;PA~600/4-87-037
United States
Environmental Protection
Agency
Office of Acid Deposition,
Environmental Monitoring and
Quality Assurance
Washington DC 20460
EPA/600/4-87/037
November 1987
Research and Development
Western Lake Survey
Phase I
Quality Assurance
Report
-------
Pacific
Northwest (4B)
California (4A)
Northern
Rockies (4C)
Central
Rockies (4D1
Southern
Rockies (4E!
Subregions of the Western Lake Survey - Phase I
-------
EPA 600/4-87/037
November 1987
Western Lake Survey
Phase1
Quality Assurance Report
A Contribution to the
National Acid Precipitation Assessment Program
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
Environmental Monitoring Systems Laboratory - Las Vegas, NV 89119
Environmental Research Laboratory - Corvallls, OR 97333
U
rvi-nl Protection Agency
.S. Environmental ProT
1S *>
Chicago, tt 60604
-------
Notice
The information in this document has been funded wholly or in part by the United States
Environmental Protection Agency under contract number 68-03-3249 and 68-03-
3050 to Lockheed Engineering and Management Services Company, Inc., No. 68-03-
3246 to Northrop Services, Inc., and Interagency Agreement Number 40-1441-84 with
the U.S. Department of Energy. It has been subject to the Agency's peer and
administrative review, and it has been approved for publication as an EPA document.
The mention of trade names or commercial products in this report is for purposes of
illustration and does not constitute endorsement or recommendation for use.
This document is one volume of a set which fully describes the Western Lake Survey -
Phase I. The complete document set includes the major data report (2 volumes), quality
assurance plan, analytical methods manual, field operations report, and quality assurance
report. Similiar sets are being produced for each Aquatic Effects Research Program
component project. Colored covers, artwork, and use of the project name in the
document title serve to identify each companion document set.
Proper citation of this document is:
Silverstein, M. E., M. L. Faber, S. K. Drouse, and T. E. Mitchell-Hall. Western Lake
Survey-Quality Assurance Report. EPA-600/4-87-037. U.S. Environmental
Protection Agency, Las Vegas, Nevada.
-------
Foreword
The primary function of this quality assurance report is to assess the data quality for the
Western Lake Survey - Phase I. A known degree of confidence in the data quality is
essential to the initial data user, who must rely on the estimates of data quality in the
determination of subregional and regional population estimates (a major goal of Phase I of
the National Surface Water Survey). Confidence in data quality is also essential to future
data users, who can use this report as a reference guide in determining levels of
performance for their own research purposes.
This document is also directed to numerous individuals, contractors, and government
agencies that were involved in the planning and the day-to-day survey operations.
Each of these participants has a unique interest in the specific performance aspects of
the survey. The U.S. Department of Agriculture Forest Service, the National Park Service,
the analytical and preparation laboratories, the field sampling personnel, and the field
laboratory personnel all have interests in specific information on performance and
participation. The detailed discussions, however, are not included solely for the benefit of
individual participants or groups; they are intended to aid program managers and future
survey designers in refining data quality objectives and methods on the basis of past
performance and sampling design.
The final goal of the document is to ask questions that do not, at present, have answers.
These questions are directed toward present and future data users and survey designers.
Thus, the document is intended as a guide for present and future data users and as a
history of events that may prove valuable to designers of similar surveys. The specific
expertise that these individuals bring to their reading of this document will be the ultimate
source of more efficient and meaningful survey designs and quality assurance programs.
-------
Abstract
The quality assurance program for the Western Lake Survey - Phase I was designed to
ensure that the data collected were of known and acceptable quality. The quality
assurance program was based on similar activities conducted for the Eastern Lake
Survey - Phase I and included the following major elements: selection of analytical
laboratories, training of field sampling and field laboratory crews, on-site evaluation of
field and analytical laboratories, daily communications with survey participants, and
verification and evaluation of data collected. Numerous quality assurance and quality
control samples (e.g., blanks, duplicates, audits, spikes, and check samples) were used
to identify, qualify, and quantify sources of sampling and analytical variability in terms of
precision, accuracy, bias, and analytical detectability. The relative importance of these
sources of variation was assessed by comparative statistical evaluations.
Until all of the phases of the National Surface Water Survey have been conducted and
their data sets are available for comparison, an assessment of Western Lake Survey -
Phase I data quality cannot be considered complete. It can be stated, however, that the
final data set represents data of high quality that can be used with confidence in the
calculation of population estimates. Precision, accuracy, and detectability estimates
generally met survey data quality objectives. Samples were complete, analyses were
performed within specified holding times, and 10 of 15 strata met sampling completeness
criteria. Quality assurance samples adequately characterized the routine lake water
samples, with the exception that field audit samples did not represent the midrange of the
lake water sample analyte concentrations.
For future surveys, refinement of data quality objectives and of the sampling design will
be necessary to improve partitioning of the components of variability and to account for
circumneutrality, differences in sample concentration, and differences in ionic strength of
lake waters. Data from the West can be compared to data from other elements of the
National Surface Water Survey; no calibration of data is necessary for procedural
differences in sampling or analytical methodology.
By its ability to identify trends and to isolate problems in the survey data, the quality
assurance program also confirmed the overall soundness of the survey design, execution,
and data generation process. The data verification process yielded numerous suggestions
for refining lake sampling, field laboratory, analytical laboratory, and data management and
analysis procedures. These suggestions are given in tabular form in the Conclusions and
Recommendations section, along with summaries of the associated findings, corrective
actions, and impact on data quality.
-------
This report was submitted in partial fulfillment of contract number 68-03- 3249 by
Lockheed Engineering and Management Services Company, Inc., under the sponsorship
of the U.S. Environmental Protection Agency. This report covers a field work period from
September 10, 1985, to November 4, 1985; data evaluation and verification were
completed as of May 14, 1986.
-------
Contents
Notice .!!
Foreword '"
Abstract IV
Figures Xl
Tables x"
Acknowledgments x'v
1. Introduction 1
Purpose 1
Organization 1
Specific Applications 1
Survey Design and History 2
Survey Participants 3
Data Quality Objectives 3
Sampling, Analytical, and Data Management Operations 6
2. Conclusions and Recommendations 11
Data Quality Objectives 11
Precision 11
Accuracy 11
Detectability 11
Representativeness 12
Completeness • 12
Comparability 12
Lake Water Characteristics 13
Extractable Aluminum 13
Total Aluminum 13
Acid Neutralizing Capacity 13
Base Neutralizing Capacity 13
Calcium 13
Chloride 14
Conductance 14
Dissolved Inorganic Carbon (air equilibrated) 14
Dissolved Inorganic Carbon (open system) 14
Dissolved Inorganic Carbon (closed system) 15
Dissolved Organic Carbon 15
Fluoride (total dissolved) 15
Iron 15
Potassium 15
Magnesium 15
VII
-------
Contents (continued)
Manganese ^g
Sodium '.'.'.'.'. 16
Ammonium ' ^
Nitrate -Ig
Phosphorus (total) '.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'. 16
pH (acidity; open system) 16
pH (alkalinity; open system) 16
pH (air equilibrated) 17
pH (closed system) '.'.'.'.'.''' 17
Silica 17
Sulfate '.'.'.'. 17
True Color ' ' 17
Turbidity '.'.'.'.'. 17
Overall Operations '.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'. 17
3. Operational Quality Assurance Program 25
Selection of Analytical Laboratories '.'.'.'.'.'.'.'.'.'.'.'.''' 25
Training of Sampling and Field Laboratory Personnel . . . . . . 25
Quality Assurance and Quality Control Procedures 25
Types of Quality Assurance and Quality Control Samples '.'.'.'.'.'.'.'. 25
Field Sampling Quality Assurance and Quality Control
Procedures 28
Field Laboratory Quality Assurance and Quality Control
Procedures 29
Analytical Laboratory Quality Assurance and Quality
Control Procedures 30
Communications 32
On-Site Evaluations '.'.'.''' 32
4. Data Base Quality Assurance 33
Data Management System ' ' 33
Raw Data Set (Data Set 1) '.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.''" 33
Verified Data Set (Data Set 2) '.'.'.'.'.'. 33
Validated Data Set (Data Set 3) '.'.'.'.'.'. 33
Final Data Set (Data Set 4) '.'.'.'.'.'.'.'.'.'.'.'.'. 34
Data Review and Verification '.'.'.'.'.'.'. 35
Review of Field Data Forms ' " ' 35
Initial Review of Analytical Laboratory Data Packages 35
Final Data Verification 35
Modifications to the AQUARIUS System '.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'. 36
Confirmation and Reanalysis Requests 37
Preparation and Delivery of Verification Tapes '.'.'.'.'. 37
Data Validation 38
5. Results and Discussion - Operational Quality Assurance Program 39
Field Sampling Activities and Protocols 39
Field Laboratory Activities and Protocols '.'.'.'.'.'. 39
Filtration Procedure 39
Receipt of Samples from Sampling Crews 40
Shipment of Samples 40
Comparison of Lake Site and Field Laboratory pH
Measurements 40
Analytical Laboratory Activities and Protocols 41
Incorrect Reporting of pH Values 42
Incorrect Use of Calibration Blanks 42
-------
Contents (continued)
Suspect Silica Values 42
Data Verification Activities 42
Review of Field Data Forms 42
Correction of Data 42
Requests for Reanalysis 43
6. Results and Discussion - Precision 45
Introduction . 45
Method of Estimating Precision 45
Estimating System Precision from Field Duplicate Pairs 46
Estimating Precision of Field Duplicate Pairs Analyzed
in the Field Laboratory 48
Estimating Field Laboratory Precision from Trailer
Duplicate Pairs 48
Estimating Intralaboratory Precision from Analytical
Laboratory Duplicate Pairs 50
Establishing the Quantitation Limit 50
System Precision Results 51
System Precision Estimated from Field Duplicate Pairs 51
Precision Estimated from Field Duplicate Pairs andTrailer
Duplicate Pairs Analyzed in the Field Laboratory 58
Intralaboratory Precision Estimated from Analytical
Laboratory Duplicate Pairs 58
Method of Estimating Precision Among Batches 59
Estimating Precision Among Batches from Field Audit
Samples 59
Use of Field Audit Samples in Estimating Precision 65
Among-Batch Precision Results 66
Among-Batch Precision Estimated from Field Audit Samples
Analyzed in the Field Laboratory 66
Among-Batch Precision Estimated from Field Audit Samples
Analyzed in the Analytical Laboratory 66
7. Results and Discussion - Accuracy 71
Introduction 71
Method of Estimating Accuracy from Field Synthetic Audit
Samples 71
Accuracy Results Estimated from Field Synthetic Audit Samples 71
Summary of Audit Sample Data for Precision and Accuracy 74
8. Results and Discussion - Detectability 79
Introduction 79
Method of Estimating System Detectability from Field Blank
Measurements 79
System Decision Limit 79
System Detection Limit 79
Detectability Results Estimated from Field Blank
Measurements 81
Comparison of Results for Field Blank Samples Collected
by Helicopter Crews and Ground Crews 82
Method of Estimating Detectability from Trailer Blank
Sample Measurements 82
Detectability Results Estimated from Trailer Blank Sample
Measurements 82
-------
Contents (continued)
Method of Estimating Detectability from Calibration Blank
and Reagent Blank Sample Measurements 83
Determining Instrument Detection Limit '.'.'.'.'' 84
Detectability Results Estimated from Calibration and
Reagent Blank Sample Measurements 84
Matrix Spike Sample Results 84
9. Special Studies g7
Calibration Study '.'.'.'.'.'"'' 87
Introduction ' g7
Sampling Design '.'.'.'.'.'.'.'.'.'.'.'.'. 87
Design Modifications \\ 88
Verification of Calibration Lake Data 90
Determination of Sampling Method Bias ] '.[ '.'. ',] go
Determination of Relative Bias Between Analytical
Laboratories 90
Determination of Calibration by Linear Regression 91
Holding-Time Effects on Sample Concentration 92
Relation of Calibration Study Sampling Times
and Locations 92
Summary '.'.'.'.'.'.'' 93
Nitrate-Sulfate Stability Study '.'.'.'.'.'.'.'.'.'.'. 93
Introduction 93
Sample Processing, Preservation, and Analysis 93
Analytical Results 94
10. References 97
Appendices
A. National Surface Water Survey Form 26, Data Confirmation/
Reanalysis Request Form 99
B. Calculation of Field Blank Sample Control Limits '.'.'.'.' 101
C. Preparation of Audit Samples 103
D. Distribution of Data for Field, Trailer, and Calibration Blank
Samples Analyzed in the Analytical Laboratories 105
E. Field Laboratory Precision Data for Audit Sample Measurements
of Dissolved Inorganic Carbon, pH, Turbidity, and True Color 109
F Estimated Precision for Audit Sample by Lot 113
G. Estimated Analytical Accuracy for Field Synthetic Audit
Samples by Lot 121
H. Field Audit Sample Control Limits and Summary of Field Audit
Samples Outside Control Limits 125
I. Relative Interlaboratory Bias in the Western Lake Survey -
Phase I 131
J. Figures Depicting Detectability Data and the Relationship
Between Precision and Mean Concentration by Analyte 155
K. Distribution of Analyte Concentrations for Routine Lake
Samples 191
L. Collection and Preparation of Nitrate-Sulfate Split Samples 193
M. Proposed Procedure for Use of Low Ionic Strength, Circumneutral,
Mid-Range pH and DIG Quality Control Check Samples 195
Glossary . . 199
-------
Figures
Number Page
1 Subregions studied, Western Lake Survey - Phase I 3
2 Overview of activities, Western Lake Survey - Phase I 10
3 Quality assurance and quality control sample flow, Western Lake
Survey - Phase I 27
4 Data base management, Western Lake Survey - Phase I 34
5 Lakes sampled versus sample types, Western
Lake Survey -Phase I 46
6 Methods of estimating precision, accuracy, and bias, Western Lake
Survey - Phase I 47
7 Ways in which quality assurance and quality control samples are applied
to estimates of precision and accuracy, Western Lake Survey -
Phase I 49
8 Proposed procedural steps that would be necessary to quantify the
collection, processing, and analytical components of variability 50
9 Relationship of duplicate pair samples to quantitation limits
and sample concentrations 52
10 Relation of statistical limits to data derived from blank samples,
Western Lake Survey - Phase I 80
11 Sample flow for the calibration study, Western Lake Survey - Phase I . . 89
12 Relative differences in nitrate concentrations, nitrate-sulfate Stability
study, Western Lake Survey - Phase I 95
-------
Tables
Number
Page
1 Summary of Results, Western Lake Survey Pilot Study 3
2 Chemical and Physical Characteristics Measured, and Associated Data
Quality Objectives for Detectability, Precision, and Accuracy, Western
Lake Survey - Phase I 4
3 Changes in Protocol between Eastern Lake Survey - Phase I
and Western Lake Survey - Phase I 7
4 Significant Findings, Conclusions, and Recommendations Concerning Lake
Sampling and Field Data Collection, WesternLake Survey - Phase I .. 18
5 Significant Findings, Conclusions, and Recommendations Concerning
Field Laboratory Activities, Western Lake Survey - Phase I 19
6 Significant Findings, Conclusions, and Recommendations Concerning
Analytical Laboratory Activities, Western LakeSurvey - Phase I 21
7 Significant Findings, Conclusions, and Recommendations Concerning
Data Management and Data Verification Activities.Western Lake
Survey - Phase I 23
8 Types and Numbers of Samples Analyzed, Western Lake
Survey - Phase I 31
9 Maximum Holding Times for Samples, Western Lake Survey - Phase I . . 32
10 Exception-Generating Programs within the AQUARIUS Data Review
and Verification System 36
11 Lakes Visited Twice by Sampling Crews, Western Lake
Survey - Phase I 40
12 Field Laboratory Holding Times for Samples Collected by
Ground Crews, Western Lake Survey - Phase I 41
13 Value Changes Incorporated into the Raw and Verified
Data Sets, Western Lake Survey - Phase I 43
14 An Example of the Relationship of %RSD to Duplicate
Pair Samples for Different Concentrations 51
15 System Precision Estimates Calculated from Field Duplicate
Pairs (Sampling Methods and Laboratories Pooled),
Western LakeSurvey - Phase I 53
16 Summary of System Precision Results by Variable (Sampling Methods and
Analytical Laboratories Pooled), Western Lake Survey - Phase I .... 55
17 System Precision Estimates Calculated from Field Duplicate Pairs (by
Sampling Method), Western Lake Survey - Phase I 56
18 System Precision Estimates Calculated from Field Duplicate Pairs
(by Analytical Laboratory), Western Lake Survey - Phase I 57
-------
Tables (continued)
Number Pa9e
19 Distribution of Field Duplicate Pairs (Helicopter and Ground) by
Laboratory, Western Lake Survey - Phase I 58
20 Checklist of Variables for which System Precision Estimates Calculated
from Field Duplicate Pairs Did Not Meet Intralaboratory Precision Goals
(Pooled and Separated by Sampling Method and by Laboratory),
Western Lake Survey - Phase I 59
21 Precision Estimates for Field Duplicate Pairs Analyzed in the Field
Laboratory, Western Lake Survey - Phase I 60
22 Precision Estimates for Trailer Duplicate Pairs Analyzed in the Field
Laboratory, Western Lake Survey - Phase I 60
23 Intralaboratory Precision Estimates Calculated from Analytical
Laboratory Duplicate Pairs (Laboratories Pooled), Western
Lake Survey - Phase I 61
24 Intralaboratory Precision Estimates Calculated from Analytical
Laboratory Duplicate Pairs (by Laboratory), Western Lake
Survey - Phase I 63
25 Summary of Intralaboratory Precision Results by Variable,
Western Lake Survey - Phase I 64
26 Precision Estimated from Field Natural Audit Samples
Analyzed Among Batches (Analytical Laboratories Pooled),
Western Lake Survey - Phase I 66
27 Precision Estimated from Pooled Field Synthetic Audit Sample
Lots (Analytical Laboratories Pooled and Separated) Analyzed
Among Batches, Western Lake Survey - Phase I 69
28 Summary of Analytes that Showed High Variability Among
Batches for Field Audit Samples, Western Lake
Survey - Phase I 70
29 Estimated Analytical Accuracy for Field Synthetic
Audit Samples Pooled, Western Lake Survey - Phase I 72
30 Summary of Variables That Did Not Meet Data Quality
Objectives for Estimated Analytical Accuracy, Western Lake
Survey - Phase I 73
31 Required Detection Limits, System Decision Limits, and System
Detection Limits for All Variables, Western Lake Survey -
Phase I 81
32 Evaluation of Field Blank Data by Sampling Method, Western
Lake Survey - Phase I 83
33 Results of Matrix Spike Percent Recovery Analysis, Western
Lake Survey - Phase I 85
34 Calibration Study Regression with and without 2-Way Interactions
of its Components, Western Lake Survey - Phase I 91
35 Holding Times for Calibration Study Samples Analyzed by the
Analytical Laboratories, Western Lake Survey - Phase I 92
36 Regression Statistics for the Differences between Routine
and Withheld Samples versus Holding Time by Laboratory,
Western Lake Survey - Phase I 93
37 Summary Statistics for Relative Differences in Analyte
Concentrations for the Nitrate-Sulfate Stability Study,
Western Lake Survey - Phase I 37
-------
Acknowledgments
Data presented in this report were produced through the Aquatic Effects Research
Program of the National Acid Precipitation Assessment Program under the sponsorship of
the U.S. Environmental Protection Agency. Members of the following organizations had
primary responsibility for survey design, sample collection and processing, sample
analyses, data verification and validation, and data management.
Environmental Monitoring and Services, Inc.
Lockheed Engineering and Management
Services Company, Inc.
Northrop Services, Inc.
Oak Ridge National Laboratory
Systems Applications, Inc.
U.S. Department of Agriculture - Forest Service
U.S. Environmental Protection Agency
Environmental Monitoring Systems Laboratory - Las Vegas,
Nevada
U.S. Environmental Protection Agency
Environmental Research Laboratory - Corvallis, Oregon
Versar, Inc.
The authors thank the following individuals for their contributions, assistance, and advice
during their involvement with the Western Lake Survey - Phase I.
Data Verification Analyst
Karen A. Cougan Lockheed-EMSCO
Quality Assurance Aquatics and Technical Support
Lynn W. Creelman Lockheed-EMSCO
Bryant C. Hess Lockheed-EMSCO
Mary D. Best Lockheed-EMSCO
Carol B. Macleod Lockheed-EMSCO
Stephen J. Simon Lockheed-EMSCO
Donald A. Hilke Lockheed-EMSCO
Daniel C. Hillman Lockheed-EMSCO
C.E. Mericas Lockheed-EMSCO
XIV
-------
Acknowledgments (continued)
Quality Assurance Aquatics and Technical Support (continued)
David V. Peck
Janice L. Engels
Charles M. Monaco
James E. Pollard
Gerald E. Byers
Eugene P. Meier
Statistical Support
Martin A. Stapanian
Forest C. Garner
Thomas Permutt
Alison K. Pollack
Mithra Moezzi
Computer Support
Daniel Allison
Brian N. Cordova
Richard K. Maul
John (In Seung) Lau
Martin A. Stapanian
Carol B. Macleod
Michael J. Pearson
Joseph Scanlan
Robert E. Enwall
Charles M. Monaco
John Fountain
David Hoff
Ganise Satterwhite
Thomas Hody
James Pendleton
Merylin Gentry
Raymond McCord
Les Hook
Paul Kanciruk
Methods Development
Daniel C. Hillman
F. Xavier Suarez
Eileen M. Burke
Logistical Support
Kenneth Asbury
Valerie A. Sheppe
David V. Peck
Kevin J. Cabbie
Gerald Filbin
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
U.S. EPA, Environmental
Monitoring Systems Laboratory - Las Vegas
Lockheed-EMSCO
Lockheed-EMSCO
Systems Applications, Inc.
Systems Applications, Inc.
Systems Applications, Inc
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Science Applications International Corporation
Science Applications International Corporation
Science Applications International Corporation
Oak Ridge National Laboratory
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed
Lockheed
Lockheed
Lockheed
Lockheed
•EMSCO
•EMSCO
•EMSCO
-EMSCO
-EMSCO
-------
Acknowledgments (continued)
Logistical Support (continued)
Alan Groeger
Mel Knapp
Ky Ostergaard
Robert Cusimano
Analytical Support
Daniel C. Hillman
Linda A. Drewes
John M. Henshaw
Molly Morison
David V. Peck
C. Hunter Nolen
F. Xavier Suarez
Richard C. Buell
Eileen M. Burke
Jerry Wagner
Kenneth Ives
Sally Ann Reed
Linda Carlin
Joe Matta
Sue Czdemir
David L. Lewis
Management Team
Dixon H. Landers
Joseph Eilers
David F. Brakke
Rick A. Linthurst
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Northrop Services, Inc.
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
U.S. EPA, Environmental Research Laboratory
Corvallis
Versar, Inc.
Versar, Inc.
Environmental Monitoring and Services, Inc.
Environmental Monitoring and Services, Inc.
Environmental Monitoring and Services, Inc.
Radian Corporation
U.S. EPA, Environmental Research Laboratory
Corvallis
Northrop Services, Inc.
Western Washington University
U.S. EPA, Aquatic Effects Research Program,
Research Triangle Park
Clerical. Editorial, and Word Processing Support
Linda K. Marks
Ramone W. Denby
Margaret E. Oakes
Annalisa H. Hall
Lynn A. Stanley
Brian N. Cordova
Kit M. Howe
John M. Nicholson
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
Lockheed-EMSCO
and the word processing staff of Computer Sciences Corporation, Inc Las Vegas
Nevada. '
XVI
-------
Acknowledgments (continued)
Graphics
Lynn A. Stanley Lockheed-EMSCO
Richard C. Buell Lockheed-EMSCO
Contributions provided by the following reviewers improved the quality and focus of this
report and are gratefully acknowledged: Stephen Bauer (Idaho Department of Health and
Welfare - Division of Environment), Donald Bogen (U.S. Department of Energy -
Environmental Measurements Laboratory), Earl Byron (University of California, Davis),
John Lawrence (Environment Canada - National Water Research Institute), Joseph
Eilers and Susan Christie (Northrop Services, Inc.), Thomas Permutt (Systems
Applications, Inc.), Merilyn Gentry and Raymond McCord (Science Applications
International Corporation, Inc.), Paul Kanciruk (Oak Ridge National Laboratory), Martin
Stapanian, David Peck, Gerald Byers, James Pollard, and Janice Engels (Lockheed
Engineering and Management Services Company, Inc.). Finally, recognition belongs to
Robert D. Schonbrod who served as project officer of this project.
XVII
-------
-------
Section 1
Introduction
Purpose
This report reviews the quality assurance (QA) and
quality control (QC) activities and the analytical data
quality estimates associated with the Western Lake
Survey - Phase I (WLS-I). It is intended to provide
baseline information on WLS-I data quality. The final
report for the survey (Landers et al., 1987; Eilers et
al., 1987) provides an overview of WLS-I activities
and results. The WLS-I QA plan (Silverstein et al.,
1987), the analytical methods manual (Kerfoot and
Faber, 1987), and the field operations report (Bonoff
and Groeger, 1987) provide detailed information about
specific aspects of WLS-I. These documents, in
turn, reference their Eastern Lake Survey - Phase I
(ELS-I) counterparts: Drouse et al. (1986), Hillman
et al. (1986), and Morris et al. (1986).
Organization
This QA report describes QA and QC activities
related to collecting, processing, analyzing, and
handling samples and data. General conclusions and
recommendations concerning data quality, as well as
supporting conclusions and recommendations
concerning QA program design and operation, are
presented in Section 2. The QA program used during
WLS-I sampling and sample analysis is described in
Section 3. Data review and data verification
procedures are described in Section 4. Results and
discussion related to the operational aspects of the
QA program are given in Section 5. Subsequent
sections present the statistical evaluations and quality
assurance results for three primary analytical data
quality objectives (DQOs): precision (Section 6),
accuracy (Section 7), and detectability (Section 8).
Sections 6 through 8 also provide guidance for using
the QA and QC data in interpreting WLS-I overall
results. Section 9 summarizes the special studies
conducted in conjunction with WLS-I and presents
QA and QC results associated with those studies.
The appendices provide supporting data, and a
glossary at the end of the document defines
abbreviations and terms used throughout.
Specific Applications
The sampling and QA designs of WLS-I were
complex. As a result, this document contains detailed
information about situations that may have affected
data quality. Readers interested solely in the impact
of data quality on population estimates are directed to
the following sections, tables, and figures:
• Section 2, "Lake Water Characteristics," is an
analyte-by-analyte synopsis of the QA and
QC data interpretation.
• Appendix J summarizes data quality analyte by
analyte. The figures illustrate data detectability
and the variability of the data used to calculate
the population estimates for all analytes over the
range of concentrations of WLS-I lake waters.
• Tables 15 and 21 present precision statistics
that indicate how variability affects the routine
lake sample results.
• Table 16 interprets the statistical results of
Tables 15 and 21.
• Table 20 summarizes the success of the major
precision components by sampling method and
by analytical laboratory.
• Table 29 presents estimated accuracy statistics.
• Table 30 summarizes Table 29 by presenting
analytes that exhibit a high degree of
inaccuracy.
• Table 31 presents the statistical relation
between detectability and the routine sample.
The system decision limit (Pgs) should be of
particular interest.
Readers interested in assessing whether or not the
DQOs were met and in determining how WLS-I
experience can be applied to future surveys are
directed to the following sections, tables, and figures:
• Tables 4 through 7 present significant findings
concerning WLS-I sampling, sample
preparation, sample analysis, and data analysis.
The problems, corrective actions, effects on the
-------
data, and recommendations for future surveys
are described for each aspect.
• Tables 22 and 23 present statistical results of
the intralaboratory precision estimates, which
relate directly to the DQOs for precision.
• Table 25 presents an interpretation of Tables 22
and 23.
• Table 29 presents a statistical analysis of
accuracy estimates.
• Table 30 summarizes Table 29 by presenting
analytes that exhibit a high degree of
inaccuracy.
• Section 8, "Estimating Detectability from
Calibration and Reagent Blanks"compares the
results of analytical instrumental detection limits
to the required detection limit, the DQO for
detectability.
• Appendix D, Table D-3 presents the statistical
results of the instrument detection and
calibration blank data discussed in Section 8.
• Section 9, "Special Studies," presents results of
the calibration study and nitrate-sulfate stability
study.
Survey Design and History
WLS-I was conducted during fall 1985 as a part of
the National Surface Water Survey (NSWS). NSWS,
which was initiated by the U.S. Environmental
Protection Agency (EPA) in 1983, is a project within
the National Acid Precipitation Assessment Program
(NAPAP). The goals of NSWS are (1) to describe and
evaluate, through a series of regional field surveys
and monitoring projects, the present chemical status
of lakes and streams in areas of the United States
that are potentially susceptible to the effects of acidic
deposition, (2) to study the temporal variability
associated with the chemical status of these waters,
(3) to identify associated biological resources, and (4)
to monitor changes over time in a representative
subset of the aquatic systems studied (Landers et al.,
1987).
