each state and
territory (where laboratory sorting methods are used). That work confirmed
the sub-sampler operated effectively to randomly sort the sample and allow
the retrieval of a representative 10% sub-sample.
•
A 5% randomised selection of residues from each survey has been kept in
storage for any possible future QA of the sorting protocol used in S.A.
•
As part of a national QA/QC program involving the identification of
macroinvertebrates, the Murray-Darling Freshwater Research Centre
independently checked samples that had been sorted and identified by the
team in S.A. Samples were assessed for the 1994, 1995 and 1997 surveys.
The results from these showed the high performance of the approach used in
S.A. and indicated that no further work was needed to improve the
identification of specimens from this State.
•
All new staff are trained in the use of the sub-sampler and identification keys
used in S.A. The experienced team members have also assisted new staff to
identify organisms that they are not familiar with and check difficult taxa.
•
The addition of new staff during the program led to the development of a
more rigorous internal training protocol in 1998. This included:
à
à
à
Random checks of sorting trays of new members to ensure all
specimens were being collected and more importantly that novel taxa
were not being overlooked.
All staff involved with the project process and identify a contrived
sorted sample to provide a check on counting and identification skills.
Random checks of identifications carried out by all operators.
Western Australia
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/WA/)
•
As a minimum, 10% of all sample residues are to be preserved for quality
assurance/quality control assessment
•
Stratified random allocation of samples must be conducted with respect to
operator, catchment and habitat
Appendix E - appendices
Page 32
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
The person with knowledge of specific sites/operators for which samples
require preservation must not be involved in sample sorting.
•
Half of these samples (ie. 5% of the total sample number) must be sorted and
an assessment made of individual operator quality assurance/quality control
performance
•
Additional internal quality assurance/quality control should be performed
Northern Territory
(AusRivAS manual: http://ausrivas.canberra.edu.au/Bioassessment/
Macroinvertebrates/Man/Sampling/NT/)
•
Quality control / quality assurance procedures are designed to establish an
acceptable standard of macroinvertebrate sorting and taxonomic
identifications. The quality control component minimises the variation in
sorting and identification efficiency. Quality assurance provides potential
users with the assurance that the accuracy of results is within controlled and
acceptable limits.
Sorting Efficiency
•
All new staff are to be trained in the NT laboratory procedures including subsampling, sorting and sample storage.
•
Once sorting has commenced, the residues of the first five samples are
checked by quality control staff.
•
An assessment is made following the first five samples whether sorting
efficiency is acceptable.
•
If not satisfactory the checking is continued for every sample until the problem
is rectified.
•
If the sorting for the first five samples is acceptable a random selection of two
samples in the following ten, two samples in the following 30 and finally 1
sample in every 20 will be checked by quality control staff.
•
The protocol is repeated for the commencement of new projects and
sampling runs as well as new staff.
•
A sorting efficiency of > 90% is deemed acceptable. Sorted samples are
resorted and missed animals are identified, enumerated and compared to the
animals collected from the original sort. If the number of animals of a
particular Family counted in the re-sort is ?10 % of the total count (original plus
re-sort count for that Family), the person is given suitable instruction to ensure
that particular Family is adequately collected in future samples.
•
It is important when sorting to ensure all different taxa in a sample are
collected. If a particular Family contributes significantly to the count (> 10% of
the total count) an error in the count of ?10% is less important than in a Family
with a lower count. Where errors occur in the count of Families which
contribute significantly to the total count, sorting staff are cautioned and made
aware of the these groups to reduce the chance of significant errors in future
samples. Taxa not collected in the original count but collected in the re-sort
are treated as a significant error with the appropriate instruction to correct the
problem.
Appendix E - appendices
Page 33
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Identification and Enumeration Efficiency
•
All new staff are to be trained in the NT laboratory procedures including subsampling, sorting, identification, enumeration, sample storage and archiving.
•
Once sample identification has commenced, the identification results of the
first five samples are checked by quality control staff.
•
An assessment is made following the first five samples whether identification
and enumeration accuracy is satisfactory.