Between mid-September and mid-November 1984,
the U.S. Department of Agriculture - Forest Service,
in conjunction with EPA Region 8, conducted a pilot
study to test the procedures that would be used in
WLS-I. The Forest Service selected 62 lakes from
among those in the Weminuche Wilderness (San
Juan Mountains, Colorado), the Uintas Wilderness
(Utah), and the Cloud Peak Wilderness (Big Horn
Mountains, Wyoming). When the pilot study lakes
proved difficult to reach from the ground within the
time constraints imposed by the sampling design,
concern arose that some of the approximately 900
lakes scheduled for sampling in WLS-I could not be
accessible from the ground. As a result, helicopters
were introduced as an alternative method of reaching
WLS-I lakes.
The pilot study helped anticipate problems that might
be encountered during WLS-I and contributed to the
refinement of procedures used to reach wilderness
lakes from the ground. Pertinent results from the pilot
study are summarized in Table 1.
In most respects, WLS-I followed the survey design
and protocols used for ELS-I. The major difference
between the two surveys was that WLS-I used two
access methods. Ground crews sampled the lakes
from boats; helicopter crews landed the helicopters
on the lakes in order to conduct sampling activities.
ELS-I crews used helicopter access only. Because
two methods of access were used, it was necessary
to develop a method for quantifying differences
between them. To provide this comparative
information, a calibration study (Section 9) was
incorporated into the WLS-I sampling design.
The mountainous areas studied were categorized as
subregions as shown in Figure 1, and the subregions
were divided into alkalinity classes. WLS-I ground
crews and helicopter crews collected samples from
757 of the 920 lakes originally scheduled for
sampling. (Bonoff and Groeger [1987] and Section 5
of this QA report discuss the reasons that some lakes
were not sampled.) This sample represented nearly
10,400 lakes in the target population. Most of the
lakes sampled were chosen randomly for use in
population estimates (Landers et al., 1987); these
lakes are referred to as the probability sample (719 of
the 757 lakes). Other lakes were chosen as special-
interest lakes; these lakes were not part of the
probability sample and were not used in population
estimates. For WLS-I, the term "population
estimate" refers to an estimate of the number of
lakes in the target population that have a particular
characteristic (i.e., alkalinity class of a subregion).
The estimate is extrapolated from the number of lakes
sampled (the probability sample).
Each lake was represented by a single routine
sample, for which 24 chemical and physical
characteristics were measured at the lake sites, field
laboratories, or analytical laboratories (see Table 2).
Descriptions of these characteristics and of the
analytical methods are given in Hillman et al. (1986)
and in Kerfoot and Faber (1987). The WLS-I
sampling design was based on the premise that
measurement of 24 variables for a single routine
sample from each lake would provide information
sufficient to evaluate the present chemical status of
the lakes studied. See Landers et al. (1987) for a
detailed discussion of population estimates.
-------
Table 1. Summary of Results, Western Lake Survey Pilot Study
Pilot Activity Situation Encountered
Application to WLS-I
Lakes Access was by ground (boat)
only
Lakes selected for proximity to
trailhead
Lake samples preserved and
processed in tents or outdoors
Lake samples processed without
electricity
Ground crews used a Hydrolab for in
situ measurements of conductance,
pH, and temperature
Samples shipped from field every
three days
Field communications
Data not collected on standardized
NSWS form; no QA/QC data
documentation
Bad weather (snow storm) closed trails in
Wyoming; 75% of lakes could not be
sampled
Phase I lakes were selected randomly;
some were far from trailhead
High risk of contamination
Unable to process extractable Al aliquot,
perform sample filtrations, or analyze for
DIG (closed system), pH (closed system), or
turbidity
An extra pack animal was needed to carry
CO2 tank for Hydrolab calibration
Protocol stated that extractable Al,
and pH had to be analyzed within 7 days;
improbable that analytical laboratory could
perform analysis within holding times
Considered inadequate; possibility of safety
hazards for sampling crews
Inability to compare pilot survey data to
other data bases confidently
Emphasized need for (1) helicopter
access and (2) coordinating sampling
time with weather forecasted
Emphasized need for helicopter
access
Emphasized need for central,
accessible field laboratory
Emphasized need for central field
laboratory
Emphasized complex logistics
necessary to obtain in situ
measurements by ground access
Emphasized need for daily shipments
as in ELS-I
Emphasized need for coordinated
communications network
Emphasized need for data base
management and QA/QC input
Figure 1. Subregions studied. Western Lake Survey
Phase I.
Northern Rocky
Mountains (40)
Pacific
Northwest (4B)
California (4A)) 1 NV
Rocky
^Mountains (4E)
Subregion Boundary
Survey Participants
The EPA Environmental Monitoring Systems
Laboratory in Las Vegas, Nevada (EMSL-LV), had
primary responsibility for the WLS-I sampling
operations and QA program. EMSL-LV received
assistance in these areas from its prime contractor,
Lockheed Engineering and Management Services
Company, Inc. (Lockheed-EMSCO). Lockheed-
EMSCO personnel performed the helicopter-access
sampling activities, and Forest Service personnel
performed most of the ground-access sampling
activities. State agencies and EPA regional offices
also were involved in the sampling activities.
Environmental Monitoring and Services, Inc., in
Thousand Oaks, California, and Versar, Inc., in
Alexandria, Virginia, provided analytical laboratory
services. The two laboratories were selected
according to procedures established for the EPA's
Contract Laboratory Program (CLP). Oak Ridge
National Laboratory (ORNL) in Oak Ridge,
Tennessee, was responsible for data base
management. The EPA Environmental Research
Laboratory in Corvallis, Oregon (ERL-C), had
primary responsibility for survey design, data
validation, and data interpretation.
Data Quality Objectives
WLS-I analytical data quality objectives (DQOs)
established the measurement criteria for the 24
variables studied. The statistical design, sampling and
analytical methods, and QA activities for WLS-I were
-------
Table 2. Chemical and Physical Characteristics Measured, and Associated Data
Survey - Phase I
Quality Objectives for Detectability, Precision, and Accuracy, Western Lake
Detectability
Intralaboratory
(Laboratory Duplicate)
Precision
Measure-
ment
Site*
A
A
A
A
A
A
A,L
A,F
A
A
A
A
A
A
Variable
(dissolved ions and
metals unless noted)
Al, extractable
Al, total
Acid Neutralizing
Capacity (ANC)
Base Neutralizing
Capacity (BNC)
Ca
cr
Conductance
(at25°C)
Dissolved Inorganic
Carbon (DIC)f
Dissolved Organic
Carbon(DOC)
F", total dissolved
Fe
K
Mg
Mn
Analytical
Method
Complexation with 8-hydroxyquinoline
and extraction into methyl isobutyl
ketone followed by atomic absorption
spectroscopy (furnace)
Atomic absorption spectroscopy
(furnace)
Titration and Gran analysis
Titration and Gran analysis
Atomic absorption spectroscopy (flame)
or inductively coupled plasma atomic
emission spectroscopy"
Ion chromatography
Conductivity cell and meter
Instrumental (acidification, CO2
generation, IR detection)
Instrumental (uv-promoted oxidation,
C02 generation, IR detection)
Ion-selective electrode and meter
Atomic absorption spectroscopy (flame)
or inductively coupled plasma atomic
emission spectroscopy d
Atomic absorption spectroscopy (flame)
Atomic absorption spectroscopy (flame)
or inductively coupled plasma atomic
emission spectroscopy''
Atomic absorption spectroscopy (flame)
or inductively coupled plasma atomic
emission spectroscopy d
Unit
rtlg/L
mg/L
ueq/L
neq/L
mg/L
mg/L
pS/cm
mg/L
mg/L
mg/L
mg/L
mg/L
mg/L
mg/L
Expected
Range (for lake
waters)
0.005-1.0
0.005- 1.0
-100 - 1,000
-10- 150
0.5 - 20
0.2- 10
10 - 1,000
0.05 -15
0.1 -50
0.01 -0.20
0.01 - 5.0
0.1 -1.0
0.1 - 7.0
0.01 - 5.0
Required
Detection
Limit
0.005
0.005
C
c
0.01
0.01
e
0.05
0.1
0.005
0.01
0.01
0.01
0.01
Percent Relative
Standard Deviation
(%RSD), Upper Limitf
10 (if Al cone. > 0.01 mg/L)
20 (if Al cone. < 0.01 mg/L)
10 (if Al cone. > 0.01 mg/L)
20 (if Al cone. < 0.01 mg/L)
10
10
5
5
2
10
5 (if DOC cone. > 5 mg/L)
1 0 (if DOC cone. < 5 mg/L)
5
10
5
5
10
Accuracy
Maximum
Absolute
Bias
m%
1 U /O
20%
10%
20%
•4 f\Q/
lU 70
1 no/
l(J/o
1 no/
I U /o
10%
CO/
O70
10%
10%
10%
10%
1 no/
lU%
10%
•i no/
10%
10%
(continued)
-------
Table 2. (Continued)
Detectability
Intralaboratory
(Laboratory Duplicate)
Precision
Measure-
ment
Site*
A
A
A
A
F,L
A
A
A
F
F
Variable
(dissolved ions and
metals unless noted)
Na
NH4 +
N 0.01 mg/L)
20 ( if P cone. < 0.01 mg/L)
±0.1 (pH unit)
± 0.05 (pH unit)
5
5
+ 5 (PCU)
10
Accuracy
Maximum
Absolute
Bias
10%
10%
10%
10%
20%
+ 0.05pH
± 0.05 pH
10%
10%
N/A
10%
a A = analytical laboratory, F = field laboratory, L = lake site.
& This limit was the %RSD at concentrations 10 times the required detection limit, unless otherwise noted.
c Absolute value of each blank had to be < 10 iieq/L
d Atomic absorption spectroscopy used by Laboratory II; inductively coupled plasma atomic emission spectroscopy used by Laboratory I.
e The mean of six nonconsecutive blank measurements had to be < 0.9 u-S/cm.
f Although more than one sample preparation procedure was used (e.g., air equilibration, closed system, open system), the data quality objectives were identical.
NOTE: No specific data quality objectives were set for in situ Secchi disk transparency and temperature measurements.
-------
structured to meet the DQOs for reporting population
estimates and chemical variability. The DQOs also
were applied to the statistical assessment of
sampling, field laboratory, and analytical laboratory
performance.
The primary DQOs were measures of precision,
accuracy, and detectability (see Table 2). Precision
was expressed as (1) standard deviation, (2) percent
relative standard deviation (%RSD), and (3) the
root-mean-square (RMS) of the %RSD, that is, as
a "pooled" precision or coefficient of variation. (See
Section 6 and Glossary for further explanation of
RMS and %RSD.) Accuracy was expressed as
maximum absolute bias, in percent. Detectability was
expressed in applicable units as an expected range of
values and as a detection limit. Each laboratory had
to meet the detection limit specification, which is
referred to throughout this report as the required
detection limit, for each analyte. For the variables
studied, measurements taken at the lake sites, in the
field laboratories, and in the analytical laboratories
were compared directly or indirectly to the values and
the ranges of values established for the DQOs.
During the survey, these comparisons were used to
locate potential sampling, analytical, and reporting
errors so that problems could be identified and
corrected early.
The values and the ranges of values originally were
determined on the basis of known instrument
performance as specified by the manufacturers,
standard laboratory practices (U.S. EPA, 1979), and
practical knowledge applied to statistical modeling of
chemical population estimates. The WLS-I values
and ranges were identical to those used in ELS-I
(Drouse et al., 1986), except that the precision
requirement for conductance was changed from 1
percent in ELS-I to 2 percent in WLS-I.
Three other DQOs, completeness, comparability, and
representativeness, also were considered in the
survey design. Completeness is a measure of the
quantity of data actually collected in relation to the
quantity that is expected to be collected. On the basis
of ELS-I results, completeness for WLS-I was set
at 90 percent or better for all variables. That is, of the
lakes selected for sampling, 90 percent or more were
expected to yield samples that would meet the QA
criteria and that could be used to estimate
populations. In addition, completeness refers to the
relation between the number of QA samples analyzed
and the number of routine samples analyzed.
Completeness also refers to the percentage of
samples that meet internal consistency checks and
that are analyzed within required holding times.
Comparability is the confidence level with which one
data set can be compared to another. For WLS-I,
comparability was ensured by requiring all sampling
crews and laboratory analysts to use uniform
procedures and by ensuring that a uniform set of
units was used for reporting the data. The calibration
study quantified the comparability of the helicopter-
access and ground-access sampling methods.
Comparability between WLS-I and other NSWS
surveys and between WLS-I and surveys conducted
under non-NSWS programs is discussed further in
Landers et al. (1987). In addition, significant design
and protocol changes that were implemented for
WLS-I as a result of ELS-I experience are qiven in
Table 3.
Representativeness, defined as the degree to which
data accurately and precisely represent a
characteristic of a population, is an important concern
of NSWS. The sampling scheme for WLS-I was
designed to maximize representativeness. A
systematic, random sample drawn within each
stratum ensured good geographic coverage without
bias (Landers et al., 1987). Other aspects of
representativeness apply to (1) the degree to which a
subset of lakes sampled represents the subregional
and regional population of lakes and (2) the degree to
which a single lake sample characterizes the
chemistry of the lake spatially or temporally.
Theseaspects of representativeness are discussed in
Landers et al. (1987). Finally, representativeness
applies to the degree to which QA and QC samples
represent routine lake samples. The ranges of analyte
concentrations in the QA and QC samples and in the
routine samples are evaluated to assess this aspect
of representativeness.
Sampling, Analytical, and Data
Management Operations
Field sampling activities conducted by ground crews
and by helicopter crews included locating and
describing lake sites, collecting lake water samples,
and collecting and recording physical and chemical
lake data at the sampling sites (see Figure 2).
Detailed field sampling procedures are given in Bonoff
and Groeger (1987). Ground-access and
helicopter-access sampling protocols are described
in Silverstein et al. (1987).
Sampling support facilities and mobile field
laboratories were located at the five WLS-I field
bases in Carson City, Nevada; Wenatchee,
Washington; Missoula, Montana; Bozeman, Montana;
and Aspen, Colorado. The primary goals of the field
laboratory operations were to receive samples, to
prepare sample batches, to perform selected
chemical analyses, and to preserve the integrity of
samples until their analysis at the analytical
laboratories. WLS-I analytical laboratories received
samples from the field laboratories, analyzed the
samples, and generated a report on the analytical
data (see Figure 2). The WLS-I analytical methods
are discussed in Kerfoot and Faber (1987); these
-------
Table 3. Changes in Protocol Between Eastern Lake Survey - Phase I and Western Lake Survey - Phase
Sampling Method and Field Data Collection
Protocol Change
ELS-I
WLS-I
Effect on Data
Recording lake site
locations (latitude and
longitude) on lake data
form
Van Dorn sampling
apparatus dimensions
In situ lake
measurements
(conductance, pH,
temperature)
Access to lakes for
sampling
Only Loran-C guidance
system coordinates recorded
Length 43 cm (volume
6.2 L)
Hydrolab used for all
measurements
Only by helicopter
Loran-C and map (USGS;
Forest Service) coordinates
recorded
Length 81 cm (volume 6.2 L)
Only helicopter crews used
Hydrolab; ground crews used
indicator strips for pH, used
thermistor for temperature,
and did not take conduc-
tance measurement
By helicopter and by boat
(ground crew)
Easier to confirm that lake
sampled was the correct lake
Shallow lakes sampled in
ELS-I could be as much as
0.5 m shallower than shallow
lakes sampled in WLS-I
No in situ conductance
measurements for 362 lakes;
questionable in situ pH
measurements for 362 lakes
No apparent effects on
population estimates
Field Laboratory Protocols
Protocol Change
ELS-I
WLS-I
Effect on Data
Sampling filtering
procedures
All aliquots of each sample
filtered in one filtration
apparatus
Preparation of aliquots Aliquots prepared (poured)
analyzed for total at workbench
aluminum
Color of labels used on All labels one color
aliquot bottles
Reanalysis of field
duplicate pair samples
when precision not
within control limits for
turbidity, true color, and
closed-system DIC
and pH
Safety check for MIBK
in ambient laboratory air
Only the duplicate sample
reanalyzed
Organic vapor monitors
Segregated aliquots filtered
for NOs" analysis from
apparatus that was washed
with 5% HNO3 (used for
aliquots filtered for metals
analyses); procedure used
first in NSS-Pilot after
development during ELS-I
Aliquots poured under
laminar-flow hood
Color-colored labels used
to distinguish aliquots
preserved with nitric acid,
with sulfuric acid, and by
refrigeration only
Routine and duplicate
samples both reanalyzed
Photoionization detector
Reduced the level of NQ3
background contamination
detectable in field blank
samples
Minimized chance of sample
contamination from dust
particles in ambient air of
field laboratory
Minimized chance of analyst
switching or improperly
preserving aliquots of one
sample or of multiple
samples
Better assessment of which
sample may have caused
the poor precision.
None; immediate response
time of photoionization
detector minimized health
risks
(continued)
-------
Table 3. (Continued)
Protocol Change
Analytical Laboratory Protocols
ELS-I
WLS-i
Effect on Data
Calculating the starting
date of analytical
laboratory sample
holding time
Began on date sample was
collected
Began on date sample was
processed and preserved in
field laboratory
Affected some ground-
access samples only; no
apparent effect on data (see
results of calibration study,
Section 9)
Data Verification and Data Analysis
Protocol Change
ELS-I
WLS-I
Effect on Data
Use of laboratory
synthetic audit samples
Synthetic audit
concentrations
Determination of field
blank control limits in
AQUARIUS program
AQUARIUS program
developed to compare
extractable and total
aluminum concentra-
tions for each sample
Anion-cation balance
program in AQUARIUS
Identifying erroneous or
unreliable data in the
verified data set (e.g.,
pH = 15.2)
Applying data qualifier
flags to raw data set
System of QA staff
requesting confirmation
and reanalysis of
analytical laboratory
data
Data tape transfer
among ORNL, EMSL-
LV, and ERL-C
Employed; possible
problems in sample
preparation
Low and high concentrations
used
Based on QA chemists'
experience with
environmental sample
analysis
Not a part of the
AQUARIUS system
All ANC values in the ion
balance calculation used as
they were reported by the
analytical laboratory
No mechanism
Employed
No systematic tracking
system used
Approx. 10 tapes used to
transfer data from raw to
verified data set
Not employed due to results
obtained in ELS-I
Only low concentrations
used; WLS-I lakes
expected to be dilute
Based on ELS-I field blank
data results (Appendix B)
Employed in WLS-I
All ANC values between
-10 neq/L and +10 jieq/L
changed to 0 ueq/L for the
ion balance calculation only
(Section 4)
Creation of the "XO" data
qualifier flag
Not employed
Application of a new NSWS
standardized form for
tracking requests (Appendix
A)
Two tapes used to create a
verified data set from the raw
data
Unable to estimate accuracy
of analytical laboratory
performance only
Without a variety of
concentrations, bias
calculations cannot be
performed
Historic NSWS data
provided a priori information
unavailable in ELS-I;
provided more confidence in
assessing blank data for
acceptable background
concentrations
Minimized possibility of
overlooking reporting or
analytical errors evident from
examining the
total/extractable aluminum
relationship
Eliminated unnecessary
flagging of data
Easier for data user to isolate
questionable data in
statistical analyses
Minimized confusion
concerning source of data
problems
Easier to track requests;
established documentation
system for data changes
Eliminated confusion in data
transfer by minimizing
number of iterations
(continued)
-------
Table 3. (Continued)
Protocol Change
Data Verification and Data Analysis (continued)
ELS-I
WLS-I
Effect on Data
Preparation of natural
audit lot as 2-L
samples at Radian
Corporation
Use of sample codes to
distinguish samples
collected by helicopter
and ground crews
Data quality objective
for conductance
(intralaboratory
precision goal)
Prepared samples as
needed
Not necessary, only
helicopter access used
1%
Prepared total lot volume en
masse as 2-L samples
Employed
2%
Ensured homogeneity of lot
by eliminating chance of
day-to-day contamination
Ease of statistical analysis to
detect potential differences in
data collected according to
different sampling methods
Probably none; 1 % may
have been too strict
methods paralleled ELS-I methods (Hillman et al.,
1986) to ensure data comparability.
Standardized, multicopy, field data reporting forms
were developed for use in recording site descriptions
and data collected at the lakes and the field
laboratories. One copy of each form was sent by
overnight mail service to ORNL for entry into the
NSWS data base, and a second copy was sent to the
EMSL-LV QA staff (see Figure 2). The field forms
are illustrated in Drouse et al. (1986) and in Bonoff
and Groeger (1987).
Data management and data review activities were
coordinated by EMSL-LV, ERL-C, and ORNL (see
Figure 2). A description of the data base management
system is given in Kanciruk (1986). Data review and
data verification procedures are described in
Silverstein et al. (1987) and are summarized in
Section 4 of this QA Report. Data validation
procedures are described in Landers et al. (1987) and
are summarized in Section 4.
-------
Figure 2. Overview of activities, Western Lake Survey - Phase I.
^^•MBM^H
NAPAP
ELS-I
Fall
1984
_^ A \ WLS-Pilol |
T^~~*| Fall 1984 |
WLS-I
Sampling Design
Lake
Selection
(ERL-C)
Develop Data
Quality Objectives,
Analytical Methods
(EMSL-LV)
Analytical
Laboratory
Selection
Field Personnel
Training
Lake Sampling
Ground Access
(Forest Service)
-M
Audit Sample
Preparation
(Radian Corp.
Austin, TX)
ICPAES
Split
Analysis
(ERL-C)
r
Logistics
Support
(EMSL-LV)
-n-
Lake Sampling
Helicopter Access
(EMSL-LV)
a
Field
Communications
Coordination
(EMSL-LV)
Field Laboratory
Operations
• Carson City, NV (4A)
• Wenatchee, WA (48)
• Missoula, MT (4C)
• Bozeman, MT (4D)
• Aspen, CO (4E)
Analytical Laboratory
Operations
•VERSAR Springfield, VA
•EMSI Newbury Park, CA
NO3 /S04
Split
Analysis
(EMSL-LV)
Field
Operations
Report
Management
Team Activities
IE
QA
Plan
Validated Data Set
(Data Set 3)
Final Data Set
(Data Set 4)
+
Sample
Tracking
(SMO)
^
r ^
Calibration
Study
Results
>
Deter
Da
Qua
Release Data to
Scientific Community
10
-------
Section 2
Conclusions and Recommendations
Data Quality Objectives
Precision
• For most analytes, system precision met the
DQOs for intralaboratory precision. This is the
only precision goal established before the survey
and, therefore, is the only gauge applicable for
comparing system precision results. Precision
for 19 of the 28 analytical laboratory and field
laboratory variables met or approached the DQO
(see Table 16 in Section 6). Poor precision for
most of the remaining analytes was attributed to
the low analyte concentration levels or the
circumneutrality of most WLS-I lake samples.
The few remaining poor precision estimates
were related to procedural (method or analytical)
problems.
• Field laboratory precision was acceptable for
analyses performed in all five WLS-I field
laboratories. Acceptable field laboratory
precision is especially critical for the closed-
system dissolved inorganic carbon and pH
measurements, which are used in population
estimates.
• Analytical laboratory precision met the DQOs for
all analytes except manganese.
• Precision differences between helicopter-
access and ground-access methods were
minimal.
• DQOs for precision must be developed to
account for different sample concentrations,
different ionic strengths, and circumneutrality of
lake water samples.
• DQOs must be developed that differentiate
between system (field related) precision and
laboratory precision.
• Audit sample precision estimates are most
useful if the mean concentrations of audit
samples are similar to the analyte
concentrations of the lake samples in the
subregion. WLS-I audit sample concentrations
did not always bracket the concentrations of the
lakes in WLS-I subregions.
Accuracy
• On the basis of field synthetic audit sample data,
accuracy could only be estimated for 15 of the
28 variables analyzed in WLS-I laboratories. Of
those 15 analytes, only calcium and total
aluminum exhibited levels of inaccuracy that
were higher than the DQO criteria.
• Accuracy estimates can be affected by analyte
concentration. WLS-I used one synthetic audit
sample at one theoretical concentration to
estimate accuracy for each analyte. Varying
analyte concentrations that represent the range
of concentrations in the routine lake samples
could improve the estimation of accuracy. For
future surveys, DQOs must account for this
relationship. Concentrations of analytes in the
synthetic audit samples and the number of
synthetic audit samples at different
concentrations should be established
accordingly.
• It is difficult to ensure the theoretical values for
the analyte concentrations in WLS-I synthetic
audit samples. Methodological changes in the
preparation of synthetic audit samples or use of
applicable samples certified by the National
Bureau of Standards (NBS) will be necessary if
future surveys require accuracy estimates for
acid neutralizing capacity, base neutralizing
capacity, dissolved inorganic carbon, and pH.
• For WLS-I, synthetic audit samples were
processed in the field laboratory only; therefore,
there is no means of isolating analytical
laboratory accuracy by using the data collected.
To provide an estimate of analytical accuracy in
future surveys, reliable audit samples (such as
those certified by NBS) must be sent directly to
the analytical laboratory. Conversely, if an
estimate of system accuracy is desired, a
synthetic audit sample must be processed
through the sampling apparatus at the lake site,
as are field blanks and field duplicates.
Detectability
• For most analytes, system background
contamination was within expected and
acceptable limits. Significant exceptions were
calcium, nitrate, and silica (see discussions later
in this section and in Section 8).
• Background contamination contributed by the
field laboratories was negligible for most
11
-------
analytes; nitrate, silica, and sulfate were
exceptions. In future surveys, trailer blanks
should be used regularly to allow estimation of
the effect of the sample processing component
on the sampling and analytical system.
• On the basis of calibration blank and reagent
blank analyses, both analytical laboratories met
the required detection limit criteria (see Table 2)
for every applicable analyte. Background
contamination and instrumental signal variability
contributed by the analytical laboratories,
therefore, was negligible.
• Helicopter crews and ground crews had similar
success in minimizing contamination in lake
water samples.
• DQOs were not set for field blank and trailer
blank concentrations prior to the survey; they
were set for analytical instrument detection of
calibration blanks and reagent blanks only.
Consequently, DQOs did not apply to field blank
or trailer blank analyses in WLS-I. Objectives
for field blank and trailer blank analyses should
be developed for future surveys.
Representativeness
• Duplicate pair samples adequately represented
the sampling methods and the ranges of
concentrations found in lake samples.
• One portion of the lake samples was not
adequately represented by field audit samples
because there were no audit samples with
concentrations of analytes in the midrange of
the routine lake water samples analyzed during
WLS-I. This lack of representativeness affected
the ability to quantify possible biases attributable
to the analytical laboratories.
• The field synthetic audit was used to estimate
accuracy, but it represented a single theoretical
concentration.
• Field blank samples adequately characterized
background contamination.
• Matrix spike percent recovery analyses indicated
that the reported concentrations were
representative of the analytes in the samples.
Completeness
• Each of the five WLS-I subregions represented
three alkalinity classes (strata) for a total of 15
strata. Fifty lakes were to be sampled within
each stratum. Of the resulting 750 lakes that
were expected to be sampled, (referred to in
Landers et al. [1987] as probability sample
lakes), 720 were sampled. (When population
estimates were performed, one of the 720 lakes
was deleted from the statistics because it was
too large.) The completion rate of 90 percent
(45 lakes) per strata was met for 10 of the 15
strata. Two strata in the 4D (Central Rocky
Mountain) subregion were undersampled to the
point that confidence in the population estimates
could be low. Most of the unsampled lakes in
these two strata were high-altitude lakes that
were frozen when visited by the sampling crews.
• Most WLS-I samples were complete in internal
consistency; 99.1 percent were within QA
criteria for anion-cation balance, and 97.6
percent met the conductance balance criteria.
• Of the 39,400 analyses performed in the
analytical laboratories, 98.6 percent were
completed within prescribed holding times.
• Each type of QA and QC sample was
represented by a large enough population to
allow statistical analyses of data quality to be
performed. Field blanks, field duplicates, field
audits, and the extra samples collected to
perform the calibration study constituted 54
percent of the WLS-I field samples analyzed.
• All on-site laboratory reviews were completed.
Both analytical laboratories, all field laboratories,
and all helicopter crews were evaluated. Five of
the sixty ground crews also were evaluated. No
criteria were set for the percentage of field
crews that should have been evaluated. This
aspect of completeness should be assessed if
future surveys warrant the use of a large
number of sampling crews.
Comparability
• The WLS-I data base can be compared to
other National Surface Water Survey data
bases. For most protocols, the field sampling
and analytical methodologies were identical to
those used in ELS-I. Where protocols differed
(i.e., helicopter access versus ground access),
no calibration of data was necessary (see
Section 9). Differences between data collected
by helicopter-access and ground-access
sampling methods were determined to be of
small enough magnitude that they do not affect
data interpretation or population estimates.
• Little difference between measurements was
indicated for samples preserved at the lake site
and at the field laboratory for nitrate and sulfate
(Section 9).
• Some biases between the two analytical
laboratories were detected for some analytes,
but the biases were relative, as well as small, in
most cases. The ability to quantify bias at
different analyte concentrations and to
compensate for those biases should be
investigated for future surveys.
12
-------
Lake Water Characteristics
Extractab/e Aluminum
• All detectability data for extractable aluminum
met the DQOs; contamination was not a
significant factor.
• The low concentrations of extractable aluminum
found in the lake water samples made it difficult
to compare the precision results to the DQOs.
Only 2 of 210 field duplicate pairs had mean
concentrations above 0.04 mg/L, and only 1 of 6
audit sample lots had a mean concentration
above 0.01 mg/L. The data user should take
note of the low extractable aluminum
concentrations when assessing data quality.
• Accuracy could not be estimated because of a
methodological problem caused by the instability
of the extractable Al species in the field
synthetic audit sample solution. Methodologies
for preparing field audit samples should be
modified, or an alternative method should be
investigated for future survey efforts.
Total Aluminum
• Most of the field routine, duplicate, and audit
samples used in calculating precision estimates
were near or below the detection limit for total
aluminum.
• Although accuracy can be estimated, the low
theoretical concentration of the synthetic audit
(0.02 mg/L) was also near the detection limit.
• Because precision and accuracy estimates are
concentration dependent (especially for low
concentrations), the DQOs did not account for
most total Al sample concentrations that were
near the detection limits. Data for concentrations
that were sufficiently above the detection limits
(usually, about 10 times the required detection
limit) are more useful for the calculation of
population estimates.
• There was good agreement in the QA check
comparing total aluminum and extractable
aluminum concentrations: 99.8 percent of the
1,642 samples analyzed for both variables had
total aluminum concentrations that were higher
than the respective extractable aluminum
concentrations.
Acid Neutralizing Capacity
• All quality assurance data estimates indicated
that results for acid neutralizing capacity are of
acceptable quality and are suitable for use in
calculating subregional population estimates.
• The analysis of field blank data indicated that
the required detection limit was met for acid
neutralizing capacity.
• For measurements of acid neutralizing capacity,
precision met the DQOs over the range of
routine sample concentrations. A method should
be developed for determining a quantitation limit
for use in assessing laboratory duplicate
(intralaboratory) precision.
• The WLS-I quality assurance program did not
include methods applicable to the estimation of
accuracy for acid neutralizing capacity. A means
of estimating accuracy should be developed for
use in future surveys.
• All computer software that the analytical
laboratories use to calculate ANC should be
checked to ensure that the programs are
calculating the titration data results correctly.
This procedure would minimize the possibility of
miscalculating ANC results, as did one analytical
laboratory during the initial stages of WLS-I.
Performing standardization checks on the
computer programs before survey analytical
activities commence will ensure consistent data
reporting and comparability among data bases.
Base Neutralizing Capacity
• Detectability estimates were higher than the
required detection limit for about 50 percent of
the field blank samples measured.
• Precision improved as concentration increased;
many of the field duplicate pair and field audit
sample mean concentrations were near or below
the detection limits.
• Accuracy could not be estimated by using the
QA samples employed in WLS-I. A means of
calculating accuracy estimates should be
developed for future surveys.
• The DQOs for base neutralizing capacity may be
too stringent. Alternatively, modifications to the
measurement system may be needed. Base
neutralizing capacity was not assessed for
population estimates. The uncertainty of the
estimation of base neutralizing capacity results
for WLS-I should be noted by the data user
concerned with this analytical measurement.
• In the future, all computer software that the
analytical laboratories use to calculate BNC
should be checked to ensure that the programs
are calculating the titration data results
consistently and correctly.
Calcium
• QA data for calcium indicated that data for the
routine samples are of acceptable quality and
can be used with confidence.
• Background contamination (as much as 0.07
mg/L) may be related to the fact that high
concentrations of Ca (mean of 3.7 mg/L) were
found in routine lake samples, which may have
resulted in the analyte carryover indicated in the
field blank sample. This carryover may relate to
residual analyte concentrations (i.e., inefficient
rinsing of the sampling apparatus or the filtration
apparatus) or to the way in which the instrument
13
-------
analyzes the sample and interprets the findings.
This slight contamination should not affect
population estimates.
• Precision estimates met the DQOs.
• A relative analytical bias of 4 percent to 8
percent was indicated for the two analytical
laboratories on the basis of calibration study
data. Field audit sample data indicate a bias of 8
percent. Measurements from Laboratory I were
higher than those from Laboratory II. When
assessing population estimates by subregion,
knowing which analytical laboratory analyzed the
samples may be important to the data user
investigating the anion deficit described in
Landers et al. (1987). Because the biases are
relative, however, no conclusion can be drawn
concerning the accuracy of one laboratory over
the other in the measurement of calcium (except
in the case of field synthetic audit samples; see
below).
• The accuracy estimates calculated from the field
synthetic audit sample data indicate that one
analytical laboratory exhibited better accuracy
than the other at the theoretical concentration of
0.19 mg/L. Laboratory It's accuracy estimate
( + 1.6%) was within the DQO, but Laboratory I
had an accuracy estimate well outside the DQO
and values that were much higher ( + 28.7%)
than the theoretical concentration. This absolute
bias (as accuracy) is consistent with the relative
bias results indicated by field natural audit
sample data and calibration study data. This bias
may be correlated with an anion deficit
described in Landers et al. (1987). However,
because the accuracy estimate for Ca was
based on only one theoretical concentration,
confidence in calculating an absolute bias as
accuracy is restricted to that concentration and
cannot be extrapolated with confidence across
the entire range of routine sample
concentrations.
Chloride
• The analytical results for the chloride
measurement indicate that the data are of
acceptable quality.
• Slight background concentrations of chloride (as
much as 0.05 mg/L, but generally lower) were
seen in field blank and trailer blank
measurements, but population estimates should
not be affected.
• Precision estimates indicate that, for samples
above the detection and quantitation limits, the
DQOs were met.
• At sample concentrations of 0.34 mg/L (the
theoretical concentration of chloride in the field
synthetic audit), accuracy estimates met the
DQO.
Conductance
• Conductance data are of acceptable quality and
can be used confidently in calculating population
estimates.
• Background concentrations were found to be as
much as 1.0 pS/cm (at 25°C) above the
required detection limit, but contamination was
at very low levels and should not affect data
interpretation.
• The distribution of field duplicate pair and field
audit mean conductance values indicated that
precision improves with the increasing ionic
strength of the sample. Because many lakes of
low ionic strength were sampled in the West,
precision estimates for such samples can be
expected not to meet the DQOs. Imprecision at
these low levels should not affect data
interpretation.
• The WLS-I QA program did not provide a
means of estimating accuracy for conductance.
A method of performing this estimate should be
incorporated in future survey designs.
Dissolved Inorganic Carbon (air equilibrated)
• The QA data for this analyte indicated that the
lake data are of high quality and can be used
with confidence.
• Background concentrations between 0.15 and
0.35 mg/L (compared to a required detection
limit of 0.05 mg/L) were found in most field
blanks and trailer blanks. Although these
measurements were above the required
detection limit, they may still be considered
acceptable for deionized blank water samples.
• Above concentrations of 1.5 mg/L, field audit
samples exhibited precision that met the DQO.
Significant imprecision at lower concentrations
may have been caused by slight differences
between samples and between laboratories in
the process used to sparge the sample. In
addition, higher precision estimates were
expected for samples at lower concentrations.
• There was no mechanism for estimating
accuracy for this analyte in the WLS-I QA
program. A means of performing the estimate
should be incorporated in future survey designs.
Dissolved Inorganic Carbon (open system)
• The QA data for this analyte indicated that the
lake data are of high quality and can be used
with confidence.
• Field blank background concentrations were
similar to those for air-equilibrated dissolved
inorganic carbon. These background
concentrations are unavoidable when the
methodology employed in the West is used, but
they should not affect data quality.
14
-------
For sample concentrations above the
quantitation limit, precision generally met the
DQO.
Although a theoretical value was calculated for
estimating accuracy, the field synthetic audit
sample exhibited sample matrix problems that
made the accuracy estimate unreliable. A means
of confidently estimating accuracy for open-
system dissolved inorganic carbon
measurements should be incorporated in future
survey designs.
Dissolved Inorganic Carbon (closed system)
• The QA data for closed-system dissolved
inorganic carbon indicated that the lake data are
of acceptable quality and can be used in
calculating population estimates.
• Background contamination could not be
assessed because field blanks and trailer blanks
were not analyzed for this measurement. Field
blanks or other means of determining field-
related background contamination should be
considered for inclusion in future sampling
designs.
• Precision was good for this measurement in
each of the field laboratories.
• No applicable accuracy checks were available
for this measurement; such checks should be
developed for use in future surveys.
Dissolved Organic Carbon
• The QA data indicated that the lake data for this
analyte are of acceptable quality.
• Background concentrations generally were
between 0.05 and 0.35 mg/L; the required
detection limit was 0.1 mg/L.
• Field duplicate pair and field audit analyses
showed a strong relationship between pooled
precision and concentration. Precision for mean
concentrations above the quantitation limit met
the DQO (except for two values). Precision for
many QA samples was above the DQO. Routine
lake sample concentrations, however, were
generally low. Thus, the precision may still
indicate high-quality data at these
concentrations.
• The accuracy estimate was within acceptable
limits.
Fluoride (total dissolved)
• The QA data indicate that the routine data are of
acceptable quality and will be useful in
calculating population estimates.
• The blank data met the DQO for detectability.
• Precision above sample concentrations of 0.08
mg/L met the DQO. Field duplicate pair mean
concentrations, field audit sample
concentrations, and most concentrations in
routine lake samples were below that level.
Some imprecision is indicated for analyses
performed by Laboratory I, where samples from
subregions 4D (Central Rocky Mountains) and
4E (Southern Rocky Mountains) were analyzed.
Iron
• Background concentrations were 0.01 mg/L
above the required detection limit.
• Mean concentrations of most field duplicate
pairs and of five of the six field audit sample lots
were below the quantitation limit and near or
below the detection limits. This observation
correlates well with the low concentrations of
iron found in lakes in the West: most concen-
trations for routine samples, field duplicate pairs,
and field audit samples were less than 0.06
mg/L. Although contamination was negligible,
precision at low concentrations did not meet the
DQO. The data user should consider that the
poor precision estimates may have been a
function of concentration and not a reflection on
sampling or analytical methods.
• The accuracy estimate was poor. It was directly
related to methodological problems associated
with the field audit sample instability and was not
related to the analytical measurements. A
different method of estimating accuracy should
be incorporated in future survey designs.
Potass/urn
• The QA data indicated that the lake data for this
analyte are of high quality and can be used
confidently in calculating population estimates.
• Contamination was negligible (0.01 mg/L);
background concentrations were near the
required detection limit.
• Precision and accuracy estimates met the
DQOs.
Magnesium
• The QA data indicated that the lake data for
magnesium are of high quality and can be used
confidently in calculating population estimates.
• The DQOs were met for detectability, precision,
and accuracy.
Manganese
• Contamination was negligible; most values for
field blanks were near the required detection
limit. Laboratory II showed some negative bias
for about 25 percent of the field blanks analyzed
there.
• Lake sample data for concentrations above
0.030 mg/L can be used confidently in
calculating population estimates. Field duplicate
pairs and field audit samples that had
15
-------
concentrations above 0.030 mg/L met the DQOs
for precision and accuracy. Because the
manganese concentrations in most lakes in the
West were below or slightly above the detection
limits, imprecision at those concentrations
should have little impact on the calculation of
population estimates.
Sodium
• The QA data indicated that the lake data are of
high quality and are suitable for use in
calculating population estimates.
• Negligible contamination (0.01 mg/L), was seen
in relation to the required detection limit.
• Precision and accuracy estimates generally met
the DQOs.
Ammonium
• Very low concentrations of ammonium were
measured in all lake and QA samples; most
were below the required detection limit.
• There was negative bias for 51 percent of the
field blanks analyzed in Laboratory I.
• Precision estimates for the field synthetic audit
samples were near the DQO at measurable
concentrations.
« Accuracy estimated from one of the two field
synthetic audit sample lots was good; for the
other field synthetic audit sample, analyte
degradation may be the cause of accuracy
estimates that did not meet the DQOs.
• At the concentrations measured, imprecision
and inaccuracy should not affect population
estimates.
Nitrate
• Measurable concentrations of nitrate (as much
as 0.071 mg/L) were detected in field blanks.
Analytical laboratory calibration showed minimal
contamination. Trailer blank measurements, on
the other hand, detected as much as 0.074
mg/L nitrate, which indicates that the
contamination may have been introduced in the
field laboratories and probably was not related to
field sampling methodology. Because con-
centrations in the field and trailer blanks were
substantially higher than the required detection
limit and because concentrations in many of the
lake samples were low, background
contamination may have been a significant
contributor to the analytical results for some lake
samples. The data user should note the possible
source of contamination. This factor, however,
may not be of concern in calculating population
estimates because the nitrate concentrations
were low in the lake samples. If contamination at
these low concentrations is of concern,
sample-processing and sample-handling
protocol modifications should be considered in
the design of future surveys.
• Precision estimates for samples above the
quantitation limit (0.342 mg/L) met the DQOs,
but imprecision was indicated in some field
duplicate pair mean concentrations below the
quantitation limit.
• Accuracy estimates met the DQO.
• The results of the nitrate-sulfate stability study
indicated that there was little difference between
nitrate concentrations in lake samples preserved
with mercuric chloride at the lake site and
concentrations in samples processed according
to NSWS protocol in the field laboratories.
• The length of time that a sample was held
before preservation had minimal effect on data
quality.
Phosphorus (total)
• Some contamination was detected at
concentrations of as much as 0.017 mg/L for
analyses performed in Laboratory I.
• Most concentrations of total phosphorus for
routine lake samples and for field duplicate pair
and field audit samples were less than 0.025
mg/L. Precision estimates have little meaning at
these low concentrations. QA samples that had
higher concentrations met the DQO for
precision.
• Estimated accuracy was within acceptable limits.
pH (acidity; open system)
• The QA data indicate that the open-system pH
measurements are of high quality. Closed-
system pH measurements made in the field
laboratory, however, are used in calculating
population estimates. The open-system pH
measurements performed in the analytical
laboratory were used as a redundant check on
the closed-system measurements.
• Field blank analyses indicated that background
contamination had minimal effect on pH values.
• Precision was greatly affected by the ionic
strength and circumneutrality of the sample.
Precision estimates improved as pH increased
or decreased from pH 7.0. A means of
calculating quantitation limits that can be related
to ionic strength and circumneutrality should be
developed for use in future surveys.
• The survey design did not allow accuracy to be
determined for pH. A means of determining
accuracy for pH should be developed for use in
future surveys.
pH (alkalinity; open system)
• Conclusions and recommendations for open-
system pH (alkalinity) are identical to those for
open-system pH (acidity).
16
-------
pH (air equilibrated)
• Conclusions and recommendations for open-
system pH (acidity) are related directly to this
pH measurement.
pH (closed system)
• QA data indicated that the field laboratory pH
measurements are of high quality and can be
used confidently in calculating population
estimates.
• Field blanks were not analyzed for this
measurement, so background contamination
could not be assessed. A means of determining
background contamination levels should be
incorporated in future sampling designs.
• The trailer duplicate precision for pH measured
in the field laboratory (0.03 pH units) met the
DQOs.
• When field duplicate pair measurements for all
five field laboratories were pooled, however, the
precision was 0.12 pH unit. Field audit sample
data indicated precision near the DQO for all
field laboratories. A quantitation limit related to
ionic strength and circumneutrality should be
considered for use in future sampling efforts.
• The WLS-I QA program did not provide a
mechanism for estimating accuracy for closed-
system pH. A means of estimating accuracy of
pH measurements should be developed for use
in future surveys.
Silica
• Although field blank measurements indicated
background contamination (as much as 0.18
mg/L) that was higher than the required
detection limit, the average SiO2 concentration
for a routine lake sample was about 3.7 mg/L.
Therefore, background contamination should
have a negligible effect on population estimates.
• For mean concentrations above the quantitation
limit, precision estimates were slightly above the
DQO. Some imprecision indicated from field
duplicate pair measurements may be related to
the digestion process used in the analytical
laboratory. Population estimates, however,
should not be affected; the precision may still be
reasonable for the specific purpose defined by
the data user.
• Accuracy estimates met the DQO.
Sulfate
The QA data indicated that the routine lake
sample data are of high quality and can be used
confidently in calculating population estimates.
Background contamination was 0.02 mg/L higher
than the required detection limit.
Precision and accuracy estimates met the
DQOs.
• A relative interlaboratory bias of 2 percent was
calculated on the basis of field audit sample
data, and a relative interlaboratory bias of 5.5
percent was calculated on the basis of
calibration study sample data. Because these
biases are relative determinations in the
evaluation of population estimates, it may be
necessary to assess the data by the subregions
for which each laboratory analyzed samples.
True Color
• The QA data indicate that the true color data for
the routine lake samples are of acceptable
quality.
• Negligible contamination was indicated for this
field-laboratory measurement.
• Precision was acceptable, considering the low
levels of color found in the routine lake samples.
• There were no applicable accuracy
measurements for true color.
Turbidity
• Turbidity QA data indicated that the lake sample
turbidity data are of acceptable quality. Many
routine lake samples, however, were very low in
turbidity.
• Background contamination was below the
required detection limit.
• Precision was acceptable, considering the low
turbidity observed in most samples. Field audit
samples should not be used to estimate
precision for turbidity; they were filtered in the
audit sample preparation laboratory and,
therefore, received different treatment than did
the routine lake samples.
• Accuracy estimates were not calculated for
turbidity. A means of estimating accuracy should
be developed for use in future surveys.
Overall Operations
• The QA data indicate that the sampling,
analytical, data management, and data analysis
activities were successful. These operational
aspects of the survey resulted in
recommendations for future survey efforts (see
Tables 4 through 7).
• The formal audit of the WLS-I data base (field
data forms through the final data set) reported a
data documentation and consistency rate of
more than 99.5 percent.
• All 1,642 samples (149 batches) were received
and analyzed by the analytical laboratories.
• Analytical differences between samples
collected by helicopter crews and by ground
crews were negligible.
• Ninety percent of the samples collected by
ground crews in wilderness areas were received
17
-------
and processed in the field laboratory within one
day of sampling.
The high quality of data generated in WLS-I
shows that personnel training and analytical
laboratory selection were effective.
ELS-I and the WLS-I pilot survey provided
information useful in the planning of WLS-I
field, analytical, and data management
operations.
Table 4.
Finding
Significant Findings, Conclusions, and Recommendations Concerning Lake Sampling and
Field Data Collection, Western Lake Survey - Phase I
Corrective Action
Effect on WLS-I Data
Conclusion or
Recommendation
Lake sampled twice (once
when stratified, then when
isothermic)
pH indicator strip
measurements proved to
be unreliable
Both samples analyzed
and evaluated during data
verification and data
validation
Data not modified during
data verification and data
validation
Sample from isothermic
lake provides data more
related to goals of WLS-I
pH indicator strip
measurements not used
in population estimates
Hydrolab pH measurement
was difficult to stabilize in
situ (in some lakes with
dilute systems)
Loran-C guidance
system malfunctioned at
many lake sites in the 48
subregion (Wenatchee,
WA)
Helicopter crew safety
training not tailored to
WLS-I sampling
protocols
Two minutes additional
time allotted to stabilize in
situ; any unstable lake
measurements tagged
None; lake data tagged
None
Minimal; closed-system
pH measurements used in
population estimates
Negligible; maps, lake
photo information are
additional checks on
lake identification and
verification
None
Final data set (No. 4)
includes data from
isothermic sampling in
population estimates
Future data users must be
alerted that this portion of
data base is unreliable;
relationship of pH
measurement method
requires further
investigation; do not use
pH indicator strip method
in future surveys
Method modification noted
for use in future surveys
Continue use of Loran-C
system
Emphasize "on-board"
field training to
complement classroom
presentations
Note: See glossary for explanation of abbreviations.
18
-------
Table 5.
Finding
Significant Findings, Conclusions, and Recommendations Concerning Field Laboratory
Activities, Western Lake Survey • Phase I
Conclusion or
Corrective Action Effect on WLS-I Data Recommendation
One presampling "practice
run" performed in field
laboratories at
Wenatchee, WA, and
Bozeman, MT (two runs
were recommended)
Laminar-flow hood not
heated; reagents, filtration
apparatuses, and the
hands of the analysts
subject to external cold air
temperatures
Field duplicate pair
samples (collected by
ground crews) that arrived
late were not inserted
randomly in batches
One batch (14 samples)
lost in transit to analytical
laboratory for 3 days
Occasionally samples
arrived at field laboratory
with ice in syringes and
Cubitainers
Container leakage in extr.
Al aliquot bottle because
of inadequate sealing rings
Field laboratory pH
(closed system)
measurement took a long
time to stabilize for some
samples of low ionic
strength; one day's
analyses began to overlap
with the next
Proposed pH and DIG
QCCS solution (prepared
by equilibrating 300-ppm
CO2 in air) was unstable
None
None
Portable heaters used in
field laboratories when
cold temperatures
warranted their use
None; it was more
desirable to process the
samples on date of receipt
than to hold them for next
sample-processing day
Samples located and
analyzed as quickly as
possible
Field sampling crews
notified to moderate
number of frozen
refrigerant gel-packs
used in sample transport;
data qualified (tagged) for
use during data
verification and validation
Removed rings; oriented
aliquots in shipping
container to minimize
leakage; tagged data
where appropriate
Etch pH electrode with
50% NaOH (Knapp et al.,
1987)
None; the procedure was
cancelled for logistical and
technical reasons (e.g.,
atmospheric pressure,
laboratory temperature);
time needed would have
hindered other required
analyses
None
None apparent
Negligible; samples
arrived at 10°C; most
sample measurements
were performed within
holding times
Unknown; approximately
2% of WLS-I lakes
involved
Tagged data inspected; no
effect on data detected
None; relieved overloaded
pH analysis schedule
None, although the trial
was time-consuming
One practice run may be
sufficient for experienced
crews
Field laboratories require
insulatory modifications if
used in cold environments
(outside air temperature
Continue to strive for
random allocation of QA
samples to each batch
None; overall (>99%)
sample shipment protocols
met in WLS-I
Investigate correlation
between lake water
sample temperature,
transport time, and
number of gel-packs
used per shipping
container
Investigate use of different
aliquot containers
Continue the etching
practice in future surveys
when applicable
If future surveys sample
many circumneutral
waters of low ionic
strength, this QCCS
method needs further
investigation and
development (see
Appendix M)
(continued)
19
-------
Table 5 (Continued)
Finding
Corrective Action
Effect on WLS-I Data
Conclusion or
Recommendation
Field duplicate pair
samples yielding turbidity
values < 2.0 NTU did not
meet 10% precision DQO;
protocol required
reanalysis of both samples
For first seven batches
processed at Missoula,
MT, field base (subregion
4C) technician used only
one pipet tip per batch for
MIBK transfer into aliquot
bottle
Single routine sample
arrived at field laboratory
without QA samples
One field duplicate sample
and one field blank arrived
at field laboratory with
obvious contamination
(high in sediment)
Field laboratory pH
measurement not in
agreement (within 0.5 pH
unit) with pH indicator strip
measurement. Field
laboratory protocol was to
reanalyze all pH readings
not in agreement
If sample < 1.0 NTU,
reanalysis not required
Correct procedure used
for all subsequent batches;
confirmation made that
other subregions
performed protocol
correctly.
Trailer blank and audit
sample added to single-
sample batch
Laboratory supervisor
discarded the duplicate
and blank samples,
processed routine sample
Part way through field
effort, protocol changed to
reanalyze only one sample
at the low end, the middle,
and the upper end of the
range of samples in the
batch
None
None; the seven affected
batches were checked,
and no contamination
carryover was found
None; determined better to
process single sample on
day sampled than to wait
for more samples the next
day
Probably none; all data
generated from the
contaminated samples
would have been deleted
from statistical analyses
None; pH indicator strip
values considered
unreliable
Continue practice; modify
DQO to consider precision
at low levels
Reinforces need for close
inspection at on-site
evaluations
None
Do not discard samples
until QA manager has
approved, even if obvious
contamination
QA programs should
remain flexible to deal with
deviations
Note: See glossary for explanation of abbreviations.
20
-------
Table 6. Significant Findings, Conclusions, and Recommendations Concerning Analytical Laboratory
Activities, Western Lake Survey - Phase I
Finding
Corrective Action
Effect on WLS-I Data
Conclusion or
Recommendation
Contingency plan needed
to accommodate possible
emergency shutdown of
an analytical laboratory
during sample analysis
("acts of God")
Dilute nature of many
WLS-I lake samples
made it difficult for
analytical laboratory to
meet intralaboratory
precision criteria (DQO)
for laboratory duplicates;
SOW stated that if criteria
not met, additional
duplicate sample analysis
was required
DIC and pH QCCS
unstable
A non-standard NSWS
aliquot (No. 1) bottle was
used for one sample
(blank); aliquot showed
gross contamination
For total Al, matrix spike
(% recovery) criteria not
met for 3 consecutive
samples in one batch
Negative bias for NH4 +
determination indicated
from field blank analyses
in Laboratory I (51 % of all
field blanks considered
excessively negative)
Negative bias for Mn
determination indicated
from field blank analyses
in Laboratory II (25% of
all field blanks considered
excessively negative)
Sporadic negative bias for
SiC>2 determination
indicated from field blank
analyses in Laboratory I
(9% of all field blanks
considered excessively
negative)
None needed
None
Modified requirement: if
precision criteria not met
and no samples in batch
had analyte concentration
at least 10 times required
detection limit, further
duplicate analyses not
required
Negligible; duplicate
precision statistical
analysis (precision
estimates) uses the
quantitation limit to
eliminate low-level
duplicate pairs near the
detection limit from
statistical evaluation
Have back-up laboratory
available when planning
any major analytical effort
New DQOs for precision
are necessary to account
for the fact that precision
depends on analyte
concentration (see
Figure 8)
See discussion in Table 5 See discussion in Table 5 See discussion in Table 5
QA staff investigated, but
source of bottle could not
be identified. Data flagged
to be deleted from any
statistical analyses;
analytical laboratory
manager noted deviant
bottle in data package
cover letter to QA staff
Standard additions
performed, data tagged,
and QA staff notified as
required by SOW
Problem investigated; raw
data inspected but cause
not isolated by QA staff or
analytical laboratory
manager; affected
samples were flagged
Problem investigated; raw
data inspected but cause
not isolated by QA staff or
analytical laboratory
manager; affected
samples were flagged.
None; affected samples
were flagged
None; data discarded; this
is the only case (1 in
15,000 aliquot bottles) in
which a non-standard
bottle was used
None; reliable results
obtained;matrix
interference negligible in
WLS-I analytical results
(see Section 8)
Probably negligible; NH41
concentrations were
extremely low for most
WLS-I lakes sampled
Probably negligible; Mn
concentrations were
extremely low for most
WLS-I lakes sampled
Probably negligible
Delete data generated
when nonstandard
protocols yield
questionable data
None; proper protocols
followed by analytical
laboratory, and
documentation provided to
QA staff
None
None
None
(continued)
21
-------
Table 6. (Continued)
Finding
Corrective Action
Effect on WLS-I Data
Conclusion or
Recommendation
Software problem in the
calculation of ANC and
BNC was discovered by
manager of Laboratory II
during survey operations
Improper formula used by
Laboratory II to calculate
laboratory duplicate
precision (%RSD) results
(1,751 duplicate pairs)
In calculating % recovery
for matrix spikes,
Laboratory II converted all
negative sample results to
0. (240 spikes affected)
The mean concentrations
of duplicate pair analyses
(Laboratory I) were being
reported as the routine
sample value (discovered
during on-site evaluation)
Calibration blanks not
analyzed as specified for
Ca, Mg, K, Na, Fe, Mn
instrument calibration;
instrument "auto-zeroed"
with these QC samples; all
calibration blanks reported
as 0.00 mg/L (for 91
batches from
Laboratory II)
All open-system initial
pH (pH, acidity; pH,
alkalinity) values reported
by Laboratory I were from
Gran analysis calculation
rather than that measured
from pH meter
Software problem
corrected; ANC and BNC
values in 46 batches
(approximately 1,000
analyses) were
recalculated and
resubmitted to ORNL
before data were entered
into raw data set
Data corrected in verified
data set
% recoveries recalculated
during data verification
Practice discontinued;
affected data corrected to
meet protocol reporting
requirements
None; problem discovered
during statistical analysis
after data verification
None; problem not
detected before final data
set generated
None; correct values in
data base
None; the intralaboratory
precision goals were met
after %RSD values were
recalculated
Minimal; negligible matrix
interference detected in
WLS-I sample data; (see
Section 8)
None; data corrected
Negligible or none for
population estimates;
detection limit QCCS
results indicate low end of
calibration curve (linear
dynamic range) adequate;
affected calculation of
laboratory precision
statistics (i.e., quantitation
limit)
No impact on population
estimates because field
laboratory (closed) pH
measurements used.