•
If not satisfactory appropriate instruction is given to the person as described
in the boxed section below.
•
If identification and enumeration for the first five samples is acceptable a
random selection of two samples in the following ten, two samples in the
following 30 and finally 1 sample in every 20 will be checked by quality control
staff.
•
The protocol is repeated for the commencement of new projects and
sampling runs as well as new staff.
•
A misidentification error of < 10 % of the total number of animals is deemed
acceptable. This is the error rate used by the Murray Darling Freshwater
Research Centre who conducted external quality control checks of all State
agencies participating in the AusRivAS program. If the error is > 10 %,
misidentifications are corrected under the guidance of quality control staff. All
misidentifications will be shown to the person and suitable instruction given to
rectify the misidentification. Other samples containing taxa that were
misidentified are then checked by the original identifier for misidentification
errors.
•
When an identification problem is encountered, a decision tree for
identifications should be followed as is figured in Hawking and O’Connor
(1997).
•
Very small, damaged, immature animals or pupae that cannot be identified
with confidence should be noted as such (eg. Trichoptera juvenile). These
animals are counted and placed in separate vials for each category. The
counts for unidentified animals are not included in the 200 organism
subsample.
•
Damaged animals should be identified if possible, with both head and tails
counted and the highest number recorded and placed in the appropriate vials.
If a specimen cannot be identified it should be noted as such (eg.
Ephemeroptera damaged) and placed in the appropriate vial. The same
procedures apply to the identification of Oligochaeta.
•
When identifying the samples, the taxa are separated into Orders and placed
in separate vials to eliminate any high level discrepancies. This is also
required for future curatorial preservation and storage.
Appendix E - appendices
Page 34
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Appendix E:7
Live Picking Methods
These descriptions of live picking methods are derived directly from each state /
territory methods manual. Two of the State protocols are described in flowchart
form below to illustrate the methodology (diagrams from WATER ECOscience,
2003).
1 Queensland Protocol
Victorian Protocol
Appendix E - appendices
Page 35
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Queensland
•
It is recommended that you initially separate small organic and substrate
material from large using a 1 cm panning sieve. Sort through these fractions
a small amount at a time, retaining the residue for QA/QC requirements. Pick
progressively through the sample (using forceps and/or pipette), replacing
picked material with remaining parts of the sample as picking progresses,
until a total of 200 animals have been collected or for 60 minutes (whichever
is completed first) attempting to pick as many different groups (families) as
possible.
•
Collect a maximum of 10 of any one type (family and in some cases order) or
animal. At least 20 midge larvae (Chironomidae) should be collected to
ensure adequate representation of the subfamilies.
•
For the first 15 minutes, collect from the predominant groups, remembering
only 10 animals per group
•
For the next 30 minutes scan for cryptic and/or rare groups. These will tend
to be the following groups:
→ Corbiculidae (juveniles
→ Chironomidae (larvae and pupae)
→ Empididae
→ Hydroptilidae (larvae)
→ Ceratopogonidae (larvae)
→ Oliogchaeta (including broken bits)
→ Elmidae (larvae)
→ Hydrophilidae (larvae and adults)
→ Simulidae (larvae)
•
For the final 15 minutes return to the more common animals
•
If it is raining or cold, or conditions of poor light exist due to cloud cover or
approaching twilight, the sample must be taken back to the
vehicle/motel/camp etc. for sorting undercover and with improved light
conditions.
Laboratory sorting
•
Ensure adequate ventilation in the workplace. Tip the contents of the jar into
a large petri dish; or pour the contents through a 250 um sieve over a sink,
wash the organisms and flush the sieve contents into a large petri dish with
water from a squeeze bottle
•
Place the petri dish under a stereomicroscope. Take a vial of suitable size to
take the collection of specimens in the petri dish with label inserted. The label
should have the following information: collection number, location code, site
name, collection date, habitat, sample identifier, and the identification date.
•
A dedicated tally sheet should be developed for recording the identities and
numbers of all taxa in a sample (see appendix). The sheet should allow
listing of the taxonomic key used for identification for each family, the person
making the identification, the site, date and sample code.