Reported (open-system)
values are consistently
about 0.05 pH unit lower
than measured values
Test software for
calculating ANC and BNC
before sample analyses
Specify the %RSD
formula clearly in SOW to
avoid misinterpretation
If matrix spike analyses
are used in future surveys,
clearly state calculation
procedure in statement of
work
Modify AQUARIUS
program to detect
misreporting; modify SOW
to avoid misinterpretation;
continue on-site
evaluations
Modify SOW to minimize
misinterpretation; modify
verification process to
detect problem
immediately; emphasize
check during on-site
evaluations
Specify reporting
procedure in SOW to
minimize misinterpretation;
modify verification
procedure and
AQUARIUS programs to
detect misreporting
Note: See glossary for explanation of abbreviations.
22
-------
Table 7. Significant Findings, Conclusions, and Recommendations Concerning Data Management and
Data Verification Activities, Western Lake Survey - Phase I
Finding
Corrective Action
Effect on WLS-I Data
Conclusion or
Recommendation
Biological growth
discovered in the bulk
sample of audit lot FN4
before audit sample
aliquot prepara-
tion(Appendix C) began
The ability to confidently
estimate accuracy with
synthetic audit samples is
in question because "true
value" not known
Decision not to use
calibration study duplicate
and triplicate samples as a
QA tool
Decision not to designate
separate flags (data
qualifiers) for samples
collected by helicopter
crews and by ground
crews
Turbidity precision
estimated from field audit
sample data misleading
(Appendix E)
AQUARIUS field duplicate
precision program did not
flag pairs when only one
sample > 10 times
required detection limit (50
pairs had poor precision)
Of 149 batches in
WLS-I, 4 did not contain
either a field blank or a
trailer blank sample
One field blank (of 236)
sample concentration for
NO3" was 11.284 mg/L
Entire lot volume refiltered
prior to use; integrity of lot
maintained; use of HgCI2
to stop biological growth
was rejected in favor of
filtration
None
Calibration study samples
can only be compared to
samples from same lake
to check for outlier data
None
If audit samples are used
to determine precision,
they must not be filtered
first
All affected pairs
inspected manually;
confirmation of sample
concentrations performed
by analytical laboratory
None; the batches were
inspected for all other QA
and QC results
Analytical laboratory
reanalyzed sample and
obtained similar results;
data flagged
None; mean analyte
concentrations measured
after refiltering did not
differ significantly from
concentrations measured
before biological growth
was detected
Accuracy calculation and
results must be qualified
Keeping calibration study
data separate from QA
data allowed QA staff to
assess calibration study
results more efficiently
None
Precision estimates should
be discarded
After confirmation,
precision still poor for 33
pairs; data not flagged, but
information provided to
validation staff for use in
calculating population
estimates
Probably none; 236 field
blanks and 22 trailer
blanks analyzed in
WLS-I provided enough
blank data for required
statistical analysis
Sample deleted from
statistical analysis;
assumption made that it
was preserved with
in field laboratory
Continue practice of
apportioning bulk audit lot
into 2-L samples in order
to maximize audit lot
consistency
Utilize NBS certified
standards to estimate
accuracy more confidently
Sampling methods
comparable, data
calibration not necessary
(see Section 9 and
Landers et al., 1987); no
need for future calibration
study
Calibration study data
confirm no need to treat
two sampling methods
differently
Use only field duplicate
precision for turbidity or
use an unfiltered audit
sample lot for this
measurement
AQUARIUS program
modified for future
surveys; new data
verification flag created
Assess NSWS data base
to determine the number
of blank sample analyses
needed to estimate system
components and system
contamination
None
(continued)
23
-------
Table 7. (Continued)
Finding
Corrective Action
Effect on WLS-I Data
Conclusion or
Recommendation
80 SiOg analyses
>14 mg/L were diluted
improperly or
concentration was
miscalculated after
dilution; trend detected by
an audit sample
For 74 samples (in 4
batches), columns were
switched in reporting initial
and air-equilibrated DIC
results (148 analyses);
discovered by comparing
DIC relationship to pH
AQUARIUS program
generated data qualifier
flags for every pH
indicator strip value that
showed poor agreement
with other pH
measurements (e. g.,
closed-system and
open-system pH)
In situ (Hydrolab)
conductance
measurement for all lakes
in the first 22 batches
(1501-1522) in
subregion 4A (California;
Carson City, NV, field
base) did not agree with
the calculated
conductance
60 values recalculated; 20
samples reanalyzed
15 key samples were
reanalyzed; provided
enough proof that the
results originally were
reported incorrectly
Flags deleted from data
base
None; probable Hydrolab
instrument problem;
validation staff notified;
data properly flagged
None; affected sample
concentrations corrected
None; affected values
corrected
Flags considered
extraneous information
because pH indicator strip
values known to be
unreliable; these pH
values not used in
population estimates
No impact on data
analysis because
analytical laboratory
conductance
measurement used in
population estimates
Reinforces need to use
audit samples that have
different concentrations at
levels of interest for
routine lake sample data
Minimal reanalysis can
achieve maximum
efficiency when
relationships between pH
and DIC are evaluated
Advise future data users
about the poor
performance of this pH
method; consider
alternative methods for
future surveys
None
Note: See glossary for explanation of abbreviations.
24
-------
Section 3
Operational Quality Assurance Program
The QA and QC aspects of WLS-I included several
major activities designed to ensure that established
survey protocols were followed for collecting,
preparing, preserving, shipping, and analyzing
samples and for reporting, verifying, and validating
sample data. The QA and QC activities included
selecting contract analytical laboratories; training the
field sampling and field laboratory personnel;
collecting and analyzing a variety of QA and QC data
in order to evaluate data quality statistically, in terms
of the DQOs; maintaining communications with
management, sampling, and analytical personnel; and
conducting on-site field and laboratory evaluations.
The WLS-I QA and QC activities summarized in this
section are described in detail in the QA plan
(Silverstein et al., 1987). Most of the QA and QC
procedures used during WLS-I were previously
applied during ELS-I as described in Drouse et al.
(1986), although some procedures were modified.
Selection of Analytical Laboratories
The objective of analytical laboratory selection was to
award contracts to the fewest number of laboratories
possible, yet to ensure that the laboratories had the
capability and the qualifications to analyze WLS-I
samples. A statement of work (SOW) that defined the
analytical and the QA and QC requirements in a
contractual format was prepared, and bids were
solicited from analytical laboratories. On the basis of
the performance-evaluation sample analyses and the
on-site evaluations, two qualified laboratories,
Environmental Monitoring and Services, Inc. (EMSI),
in Newbury Park, California, and Versar, Inc., in
Springfield, Virginia, were selected from among the
respondents. The laboratory-selection process
paralleled the process EPA uses to select CLP
laboratories.
Training of Sampling and Field
Laboratory Personnel
Before field sampling activities began, Lockheed-
EMSCO field sampling and field laboratory personnel
received an extended training course in Las Vegas,
Nevada. These personnel, most of whom had
received extensive sampling experience during ESL-
I, were sent to the field bases where they provided
EPA and Forest Service personnel with the training
necessary to ensure that field activities were
performed consistently and according to approved
procedures. Time constraints for training limited the
curriculum to protocol and procedural information. All
personnel received hands-on experience with the
activities that they would be expected to perform in
the field; all were given practical and written tests on
their understanding of pertinent methods (Bonoff and
Groeger, 1987). Simulated sampling and field
laboratory activities were conducted at the five field
bases before routine sampling began. The high
quality of the data collected indicates that training was
adequate.
Quality Assurance and Quality Control
Procedures
The WLS-I sampling design was intended to provide
a data set that contained information sufficient for
assessing potential sampling, analytical, and
methodological bias; contamination; and detection and
precision differences related to sampling method.
Specified QA and QC procedures and samples were
used to maintain data quality and to ensure that data
quality could be characterized accurately. Rigid
requirements for instrument calibration ensured that
measurements were accurate and that instrument
malfunctions and drift were readily detected. QA and
QC sample data were compared to the expected
values and ranges established for the survey (Table
2). The results of these comparisons were used
during the survey to correct sampling and analytical
errors and after the survey to evaluate overall data
quality.
Types of Quality Assurance and Quality Control
Samples
The success of the QA program and the evaluation of
overall data quality required the appropriate use of QA
and QC samples to ensure that sampling and
analytical activities were performed according to the
QA plan (Silverstein et al., 1987) and the Statement
of Work. For this report, QA samples are defined as
control samples received by the analyst, who does
not know what the analytical results should be. QC
samples are defined as control samples for which the
analyst knows the theoretical or true analyte
25
-------
concentrations or values. QA samples were used by
the QA staff to evaluate overall method performance
for field sampling, field laboratory, and analytical
laboratory procedures and to estimate overall data
quality. QC samples allowed field sampling personnel,
field laboratory personnel, and analytical laboratory
personnel to identify and correct local problems (e.g.,
get immediate feedback on instrument malfunction or
reagent contamination) before routine samples were
analyzed. Additional QA samples were employed in
the calibration study (Section 9) to ensure the quality
of the ground-access sampling method. A sample
flow diagram (Figure 3) shows the types of QA and
QC samples used and delineates their progression
through the sampling, processing, and analytical
steps of WLS-I. QA and QC sample types are
described below.
Quality Assurance Samples-
QA samples collected at the lake site or introduced at
the field laboratory were analyzed at the field
laboratory and the analytical laboratory. The QA
samples comprised field and trailer blanks, field
duplicates, and field audits.
Field Blank-Field blanks were prepared at the field
laboratory from deionized water that met American
Society for Testing and Materials (ASTM)
specifications for Type I reagent-grade water
(ASTM, 1984). The sampling crew transported the
blank water in Cubitainers to the lake sampling site
and processed the water through a Van Dorn sampler
as if the blank were a routine lake sample. The action
of pouring the blank water through the Van Dorn
sampler could change the CO2 concentration in the
sample, thereby affecting the pH and dissolved
inorganic carbon (DIG) values of the field laboratory
measurements. Consequently, for field blanks, the
field crews did not collect syringe samples for pH and
DIG analysis in the field laboratory. Each helicopter
crew collected one field blank on each operating day;
each ground crew collected two field blanks during
the entire survey.
At the field laboratory, field blank samples were
analyzed only for true color and turbidity. Field blanks
were inserted into the sample batches and were
processed along with the routine lake samples that
were sent to the analytical laboratories. The blank
samples were used initially to identify possible
contamination problems resulting from sampling,
sample handling and transportation, and analytical
processes. Subsequently, they were used to estimate
the background contamination levels, referred to as
system decision and detection limits (see Section 8
and glossary). They were also used to estimate the
quantitation limit, which is helpful in evaluating
precision data (see Section 6 and glossary). For data
interpretation, any routine lake data point above the
expected value for the field blank was considered to
be a positive response for a given variable; any point
at or below the expected value is not reliably
discernible from a field blank.
Trailer Blank Samp/e-Occasionally, the complex
WLS-I sampling design yielded a situation in which a
field blank was not scheduled to be processed at any
lake site for a particular sampling day. In such
instances, a deionized water sample was processed
in the field laboratory trailer as if it were a field blank.
This "trailer blank" was not processed through the
Van Dorn sampler; however, the trailer blank was
substituted for the missing field blank in the sample
batch that was sent to the analytical laboratory.
Field Duplicate Samp/e-A field duplicate was a
second sample collected at the lake site immediately
after the routine sample was collected. The sampling
crew used the same procedure to collect the routine
sample and its duplicate. For each field base, one
helicopter crew collected one field duplicate on each
sampling day. The ground crews did not collect field
duplicates as frequently; each ground crew collected
two field duplicates during the entire survey. Field
duplicates were processed by the field laboratory and
were inserted into the sample batches sent to the
analytical laboratories. The routine sample and its
field duplicate (referred to in this report as a field
duplicate pair) provided the basis for estimating the
cumulative variability of field sampling, field laboratory
processing, and analytical laboratory analyses. This
cumulative variability is referred to in this report as
system precision.
Field Audit Samp/e-Field audit samples were used
(1) to determine bias between the two analytical
laboratories so that measurements made by the two
laboratories could be compared and (2) to indicate
the precision and accuracy of those measurements
through repeated analysis of the same sample type.
Two types of audit samples (field natural and field
synthetic) were, used to establish overall field
laboratory and analytical laboratory performance. Field
natural audit samples consisted of natural lake water
that was passed through a 0.45-pm filter and was
stored at 4°C until use. Field synthetic audit samples
were prepared samples that included analytes of
interest at specified theoretical concentrations. The
concentrations of analytes in the synthetic audit
samples were intended to simulate the concentrations
in natural lake water (see Appendix C).
Natural and synthetic field audit samples, received by
the field laboratory in 2-L aliquots from Radian
Corporation Laboratory in Austin, Texas, were subject
to the same filtering and aliquot preparation
procedures as routine lake samples. These samples
were incorporated into the batches and were shipped
to the analytical laboratories without any identification
that would distinguish them from routine samples.
There were four field natural audit samples
(designated FN3, FN4, FN5, and FN6) and one low-
26
-------
Figure 3. Quality assurance and quality control sample flow, Western Lake Survey - Phase I.
Field Sampling Field Laboratory Analytical Laboratory
Personnel Personnel Personnel
X "1
Trailer Blank
(in lieu of
Field Blank)
^| Field Blank
I
1
I Field Duplicate r
QCCS
Hydrolab pH, Cond.
Natural
Audits
Lake Superior
(FN3)
Big Moose Lake
(FN4)
Bagley Lake
(FN5, FN6)
Field Duplicate
Trailer Duplicate
(Split of a randomly
selected routine
lake sample)
Field Audits
QCCS
pH, DIG, Turbidity
Audit Sample
Preparation Laboratory
Prepared Natural Audits
(FN3, FN4, FN5, FN6)
and Synthetic Audits
(FL11.FL12)
Calibration Blank
DIG
Field Duplicate I
Field Audits
QCCS
Calibration/
Reagent Blank
J
Matrix Spike
(on Field Sample)
Laboratory
Duplicate
(Split of Field
Sample)
concentration field synthetic audit sample (two lots,
designated FL11 and FL12). Three of the natural audit
samples (FN4 from Big Moose Lake in the
Adirondack Mountains of New York and FN5 and FN6
from Bagley Lake in the North Cascade Mountains of
Washington) represented surface waters low in ANC
and in ionic strength, which were expected to be
encountered during the survey. The fourth natural
audit sample (FN3 from Lake Superior) had high acid
neutralizing capacity (ANC) and high ionic strength.
Through daily QA communications with the analytical
laboratories, the QA staff requested preliminary data
on field audit samples. The data were checked for
trends, and data for each audit sample were
compared to data for other samples of the same lot
(within WLS-I and from ELS-I and NSS Phase I
Pilot Survey historical data). For synthetic audits, the
preliminary data were compared to the theoretical
values provided by the preparation laboratory. When
aliquots of FN6 were prepared in the middle of the
field sampling operations, the Radian Corporation's
analytical laboratory analyzed three samples that the
QA staff used as references for comparison with the
preliminary data generated from the two analytical
laboratories. When all of the analytical laboratory data
(149 batches) had been entered into the raw data set,
final audit control limits were generated. The formula
for generating the control limits is given in Drouse et
al. (1986). Appendix H presents the limits and the
numbers of audit samples that did not fall within them.
Values that were outside the limits were considered
27
-------
suspect, and the QA staff requested confirmation.
Outlier values also were detected, which indicated
reporting error, analytical error, or contamination.
Appropriate corrective action then was taken to
resolve issues related to suspect data. If any audit
data remained outside the control limits after all
corrective action had been taken, data qualifier flags
were placed on all the samples in the affected batch.
The audit data, along with the data for other QA and
QC samples, were used then in determining the data
quality of each analytical batch, and the cumulative
results were used to determine overall data quality.
Caution should be taken in assessing data quality in
terms of the numbers of samples that were either
within or outside the control limits. In addition, sample
concentration levels must be assessed before the
control limits are categorized. These distributions can
be quantified in terms of the precision estimates,
expressed as %RSD, derived from pooled data for an
audit sample type (see Section 6). These %RSD
results are referred to as precision estimated from
field audit samples among batches.
Field Sampling and Field Laboratory Quality
Control Samples-
The helicopter crews used quality control check
samples (QCCSs) to calibrate Hydrolab pH,
temperature, and conductance measurements in the
morning, prior to sampling activity. In the evening,
after sampling activity was completed for the day, the
QCCSs were used to check instrumental drift over
time.
The field laboratory staff used three types of QC
samples to ensure that instruments and data
collection were within specified control limits. Before
samples in the batch were analyzed, a calibration
blank was analyzed to check for baseline drift of the
carbon analyzer and to check for contamination.
QCCSs were analyzed for pH, DIG, and turbidity to
check initial instrument calibration and, during sample
analysis, to check instrumental drift. The trailer
duplicate (a subsample or "split" of a lake sample)
was used to check the precision of measurements
made in the field laboratory. The field laboratory
supervisor randomly selected one lake sample per
trailer operating day; this sample was analyzed in
duplicate for pH, DIG, true color, and turbidity.
Analytical Laboratory Quality Control
Samples-
The analytical laboratories used six types of QC
samples to ensure that instrument calibration and
data collection were within control limits: (1)
calibration blanks, (2) reagent blanks, (3) detection
limit QCCSs, (4) low-concentration and high-
concentration QCCSs, (5) matrix spikes, and (6)
laboratory duplicates.
Calibration Blank--Jhe analytical laboratory
analyzed one calibration blank for each analyte in
each batch of samples. The calibration blank, a 0-
mg/L standard, was analyzed after the initial
instrument calibration to check for drift in the
measured signal and to check for potential
contamination during the analytical process.
Reagent Blank-A reagent blank was analyzed for
dissolved SiO2 and total Al because additional
reagents were added to the samples as part of the
digestion step required for analysis of these variables.
The reagent blank sample was composed of all the
reagents (in the same volumes) used in preparing a
lake sample for analysis. The reagent blank was
carried through the routine preparation steps (e.g.,
digestion) prior to analysis.
Detection Limit Quality Control Check Samp/e-A
detection limit QCCS was analyzed for specified
variables to determine and verify the low end of the
linear dynamic range and the values for the samples
near the detection limits. The detection limit QCCS
was analyzed once per batch, prior to analysis of the
lake samples.
Low-Concentration and High-Concentration Quality
Control Check Samp/es-The analytical laboratory
QCCS was a commercially prepared or laboratory-
prepared sample that was made from a stock solution
separate from the one that was used for the
calibration standards. The QCCS was analyzed to
verify calibration at the beginning of sample analysis,
after each specified number of sample analyses, and
after analysis of the final sample in the batch.
Matrix Spike-ft. matrix spike, which was analyzed
with each sample batch, was a check to determine
the effect that the sample matrix had on the analytical
response. The analyst spiked a known concentration
of analyte into a sample of known measured
concentration, then analyzed the spiked sample. Then
the percentage of spiked analyte recovered (percent
recovery) was calculated in order to determine
whether or not there was a significant matrix effect on
the analytical results of the original, unspiked sample.
Laboratory Duplicate--An analytical laboratory
duplicate was analyzed with each batch of samples. A
duplicate analysis was performed on one sample for
each specified variable in each batch to estimate
intralaboratory precision.
Field Sampling Quality Assurance and Quality
Control Procedures
Field sampling QA and QC procedures consisted of
calibrating all instruments before and after specified
sampling activities and of monitoring changes in
instrument performance (Bonoff and Groeger, 1987).
All measurements and QC data were recorded on the
lake data form. Helicopter crews used the Hydrolab to
determine in situ temperature, conductance, and pH.
Calibration and a QC check of the Hydrolab for these
28
-------
three determinations were performed at the field base
or remote site at the beginning of each sampling day.
Ground crews used a field temperature meter
equipped with a thermistor to determine in situ
temperature at the beginning of the sampling day. At
the lake site, before any scheduled samples were
taken, the ground crew checked the temperature
recorded by the thermistor probe against the
temperature recorded by a thermometer certified by
NBS. The ground crews did not measure in situ
conductance. They used indicator strips to measure
pH; therefore, they were not required to perform QC
checks for this variable.
Before WLS-I sampling began, several field
sampling protocol changes were made to
accommodate logistical difference between ELS-I
and WLS-I or to improve data quality in response to
recommendations derived from ELS-I experience.
The changes instituted pertained to the method of
recording lake site location, the model of the Van
Dorn sampler used, and the method that the ground
crews used to measure pH. The most significant
changes are described here; a complete list appears
in Table 3 (Section 1).
Size of the Van Dorn Sampler-
Necessary additions to the ELS-I field equipment,
including Van Dorn samplers, were ordered before
the WLS-I field season began. The dimensions of
the Van Dorn samplers delivered differed from the
dimensions of the Van Dorn samplers used for ELS-
I. Although each sampler had equal volume (6.2 L),
the new samplers were almost twice as long (about
81 cm) as the ones used during ELS-I (about 43
cm). Reordering and equipping all helicopter and
ground crews with the shorter Van Dorn samplers
would have been time and cost prohibitive, so all
crews were equipped with the longer model. This
action eliminated one possible source of sampling
bias within WLS-I.
Use of the longer Van Dorn sampler for WLS-I
called for minor procedural changes because
shallower lakes (i.e., lakes where a debris-free water
sample could not be obtained 1.5 m below the
surface) had to be sampled at 0.75-m depth rather
than at 0.5 m as in ELS-I. The change was required
so that the stopper mechanism on the longer sampler
would have enough clearance below the water
surface to prevent the introduction of air into the
sample. Consequently, lakes classified as shallow in
ELS-I may have been as much as 0.5 m shallower
than their WLS-I counterparts. Conversely, it is
possible that some lakes in the West that were
classified as too shallow would have been sampled if
the shorter Van Dorn sampler had been used.
Measurement of pH with pH Indicator Strips-
The sampling protocol called for WLS-I ground
crews to take in situ pH measurements with pH
indicator strips (Bonoff and Groeger, 1987). This
method was determined to be the most practical
means for the ground crews to use, although it was
not a standard NSWS protocol. It was selected
because the cost of equipping 60 ground crews with
Hydrolabs or portable meters was prohibitive,
especially when the possibility of damaging the
Hydrolabs in ground transit and the need for back-
up units was considered.
Field Laboratory Quality Assurance and Quality
Control Procedures
Field laboratory personnel processed and preserved
aliquots of samples collected in the field; analyzed
water samples for pH, DIG, turbidity, and true color;
and prepared and shipped sample batches to the
analytical laboratories. Because DIG and pH
measurements can be affected by the loss or gain of
CO2 over time, the closed-system field laboratory
measurements provided QA and QC data that were
helpful for later comparison with air-equilibrated
measurements taken in the analytical laboratories.
Because turbidity and true color are physical
measurements, they could be performed relatively
quickly in the field laboratory.
Specified aliquots were stabilized to inhibit biological
and chemical activity and to prevent changes that
could result from volatility, precipitation, or adsorption.
Field laboratory personnel filtered designated aliquots
of each sample to remove suspended material and
other contaminants that might affect analytical results.
Suspended material was removed to reduce biological
activity and to eliminate surfaces that could adsorb or
release dissolved chemical species. Filtered and
unfiltered samples were processed into aliquots. Acid
was added to some aliquots to minimize loss of
dissolved analytes through precipitation, chemical
reaction, or biological action. Aliquots were stored and
shipped at 4°C to minimize biological activity and, in
the case of extractable Al aliquots, to minimize
volatilization of solvent. Silverstein et al. (1987)
provide detail of aliquot preparation.
Sample Batching and Shipping-
Field laboratory personnel organized the samples into
batches for shipment to the analytical laboratories. A
sample batch consisted of a group of routine lake
samples and related QA samples collected in the
field, processed in the field laboratory on the same
date (within 12 hours of sampling, when possible),
and shipped as a unit to one analytical laboratory on
the following day. Because the WLS-I sampling
operation was more complex than its ELS-I
counterpart, the number of sample types (and
corresponding sample codes) increased from 5 in
ELS-I to 24 in WLS-I (see Table 8). Ideally, each
WLS-I sample batch contained at least one field (or
trailer) blank, one field duplicate, and one field audit
sample. Each routine, blank, duplicate, and audit
29
-------
sample was randomly numbered in the batch. Each
sample could be identified by a unique batch ID and
sample ID, and thereby could be distinguished from
any other sample in the survey. The field laboratory
and the analytical laboratory also analyzed QC
samples with each batch, but these samples were not
associated with any individual lake sample in the
batch.
There were occasional deviations from this standard
batch structure as the result of the sampling design
and of field conditions that altered the sampling
pattern on a given day. Two situations led to the
preparation of batches that contained only one lake
sample each: (1) when, on a given day, helicopter
crews were not sampling and only one ground crew
delivered one routine lake sample (with no field
blanks or duplicates), and (2) when a single
calibration lake sample was shipped to the alternate
laboratory (see Section 9). When one of these
situations occurred, one field audit sample and one
trailer blank sample were processed with the routine
sample, and the three samples were shipped to the
analytical laboratory. In this way, some QA samples
were included in each batch, and the data quality of
the routine sample could be assessed.
All data and shipping forms were reviewed by the field
laboratory coordinator. Copies were sent to the QA
staff at EMSL-LV, where the forms were reviewed
for data completeness and consistency. Copies of the
lake data and batch/QC field data forms were sent to
the data base manager at ORNL, where the forms
were used for data entry.
Analytical Laboratory Quality Assurance and
Quality Control Procedures
Analytical laboratory personnel were responsible for
receiving the samples shipped by overnight courier
service from the field laboratory, inspecting the
samples for damage, logging in the sample batches,
analyzing the samples, and preparing and distributing
data packages on the analyses performed (Hillman et
al., 1986; Kerfoot and Faber, 1987).
After samples were logged in, they were analyzed
according to the analytical and QA and QC
procedures specified in Kerfoot and Faber (1987) and
in the SOW. Each variable (Table 2) had to be
measured within a specified holding time (Table 9).
As a part of the contract requirements, the analytical
laboratories agreed to follow standard laboratory
practices for laboratory cleanliness and for the use
and storage of reagents, solvents, and gases. For
standard guidelines regarding general laboratory
practices, the analytical laboratories were directed to
procedures in the Handbook for Analytical Quality
Control in Water and Wastewater Laboratories (U.S.
EPA, 1979). The analytical laboratories operated
according to a uniform set of internal QC procedures
that served as checks on data consistency (see
Silverstein et al., 1987; Kerfoot and Faber, 1987). The
laboratories also documented method performance.
Data Packages-
For each batch of samples, analytical laboratory
personnel completed a data package that included a
set of NSWS forms containing the following
information (see Drouse et al., 1986; Silverstein et.
al., 1987):
• sample concentration for each variable
• for ANC and BNC, titrant concentrations and
titration data points for each sample
• percent conductance difference calculation for
each sample (optional; this calculation is an
initial check made in the analytical laboratory to
ensure data consistency, but it is also performed
during data verification under the direction of the
EMSL-LV QA manager)
• percent ion balance difference calculation for
each sample (optional; this calculation is an
initial check made in the analytical laboratory to
ensure data consistency, but it is also performed
during data verification under the direction of the
EMSL-LV QA manager)
• ion chromatograph specifications for CI",
NO3', and SO42'
• instrument detection limits for applicable
variables
• sample holding times and date of sample
analysis for each analysis of each sample
• calibration blank, reagent blank, and QCCS
concentrations for each applicable variable
• matrix spike percent recovery calculations for
each applicable variable (once per variable; two
additional calculations if recovery of the first
spike not within criteria)
• internal (laboratory) duplicate precision as
%RSD, or, for pH, absolute difference (one for
each variable analyzed in the batch; one
additional duplicate measurement on a different
sample if the first measurement did not meet
criteria)
• standard additions analysis results, when
applicable (on the basis of unacceptable matrix
spike percent recovery results)
30
-------
Table 8. Types and Numbers of Samples Analyzed, Western Lake Survey - Phase I
Sample
Code
Sample Type
Number of Samples
Analyzed in the
Analytical Laboratories
Routine Samples
RH routine sample (helicopter)
RG routine sample (ground)
RH2 routine sample (helicopter), lake sampled a second time
RG2 routine sample (ground), lake sampled a second time
Duplicate Samples
DH duplicate sample (helicopter)
DG duplicate sample (ground)
DH2 duplicate of an RH2 sample
DG2 duplicate of an RG2 sample
TD trailer duplicate
Blank Samples
BH field blank sample (helicopter)
BG field blank sample (ground)
BH2 field blank associated with an RH2 sample
BG2 field blank associated with an RG2 sample
TB trailer blank
Audit Samples
FN3 field natural, lot 3, Lake Superior
FN4 field natural, lot 4, Big Moose Lake, New York
FN5 field natural, lot 5, Bagley Lake, Washington (1st sampling)
FN6 field natural, lot 6, Bagley Lake, Washington (2nd sampling)
FL11 field synthetic, lot 11
FL12 field synthetic, lot 12
Calibration Lake Samples
RHC routine calibration sample (helicopter)
RGC routine calibration sample (ground)
DHC duplicate calibration sample (helicopter)
DGC duplicate calibration sample (ground)
THC triplicate calibration sample (helicopter)
RHCW RHC sample withheld for holding-time study
DHCW DHC sample withheld for holding-time study
THCW THC sample withheld for holding-time study
BHC field blank collected at calibration lake (helicopter)
BGC field blank collected at calibration lake (ground)
Miscellaneous
SG special sample (ground) - not sampled according to NSWS protocols
TOTAL
395
317
5
4
88
128
1
1
118
116
1
1
22
38
20
68
37
21
26
32
45
29
38
29
13
16
16
10
6
1,642
a Not analyzed in analytical laboratory.
b This sample was deleted from the data base.
The data package included a cover letter from the
analytical laboratory manager to the QA manager.
The letter specified the batch ID number and the
number of samples analyzed, identified all problems
associated with the analyses, described all deviations
from protocol, and contained other information that
the laboratory manager considered pertinent to a
particular sample or to the entire batch. Copies of the
completed data package were sent to the QA staff for
initial review and to ORNL for entry into the raw data
set (see Section 4).
On the basis of the analytical results reported for all
QA and QC samples, the QA staff, with the approval
of the QA manager, could direct the analytical
laboratory to confirm reported values or to reanalyze
selected samples or sample batches. A tracking form
for data confirmation and sample reanalysis requests
(NSWS Form 26, Appendix A) was developed and
implemented for WLS-I. This provided a standard
documentation format for data transfer between the
QA staff and the analytical laboratories.
31
-------
Table 9. Maximum Holding Times for Samples,
Western Lake Survey - Phase I
Holding Time3
7 days
14 days
28 days
Variable
NOs", air-equilibrated pH, extractable Al
ANC, BNC, conductance, DIG, DOC
Total P, NH4*, CI", SO4a', total
28 days6
dissolved F", SiO2
Ca, Fe, K, Mg, Mn, Na, total Al
a Holding time commenced on the day that the field laboratory
processed the sample.
b Although holding time has been established at 6 months,
samples had to be analyzed within 28 days to conform with
WLS-I data reporting restrictions.
Communications
The QA staff communicated regularly with the
logistics staff, field and analytical laboratory
personnel, data base manager, and EPA management
team throughout the survey to confirm progress,
resolve protocol problems, and modify procedures.
During the sampling and analytical phases, the
Lockheed- EMSCO QA staff made daily calls to the
field bases and to the analytical laboratories (1) to
ensure that QA and QC guidelines were being
followed, (2) to ensure that samples were being
processed and analyzed properly, (3) to obtain
current sample data and QA and QC data, and (4) to
discuss sampling, processing, and analysis issues so
that problems could be resolved quickly and
efficiently, before they affected data quality or
interfered with the completion of the survey.
Throughout the data verification process, QA and
analytical laboratory personnel communicated as
necessary to confirm reported values, and to make
sample reanalysis requests, and to receive results for
reanalyzed samples. All communications were logged
on appropriate field communications forms and in
bound notebooks.
On-Site Evaluations
On-site evaluations of field sampling activities, field
laboratories, remote sites, and analytical laboratories
were conducted during WLS-I to ensure that
sampling and analytical activities were being
performed according to survey protocol. The results
of these evaluations were documented in site-
evaluation reports prepared by the QA staff, and the
reports were submitted to the QA manager at
EMSL-LV. Significant results of the on-site
evaluations are discussed in Section 5, and overall
results are summarized in Tables 4 through 6
(Section 2).
32
-------
Sect/on 4
Data Base Quality Assurance
Data Management System
The data base management system (DBMS)
incorporates the results from data collection,
evaluation, verification, and validation. By means of
the DBMS, data generated during WLS-I and other
NSWS surveys can be assembled, stored, and
edited. The DBMS also provides basic reports of the
survey results, performs certain statistical analyses,
and provides data security. A detailed description of
the system is given in Kanciruk (1986). The
relationship of data base management to other survey
activities is shown in Figure 4.
The WLS-I data base comprises four major data
sets as summarized below. See Kanciruk et al. (1987)
and Silverstein et al. (1987) for further discussion.
Raw Data Set (Data Set 1)
The raw data set includes all analytical results and
data qualifiers (Silverstein et al., 1987). Data entry
operators at ORNL employed the Statistical Analysis
System (SAS; SAS Institute, 1982) to enter the field
data from the lake data and batch forms and the
analytical laboratory data from the analytical data
forms (see Appendix A in Drouse et al., 1986) into the
raw data set. All data were entered into two separate
data sets by two different operators. A custom
program was developed to compare the two data sets
and to identify inconsistencies. Copies of the field
forms and analytical data packages were sent to the
EMSL-LV QA staff for concurrent data analysis and
as confirmation that all forms were received by
ORNL
Field data errors identified through daily
communication between QA and field personnel were
corrected immediately. If the data in question had not
been entered by ORNL, the changes were included in
the raw data set; otherwise, the data changes were
included in subsequent data sets. Documentation
accompanied each instruction to make changes in the
raw data set.
Verified Data Set (Data Set 2)
Because the numerical, tag, and flag changes were
never applied to the raw data set, a changed data set
(the verified data set) was generated. Through
magnetic tape transfer to EPA's IBM 3081 computer
at the National Computer Center (NCC) in Research
Triangle Park, North Carolina, the raw data set was
made available to the EMSL-LV QA group for
review. To produce the verified data set, the raw data
were processed by the Automated Quality Assurance
Review, Interactive Users System (AQUARIUS),
which is an on-line QA system developed by the
EMSL-LV QA staff. AQUARIUS generated "tuples"
that directed the flagging of problem data in the raw
data set. Tuples are generated by an exception
program (a computer program in AQUARIUS that
indicates data anomalies) or that are manually created
by an auditor. In order to generate the verified data
set, the QA computer support staff applied tuples (as
SAS observations) generated by the EMSL-LV QA
staff to a copy of the raw data set. The data were
sent to ORNL via magnetic tapes to be checked for
anomalies.
AQUARIUS also generated reports helpful in
evaluating intralaboratory biases and field and
analytical interlaboratory biases, as well as reports on
discrepancies in blanks, duplicates, audits, and other
QA and QC samples. The result was a verified data
set in which each of the 1,642 samples was
inspected carefully and any suspicious value or
observation was qualified with appropriate flags.
AQUARIUS data qualifier flags and their definitions
are given in Silverstein et al. (1987) and in Kanciruk
etal. (1987).
Validated Data Set (Data Set 3)
The validation process increased the overall integrity
of the data base by evaluating all data for internal and
regional consistency and by using data provided by
QA and QC information to assess possible analytical
inconsistencies. The validation process began in
tandem with the verification process. When a
computerized version of the verified data set was
provided by ORNL to the ERL-C staff through NCC,
the validation review process could be completed.
After undergoing this review process, the data were
transferred to the validated data base. The validation
process is discussed further in Landers et al. (1987)
and in Silverstein et al. (1987).
33
-------
Figure 4. Data base management, Western Lake Survey - Phase I.
Site
Selection
)f Field Base Sites/ \ / A . .. . \
( Field J ( Anneal ]
V Laboratories/ VLaboratones/
1
>v s
Raw
Data
Set
1 *
_^^
s. X
Verified
Data
Set
^ '
C~^
Validated
Data
Set
-x-
f ^^
Final
Data
Set
|
^ / Data Entrv by / i 1
^ / ORNL / Verification by
/ / fc EMSL LV QA 1
fc_ D in
Preliminary
1 ^ Validation by
/ / ERL-C
* / /
Maps ERL-C
and by EMSL-LV QA
/Data Editing, /
Questionable /
Data /
^.
Treatment
of Data
*
w .-,
Maps,
Statistics
* Data Tracking System
Final Data Set (Data Set 4)
Linthurst et at. (1986) noted that the calculation of
population estimates is difficult if the data set contains
missing values. To minimize these difficulties, a final
data set was prepared for use in calculating
population estimates. This data set was modified by
averaging the field duplicate pair values that were
within desired precision limits. Negative
concentrations that were reported by the analytical
laboratory as resulting from instrumental drift (i.e., a
negative y-intercept on the calibration curve) were
converted to zero (except for ANC and BNC), and
analytical values that the validation review had
identified as questionable were replaced. The
substituted values were determined according to
procedures described in Landers et al. (1987) and in
Silverstein et al. (1987).
34
-------
Data Review and Verification
The objectives of the data verification process were
to identify, correct, or flag raw data of questionable or
unacceptable quality and to identify data that might
need to be eliminated during or after validation. The
WLS-I verification process was modified
considerably on the basis of ELS-I experience and
as a result of the need to accommodate the large
number of sample types (Table 8) added to the
WLS-I sampling design. Many WLS-I QA personnel
had prior NSWS experience in QA, field sampling,
and field laboratory operations, as well as previous
experience in wet chemical and instrumental analysis
of water samples in the analytical laboratories. This
background expedited the modification of ELS-I
field, laboratory, and data verification protocols; the
identification and correction of sample collection,
processing, and analytical problems; and the
identification of data trends.
Preliminary sample data were obtained verbally, by
computer, or by telefacsimile, depending on the
laboratory. The preliminary data were evaluated by
comparing the QA sample data against the
acceptance criteria. Responsible parties were notified
of problems, and all interactions were recorded in
bound notebooks. If necessary, memoranda were
sent as documentation.
Data verification began when the field and analytical
laboratory data were received by the EMSL-LV QA
staff. All data were evaluated on the basis of the QA
and QC information and knowledge of lake water
chemistry. AQUARIUS computer programs automated
much of the verification process. For each analytical
data package (representing one batch of samples),
the QA audit team performed a sample-by-sample
evaluation. The audit team reviewed comments and
questions associated with the batch; performed QA
checks for data consistency and reasonableness;
reviewed QA sample data; obtained confirmation,
correction, and reanalysis data from the analytical
laboratories; and provided a verified data set to
ORNL. For each batch, the audit team prepared a
summary of the reporting errors found and of the data
confirmation and sample reanalyses required.
Review of Field Data Forms
When the lake data and batch/QC field data forms
arrived from the field, the auditor reviewed the forms
for data inconsistencies and for adherence to
procedures. Data anomalies were reported to the field
laboratory coordinator for corrective action, and when
possible, data reporting errors were corrected before
the data were entered into the raw data set. Changes
made to the raw data set were sent to ORNL by
telefacsimile for immediate action. A detailed
discussion of the field data review procedure appears
in Silverstein et al. (1987).
Initial Review of Analytical Laboratory Data
Packages
The analytical laboratory submitted a data package to
the EMSL-LV QA staff for each batch of samples.
The QA staff used the NSWS verification worksheet
to review each package for completeness, internal
QC compliance, and appropriate use of data
qualifiers. The verification worksheet was designed to
guide the auditor systematically through the data
verification process by explaining how to flag data,
track data resubmissions and requests for reanalysis
and confirmation, list the steps that lead to
identification of QA exceptions, and summarize
modifications to the raw data set (prepare records of
flag and numeric changes). Written comments
submitted with the data package also were reviewed
to determine their impact on data quality and to
determine any need for follow-up action by the
laboratory. Auditors reported problems to the
analytical laboratory manager for corrective action.
Final Data Verification
While the EMSL-LV QA staff were conducting the
initial review of analytical data packages, the data
were also being entered into the raw data set at
ORNL. ORNL sent a magnetic tape containing the
data to NCC. Through telecommunication, the
EMSL-LV QA staff had access to the raw data set
and could complete the data verification process.
Each sample was verified individually and by
analytical batch. AQUARIUS programs were used to
identify or flag results that were classed as
exceptions, i.e., results that did not meet the
expected QA and QC criteria (Table 10). Additional
data qualifiers were added to a given variable when
the QA samples (field blanks, field duplicates, or field
audit samples) in the same analytical batch did not
meet the acceptance criteria. Data also were qualified
with flags if internal consistency checks (anion-
cation balance, calculated conductance), QC checks,
or holding time requirements were not met. The
protolyte analysis program flagged field laboratory and
analytical laboratory measurements of pH, DIG, ANC,
BNC, and DOC when carbonate equilibria, corrected
for organic protolytes, were not in internal (within-
sample) agreement. A flag was not assigned if the
discrepancy could be explained by the presence of
organic species (as indicated by the protolyte analysis
program) or by an obvious and correctable reporting
error.
In all cases, each flag generated by the AQUARIUS
system was evaluated by the auditor for
reasonableness and consistency before it was
entered into the verified data set. These programs
automated much of the QA review process and
enabled the auditor to concentrate more effort on the
substantive tasks of correcting and flagging
questionable data. These programs also identified
35
-------
Table 10. Exception-Generating Programs
Verification System
Program
within the AQUARIUS Data Review and
Sample (Data) Type
Field Audit Sample Summary
Field/Trailer Blank Summary
Field Duplicate Pair Precision Summary
Instrumental Detection Limit Summary
Holding Time Summary
Conductance Check Calculations
Anion/Cation Balance Calculations
Batch QA/QC Summary
Comparison of Form 1 and Form 2
Comparison of Form 2 and Form 11
Protolyte Analysis
Comparison of Total Aluminum and Extractable
Aluminum
Audit Sample Window Generation
Raw Data Listing
QA/QC Flag Summary
Reagent/Calibration Blanks and QCCS
Calculation of Laboratory Penalties
Matrix Spike Summary
Gran Analysis
Field Natural (FN),
Field Synthetic (FL)
Field Blank (BH, BG), Field Laboratory
Blank (TB)
Routine/Duplicate Pairs (RH/DH, RG/DG)
All Species
All Species
All Species
All Species
All Exceptions
pH and DIG
pH and DIG
DIG, DOC, pH, ANC, and BNC (data
evaluation)
Total Al and Extractable Al
All Species
All Field/Laboratory Data
All Exceptions
All Species (except pH)
All Species
Applicable Species
ANC and BNC
outlier data based on QA and QC sample data. The
outlier data were the basis for requesting confirmation
of data from the analytical laboratories and for
requesting reanalysis of suspicious samples. Values
were confirmed before reanalysis requests were
issued.
The auditor used the output from the AQUARIUS
programs (along with original data and field
notebooks) to complete the NSWS verification report
form (Silverstein et al., 1987).
Modifications to the AQUARIUS System
The QA staff made several changes to the
AQUARIUS data verification programs between
ELS-I and WLS-I. These changes are summarized
in Table 3 (Section 1), and the most significant
changes are discussed below.
Determination of Control Limits for Blank
Samples-
In ELS-I, contamination levels for field blanks were
determined on the basis of previous knowledge of
how field sampling and analytical methodology may
affect water samples. The WLS-I verification
process, however, benefited from the use of historical
field blank data generated during ELS-I. The values
for the 245 ELS-I field blank samples were used to
calculate control limits which, in turn, were used to
check for contamination, to determine the necessity
of data confirmation or reanalysis, and to generate
flags that qualified the data by batch or by sample.
Calculation of WLS-I control limits for blank samples
and a comparison of ELS-I and WLS-I control limits
are given in Appendix B.
Comparison of Extractable Al and Total AI-
The EMSL-LV QA staff developed a computer
program to compare the extractable Al and total Al
values for each sample. By definition, the extractable
Al concentration for a sample could not exceed the
total Al concentration. The program generated a flag
when the value for extractable Al was higher than the
value for total Al by more than 0.010 mg/L (twice the
required detection limit; Table 2).
This qualification was intended to account for
background noise (especially at low concentrations)
and for minor fluctuations in instrument reading and
calibration.
Calculation of Percent Ion Balance
Difference-
Percent ion balance difference (%IBD) is calculated
by
S onions — S cations + ANC
S onions + S cations + ANC + 2[H+]
100
36
-------
where:
2'
S anions = [Cf] + [F'] + [NO3'] + [SO4']
E cations = [Na + ] + [K + ] + [Ca2 + ] + [Mg2 + ] +
[NH4 + ]
ANC = Alkalinity (the ANC value is included in
the calculation to account for the
presence of unmeasured ions such as
organic ions)
[H+] = (10-PH) x 106 neq/L
Note: Brackets indicate concentration of
an ion in microequivalents per liter.
The anion-cation balance limits are different
depending on the total ionic strength (the
denominator in the calculation) of the sample. If the
sum of the ions were calculated to be less than 50
peq/L, the difference could be ±60 percent. If the
sum were between 50 peq/L and 100 neq/L, the
difference could be ±30 percent. If the sum were
greater than 1 00 ueq/L, the difference could be + 1 5
percent. Any routine lake sample or QA sample that
did not fall within the applicable criterion was qualified
with a flag.
The calculation program was modified so that in all
instances where the absolute value of ANC was less
than or equal to 10 peq/L, the value zero was
substituted for ANC in the equation. The equation is
sensitive to slight variations in ANC for samples that
have very low ionic strength.
Calculation of Percent Conductance Balance
Difference-
Although no adjustment was necessary for the
conductance balance difference (%CD) calculation,
the criteria are presented here because of the
importance of this internal sample consistency check
to the QA program. The formula for determining
conductance balance is:
calculated conductance — measured conductance
measured conductance
100
The ions used to calculate conductance are Ca2 +,
Cr, C032-, H + , HC03-, K + , Mg2 + , Na + , NO3',
OH", and SO42'. The limits for the difference
between analytical laboratory measured conductance
and calculated conductance is ± 50 percent if
measured conductance was less than 5 nS/cm, ± 30
percent if measured conductance was between 5
pS/cm and 30 pS/cm, and ± 20 percent if measured
conductance was greater than 30 yS/cm. Any routine
lake sample or QA sample that did not fall within the
applicable criterion was qualified with a flag.
Confirmation and Reanalysis Requests
Completing the verification process often required
communication with the analytical laboratory to obtain
confirmation or correction of reported data and to
request sample reanalyses. The follow-up
communication was time-consuming, particularly
when the type of request made to the laboratory had
not been specified in the original SOW or when the
laboratories concerned were involved in other
analytical activities at the time of WLS-I verification.
Typically, responses to requests for confirmation or
correction of reported data were completed within 2
to 4 weeks. Generally, reanalyses were requested
when at least three different QA/QC samples
generated flags for a particular variable in a particular
batch. Three flags were enough to classify a result as
suspect.
Preparation and Delivery of Verification Tapes
Constructing the verified data set required a
consistent and trackable method for transferring the
change records to ORNL. The method chosen to
accomplish this transfer employed tuples to identify a
change to the data set. Tuples generated by
computer programs and those generated by QA
personnel were stored in separate data files until the
tuple listing was applied to a copy of the raw data set.
At that time, a computer program combined all tuple
areas (flag, tag, or value changes) and applied the
combined tuples to the data set only if the batch ID,
sample ID, variable name, and originally reported
values matched. Tuples that could not be applied to
the data set were reexamined by the QA staff, were
corrected, and were reapplied. The final verified data
set was generated by the EMSL-LV QA computer
support staff. The tape was sent to ORNL where it
was checked for consistency before it was used in
data validation. ORNL also was responsible for
archiving the tape.
At the conclusion of the verification process, a data
base audit was performed by an independent firm that
did not participate in other WLS-I activities. The
audit consisted of reviewing the written verification
records, evaluating for accuracy the results generated
by AQUARIUS and other computer programs,
reviewing the procedures used to substitute for
missing values, and determining the error rates
associated with each aspect of the verification
procedure. The audit identified an error rate of 0.05
percent for data entry at ORNL. In the verified data
set no incorrect value changes were detected, and all
of the value changes were documented (IS&T, Inc.,
1986).
37
-------
Data Validation
Validation is a functional term for describing the
continuing process of defining the quality of the data
so that each step results in increased knowledge of
and presumably confidence in the data. This is
accomplished by reviewing the data for errors; data
known to be erroneous are identified so that correct
data can be substituted, and possible errors are
flagged to alert the user to their questionable status.
The system of data validation used for ELS-I was
also used for WLS-I. In the verification step, the
quality of the analytical chemical data was determined
through a rigorous protocol based on known
principles of chemistry. Not all potential sources of
error, however, can be evaluated in the verification
process. The validation process, then, investigated
errors in the chemical analyses not detected in
verification and provided a review of the quality of
nonchemical variables.
Two aspects of the data validation process were the
identification of outliers and the evaluation of possible
systematic errors in the measurement process. The
methods selected for detecting outliers and
systematic errors stressed visual presentations and
conservative, subjective selection procedures. They
were chosen for their simplicity of implementation and
employed pre-existing software whenever possible.
An audit performed on the validated data set (IS&T,
Inc., 1986) identified an error rate of 0.01 percent for
data values.
Data validation procedures are discussed fully in
Drouse et al. (1986) and in Linthurst et al. (1986).
WLS-I data validation design and results are
discussed in Landers et al. (1987).
38
-------
Section 5
Results and Discussion -
Operational Quality Assurance Program
The results presented in Sections 5 through 8 are
limited to QA and QC observations and data. A
discussion of regional population estimates for WLS-
I appears in Landers et al. (1987).
Field Sampling Activities and Protocols
The field auditors conducted on-site evaluations of
all field bases and of selected remote sites. The
auditors found that all field sampling and laboratory
crews performed their duties professionally and
cooperatively, in spite of the tight schedules required
to complete seasonal sampling activities. In some
cases, severe weather conditions contributed to
logistical problems, but the crews adapted well,
documented problems accurately, and often proposed
effective solutions (Bonoff and Groeger, 1987).
Sampling activities commenced on September 11,
1985, and were completed on November 4, 1985. In
the WLS-I sampling design, 973 lakes originally
were selected for sampling and for use in estimating
populations; 95 of those randomly selected lakes
were eliminated from the design before sampling
began. Forty-two other lakes also were scheduled
for sampling because of special interest, but they
were not part of the random selection process and
they were not used in calculating population
estimates. (Landers et al. [1987] details the process
of selecting the lakes used in the population
estimates and the application of the data derived from
those lakes.) Therefore, at the start of field
operations, 920 lakes were scheduled to be visited.
Sampling crews attempted to sample 912 lakes; they
collected 811 routine lake samples from 757 of the
912 lakes. Data from 719 of the lakes were used in
calculating population estimates. Some lakes were
not sampled because they were frozen, thermally
stratified, or too shallow. Other lakes were not
sampled because access permission could not be
obtained, because weather conditions or hazardous
conditions prevented access, or because the lakes
had dried up since they were mapped. Some lakes
were sampled twice, either by different sampling
crews (Section 9) or by the same crew (see below).
Bonoff and Groeger (1987) describe sampling
activities in detail.
In addition to the changes in field sampling protocol
that were made prior to WLS-I (Table 3 in Section
1), some changes were made in response to
situations that arose during the survey. These
changes are summarized in Table 4 (Section 2); the
most significant change is also discussed here.
During WLS-I sampling, some lakes were found to
be thermally stratified. When a helicopter crew
determined that a lake was stratified, they did not
always sample the lake, but returned at a later date to
sample when there was a better chance that the lake
was isothermal. The field base coordinator
incorporated the second visit into the sampling
schedule. Ground crews were not constrained by this
protocol because of the time and distance involved in
returning to the lake; however, they did return to
these stratified lakes when possible. The lakes visited
twice are categorized in Table 11.
If a lake was sampled twice by the same crew, a "2"
was added to the sample code for samples collected
on the second visit. For example, RH2 is the code for
a routine helicopter sample collected on the second
visit (see Table 8 in Section 3). For cases where a
lake was stratified on the first visit but not stratified on
the second visit, only the analytical results from the
sample collected when the lake was unstratified were
used in estimating populations.
Field Laboratory Activities and Protocols
In general, WLS-I field laboratory operations were
conducted without major difficulties (Bonoff and
Groeger, 1987). Numerous field laboratory protocol
changes and deviations were instituted in response to
WLS-I field situations. These issues are summarized
in Table 5 (Section 2); the most significant issues are
discussed here also.
Filtration Procedure
Because of residual nitrate contamination, separate
filtration apparatuses for the 0.45-nm filters and
filtrators rinsed with nitric acid were used in
processing certain aliquots during ELS-I. (A nitric
39
-------
Table 11. Lakes Visited Twice by Sampling Crews, Western Lake Survey - Phase I
No. of Lakes Sampled on Second Visit
No. of Lakes Sampled on Both Visits Only
Sampling Crew Stratified Other Reasons Stratified Other Reasons
Helicopter
Ground
5
2
0
2
15
2
2
0
Total
17
acid solution [5 to 10%] rinse is standard laboratory
procedure for washing off residual metals [e.g., Ca,
Mg] that may be adsorbed onto filters and onto the
walls of filtration apparatuses.) This segregation of
filtration apparatuses became standard protocol for
the NSS Phase I Pilot and WLS-I.
Receipt of Samples from Sampling Crews
WLS-I protocols established a daily cut-off time for
sample receipt. Samples were not incorporated into
the daily batch later than 2 hours after field laboratory
start-up. Samples that arrived after that time were
refrigerated until the next day and were included in
the next day's batch. Exceptions were allowed when
the following three conditions were met:
• Field communications alerted the laboratory that
samples would be received by a specified time.
• The DIG analyzer had not been turned off
(recalibration of the instrument would take
analyst's time away from normal pH analysis).
• Filtrations and Al extraction were not completed
on the other samples.
The Forest Service manager, field base coordinator,
and field laboratory personnel evaluated each
situation to accommodate delivery-schedule
deviations, usually brought about by adverse weather
conditions.
During WLS-I, ground crews collected 366 routine
samples from 362 lakes (see Table 8). Of the
samples collected at these lakes, 62 percent were
processed at the field laboratory on the day that they
were collected, 28 percent were processed within one
day after sampling, 7 percent were processed within
two days after sampling, and 3 percent were
processed more than two days after sample collection
(see Table 12).
Shipment of Samples
Of the 149 batches analyzed during WLS-I, one
batch (ID 1117; 14 samples) was detained in transit
for three days between the field laboratory and the
analytical laboratory. Upon arrival at the analytical
laboratory, the internal temperature of the shipping
containers was below 10°C, which indicated that the
integrity of the samples was maintained and that the
shipping container insulation was effective.
Subsequent analyses did not exceed specified
analytical laboratory holding times by more than two
days for any analyte. The data for the samples are
flagged appropriately in the data base.
In a second situation, the Carson City field base sent
an unscheduled sample shipment to one analytical
laboratory. This flexibility had been given to the
Carson City field base before routine sampling began.
When the field base's regular analytical laboratory (II),
which was also used regularly by two other field
bases, received too many samples at one time, the
Carson City base shipped one batch (ID 1504) to the
other analytical laboratory (I) for analysis. This
situation illustrates the role that the Communications
Center in Las Vegas played in tracking sample flow to
ensure that the analytical laboratories were not
overloaded with samples on a given day.
Comparison of Lake Site and Field Laboratory pH
Measurements
ELS-I and WLS-I field laboratory protocols included
a comparison of Hydrolab pH (in situ) measurements
and field laboratory pH meter measurements. If the
difference between the two measurements was
greater than 0.5 pH unit, field laboratory pH was to be
remeasured. The majority of Hydrolab versus pH
meter readings, however, were within the ± 0.5 pH
QC requirement.
Initially, the parallel comparison between field pH
(indicator strips) and field laboratory pH (meter) was
conducted. At several field bases, a pattern
developed among the first few batches. Most pH
indicator strip measurements were at least 0.5 pH
unit lower than the laboratory pH meter reading, and
many readings were at least 1.0 pH unit lower. When
40
-------
Table 12. Field Laboratory Holding Times9 for Samples Collected
by Ground Crews, Western Lake Survey - Phase I
Number of Days Elapsed Number of Lakes Percent
Since Sampling Date4' Sampled (to nearest whole)
0
1
2
3
4
5
6
7
8
227
102
25
4
3
2
1
1
1
366
62
28
7
1
1
<1
<1
<1
_ _.!
100
These holding times refer to the time differences between sampling and
sample preservation. They are not related to holding time requirements
for the analytical laboratories.
> An estimated 5 to 10 percent of holding time delays beyond the
sampling date were because the field laboratories did not operate on
some days.
the discrepancies were first noted, field pH indicator
strip readings were cross-checked with indicator
strip readings taken in the field laboratory. Results of
this cross-check confirmed that the ground crews
were reading the pH paper color development
correctly. Subsequently, the protocol for the pH paper
comparison was changed so that reanalysis of only
two to four laboratory measurements per batch was
necessary when indicator strip pH was outside
criteria. In these cases, one measurement was taken
at each extreme of the pH readings. Depending on
the size of the batch and the pH range, a reading also
was taken at reasonable intervals within the range.
Theories about the origin of the problem are being
explored, but are not within the scope of this report.
The data user, however, should note that pH indicator
strips provide a coarse reading at best. The data user
also should be aware that the only way to distinguish
Hydrolab pH measurements from pH indicator strip
measurements, both of which are reported on the
lake data form (entered into the data base as
"PH TOP"), is to check whether the lake was
sampfed by helicopter-access or ground-access.
The closed-system pH readings determined in the
field laboratory provide data most suitable for use in
estimating populations.