Appendix E - appendices
Page 36
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
Organisms are identified to family level except for Porifera, Nematoda,
Nemertea, Oligochaetes, Acarina and microcrustacea for which family level
identification is optional. Chironomids should be identified to subfamily level.
•
Select specimens and follow the appropriate taxonomic keys to family level.
Follow the keys to family level. If uncertain about the identity obtain a second
opinion from a colleague/local specialist. If a new family is suspected or other
significant problems arise in taxonomic identification, contact a national
specialist.
•
Identify each specimen, place in the vial and mark the tally sheet. When all
specimens have been counted record the total tally for each taxa. Place the
vial, filled with preservative, in an evaporation proof box in a suitable storage
location.
Victoria
•
The sample is washed in the net and then spread out onto a large white tray
where it is sorted for a standard time between 30 and 60 minutes. Animals
are collected with the aid of forceps, pipettes and spoons. Many animals can
be quickly collected from the corners of the tray, particularly when using
pipettes. Once or twice during the sorting it is advisable to strand animals by
tilting the tray to one side, thus exposing a third or half of the bottom. Rapidly
moving animals can be collected in this way, as well as molluscs, which
adhere to the bottom of the tray.
•
If large amounts of leaves, wood or aquatic vegetation are collected, these
should be rinsed and removed before sorting. If the water is cloudy due to
clays or fine sediment in suspension, put the sample back into the net and
rinse it in the stream again.
•
If large amounts of sand or coarse organic material are collected, put only a
proportion of the whole sample into the tray at one time, bearing in mind that
you need to completely sort the whole sample within the allotted time.
•
Good practice requires working in relatively high light levels. If ambient light
is low then artificial lighting will be required. Rain drops also adversely affect
sorting ability and umbrellas or tarps should be used.
•
The main objective of sorting is to collect as many different taxa as possible.
Care must be taken not to take too much time picking out large numbers of
very abundant species, as this will result in the less common species being
under represented or not collected. Only about 30 of each taxa need to be
picked out, they can then be ignored and the remaining sorting effort applied
to collecting other species. Considerable effort needs to be directed towards
searching for small or cryptic species.
•
The number of animals collected in 30 minutes is typically about 200, grossly
impacted sites are likely to have fewer than 50. The live sorting should aim to
collect about 200 animals in the allotted time. If after 30 minutes fewer than
100 animals have been collected, sorting should continue. If no new taxa are
found in the next ten minutes, cease sorting. If new taxa are found continue
up to a maximum of 60 minutes.
Appendix E - appendices
Page 37
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
Laboratory sorting
•
The sample is preserved in 80% ethanol and returned to the laboratory for
identification.
•
All identifications should be carried out with a stereomicroscope using
appropriate keys and up-to date texts. For a listing of invertebrate keys, see
Hawking (1994). Animals should be identified to family level with the
exception of Oligochaeta, Hydracarina, Chironomidae (subfamily), Hirudinea,
Platyhelminthes.
Western Australia
•
Empty the contents of the net into a bucket of water and swirl vigorously to
separate the mineral substrate from the organic component of the sample.
Pour the water fraction onto a stack of four sieves. The four sieves should
have mesh apertures of 16mm, 2mm, 500 um, 250 um, with the top sieve
having the coarsest mesh size progressing down to the 250 um mesh sieve at
the bottom of the stack. Continue to elutriate the sample in this manner until
confident that all animals have separated from the mineral component of the
sample. After elutriating the sample, agitate the stack of sieves in the water
column to separate the contents into the four size fractions. Empty each of
the four size fractions into separate sorting trays. Separating the samples into
the various sizes fractions facilitates the picking process by refining the
search image of the collector.
•
The sample must be picked by two collectors for a total of 60 mins ie. 30
mins by each collector. Allocate picking effort to each tray proportional to the
amount of material it contains. Typically the larger fractions (16mm and 2mm)
are picked for longer (eg. 9 minutes by each collector) with less time being
allocated to the smaller fractions (500 um and 250 um – 6 minutes each).