Analytical Laboratory Activities and
Protocols
On-site evaluations were conducted at the analytical
laboratories after sample analysis was underway. QA
and QC data that had been provided up to the time of
the visit were reviewed, and the QA and QC issues
were identified and discussed. The on-site
evaluations were a contractual part of the QA
program used to observe laboratory operations and to
check for protocol deviation. The evaluations
permitted the QA and analytical laboratory staff to
discuss concerns about contract interpretation and
questions about sample analysis. Topics of particular
interest to the WLS-I laboratories included
preliminary QA and QC data, lost samples,
contingency plans for unexpected laboratory shut-
down, QA and QC acceptance criteria, and
improvement and documentation of protocol decisions
and procedural changes. The meetings were helpful
in solving problems and clarifying previous telephone
communications.
At Laboratory I, operations were satisfactory and
QCCS control charts were current. To meet analytical
holding time requirements, the laboratory was
operating two shifts, both of which were observed
during the visit. It was noted that reagent bottles
needed to be labeled more carefully in accordance
with good laboratory practices.
The visit to Laboratory II showed that communication
between QA and laboratory staff members was
adequate and that sample receipt was progressing
smoothly. Sample analysis trends indicated that the
laboratory's preliminary data for NH4 + , total P, and
dissolved organic carbon (DOC) were near the
detection limits for routine samples, but
corresponding synthetic audit sample data did not
indicate any analytical problems.
41
-------
The analytical laboratories implemented several
protocol changes during WLS-I. Some of these
changes, as well as a contingency plan for sample
analysis in the event a laboratory became inoperable,
applied to both laboratories. In addition, each
laboratory had specific questions and analytical
problems that required the direction of the
management team and QA staff. These issues are
summarized in Table 6 (Section 2). The most
significant issues are also discussed here.
Incorrect Reporting of pH Values
After data verification, a consistent data reporting
problem was identified concerning the initial pH
values for all samples in the 58 batches analyzed by
Laboratory I. Instead of reporting the measured pH
value as required, the laboratory reported the
calculated pH value used in the Gran analysis. In
most cases, the calculated pH was approximately
0.05 pH unit less than the measured pH. The
calculated pH values appear in the data base.
However, population estimates are not affected
because only the field laboratory (closed-system) pH
measurements are used in their preparation.
Modifications to the QA and QC procedures, the
SOW, and the data verification process are under
consideration to ensure that such misreporting
problems are not encountered in future surveys.
Incorrect Use of Calibration Blanks
At Laboratory II, the calibration blanks required for the
atomic absorption analysis for Ca, Fe, K, Mg, Mn, and
Na were not used properly or were not reported as
protocol required. The proper procedure for use of
the calibration blank was (1) to analyze calibration
standards, (2) to fit the calibration curve and the
linear dynamic range to those standards, and (3) to
analyze the calibration blank and detection limit
QCCS. Instead, after fitting the calibration curve, the
laboratory analyst analyzed the calibration blank, then
"auto zeroed" the instrument before the detection
limit QCCS was analyzed. This fact was not revealed
until initial data verification was complete and
statistical analyses began. Fortunately, the detection
limit QCCSs provided an additional check of the
extremely low end on the linear dynamic range of the
calibration curve. Because these QCCSs were
consistently within the QA limits, the incorrect use of
the calibration blank sample appeared to have a
negligible effect on data reporting for these metals.
Laboratory II did use the calibration blank sample
correctly for all other applicable analytes. The NSWS
verification process has been modified to ensure that
similar situations do not occur in future surveys.
Suspect S/7/ca Values
From a trend indicated in field natural audit lot FN5
during data verification for Laboratory II, all the SiOa
values that represented concentrations greater than
14 mg/L (i.e., about 80 samples) were suspect.
Approximately 75 percent of these values represented
reporting errors that resulted from incorrect dilution-
factor calculations or data-reporting errors. The
remaining suspect samples had to be reanalyzed
because the proper dilution and digestion procedures
were not implemented when these samples originally
were analyzed. This case shows the need for
different audit lots with varying concentrations that
cover the range of the routine lake sample
concentrations.
Data Verification Activities
The QA staff reviewed field data forms and analytical
data packages to identify and correct data reporting
errors, to evaluate data trends, and to identify which
samples needed reanalysis. These reviews resulted in
changes to the raw data set and created the verified
data set. The types and quantities of changes made
to create the data sets are given in Table 13. The
results of each data verification step are discussed
below and are summarized in Table 7 (Section 2).
The QA staff also identified several necessary
modifications to the flag-assignment process. These
changes also are presented in Table 7.
Review of Field Data Forms
The first step in the confirmation and reanalysis
process was to check the lake data forms, batch/QC
field data forms, and, for the ground-access
samples, the chain-of-custody forms. The QA staff
identified more than 1,000 possible field data
problems involving analytical values, QA and QC
sample values, data tags, and preparation of data
forms. The field data review resulted in 770 changes
that affected approximately 1.5 percent of the
reported data. Because the field personnel responded
quickly (usually within one day) to requests for
information concerning the field data forms, the
changes were made on the forms before the data
were entered into the raw data set. Discrepancies
that were found in the field data after the raw data set
was completed were resolved in the verified data set.
Correction of Data
Review of the sample data packages submitted by
the analytical laboratories took much longer than the
review of the field data. The analytical laboratory data
were more extensive and more complex, and the
values for QA and QC samples had to be assessed
for each batch. Data for special-studies samples
(see Section 9) also were checked for data
consistency and outlying values. The analytical
laboratories typically took 2 to 4 weeks to confirm
questionable data. Therefore, these data could not be
corrected before the raw data set was created. The
corrections were made in the verified data set.
More than 75 percent of the approximately 1,900
requests for data confirmation were tracked on the
42
-------
TTable 13. Value Changes Incorporated into the Raw and Verified Data Sets, Western Lake Survey
Phase I
Data Source (Data Set)
Number of Changes
Made to Data Set
Percent (approx.) of
Changes to Total
Number of Values in
Data Set
Comments
Field data forms 770
(raw data set)
Lake sample data from analytical
laboratory measurements
confirmed data 4-51
reanalyzed data 141
(verified data set)
Analytical laboratory quality 3,168
control data (verified data
set)
EMSL-LV split sample data 124
(split sample data set)
All verified data (verified data set 4,654
and EMSL-LV split
sample data set)
1.5
1.1
0.4
5.5
6.8
2.7
Changes made to data tags
before entry into raw data
set
Changes made because of
data reporting errors
(confirmed data)3
2,229 changes on laboratory
duplicates and matrix
spike from improper data
reporting practices from
Laboratory II; 541 data tag
changes
Changes made because of
reporting errors
Approximate number of
changes made to data
and data tags (not
number of data flags)
aa This does not include the approximately 1,000 ANC and BNC recalculations that Laboratory II performed before
the values were entered into the raw data set.
data confirmation/reanalysis request form (Appendix
A). The rest were transmitted by telephone or by
letter.
In some cases, the QA staff requested that the
analytical laboratories submit raw instrumental data
(e.g., instrument readouts, strip charts) with changes
in analytical values. As a result of analytical laboratory
data verification, 451 sample values (about 1.1
percent of the analyses) were changed. These
changes were made to correct transcription, decimal
place, and dilution-factor errors, and to include
previously omitted data. Approximately 5.5 percent of
the QC sample data (3,168 of the 57,000 values) had
to be corrected. Most of the changes (70%) to the
QC sample data were required because of consistent
errors in the method of calculation used to derive
some matrix spike and laboratory duplicate data (see
Table 6 in Section 2).
Requests for Reanalysis
The purpose of reanalysis was to improve the quality
of suspect data or to substantiate the value from the
first analysis. If it was not evident that better
information on data quality could be obtained from the
reanalysis, the request was not made. Before any
reanalyses were requested, all suspect values were
confirmed by analytical laboratory personnel. If, after
confirmation, the values were still suspect, reanalyses
were requested. However, the analytical laboratories
were not asked to reanalyze samples unless the
verification procedure generated three data qualifier
flags for either a single variable within a batch or an
individual sample. Usually, one request was
generated for all reanalyses that pertained to a
particular batch so that all reanalyses for the batch
could be performed at the same time, but there were
exceptions to this policy. The laboratories were not
normally asked to reanalyze when all the flags on the
batch were related to a single, outlying sample value;
nor were reanalyses requested (for analytes that were
subject to high variability over time) if analytical
laboratory holding times had been exceeded by
weeks or months by the time the need for reanalysis
was determined.
During WLS-I, 237 reanalyses were requested and
211 were performed. The 26 reanalyses requested
but not performed were flagged to indicate that they
were highly suspect and that they should not be used
in statistical analyses. Of the reanalyses requested,
40 percent were for SiO2 and 15 percent were for
NO3~. SO42", CI", BNC, total P, conductance, and
air-equilibrated pH together were responsible for 20
percent of the requests. The remaining 25 percent of
the reanalysis requests were related to suspect DIG
values that had been identified by the use of the
protolyte program (see Table 10 in Section 4). The
analytical laboratories performed all reanalyses; 141
of the reanalyzed samples (0.4 percent of all sample
data) were used in place of the original values.
43
-------
Seventy reanalyses values were not substituted for the quality of the data. The new values, however,
the original values because they did not decrease the were relayed to the validation staff for possible future
number of flags; therefore, they would not increase use.
44
-------
Section 6
Results and Discussion - Precision
Introduction
During WLS-I, 757 lakes were sampled (5 of these
later were determined to have been missampled or
otherwise were not useful in the survey design);
1,642 lake water and QA samples in 149 batches
were analyzed in the analytical laboratories. Figure 5
shows the total number of lakes sampled as
compared to the number of QA and calibration study
samples taken. Table 8 (Section 3) shows a
breakdown of the number of samples collected by
sample type. Table K-1 in Appendix K characterizes
the distribution of analyte concentrations for WLS-I
routine lake samples. For each batch, WLS-I
averaged 11 samples associated with 5 lakes,
whereas ELS-I batches averaged 19 samples
associated with 14 lakes. The smaller number of
samples and lakes per batch in WLS-I can be
attributed to the greater distance between lakes in the
West than between lakes in the East, severe weather
conditions, and the fact that ground crews could
sample only one or two lakes per day. The
percentage of QA samples used during WLS-I was
considerably larger than the percentage used during
ELS-I because the batch sizes and the number of
lakes represented in each batch were much smaller
in WLS-I. Of the WLS-I samples collected, 46
percent were routine lake samples, and 54 percent
were QA-related or calibration study lake samples.
In contrast, 75 percent of ELS-I samples were
routine lake samples, and 25 percent were QA-
related samples (Best et al., 1987). QC samples and
split samples (i.e., matrix spikes, laboratory and trailer
duplicates, calibration/reagent blanks, detection limit
and low and high QCCS samples) are not included in
these QA/routine sample percentages.
QA samples were analyzed during WLS-I so that
estimates of precision, accuracy, detectability, and
bias could be made. Sampling and analytical variance
can arise from three major sources apart from
seasonal variations in lake chemistry:
• a field component associated with sample
collection or with short-term, localized
variability in lake chemistry
• an analytical component associated with aliquot
preparation or with variation in instrument
response within an analytical batch
• an analytical component associated with batch-
to-batch variation in instrument calibration and
response
The relative importance of these sources of variation
was assessed by comparative statistical evaluations.
Evaluations of field audits, field duplicates, trailer
duplicates, and laboratory duplicates provided
estimates of precision and bias. Field synthetic audit
samples were used to estimate accuracy. Field blank
samples were used to estimate system detectability
and levels of potential contamination introduced at the
lake site, during field laboratory processing and
sample handling, and during analysis at the analytical
laboratory.
Estimating precision and accuracy are important
facets in determining WLS-I data quality and
reliability of the lake water samples measurements
(see Figure 6 for an explanation of the ways in which
these estimates are calculated). These estimates, in
turn, aid the data user in estimating subregional and
regional populations. This section discusses the ways
in which precision has been estimated for WLS-I. It
provides estimates of precision on the basis of the
QA and QC information gathered. Similar discussions
in subsequent sections describe accuracy (Section
7), detectability (Section 8), and relative bias (Section
9 and Appendix I).
Method of Estimating Precision
The most important aspect of estimating precision is
to determine the overall (system) precision, which
accounts for the cumulative effect on analyte
concentration of all activities from sample collection to
final sample analysis and data reporting. This is the
precision estimate of most interest to the data user
concerned with calculating population estimates
(Landers et al., 1987). Subsets of system precision
can be used to segregate total variability into
components: sample collection and handling,
45
-------
Figure 5. Lakes sampled versus sample types. Western Lake Survey - Phase I.
800'
a
E
aj
in
•s
to
£
3
700 ->
600 -
500-
400-
300 -
200-
100 ~
[
Samples Collected by Helicopter Crews
Samples Collected by Ground Crews
Synthetic Audit Samples
Natural Audit Samples
L*.*.*.*.I Field and Trailer Blanks
Total
Lakes Sampled
Samples
Collected
in
Calibration
Lake Study
Field Audit
Samples
Field
Duplicate
Samples
Field/Trailer
Blanks
Misc
processing and preservation, and analysis. These
components, however, cannot be separated easily.
The following discussions describe the ways that the
different components of variability are addressed in
the WLS-I sampling design. Aspects of variability
that the sampling design does not address are
discussed also. Where applicable, recommendations
are given for refining the methods used to determine
the components of variability.
For WLS-I, two principal types of samples (duplicate
pairs and audit samples) were used to estimate
precision. Figure 7 illustrates the ways in which
duplicate pairs (field, trailer, and analytical laboratory)
and audit samples (natural and synthetic) were used
to estimate precision.
Estimating System Precision from Field Duplicate
Pairs
System precision was estimated from data on field
duplicate pair samples because (1) they were the only
precision-related QA sample type that was carried
through the entire system, that is, every process
applicable to the routine lake water samples from
collection at the lake site to analysis at the analytical
laboratory (see Figure 7), and (2) they had varying
analyte concentrations, depending on the lake from
which they were collected. This system precision
estimate differed from the precision estimated from
field audit samples, because field audit samples were
introduced at the field laboratory; therefore, they did
not include the field sampling component of
variability. Field audit samples also reflected the
"among-batch" variability caused by day-to-day
differences introduced at the field laboratory and by
instrument calibration, but only at a single
concentration for each analyte. (Precision estimates
derived from field audit sample data are discussed
later in this section.) Because field duplicate pairs
were analyzed within one batch (i.e., on the same
calibration curve), they do not include the variability of
day-to-day instrument calibration. Therefore, the
46
-------
Figure 6. Methods of estimating precision, accuracy, and bias. Western Lake Survey - Phase I.
Routine (R) Sample and Its Duplicate (D) Audit Samples
Precision*
Natural (FN)
Lab Bias, Precision
Synthetic (FL)
Lab Bias, Accuracy
Lake* [R],[D]
Lake3:
%RSD2
%RSD3
Calculate RMS, the
Pooled Set of Variances
About Different [x]
[FNY-1]
[FNY-2]
[FNY-3]
[FNY-n]
Y=Lot #
[FLZ-1]
[FLZ-2]
[FLZ-3]
Z=Lot #
[FLZ-n]
System Precision*
Calculate
[x]
Calculate
%RSD
Compare [x] from
One Laboratory to
[x] from Another
Interlaboratory
Bias
I
Calculate
%RSD
Variance
About [x] of
Audit Sample
Lot
Audit Sample
Precision
Calculate
[x]
Difference
of [x] from
True
Concentration
Accuracy
*lf Laboratory Duplicates are Used Instead of Lake Duplicates,
the Estimate is Termed Intralaboratory Precision
[ ]=Concentration
data user should assess the information derived from
the field duplicate pair system precision in conjunction
with the precision estimates derived from the audit
sample data (see Appendix J).
For each analyte, the precision of the field routine
sample and its duplicate (termed a field duplicate
pair), analyzed in the same batch, represented the
precision within that batch. This within-batch
variability was expressed as the percent relative
standard deviation (%RSD; also known as the
coefficient of variation), which is calculated as follows:
SD
%RSD = 100 —
X
where:
SD = standard deviation of the field duplicate pair
X = mean concentration of the field duplicate pair
(For pH, the within-batch variability was expressed
as the standard deviation of the field duplicate pairs.)
The system precision was calculated by pooling all
the %RSD values from all the duplicate pair samples
(which represent the unique analyte concentrations of
each lake). This "pooling" procedure was
accomplished by calculating the root-mean-square
(RMS) of the %RSD values of the field duplicate pair
47
-------
samples. The formula for calculating the RMS%RSD
is:
RMS
where:
%RSD =
[SD2
X = the mean of the %RSD values
SD = the standard deviation of the
%RSD values
n = the number of duplicate pairs
RMS%RSD = root-mean-square of the
%RSD values
For pH, the precision estimate is calculated as the
RMS of the standard deviation of the field duplicate
pairs. A statistical discussion on the use of RMS to
calculate precision estimates with duplicate pair data
is provided by Permutt and Pollack (1986) in Best et
al. (1987).
For a given analyte, the RMS is a single value whose
square estimates the mean variance for all duplicate
pair measurements. Because the %RSDs for all
duplicate pairs are used in the RMS calculation, it has
become necessary to segregate pairs whose mean
concentrations are near the detection limit, as
precision is a function of analyte concentration (see
Figure 9 later in this section). This segregation
process is accomplished with the use of a quantita-
tion limit, which is discussed below.
Although RMS is the calculation for this pooled
%RSD, as an aid to the reader all tables and
discussions in this report that refer to duplicate pair
(field, trailer, laboratory) precision estimates use the
term pooled %RSD, not RMS.
To estimate sampling method precision and
laboratory precision, the system precision estimates
also can be separated by field duplicate pairs
collected according to each sampling method and
analyzed by each analytical laboratory. Just as the
term "system precision" applies to the estimation for
all field duplicate pairs, it applies when only laboratory
values are pooled, and when only collection methods
are pooled. Because samples collected by helicopter
crews and by ground crews were distributed to both
analytical laboratories, it was not necessary to
evaluate each collection method in each laboratory.
Therefore, when the collection methods were
compared, the laboratories were pooled, and, when
the laboratories were compared, the collection
methods were pooled.
Estimating Precision of Field Duplicate Pairs
Analyzed in the Field Laboratory
Field duplicate pairs can be used to determine
precision in the field laboratory for the measurements
of pH, DIG, turbidity, and true color. Because field
duplicate pairs are analyzed for these variables,
precision can be assessed for samples that are
analyzed in the field laboratory. This precision
estimate, however, does not isolate the precision
attributable to the field laboratory alone, because the
field sampling variability is included in the estimate.
For WLS-I samples, trailer duplicate pairs are the
only samples that can be used to quantify precision
for field laboratory measurements.
Estimating Field Laboratory Precision from Trailer
Duplicate Pairs
The trailer duplicate sample pair is created by splitting
a lake water sample in the field laboratory. The
precision estimate for the trailer duplicate pair is
different from the estimate for the field duplicate pair
analysis in the field laboratory, because the effect of
sample collection is eliminated from the trailer
duplicate pair. This precision estimate, termed the
field laboratory precision estimate, is calculated from
trailer duplicate pairs. It applies only to the four
variables analyzed in the field laboratory (closed-
system pH, closed-system DIG, true color, and
turbidity) and can be related directly to the DQOs for
intralaboratory precision (see Table 2 in Section 1). In
order to check the precision for other analytes,
especially for those filtered or preserved in the field
laboratory, an additional QA step that was not in the
WLS-I sample design or QA design would be
needed. This step would consist of splitting a routine
sample in the field laboratory, processing the split
samples, and including them in the batches sent to
the analytical laboratory. This design modification
would impact other aspects of sample collection and
analysis; the volume of sample required for the
performance of all analyses would exceed the amount
of sample collected according to the WLS-I design
(see Figure 8).
As in ELS-I, the design of the QA program provided
one trailer duplicate to be run for each sample batch.
In fact, 137 trailer duplicates were analyzed for the
149 WLS-I batches. The discrepancy between the
number of trailer duplicates and the number of
batches resulted from the complex sampling design of
the calibration study. Because the trailer duplicate
was designed to be a daily check on the variability of
analyses within the field laboratory, it was necessary
to analyze only one trailer duplicate per processing
day. On 12 occasions, the field laboratories
processed two separate batches in one day. One
was the normal batch that contained a trailer duplicate
and one was a batch that contained calibration study
samples only. In these instances, a single trailer
duplicate was used to check for field laboratory
precision in conjunction with the two batches.
48
-------
Figure 7. Ways in which quality assurance and quality control samples are applied to estimates of precision and accuracy, Western Lake Survey - Phase I.
Sample
Type
Field
Sampling
Activities
Field
Laboratory
Activities
Analytical
Laboratory
Activities
Field Duplicate Pairs
-fr.
CO
Trailer Duplicate Pairs
Analytical Laboratory
Duplicate Pairs
Field Audits
(Natural and Synthetic)
Field Audits
(Synthetic Only)
Effects
Determined
(As Estimates)
Externally
(Cooperative
Laboratory)
Prepared
Samples
"Collected
Externally
(Cooperative Lab)
Prepared Samples
"Collected"
All Variability
Components
All Analysis
Varied Concentrations
for Population
Estimates
• Sampling + Field Lab
Components
• Field Lab Analytes Only
• Varied Concentrations
for Population
Estimates
Field Lab Component
Field Lab Analytes Only
Varied Concentrations
Intralaboratory
Precision DQO
• Analytical Lab Component
• All Analytes
• Varied Concentrations
Intralaboratory
Precision DQO
Field Lab + Analytical
Lab Components
All Analytes
One Concentration
over Time
Precision; Relative
Interlaboratory Biases
Field Lab Component
Field Lab Analytes Only
One Concentration
over Time
Precision; Relative
Interlaboratory Biases
Field Lab + Analytical
Lab Components
Most Analytes
One Concentration
over Time
Accuracy DQO
Name
Of
Table
System Precision
Estimates
(Calculated from Field
Duplicate Pairs)
Precision Estimates
of Field Duplicate Pairs
Analyzed in the
Field Laboratory
Field Laboratory
Precision Estimates
(Calculated from
Trailer Duplicate Pairs)
Intralaboratory
Precision Estimates
(Calculated from
Analytical Laboratory
Duplicate Pairs)
Precision Estimates of
Audit Sample Lots
Analyzed Among Batches
Precision Estimates of Audit
Sample Lots Analyzed
Among Batches in the
Field Laboratory
Estimated Analytical
Accuracy
(Calculated trom Field
Synthetic Audit Samples)
Category;
See
Table
All Samples;
Tablets
By Lab;
Tablets
By Sampling
Method;
Tablet?
All Field Labs;
Table 21
All Field Labs;
Table 22
Both Labs;
Table 23
By Lab;
Table 24
Both Labs;
Table 26, 27
By Lab;
Appendix F, I
All Field Labs;
Appendix E
By Field Lab:
Appendix E
All Field Labs.
Table 29
Appendix G
By Lab:
Table 29
Appendix G
-------
Figure 8. Proposed procedural steps that would be necessary to quantify the collection, processing, and analytical components
of variability.
Split Performed
in Analytical Laboratory
Split Performed
in Field Laboratory
Sampling
Apoparatus
Lake
Water Spliti
or
Audit
Sample
Split Performed
at Lake Site
Analytical Laboratory
Precision (%RSD)
Total Laboratory
Precision (%RSD)
System
Precision
(%RSD)
%RSD = Percent Relative Standard
Deviation
R = Routine Sample
D! = Duplicate Sample of 1st Split
D2 = Duplicate Sample of 2nd Split
Da = Duplicate Sample of 3rd Split
%RSD Split! (R, D,) >%RSD Split2 (R, D2) >%RSD Splits (R, D3)
Estimating Intralaboratory Precision from
Analytical Laboratory Duplicate Pairs
Although field duplicate pairs are analyzed in the
analytical laboratory, the precision estimates derived
from these samples include the overall effects on the
sample (sampling, handling, and analysis at all
stages). In order to quantify only the analytical
laboratory precision, analytical laboratory duplicate
pairs are used. These pairs are created in the
analytical laboratory by splitting one sample from
each batch. Precision for these pairs is termed
intralaboratory precision. It can be compared directly
to the intralaboratory precision goal (the DQO).
Statistical manipulation of analytical laboratory
duplicate pairs is similar to that for field duplicate
pairs. A %RSD is calculated for each pair to
represent variability within the batch, and a pooled
%RSD is calculated for all duplicate pairs.
Estimates can be calculated for the laboratories
combined, also termed "pooled," and for each
laboratory individually. The interpretation of the results
for analytical laboratory duplicate pairs and field
duplicate pairs, however, differs significantly: The
pooled %RSD for analytical laboratory duplicate pairs
estimates intralaboratory precision (a DQO); the
pooled %RSD for field duplicate pairs estimates
system precision.
Establishing the Quantitation Limit
To ensure that mean sample concentrations of the
field duplicate, trailer duplicate, and analytical
laboratory duplicate pairs are sufficiently above the
level of background contamination to estimate
precision reliably, a quantitation limit is used for all
variables except pH. For WLS-I, the quantitation limit
was calculated as 10 times the standard deviation (10
SB) of the concentrations of the corresponding blanks
(field, trailer, or analytical laboratory). Precision
50
-------
estimates can be calculated from all sample pairs
(pairs that have mean concentration greater than
zero). Some of these pairs are affected greatly by
background (pairs that have mean concentration near
zero, or the detection limit); other pairs are affected
minimally by background (those pairs with mean
greater than 10 SB)- Therefore, the quantitation limit
is the level above which duplicate precision is
expected to stabilize. The relationship of duplicate
pair samples to quantitation limits and to sample
concentrations is illustrated in Figure 9; supporting
sample-concentration data are given in Table 14.
System Precision Results
System Precision Estimated from Field Duplicate
Pairs
Sampling Methods and Analytical Laboratories
Pooled-
System precision estimated from WLS-I field
duplicate pairs is shown in Table 15. The
intralaboratory precision DQOs that were the check
on analytical laboratory precision are inappropriate for
rigid application to field duplicate precision, but these
DQOs are useful as a gauge for assessing the field
duplicate precision estimate. Field duplicate precision
estimates that are within these intralaboratory
precision goals (using the quantitation limit as a
cutoff) should be considered better than the precision
that was anticipated before the survey began. For
some variables, precision estimates calculated from
field duplicate pairs did not meet the intralaboratory
precision goals, but they may be reasonable
estimates when the additional handling of the samples
in the field is considered. Still other variables may be
considered to have poor precision based on the
intralaboratory precision goals. Table 16 lists variables
for which field duplicate pair precision was within or
slightly above the DQO for intralaboratory precision. It
also lists variables for which field duplicate pair
precision was poor (well above the DQO for
intralaboratory precision).
Sampling Methods Separated-
The ability to sample lakes in a precise manner was
essential to meeting the goals of WLS-I. Duplicate
pairs were compared for precision as one way of
assessing potential differences between the ability of
the helicopter crew and that of the ground crew to
collect routine samples. Table 17 separates the two
sampling methods and shows quantitation limits that
are calculated from the appropriate field blanks (e.g.,
field blanks collected by the ground crews were used
to establish the quantitation limit for precision of the
field duplicate pairs they collected). By separating the
two methods, the precision of each sampling method
and the differences in precision between methods
can be assessed. For both sampling methods, most
analytes show excellent precision for duplicate pairs
that have mean values above the quantitation limit.
Precision met or was near the DQO for all analytes
that have a large enough sample size (n) to yield
reliable precision estimates.
Analytical Laboratories Separated-
Table 18 presents system (field duplicate) precision
separated by analytical laboratory. Comparing these
two sets of results may be useful in determining
Table 14. An Example of the Relationship of %RSD to Duplicate Pair Samples
for Different Concentrations9
Differences between the Routine
and its Duplicate
Pair
Number
1
2
3
4
5
6
7
8
g
10
11
12
13
Routine
Sample
Concentration
(mg/L)
0.001
0.011
0.021
0.031
0.041
0.051
0.061
0.071
0.081
0.091
0.101
0.501
1.001
Duplicate
Sample
Concentration
(mg/L)
0.006
0.016
0.026
0.036
0.046
0.056
0.066
0.076
0.086
0.096
0.106
0.506
1.006
Practical
Difference
(absolute,
mg/L)
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
0.005
Statistical
Difference
(relative,
%RSD)
101.0
26.2
15.0
10.6
8.1
6.6
5.6
4.8
4.2
3.8
3.4
0.7
0.4
aThese are not actual WLS-I data They are companion material to Figure 9.
51
-------
Figure 9. Relationship of duplicate pair samples to quantitation limits and sample concentrations.
n X<0
x>0
Precision estimate can be calculated'with all duplicate
pair samples that have mean concentrations
greater than 0 mg/L* (this includes low-level
sample pairs greatly affected by background
influences).
x< 10SB
Duplicate pair samples that have mean
concentrations below or equal to the quantitation
limit can be excluded from calculating precision
estimates because of background influences at
low levels.
Quantitation
Limit
10 SB
x> 10SB
Duplicate pair samples that have mean
concentrations above the quantitation limit
are used to calculate precision estimates
where background influences are minimal.
All
Duplicate
Pairs
Less than
or Equal to.