•
Remove animals from the trays using forceps and /or a pipette and store in a
vial containing 70 % alcohol.
•
The aim of the sampling is to maximise the diversity of the animals collected.
Start by collecting common, abundant taxa for the first five minutes. After that
the major picking effort should be directed at finding the less common,
inconspicuous taxa. Avoid over-picking large or colourful taxa. Aim for a total
of 200 animals (use a hand held counter) with maximum diversity. There is
no need for large numbers of any single taxon but a minimum of 30
chironomids should be picked from every sample to ensure that all the subfamilies are represented in the vial.
•
Particular care should be taken to search for the groups that can be
commonly missed when live sorting (ie. cryptic taxa) : elmid larvae,
Oligochaeta, Empididae, Hydroptilidae, small molluscs, Ceratopogonidae
•
If it is a really poor sample with very few animals in total, then stop at 60
mins. Make it clear on the field sheet that it is a poor site and why that was
so.
Laboratory Procedures
•
Tip the contents of the vial into a large petri dish. Place the petri dish under a
stereomicroscope which is correctly adjusted for your vision and work
posture. Use a vial of suitable volume to store the specimens in the petri
dish. Label the vial with the following information: site code, site name,
Appendix E - appendices
Page 38
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
collection date and habitat. Use a pencil or waterproof ink to fill in details.
Store the specimens in 70% ethanol.
•
A dedicated tally sheet should be developed for recording the identities and
numbers of taxa in a sample. (see appendix) The sheet should allow listing
of the key used for identification for each family the person making the
identification the site and date and sample code.
•
Organisms should be identified to familiy level with the exception of
Oligochates, Hydracarinids,
•
Ignore all microcrustacea (ostracods, copepods, cladocerans) but include
conchostracans and anostracans. Chironomids should be identified to subfamily level
•
Follow the keys to family level. If uncertain about the identity obtain a second
opinion from a colleague/local specialist. If a new family is suspected or other
significant problems arise in taxonomic identification, contact a national
specialist.
•
When each specimen has been confidently identified, mark one stroke in the
tally column of the data sheet, place the specimen in the vial and examine the
next specimen. When all specimens have been counted, add the tally for
each taxon and write the total in the appropriate column on the data sheet.
Place the vial, filled with preservative, in an evaporation proof box in a
suitable storage location.
Northern Territory
•
Whole samples are preserved in 90% ethanol and taken back to the
laboratory for processing.
Laboratory procedure
•
Set up a 250 um mesh sieve in a large container. Place the sample in the
sieve and allow the ethanol to drain into the container. Remove and save the
ethanol. Rinse the sample pot containers and lids with water into the sieve.
Refill washed containers with recycled ethanol. Rinse the sample and wash
out any fine sediment.
•
Place the sample into a subsampling device. In the NT a waterproof ‘Pelican’
case is set up with a perspex rack containing small plastic vials. A fitted wire
mesh is used to hold the vials and perspex rack in place. This type of box
subsampler is based on a subsampler designed by Marchant (1989)
•
Wash the sample into the subsampler box and fill with water until the case is
full. Lift the case up and invert so that the sample always falls away from the
vials. Agitate the case vigorously while inverted. While the sample material is
still being agitated flip the case upright onto the bench top. This should
ensure the sample is evenly distributed throughout the vials in the box.
•
A 200 organism subsample is required. Randomly select vials from the case
and place into a rack. Sort samples under a stereo microscope and remove
and count all macroinvertebrates. Record the number of vials required to
obtain the 200 organisms. Extracted vials must be completely sorted even if
the 200 organism count is reached, to enable estimates of total numbers.
Appendix E - appendices
Page 39
WATER ECOscience: QA/QC Project – Year 2 Milestone and Final Report.
Appendix E: Macroinvertebrate Sample Processing Error Report
•
All macroinvertebrates are to be identified to family level except for the
following, using the keys recommended by Hawkings (1999). Adults and
larvae are combined for the purpose of data entry and analysis.
→ Oligochatea (Class)
→ Acarina (Order)
→ Chironomidae (sub-family)
Appendix E - appendices
Page 40