} mg/L* are'
Excluded
Required
Instrument
Detection Limit
Calculations
40 (Except forlj.
ANCandBNC)
I ntra laboratory
Precision
Goal
-0.010 0.000 0.010 0.020
0.030 0.040 0.050 0.060 0.070
Mean Concentration (mg/L*)
0.080 0.090
3.4%0.7% 0.4%
i i-
0.1000.5001.000
• = The mean concentration of a routine
sample and its duplicate; each pair has an
absolute difference of 0.005 mg/L.
10 SB = 10 times the standard deviation of the
blank sample concentrations.
*mg/L is used for illustrative
purposes; other units can apply.
whether imprecision is associated with sampling
technique or with analytical performance.
Tables 15, 17, and 18 must be evaluated with the
understanding that, except for calibration study lakes,
each laboratory analyzed samples from different
subregions (Landers et al., 1987). Thus, the field
duplicate pairs for each subregion also were
segregated by laboratory. Consequently, because
precision is concentration dependent (Figure 9),
differences in the precision estimates for the two
laboratories may be in part the result of subregional
differences in concentrations of some analytes.
Precision also may depend on the chemical matrix of
the lake water samples, which may be a subregional
characteristic. (See Landers et al., 1987, for further
discussion of sub- regional lake chemistry.)
Another consideration is the distribution of duplicate
pairs sent to each laboratory (see Table 19).
Laboratory II analyzed about 60 percent of the WLS-
I field duplicate pairs sampled by helicopter crews
and by ground crews, and Laboratory I analyzed the
other 40 percent. Because the pooled system
precision (Table 15) may mask poor precision
associated with one laboratory or with one method, it
is essential that all of these variability issues are
accounted for when field duplicate precision
estimates are assigned to particular components of
laboratory and method.
Table 20 summarizes the system precision results
(pooled and by method or analytical laboratory
component) that showed a high degree of variability.
This table illustrates that, for a given analyte, isolating
a particular source of variability (sampling method,
laboratory, lake chemistry, subregion, or quantitation
limit) from other potential sources is often difficult.
Table 34 (Section 9) illustrates the interactions
between the major components that contribute to
variability. This table was constructed on the basis of
calibration lake sample data.
In a few cases, field duplicate pairs produced results
far outside the precision goals across many analytes.
For instance, the field laboratory pH readings for the
field duplicate pair collected from Lake ID 4A3-044
(Twin Lakes-North) in California were quite different
52
-------
Table 15. System Precision Estimates Calculated from Field Duplicate Pairs (Sampling Methods and Laboratories Pooled), Western Lake
Survey - Phase I
s
All pairs (mean > 0)
Pairs that have a mean > Quantitation
Limit
Variable
(in mg/L
unless
noted)
Al, extractable
Al, total
ANC (iieq/L)
BNC (neq/L)
Ca
cr
Conductance
(nS/cm)
DIC, air
equilibrated
DIC, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH/
N03
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
Intralaboratory
Precision Goal
(in %RSD unless noted)
10 (if Al cone. > 0.01 mg/L)
20 ( if Al cone. < 0.01 mg.L)
10 (if Al cone. > 0.01 mg/L)
20 (if Al cone. < 0.01 mg/L)
10
10
5
5
2
10
10
5 (if DOC cone. > 5 mg/L)
10 (if DOC cone. < 5 mg/L)
10
5
5
5
10
5
5
10
10 (if P cone. > 0.01 mg/L)
20 (if P cone. < 0.01 mg/L)
0.05 (pH units)
0.05 (pH units)
n
189
215
215
203
215
215
215
215
215
215
213
197
215
215
137
215
83C
205
210
215
215
Estimated
Precision
(Pooled %RSD)t>
299.7
33.4
4.5
190.4
2.3
15.7
6.5
6.9
7.7
16.7
24.7
112.5
9.6
8.1
139.2
7.2
324.7
61.3
77.4
0.08
0.08
Quantitation
Limit3
0.021
0.085
24.9
92.3
0.30
0.21
6.1
0.71
1.32
2.0
0.016
0.07
0.08
0.03
0,10
0.17
0.13
0.342
0.043
--
--
n
11
11
204
3
208
65
192
173
103
49
90
16
199
212
1
206
3
8
9
—
--
Estimated Precision
(Pooled % RSD)b
44.2
10.2
4.1
72.4
2.3
8.6
6.4
6.9
2.7
12.7
26.2
24.6
4.5
8.2
N/A
7.3
7.9
1.1
37.5
-
--
(continued)
-------
Table 15. (Continued)
All pairs (mean > 0) Pairs that have a mean > Quantitation
variable
unless
noted)
pH, air
equilibrated (pH
units)
SiO2
S042-
Precision Goal
(in %RSD unless noted)
0.05 (pH units)
5
5
n
214
215
215
Estimated
Precision
(Pooled %RSD)b
0.17
16.1
14.5
Quantitation
Limita
_.
2.07
0.56
Estimated Precision
n (Pooled % RSD)b
--
119 7.0
124 4.3
a The quantitation limit is 10 SB (10 times the standard deviation of the field blank measurements). Quantitation limits are not calculated for pH
measurements.
b Pooled standard deviation used for pH.
c Number of observations is smaller because concentrations of NH4 + were low in most WLS-I samples and because of instrumental drift (i.e., mean
concentrations of NH4 < 0).
d N/A = not applicable.
-------
Table 16. Summary of System Precision Results by Variable (Sampling Methods and Analtytical
Laboratories Pooled), Western Lake Survey - Phase I
Variable that met or
was near DQO; duplicate
pairs above the
quantitation limit3
DQO (in
Variables that did not
meet DQO; duplicate
pairs above the
quantitation limit3
Comments
Al, total
ANC
Ca
DIG (air equilibrated)
DIG (closed system)
DIG (initial; open system)
K
Mg
Na
pH (acidity; open system)
pH (alkalinity; open system)
pH (air equilibrated)
SiO2
SO42-
True Color
Turbidity
10 or 20
10 or 20
10
10
5
5
2
10
10
10
5 or 10
10
5
5
10
5
5
10
10 or 20
0.05 (pH unit)
0.05 (pH unit)
0.05 (pH unit)
0.05 (pH unit)
5
5
5 (PCU)
10
Al, extractable
BNC
Conductance
DOC
F", total dissolved
Fe
Mn
P, total
pH (closed system)
Low concentrations0 found in all
WLS-I lake samples.
Only 3 duplicate pair samples above
the quantitation limit.
Mostly low-conductance0 lakes
sampled in WLS-I; even so,
precision estimate was 6.4%.
Low-DOCc routine lake samples
analyzed in WLS-I.
Most of the poor precision indicated in
Laboratory I; very low (0.016 mg/L)
quantitation limit.
Low concentrations0 found in WLS-I
lake samples.
Very low concentrations0 found in
WLS-I lake samples.
Low quantitation limit and very few
duplicate pairs that have mean
concentrations above that limit.
Possibly a result of low ionic strength,
circumneutral lake samples0; no
quantitation limit calculated for pH
measurements.
WLS-I lakes very low in turbidity.
a Note: The field duplicate pair mean concentrations are often below the quantitation limit; therefore, they are not
included when precision goals and estimates are discussed. Figures J-lb to J8b, J-9, J-10b to J-23b, J-
24, and J-25b to J-26b present all field duplicate pairs plotted by mean concentration and %RSD. This allows
the data user to examine the relationship between precision and concentration and the quantitation limits.
b Note: DQO is used as a gauge, but is not directly applicable for field duplicate samples.
c Note: Poor system precision probably is attributable to the fact that low concentrations of the analyte were
measured for most WLS-I lake samples (see Table K-1 in Appendix K.
(pH 9.55 for the routine and 8.16 for the duplicate).
The analytical laboratory results for this pair showed
very poor precision for BNC, CI", DOC, initial DIG,
pH, and Si02 as well. Written comments on the field
forms indicated that aquatic vegetation was present at
the sampling location. The presence of vegetation
could account for the heterogeneity of the sample
pair. This situation, which was assessed during data
validation and during calculation of population
estimates, indicates that thorough, detailed
documentation is essential to explaining possible data
inconsistencies and anomalies.
55
-------
Table 17. System Precision Estimates Calculated from Field Duplicate Pairs (by Sampling Method) Western Lake Survey - Phase I
Helicopter
Ground
All Pairs (mean > 0)
Variable (in
mg/L unless
noted)
Al, extractable
Al, total
ANC (neq/L)
BNC (neq/L)
Ca
cr
Conductance (nS/cm)
DIG, air equilibrated
DIG, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH4 +
NO3
P, total
pH, acidity (pH units)
pH, alkalinity (pH units)
pH, air equilibrated (pH
units)
SiO2
SO42'
n
75
87
87
79
87
87
87
87
87
87
87
81
87
87
60
87
39
81
84
87
87
87
87
87
Estimated
Precision
(Pooled
150.9
28.7
2.9
300.4
1.6
8.1
8.0
9.7
4.4
18.2
20.6
110.0
13.4
1.6
152.1
3.0
449.4
72.3
58.3
0.10
0.10
0.09
11.4
5.7
Quantitation
Limit3
0.022
0.099
19.2
80.0
0.28
0.12
5.5
0.77
0.93
2.4
0.010
0.07
0.09
0.03
0.11
0.22
0.11
0.334
0.040
--
--
--
1.25
0.40
Pairs that have a
mean > Quantitation
Limit
n
4
5
84
2
84
54
81
70
68
19
63
10
77
87
1
82
3
3
8
--
--
--
70
62
Estimated
Precision
(Pooled
12.2
12.8
2.9
25.0
1.7
7.7
8.2
9.9
4.3
17.1
23.5
29.1
4.1
1.6
N/A
2.7
7.9
1.2
8.2
--
--
--
9.1
5.6
All Pairs (mean > 0)
n
114
128
128
124
128
128
128
128
128
128
126
116
12a
128
77
128
44
124
126
128
128
127
128
128
Estimated
Precision
(Pooled
366.0
36.2
5.4
43.3
2.6
19.3
5.1
3.9
9.4
15.6
27.1
114.2
5.7
10.4
128.3
9.0
141.0
52.9
87.8
0.06
0.06
0.20
18.6
18.1
Quantitation
Limit3
0.020
0.067
30.2
104.7
0.32
0.28
6.7
0.58
1.61
1.40
0.021
0.07
0.08
0.04
0.10
0.09
0.14
0.352
0.046
--
--
--
2.71
0.70
Pairs that have a
mean > Quantitation
Limit
n
7
8
117
1
124
16
107
112
42
44
40
6
121
124
0
127
0
5
1
-
—
--
45
68
Estimated
Precision
(Pooled
54.7
10.1
4.7
N/A
2.6
12.5
4.2
3.7
3.7
8.3
34.9
14.2
4.7
10.6
N/A
9.1
N/A
0.9
N/A
-
--
--
4.0
3.7
a The quantitation limit is 10
for pH measurements.
b Pooled standard deviation
N/A = not applicable.
SB (10 times the standard deviation of the field blank measurements [helicopter or ground]). Quantitation limits are not calculated
was used for pH.
-------
Table 18. System Precision Estimates Calculated from Field Duplicate Pairs (by Analytical Laboratory) Western Lake Survey - Phase I
Laboratory I
Laboratory II
en
All Pairs (mean > 0)
Variable (in
mg/L unless
noted)
Al, extractable
Al, total
ANC (peq/L)
BNC (neq/L)
Ca
cr
Conductance (pS/cm)
DIG, air equilibrated
DIG, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH4 +
N03
P, total
pH, acidity (pH units)
pH, alkalinity (pH units)
pH, air equilibrated (pH
units)
SiO2
so/-
n
63
86
86
78
86
86
86
86
86
86
84
76
86
86
65
86
29
83
85
86
86
85
86
86
Estimated
Precision
(Pooled
%RSD)t)
507.4
49.2
6.5
302.5
1.4
15.2
2.3
4.0
3.6
11.0
37.6
141.7
14.3
12.7
70.9
11.1
424.8
56.4
99.0
0.07
0.07
0.05
24.9
6.4
Quantitation
Limit3
0.024
0.111
34.5
85.4
0.23
0.19
4.3
0.41
1.58
2.6
0.021
0.04
0.06
0.03
0.01
0.24
0.18
0.196
0.063
--
--
—
2.90
0.52
Pairs that have a
mean > Quantitation
Limit
n
4
2
82
2
86
33
86
85
44
13
45
16
83
86
8
71
2
15
1
--
--
—
20
77
Estimated
Precision
(Pooled
%RSD)*>
14.2
3.9
5.6
87.0
1.4
13.2
2.3
4.0
3.6
20.4
36.3
20.9
14.1
12.7
31.4
12.1
9.5
11.5
N/A
--
--
—
14.7
4.4
All Pairs (mean > 0)
n
126
129
129
125
129
129
129
129
129
129
129
121
129
129
72
129
54
122
125
129
129
129
129
129
Estimated
Precision
(Pooled
"/oRSD)'3
77.9
15.6
2.5
42.6
2.7
16.1
8.1
8.3
9.5
19.6
9.2
89.4
4.1
1.4
179.9
2.2
255.2
64.4
58.3
0.09
0.09
0.21
3.9
17.9
Quantitation
Limita
0.016
0.042
12.8
54.4
0.28
0.22
5.9
0.7
0.94
0.8
0.009
0.07
0.09
0.04
0.13
0.08
0.05
0.415
0.021
--
„
1.20
0.57
Pairs that have a
mean > Quantitation
Limit
n
10
35
128
7
122
39
107
103
84
90
77
6
113
125
1
129
1
1
3
--
--
__
115
46
Estimated
Precision
(Pooled
%RSD)i)
53.0
17.5
2.5
11.2
2.7
2.6
8.3
8.4
4.3
14.9
8.5
37.1
3.5
1.3
N/A
2.2
N/A
N/A
8.9
--
--
..
3.5
4.2
a The quantitation limit is 10
for pH measurements.
b Pooled standard deviation
N/A = not applicable.
SB (10 times the standard deviation of the analytical laboratory blank measurements ). Quantitation limits are not calculated
was used for pH measurements.
-------
Table 19. Distribution of Field Duplicate Pairs (Helicopter
and Ground) by Laboratory, Western Lake
Survey - Phase I
Duplicate Pairs
Collected by Duplicate Pairs Total
Helicopter Collected by Duplicate
Laboratory Crews Ground Crews Pairs
I
II
Total
31
56
87
55
73
128
86
129
215
Precision Estimated from Field Duplicate Pairs
and Trailer Duplicate Pairs Analyzed in the Field
Laboratory
Precision estimates for field duplicate pairs analyzed
in the field laboratory (Table 21) and for trailer
duplicate pairs analyzed in the field laboratory (Table
22) are given for pH, DIG, true color, and turbidity. All
pH and DIG measurements were within desired
precision goals, except that the precision estimate for
the pH of field duplicate pairs was calculated at 0.12
pH unit. Although the intralaboratory precision goal
was ±0.05 pH unit, on the basis of ELS-I
experience, the EMSL-LV QA staff considered
±0.10 pH unit acceptable when assessing daily QA
precision for field duplicate pairs (see Table 2).
Precision goals for trailer duplicate pairs for turbidity
and true color were met for mean values above the
quantitation limit. Quantitation limits were not
calculated for pH and DIG because field blank and
trailer blank samples were not analyzed in the field
laboratory for these parameters.
Intralaboratory Precision Estimated from
Analytical Laboratory Duplicate Pairs
On the basis of intralaboratory precision estimated
from pooled values for analytical laboratory duplicate
pairs, the two analytical laboratories exhibited
excellent reproducibility. The intralaboratory precision
estimated from analytical laboratory duplicate pairs
above the quantitation limit met the DQOs for every
analyte except Mn (see Table 23). Concentrations of
Mn in WLS-I routine lake samples generally were
below the levels at which good precision is expected.
The median concentration for Mn was 0.001 mg/L
(see Appendix K, Table K-1). In addition, 95 percent
of the routine lake samples had concentrations less
than 0.03 mg/L. The less than acceptable precision
estimate (16.6%) for samples above the quantitation
limit (0.02 mg/L) should be of little concern to the
data user.
Table 24 presents the intralaboratory precision by
laboratory. This analysis of individual laboratories
reveals results similar to those obtained from the
analysis of pooled values, except that for Laboratory
II, Fe as well as Mn was outside the DQO.
Mn is the only analyte that had precision estimates
higher than the intralaboratory precision goals for
duplicate pairs above the quantitation limit. ANC and
BNC, however, were not assessed in this manner
because a quantitation limit could not be calculated.
For the field duplicate determinations, field blanks
were used in the calculation of the quantitation limits;
however, in the laboratory, calibration blanks were not
analyzed for ANC or BNC. Because a quantitation
limit was not used for these analytes, precision
estimates for all laboratory duplicate pairs are shown
in Tables 23 and 24 regardless of how close the
mean values are to 0 jieq/L.
Evaluating data from the laboratories separately also
revealed a procedural problem with Laboratory ll's
analysis of the cations (Ca, Fe, K, Mg, Mn, and Na). It
was not discovered that Laboratory II reported all of
the calibration blanks (n = 91) as 0.000 mg/L until
statistical analyses and data verification had been
completed on the data for these analytes. In a
subsequent discussion, the laboratory manager
indicated that the calibration blanks were used to
"auto-zero" the spectrophotometer before the
detection limit QCCS was analyzed. This procedure
was a misinterpretation of the SOW (contract) and a
deviation from the laboratory's analytical performance
in ELS-I.
Two problems resulted from this action. First,
because the true calibration blank was not reported
and because the instrument calibration was improper,
some concern arose that there might be enough bias
to create statistical problems for the data user.
Fortunately, the detection limit QCCS provides an
additional check for values at or near the detection
limit. Because Laboratory II had no difficulty analyzing
these QCCSs within the criteria required (±20% of
the true value} for any of the metals, it was
determined that any bias created was negligible and
did not affect the statistical evaluation for popu-
lation estimates. Second, proper quantitation limits
could not be calculated from the calibration blanks for
these metals (i.e., the standard deviations of the
blanks were all equal to 0.000; therefore, the
quantitation limit equalled 0.000). Although
quantitation limits could not be calculated, Ca, K, Mg,
and Na still met the DQOs when all the pairs that had
mean concentrations greater than zero were used;
only Fe and Mn did not. To calculate overall
intralaboratory precision of the pooled laboratory
duplicate data (Table 23), quantitation limits for these
six cations were calculated from Laboratory I's
calibration blank values only.
In spite of the protocol changes and the procedural
and reporting difficulties noted above, the
intralaboratory precision goals for WLS-I were met
58
-------
Table 20. Checklist of Variables for Which System9 Precision Estimates Calculated from Field Duplicate Pairs Did
Not Meet Intralaboratory Precision Goals (Pooled and Separated by Sampling Method and by
Laboratory), Western Lake Survey - Phase I (Note: X indicates high variability above the quantitation
limit.)
Variable
Al.extract-
able
Al, total
ANC
BNC
Ca
Cl-
Heli-
DQO (%)6 Pooled copter Ground Lab. 1
10 or 20 X X
10 or 20
10
10 X X X
5
5 XX
Lab. II Comment
X Only 1 duplicate pair sample
had poor precision.
Very few samples above the
quantitation limit.
Only 1 duplicate pair sample
Conductance
had poor precision.
WLS-I samples had low
conductance.
DIC (air
equilibrated)
DIC (initial)
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH4 +
N03
P, total
pH (acidity)
pH (alkalinity)
pH (air-
equilibrated)
SiO2
SO42'
10
10
5 or 10 X
5 X
10 X
5
5
10
5
5
10
10 or 20 X
0.05 (pH
unit)
0.05 (pH
unit)
0.05 (pH X
unit)
5
5
X X X X Low DOC in WLS-I lakes.
XXX Poor precision result of
Laboratory I performance.
X X X X WLS-I samples had low Fe
concentrations.
X Only 1 duplicate pair had poor
precision.
X X Only 1 duplicate pair had poor
precision.
X Low Mn in WLS-I lakes.
X X
Only 1 duplicate pair had poor
precision.
X X Only 1 duplicate pair had poor
precision.
X Two duplicate pair samples had
poor precision.
a System precision includes variability from lake sampling, sample processing, and sample analysis.
b DQO is for intralaboratory precision and is not directly applicable to system precision.
(see Table 25). This observation indicates that any
lack of precision in sample analysis was likely to have
come from sources outside the analytical laboratories.
However, even the small amount of imprecision
(expressed as intralaboratory precision estimates)
shown by the analytical laboratory must be
considered as a component of the (system) precision
estimates.
Method of Estimating Precision Among
Batches
Estimating Precision Among Batches from Field
Audit Samples
Field audit samples (natural and synthetic) were used
to estimate precision at specific concentrations over
time (i.e., among batches) in WLS-I. Field audit
sample precision estimates (as %RSD or, for pH, as
standard deviation) also indicate variability of the
59
-------
Table 21. Precision Estimates for Field Duplicate Pairs Analyzed in the Field Laboratory, Western
Lake Survey - Phase I
All Pairs
Variable
pH
DIC
True Color
Turbidity
Intralaboratory
Precision Goal
±0.1 pH unit
10 %RSD
±5PCU
10 %RSD
(mean
n
208
208
165
206
> 0)
Estimated
Precision
(Pooled Quantitation
%RSD) Limit3
0.1 2&
6.8
61. 3= 32
24.4 0.8
Pairs that have a mean >
Quantitation Limit3
Estimated
Precision
(Pooled
n %RSD)
..
-
6 4.7<*
37 16.5
a The quantitation limit is 10 SB (10 times the standard deviation of the field blank measurements).
Quantitation limits cannot be calculated for pH and DIC because field blanks were not analyzed.
b Pooled standard deviation was used for pH.
c This value is equivalent to an average standard deviation of ± 6.0 PCU.
d This value is equivalent to an average standard deviation of + 2.2 PCU.
analytical and sample preparation methods; they
exclude variability associated with lake sampling. Data
from field audit samples play a key role in maintaining
a credible data base. Such samples are useful in
estimating relative biases between laboratories
(interlaboratory bias).
Description of Field Natural Audit Samples-
The field natural audit sample is obtained by sampling
a natural lake system in bulk (200 to 400 L). The bulk
sample receives a unique lot number which
distinguishes it from audit samples collected at other
lakes and from audit samples collected at the same
lake but at other times. The bulk sample is filtered
and is apportioned into 2-L subsamples.
As in ELS-I and the NSS Phase I Pilot Survey, EPA
contracted with Radian Corporation to prepare and
distribute the field audit samples. To ensure that all
audit samples of a particular lot were uniform, Radian
was instructed by EMSL-LV to follow a specified
protocol (see Appendix C) for preparing the 2-L field
natural audit aliquots. The procedure called for
preparing all aliquots from the sample lots at the
same time.
In contrast, ELS-I field audit samples were prepared
just before daily shipment to the field laboratories. It
can be argued that preparing aiiquots for an entire lot
of a natural audit sample by separating the lot into
2-L bottles at one time (as many as 100 aliquots,
depending on the bulk volume) creates separate
populations (i.e., each container) over time. For
instance, biological action may occur in one aliquot,
changing the chemical composition, yet may not
occur in all aliquots. WLS-I field activities had a 2-
Table 22. Precision Estimates for Trailer Duplicate Pairs Analyzed in the Field Laboratory,
Western Lake Survey - Phase I
Variable
PH
DIC
True Color
Turbidity
Intralaboratory
Precision Goal
±0.1 pH unit
10 %RSD
±5 PCU
10 %RSD
All Pairs
(mean > 0)
Estimated
Precision
(Pooled
n %RSD)
132 0.03&
134 3.4
99 25. 1C
134 11.8
Quantitation
Limit3
--
--
15
0.4
Pairs that have
Quantitation
n
-
--
25
65
a mean >
Limita
Estimated
Precision
(Pooled
%RSD)
-..
..-
3.8d
7.8
a The quantitation limit is 10 SB (10 times the standard deviation of the field blank measurements).
Quantitation limits cannot be calculated because field blanks were not analyzed.
b Pooled standard deviation was used for pH.
c This value is equivalent to an average standard deviation of ± 2.7 PCU.
d This value is equivalent to an average standard deviation of ± 0.8 PCU.
60
-------
Table 23. Intralaboratory Precision Estimates Calculated from Analytical Laboratory Duplicate Pairs (Laboratories Pooled) Western Lake Survey
- Phase I
All pairs (mean > 0)
variable
(in mg/L
unless
noted)
Al, extractable
Al, total
ANC (peq/L)
BNC (neq/L)
Ca
Cl"
Conductance
(liS/cm)
DIC, air
equilibrated
DIC, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH/
IMO3"
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
Intralaboratory
Precision Goal
(in %RSD unless noted)
10 (if Al cone. > 0.01 mg/L)
20 ( if Al cone. < 0.01 mg/L)
10 (if Al cone. > 0.01 mg/L)
20 (if Al cone. < 0.01 mg/L)
10
5
5
2
10
10
5 (if DOC cone. > 5 mg/L)
10 (if DOC cone. < 5 mg/L)
5
10
5
5
10
5
5
10
10 (if P cone. > 0.01 mg/L)
20 ( if P cone. < 0.01 mg/L)
0.05 (pH units)
0.05 (pH units)
n
140
148
139
148
149
149
149
149
149
146
148
144
149
149
120
149
5$d
148
137
149
149
Estimated
Precision
(Pooled %RSD)6
13.3
3.3
121.6
16.4
0.7
2.3
1.1
6.4
4.3
11.1
2.7
56.1
1.4
0.6
105.1
1.1
95.3
8.8
41.7
0.03
0.03
Quantitation
Limit3
0.014
0.029
--
--
0.04C
0.06
3.4
0.33
0.26
0.6
0.008
0.04C
0.08C
0.02C
0.02C
0.1 3C
0.07
0.052
0.010
..
i an o 11 id i i lave
n
22
64
—
_.
148
145
147
126
131
119
148
63
147
148
45
147
6
118
46
a n icau ." ^xuai imauui i (_iillil
Estimated Precision
(Pooled % RSD)*>
4.3
3.3
__
0.7
2.3
1.1
5.8
2.7
4.1
2.7
4.9
1.4
0.6
16.6
1.0
0.9
2.8
7.2
(continued)
-------
Table 23. (Continued)
All pairs (mean > 0)
Variable
(in mg/L
unless
noted)
pH, air
equilibrated (pH
units)
Si02
so42-
Intralaboratory
Precision Goal
(in %RSD unless noted)
0.05 (pH units)
5
5
n
149
148
149
Estimated
Precision
(Pooled %RSD)b
0.02
25.9
1.8
Quantitation
Limit3
--
0.28
0.12
Estimated Precision
n (Pooled % RSDp
—
125 2.9
143 1.7
a The quantitation limit is 10 SB (10 times the standard deviation of the calibration or reagent blank measurments). Quantitation limits are not calculated for
ANC, BNC, or pH measurements.
b Pooled standard deviation values were used for pH measurements.
c Quantitation limit calculated from Laboratory I's calibration blanks only.
d Reported concentrations of many NH4 + pairs were < 0. As a result, the n for this variable is small relative to the n for other variables.
en
-------
Table 24. Intralaboratory Precision Estimates Calculated from Analytical Laboratory Duplicate Pairs (by Laboratory), Western Lake Survey •
Phase I
Laboratory I
Laboratory II
O)
CO
All Pairs (mean > 0)
Variable (in
mg/L unless
noted)
Al, extractable
Al, total
ANC (neq/L)
BNC (neq/L)
Ca
or
Conductance (jiS/cm)
DIC, air equilibrated
DIC, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH/
N03-
P, total
pH, acidity (pH units)
pH, alkalinity (pH units)
pH, air equilibrated (pH
units)
SiO2
SO42'
n
52
57
48
57
58
58
58
58
58
58
57
53
58
58
49
58
16d
57
55
58
58
58
57
58
Estimated
Precision
(Pooled
%RSD)0
21.2
5.0
206.9C
10.6
0.7
2.5
1.3
4.3
5.9
3.0
3.0
74.1
1.4
0.6
43.7
1.3
44.3
2.2
10.7
0.02
0.02
0.02
34.9
1.9
Quantitation
Limit3
0.021
0.026
--
--
0.04
0.09
1.0
0.29
0.25
0.4
0.002
0.04
0.08
0.02
0.02
0.13
0.11
0.068
0.013
--
--
--
0.43
0.11
Pairs that have a
mean > Quantitation
Limit
n
12
34
--
--
58
54
57
50
50
58
57
25
58
58
12
58
1
52
27
--
--
--
47
57
Estimated
Precision
(Pooled
%RSD)
5.3
4.7
--
--
0.7
2.4
1.4
2.9
2.9
3.0
3.0
4.0
1.4
0.6
1.9
1.3
N/A
2.1
6.1
--
--
--
2.5
1.8
All Pairs (mean > 0)
n
89
91
91
91
91
91
91
91
91
88
91
91
91
91
71
91
43d
91
82
91
91
91
91
91
Estimated
Precision
(Pooled
%RSD)b
12.5
1.4
2.2
19.2
0.7
2.2
0.9
7.5
3.0
14.1
2.5
42.3
1.4
0.5
131.8
0.9
108.3
11.0
53.2
0.04
0.04
0.02
18.2
1.8
Quantitation
Limit3
0.007
0.013
--
--
--
0.03
1.8
0.22
0.13
0.5
0.010
--
--
--
--
--
0.03
0.037
0.006
--
--
--
0.08
0.06
Pairs that have a
mean > Quantitation
Limit
n
32
77
--
--
--
91
91
82
91
67
91
--
--
-
--
--
6
71
25
--
--
--
80
90
Estimated
Precision
(Pooled
20.0
1.4
--
--
--
2.2
0.9
7.4
3.0
5.5
2.5
--
--
--
--
--
0.9
3.4
15.8
--
--
--
3.1
1.8
a The quantitation limit is 10 SB (10 times the standard deviation of the calibration or reagent blanks). Quantitation limits are not calculated for ANC, BNC, and
pH because blanks were not analyzed. Nor were blanks analyzed at Laboratory II for Ca, Fe, K, Mg, Mn, or Na.
b Pooled standard deviation values were used for pH measurements.
c If eight pairs that have means < + 3 peq/L are removed from the precision estimate calculation the estimate would be 8.0 percent.
d Reported concentrations of many NH4+ pairs were < 0. As a result, the n for this variable is small relative to the n for other variables.
N/A = not applicable.
-------
Table 25. Summary of Intralaboratory Precision Results by Variable, Western Lake Survey
Phase I
Variables that met
DQO
Al, extractable
Al, total
DQO (%)
10 or 20
10 or 20
10
Variables
that did
not meet
DQO
ANC
Comments
Quantitation limits were not calculated for
Ca
or
Conductance
DIC (air equilibrated)
DIC (closed system)
DIC (initial; open
system)
DOC
F", total dissolved
K
Mg
Na
NH/
P, total
pH (acidity; open
system)
pH (alkalinity; open
system)
pH (air equilibrated)
pH (closed system)
SiO2
SO42'
True Color
Turbidity
10
5
5
2
10
10
10
5 or 10
5
10
5
10
5
5
10
10 or 20
0.05 (pH unit)
0.05 (pH unit)
0.05 (pH unit)
0.05 (pH unit)
5
5
5 (PCU)
10
BNC
ANC. If a quantitation limit of 0 ± 3 neq/L
were applied to the data, the precision
estimate would meet the DQO.
Quantitation limits were not calculated for
BNC; even so, the precision estimate is
close to the DQO.
Fe
Mn
Only the precision estimate for Laboratory II
did not meet the DQO, because a
quantitation limit could not be calculated.
Very low concentrations in most WLS-I
samples (95% of routine lake samples <
0.030 mg/L); only one duplicate pair value
was above quantitation limit.
month duration, which would have been long enough
to produce significant biochemical changes in the 2-
L subsamples. However, the initial homogeneity of
each bulk audit sample lot was ensured by filtering
the bulk sample lot and by storing it in the dark at 4
°C until the 2-L aliquots were prepared. Therefore, it
was determined that having aliquots of all field audit
samples prepared at one place and time by the same
technician (or the same team of technicians) would
provide better overall audit sample consistency than
would preparing aliquots as needed.
When the mean analyte concentrations of the FN4
samples analyzed during WLS-I (see Appendix F)
are compared to those analyzed during the NSS
Phase I Pilot Survey (Drouse, 1987), no significant
difference is observed between the analytical results
from the two surveys. Like FN4, FN3 and FN5 have
64
-------
been used as audit samples in more than one NSWS
survey. The FN3 sample collected from Lake Superior
(see Appendix F), was used in the ELS-I audit
program (Best et al., 1987) and was stored for almost
one year before it was used for WLS-I. The mean
analyte concentrations between the two surveys show
excellent agreement across all analytes, in spite of
the potential laboratory bias and time factors involved.
There are some differences in these audit samples
between surveys, but determining whether time
factors and laboratory bias contribute to these
differences is not within the scope of this report.
Observing differences in analyte concentration for the
same audit sample lot across surveys, however, can
be a useful daily QA tool. FN5, for example, was used
for the NSS Phase I Pilot Survey as well as for
WLS-I, and it showed very good agreement across
surveys. One change, however, was of concern to
the QA staff. The FN5 mean nitrate concentration
during the NSS Phase I Pilot Survey was 0.085 mg/L
(Drouse, 1987). Through daily QA communications
during the WLS-I sample analysis phase, preliminary
analytical laboratory data showed routine lake sample
concentrations of approximately 0.140 mg/L. The
higher concentrations were of concern to the QA staff
because nitrate contamination had been a problem
during ELS-I (Linthurst et al., 1986). A check of field
blank values and of other QA samples did not indicate
systematic contamination from nitrate. Continual
monitoring of FN5 nitrate concentrations indicated a
possible increase in concentration in the bulk sample
over time, but did not indicate contamination. A slight
decrease in ammonium levels indicates that oxidation
of nitrogen may have been responsible for converting
ammonium to nitrate and may account for the
elevated nitrate concentrations. There is no
conclusive evidence, however, that explains why the
NOa" concentrations increased between surveys for
this natural audit sample.
Midway through the survey, the reserve amounts of
natural audit samples were critically low, so a new
Bagley Lake sample (approximately 80 gallons) was
collected, shipped to Radian, and prepared as
aliquots of FN6. Radian also analyzed three FN6
samples before initial shipment to the field bases so
that the EMSL-LV QA staff could compare them to
the field laboratory and analytical laboratory values. It
was assumed that the chemical composition of the
two Bagley Lake samples (FN5 and FN6) would be
similar, even though they were collected during
different seasons of 1985. FN5 was collected in
January and FN6 was collected in September;
therefore, temporal differences between the two
sample types were expected. The concentrations of
FN5 and FN6 are given in Table 26 (later in this
section) and in Appendix E.
Description of Field Synthetic Audit Samples-
The lakes selected as the sources of field audit
samples contain a matrix of analytes considered to be
important in acid precipitation research. It is useful to
select a suite of audits that contain different
concentrations that bracket the predicted ranges of
concentrations for the key variables to be analyzed in
a survey. Because it was difficult to find lakes that
contained all the desired concentration levels of all
the variables measured in the analytical laboratories,
field synthetic audits also were employed during
WLS-I. A synthetic audit is ASTM Type I reagent-
grade water spiked with analytes at a specific
concentration. Field synthetic audit samples simulate
natural lake systems, but the analyte concentrations
in them can be artificially adjusted. Because analytical
results could be compared immediately to the
theoretical concentrations of the analyte, field
synthetic audit samples also gave the QA staff rapid
feedback on analytical performance during the
analytical phase of the survey.
Two low-concentration synthetic field audit lots
(FL11 and FL12) were used in WLS- I. No major
problems were encountered with the use, preparation,
or stability of these audit samples. These samples,
which were ionically balanced to simulate natural lake
water, were prepared by Radian at the predetermined
concentration ranges specified in Silverstein et al.
(1987). The two audits had the same theoretical
concentrations for all analytes.
The four stock concentrates used during WLS-I
comprise one synthetic lot (see Section 2 of Appendix
C for description). Each concentrate volume was
designed to last only as long as the volumes of the
other concentrates of that lot. When the concentrate
volumes were depleted, a new set of stock
concentrates was prepared. The new set was given a
sequential lot number to indicate that it was from
different stock and that it was prepared at a later
date.
Use of Field Audit Samples in Estimating
Precision
At least one field audit sample was to be incorporated
into each batch of WLS-I lake samples, and the
precision of the audits was calculated after all of the
audits from all of the batches had been analyzed. The
cumulative field audit precision estimates are
calculated as %RSD among all the batches for each
audit lot. For field audits, the precision is calculated
for many measurements of a single concentration.
Field audit samples are processed and analyzed in
the field laboratory in the same manner as routine
samples. The precision estimates calculated from
field audit samples are used to estimate the variability
of the field laboratory measurements over time.
Precision also can be estimated by the variability of
the field audit sample in the analytical laboratory. This
65
-------
Table 26. Precision Estimated from Field Natural Audit Samples
Laboratories Pooled), Western Lake Survey - Phase la
Field Natural Lot 3
(FN3, Lake Superior)
Analyzed Among Batches (Analytical
Field Natural Lot 4
(FN4, Big Moose Lake, NY)
Variable (in mg/L
unless noted)
Al, extractable
Al, total
ANC (ueq/L)
BNC (neq/L)
Ca
cr
Conductance (nS/cm)
DIG, air equilibrated
DIG, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH/
N03
P, total
pH, acidity (pH units)
pH, alkalinity (pH units)
pH, air equilibrated (pH
units)
SiO2
so42-
Mean Concen-
tration
0.002
0.012
846.1
21.9
13.84
1.43
95.5
9.90
9.86
1.4
0.035
0.005
0.52
2.90
-0.002
1.36
-0.010
1.418
0.001
7.86
7.85
8.13
2.51
3.24
Estimated Precision
as %RSD (n = 38)
116.1
51.8
5.0
59.5
4.8
6.9
1.9
9.7
10.7
20.6
21.5
155.8
4.4
2.6
439.2t>
2.5
221. 2b
4.7
296.7
0.08C
0.07C
O.OQC
18.3
6.4
Mean Concen-
tration
0.195
0.352
-24.1
119.8
2.10
0.54
32.2
0.32
0.51
8.1
0.074
0.07
0.68
0.36
0.078
0.74
-0.001
2.351
0.002
4.68
4.68
4.70
4.45
6.68
Estimated Precision
as %RSD (n = 37)
31.1
13.2
10.2*>
10.9
3.6
6.1
3.6
80.9
30.4
2.1
11.7
10.2
2.7
1.3
9.9
3.7
1770.4&
4.8
141.4
0.03C
0.02C
0.03C
11.0
5.5
(continued)
estimate includes the effect of sample processing in
the field laboratory. These precision results can be
compared to the DQOs for precision shown in Table
2 (Section 1); however, they should be used only as
a gauge in that comparison of data quality because
the DQOs apply directly to analytical laboratory
performance only and do not apply to the other
components (sources) of variability.
Among-Batch Precision Results
Among-Batch Precision Estimated from Field
Audit Samples Analyzed in the Field Laboratory
Tables E-1 through E-4 in Appendix E show the
precision estimates that are based on all the audit
sample types for each field laboratory. These tables
show the precision estimates separately and pooled
for DIG, pH, turbidity, and true color. The tables also
present the comparable values for DIG and pH from
the analytical laboratory. (Turbidity and true color
were not analyzed in the analytical laboratory.) Across
all six lots of audit samples for all four field
determinations, where population and concentration
were high enough to determine statistical confidence,
the precision for each field laboratory was acceptable.
The pooled values for the five field laboratories show
good precision across audit lots and measurements.
The only exception to good pooled precision is the
true color value for the FN4 Big Moose lake audit
sample; however, the values are sufficiently low that
this imprecision should not be of concern to the data
user.
Among-Batch Precision Estimated from Field
Audit Samples Analyzed in the Analytical
Laboratory
Table 26 presents precision data for the field natural
audit samples and Table 27 presents precision data
for the pooled field synthetic audit samples. It is
legitimate to pool the data for the two field synthetic
66
-------
Table 26. (Continued)
Field Natural Lot 5
(FN5, Bagley Lake, WA,
1st Sampling)
Field Natural Lot 6
(FN6, Bagley Lake, WA,
2nd Sampling)
Variable (in mg/L
unless noted)
Al, extractable
Al, total
ANC (ueq/L)
BNC (iieq/L)
Ca
cr
Conductance (nS/cm)
DIC, air equilibrated
DIG, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH/
NC-3
P, total
pH, acidity (pH units)
pH, alkalinity (pH units)
pH, air equilibrated (pH
units)
Si02
SO-42'
Mean Concen-
tration
0.002
0.010
146.7
37.1
1.99
0.24
17.8
1.83
1.91
0.4
0.025
0.004
0.36
0.24
0.003
1.06
0.011
1.147
0.004
7.00
7.02
7.29
11.37
0.97
a The among-batch precision estimate represents
Estimated Precision
as %RSD (n = 68)
120.4
43.1
3.5
16.1
3.2
11.8
6.4
11.3
13.3
79.2
8.0
177.8
5.5
2.0
351.7
5.2
106.1
23.1
197.2
0.1 QC
0.09C
0.1 3<=
8.6
7.6
the variability introduced
laboratory, and the audit-sample preparation laboratory. It cannot indicate
b The absolute value of the
Mean Concen-
tration
0.006
0.015
121.3
28.5
1.59
0.16
14.2
1.48
1.47
0.2
0.021
0.01
0.29
0.17
0.003
0.81
-0.001
0.016
0.002
7.07
7.09
7.25
9.36
0.63
in the field laboratory,
variability introduced
Estimated Precision
as %RSD (n = 37)
39.1
18.6
1.6
20.3
2.6
5.7
3.5
5.9
7.0
50.4
9.9
75.6
2.9
2.0
261.0
2.1
629.6b
234.2
191.6
0.08C
0.07<=
0.1 5C
7.2
4.5
, the analytical
during sampling.
percent relative standard deviation (%RSD).
c Standard deviation (SD) values were calculated
for pH measurements.
lots (FL11 and FL12) because their theoretical
concentrations are the same. Appendix F (Tables F-
5 and F-6), present the field synthetic audit data
separated by lot. For the field synthetic audits,
separate sets of values for each analytical laboratory
can be evaluated also (Table 27). These laboratory
subsets of among-batch precision indicate whether
or not an analytical method problem was inherent
throughout the survey, or if a problem resulted from
the number of audit samples analyzed by each
analytical laboratory. It is also important to check the
precision of each analytical laboratory separately
because all samples from each subregion were
analyzed by only one of the analytical laboratories.
Bias in one laboratory's measurements, therefore,
could affect population estimates (see Appendix I).
For all audit sample lots, the precision estimates for
most analytes met or were near the DQOs. The
tables in Appendix J provide detailed information
about each audit lot and are useful for the data user
who is interested in the components of variability for
each audit sample.
Table 28 summarizes among-batch precision data
for only those analytes that did not meet the DQOs or
that had concentrations so close to the detection limit
that the precision estimates have little meaning.
It is evident that both laboratories had high variability
in measuring initial and air-equilibrated DIC, DOC,
NH4+ and, at a low concentration, total P. All pH
precision estimates are about 0.1 pH unit. These
estimates did not meet the DQO of 0.05 pH unit for
intralaboratory precision, but they should still be
considered reasonable because the 0.05 pH unit goal
does not apply to field audit sample precision.
Laboratory I's variability contributed significantly to the
higher precision estimate for the pooled Ca, K, Na,
and SiO2 measurements. Laboratory II was the major
contributor to the high variability of the pooled Mn
values. For FL11, Laboratory ll's mean value was
0.077 mg/L; for FL12 the mean was 0.109 mg/L (see
Appendix F). This shows variability over time for
Laboratory II, whereas for Laboratory I the means
were 0.097 (FL11) and 0.092 (FL12). Conductance
imprecision for Laboratory II (5.0%) was greater than
that for Laboratory I (3.3%). For extractable Al, total
67
-------
Al, BNC, and Fe, the concentrations were too low for shows the relationship of precision bias and
determining precision estimates confidently. accuracy in the assessment of data quality. A
A.,HI* oam«iQo A * v * ,, • , , discussion at the end of Section 7 summarizes the
Audit samples and duplicates are used ,n calculating overall performance of the audit sample program and
precision es .mates. Audit samples also are used to discusses suggestions for modifying the use of these
assess mtralaboratory bias and accuracy. Figure 6 samples in future surveys
68
-------
Table 27. Precision Estimated from Pooled Field Synthetic Audit Sample Lots, (Analytical Laboratories Pooled and
Separated), Analyzed Among Batches, Western Lake Survey - Phase la
Laboratories Pooled
Laboratory I
Laboratory II
Variable (in mg/L Theoretical
unless noted) Concentration
Al, extractable
Al, total
ANC (ueq/L)
BNC (jieq/L)
Ca
cr
Conductance
(uS/cm)
DIC, air equilibrated
DIC, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH/
N03"
P, total
pH, acidity (pH units)
pH, alkalinity (pH
units)
pH, air equilibrated
SiO2
SO42-
0.020
0.020
c
c
0.194
0.343
c
c
0.959
1.0
0.042
0.059
0.203
0.447
0.098
2.74
0.168
0.464
0.027
C
c
C
1.07
2.28
Mean
Concentration
0.005
0.027
111.0
30.0
0.22
0.36
19.7
1.44
1.54
1.0d
0.043
0.006
0.21
0.45
0.097
2.77
0.1 6d
0.483
0.025
6.94
6.95
7.24
1.10
2.30
Estimated
Precision
(%RSD) Mean
(n = 47) Concentration
48.0"
29.3
6.1
29.6
16.1
5.6
4.4
18.3
20.4
25.0
7.1
153.3"
8.7
3.8
14.0
8.5
16.2"
6.5
22.0
0.1 3f
0.11f
0.14f
10.5
5.4
0.005
0.020
110.6
21.3
0.25
0.36
19.6
1.26
1.42
1.1e
0.044
0.001
0.22
0.46
0.095
2.78
0.1 5e
0.468
0.025
6.99
6.96
7.19
1.07
2.23
Estimated
Precision
(%RSD)
(n = 17)
32.7"
26.5
6.4
18.8
15.5
7.5
3.3
17.7
22.8
20.9
7.9
465.8"
9.7
3.2
3.4
13.5
21.2"
5.7
18.1
0.13'
0.1 3f
0.08f
14.6
5.0
Mean
Concentration
0.004
0.031
111.3
35.0
0.20
0.36
19.7
1.54
1.60
1.0
0.042
0.009
0.21
0.45
0.098
2.77
0.16
0.492
0.024
6.92
6.94
7.27
1.12
2.34
Estimated
Precision
(%RSD)
(n = 30)
56.2"
20.0
6.0
19.5
4.6
4.4
5.0
15.0
18.3
26.4
5.9
112.2"
6.7
3.9
17.1
3.7
13.5"
6.3
24.2
0.1 2f
0.10f
0.16'
7.5
4.8
a The among-batch precision estimate represents the variability introduced in the field laboratory, the analytical laboratory and the audit-
sample preparation laboratory. It cannot indicate variability introduced during sampling.
" Poor precision may be the result of sample instability, sample mixing error, or both (Best et al., 1987).
c Although theoretical values can be calculated, the theoretical value depends on the concentration of chemicals added to the synthetic
audit sample.
d n = 45.
« n = 15.
' Standard deviation (SD) values were calculated for pH measurements.
69
-------
Table 28. Summary of Analytes That Showed High Variability Among Batches for Field Audit Samples. Western Lake
Survey - Phase I
Variables
Variables Not Within or Near DQO for Precision with Extremely
Audit Sample
FN3
Lake Superior
FN4
Big Moose Lake
FN5
Bagley Lake
(1st sampling)
FN6
Bagley Lake
(2nd sampling)
FL11
(Synthetic)
FL12
(Synthetic)
FL11 and FL12
(Synthetics Pooled)
Laboratories Pooled
DOC, total dissolved
F', SiO2
Extractable Al, total Al,
total dissolved F",
SiO2
Cf, Conductance,
DIC (initial), NO3', pH
(air equilibrated)
pH (air equilibrated)
Ca, DIC (initial and air
equilibrated), DOC, K,
Na, NH4 +
Ca, Conductance, DIC
(air equilibrated), total
P, pH (air equilibrated),
SiO2
Ca, Conductance, DIC
(initial and air
equilibrated), DOC,
NH4 + , total P, pH
(acidity and air
equilibrated), Si02
Laboratory |a
CI", total dissolved
F,' SiO2
Extractable Al, Si02
CI", Conductance,
DIC (initial and air
equilibrated), NO3~,
pH (air equilibrated),
SiO2
—
Ca, DIC (initial and air
equilibrated), K, Na,
NH4 +
Ca, DIC (air
equilibrated), total P,
SI02
Ca, K, Na, SiO2
Laboratory \\a
DOC, total dissolved
F', SiO2
extractable Al, total Al,
total dissolved F", Mn
Conductance, NOs",
pH (air equilibrated),
SiO2
pH (air equilibrated)
DIC (initial and air
equilibrated), NH4 *
Conductance, DIC (air
equilibrated), total P, pH
(acidity, alkalinity, air
equilibrated)
Conductance, Mn
Low Mean
Concentrations4'
Extractable Al, total Al,
BNC,' Fe, Mn, NH4 + ,
total P
DIC (initial and air
equilibrated) , NH4 + ,
total P
Extractable Al, total Al,
BNC, DOC, total
dissolved F", Fe, Mn,
NH4 + , total P
Extractable Al, total Al,
BNC, DOC, total
dissolved F", Fe, Mn,
NH4 + , NO3", total P
Extractable Al, total Al,
BNC, Fe
Extractable Al, total Al,
BNC, Fe
Extractable Al, total Al,
BNC, Fe
a See Appendix F.
*> These variables had concentrations that were too low to allow precision to be compared confidently to the DQO. Not applicable
Tor pri nr)G3sursffl6nts.
70
-------
Section 7
Results and Discussion - Accuracy
Introduction
Accuracy is a measure of the bias in a system. It is
the degree of agreement of a measurement (or, as
used in this report, an average of measurements of
the same variable, X) with an accepted reference or
true value, T. Accuracy usually is expressed as the
difference between the two values, X-T, or the
difference as a percentage of the reference or true
value, [(X-T)/T] 100. This percentage value is used
in this report to relate accuracy to the WLS-I DQOs.
Method of Estimating Accuracy from
Field Synthetic Audit Samples
In the WLS-I sampling design, field synthetic audit
samples were used to estimate accuracy (expressed
as absolute bias) for the analytical laboratory
measurements (see Figures 6 and 7). Although there
may be several ways to analyze the two field
synthetic audit samples for accuracy, there is only
one formula for estimating accuracy as percent, as
specified in the DQOs (see above). For the statistical
analyses presented here, the theoretical concen-
trations of the field synthetic audit samples are
considered to be true values (T); the average of the
measurements for the same variable (X) is the mean
concentration of the synthetic audit samples. [Note:
In the WLS-I QA program, natural audit samples
could not be used in determining accuracy because a
theoretical concentration for each analyte could not
be confirmed.] These synthetic audit samples are as
close in composition to the WLS-I lake water
samples as could be anticipated before sampling
began. The WLS-I field synthetic audits are not
certified standards (e.g., by NBS). In addition, NBS
prepares only certain standards; none covers in a
single sample the entire range of analytes required for
WLS-I.
Accuracy Results Estimated from Field
Synthetic Audit Samples
Because the theoretical concentrations of FL11 and
FL12 are identical (see Appendix C), and because
once each week each field laboratory incorporated
synthetic audit samples into its batches, accuracy
data for the two synthetic audit samples can be
pooled to determine an overall estimate of accuracy
across the survey. The FL11 and FL12 audit stock
concentrates were prepared on different days, about
one month apart. As a result, there may be slight
differences in lot composition that can be attributed to
chemical degradation or to preparation variability.
Because subsets of each audit lot were analyzed by
each analytical laboratory, evaluating the FL11 and
FL12 values pooled by laboratory and separately by
laboratory provides an indication of whether or not
laboratory differences or lot differences significantly
affected accuracy estimates. Accuracy estimates, like
those for precision, depend on sample concentration;
therefore, the data user should observe the
theoretical concentration of the analyte when
assessing the accuracy estimates.
Table 29 presents the estimated analytical accuracy
for FL11 and FL12 synthetic audit samples pooled
and shows the accuracy by laboratory pooled and
separate. [Note: Data for FL11 and FL12 separated
are given in Appendix G.] The only analytes for which
accuracy estimates exceeded their DQOs were Ca at
+ 11.3 percent and total Al at -35.5 percent.
Laboratory I's accuracy of +28.7 percent for Ca,
compared to Laboratory ll's value of +1.6 percent,
identifies the apparent cause of the inaccuracy.
Similarly, the +55.5 percent bias of Laboratory II for
total Al overshadows the + 0.5 percent bias of
Laboratory I. No other pooled data showed this
inaccuracy. However, Laboratory ll's value was
slightly outside the DQO for total P (-11.1%), as
was Laboratory I's value for DOC ( + 13.5%). The
accuracy of DOC suggests another source of
variability: Although the theoretical DOC concentration
is 1.0 mg/L added in the form of CeH4(COOH)2 and
KHCsH4O4, there may have been as much as 0.3
mg/L DOC as background in even ASTM Type I
reagent-grade water used in the synthetic audit
preparation (personal commun. to Silverstein from
David Lewis, Radian Corporation). These background
concentrations were also observed in field blank data
(see Section 8). Table 30 summarizes the analytes
that did not meet the DQOs for accuracy.
Ten variables that were measured in synthetic audits
as part of the WLS-I protocol either have no
71
-------
Table 29. Estimated Analytical Accuracy for Field Synthetic Audit Samples Pooled, Western Lake Survey - Phase I
Laboratories Pooled
Laboratory I
Laboratory I
Variable3
Al, extractable
Al, total
ANC (neq/L)
BNC (neq/L)
Ca
cr
Conductance (uS/cm)
DIG, air equilibrated
DIG, initial
DOC
F", total dissolved
Fe
ro K
Mg
Mn
Na
NH4 +
N03
P, total
pH, acidity (pH units)
pH, alkalinity (pH units)
pH, air equilibrated
SiO2
SO42'
Theoretical
Concentration
0.020
0.020
--
--
0.194
0.343
--
--
0.959
1.0
0.042
0.059
0.203
0.447
0.098
2.74
0.168
0.464
0.027
--
--
--
1.07
2.28
FL11 and FL12
Combined Mean
Concentration
(n = 47)
0.0046
0.0271
111.1
30.1
0.216
0.358
19.7
1.435
1.537
1 .0426
0.0429
0.0059
0.211
0.449
0.097
2.772
0.1 556
0.483
0.0245
6.94
6.95
7.24
1.100
2.300
Accuracy1*
(%)
-77.0
+ 35.5
--
--
+ 11.3
+ 4.4
--
--
+ 60.3
+ 4.2
+ 2.1
-90.0
+ 3.9
+ 0.4
-1.0
+ 1.2
-7.7
+ 4.1
-9.3
--
--
--
+ 2.8
+ 0.9
FL11 and FL12
Combined Mean
Concentration
(n = 17)
0.0054
0.0201
110.6
21.3
0.250
0.356
19.6
1.258
1.424
1.135°
0.0443
0.0012
0.221
0.455
0.095
2.780
0.154
0.468
0.0255
6.99
6.96
7.19
1.067
2.228
Accuracy^
(%)
-73.0
+ 0.5
-
--
+ 28.7
+ 3.8
-
--
+ 48.5
+ 13.5
+ 3.1
-98.0
+ 8.9
+ 1.8
-3.0
+ 1.5
-8.4
+ 0.8
-5.6
-
--
-
-0.3
-2.3
FL11 and FL12
Combined Mean
Concentration
(n = 30)
0.0042
0.0311
111.3
35.0
0.197
0.359
19.72
1.536
1.600
0.955°
0.0421
0.0085
0.205
0.445
0.098
2.770
0.156
0.492
0.0240
6.92
6.94
7.27
1.118
2.342
Accuracy0*
(%)
-79.0
+ 55.5
_.
__
+ 1.6
+ 4.7
__
+ 66.8
-4.5
+ 0.2
-85.6
+ 1.0
-0.3
+ 0.1
+ 1.1
-7.3
+ 6.1
-11.1
_.
-.
__
+ 4.5
+ 2.7
' All variables are measured in mg/L unless otherwise noted. Mean concentrations are presented in as many significant figures as possible for the purpose of calculating accuracy
estimates.
'n = 45.
: n = 15.
A plus sign (+) indicates that the mean concentration was higher than the theoretical concentrations; a minus sign (-) indicates that the mean concentration was lower than the
theoretical concentration.
-------
Table 30. Summary of Variables9 that did not Meet Data Quality Objectives for
Estimated Analytical Accuracy, Western Lake Survey - Phase I
Components of Inaccuracy (Source)
Audit Lot
FL11
FL12
FL11 and
FL12 pooled
Laboratories
Pooled
Total Al, Ca, Mn, NH4 + ,
Total P
Total Al, DOC
Total Al, Ca
Laboratory I
Ca, NH4*, total P
Ca, DOC, SiO2
Ca, DOC
Laboratory II
Total Al, DOC, Mn,
Total Al, Mn, Total P
Total Al, total P
aNot included in accuracy determinations because of sample matrix problems are ANC, BNC,
conductance, DIG, and pH. Not included in accuracy determinations because of sample chemical
instability are Fe and extractabale Al.
theoretical concentration or are subject to inherent
methodological problems that result in poor accuracy
or in the inability to measure for accuracy reliably. Six
of the ten variables (ANC, BNC, conductance, QIC
[air equilibrated], and pH [initial and air equilibrated])
have no reliable, theoretical concentrations, and thus
no reliable accuracy determinations can be made.
The concentrations of each of these analytes
depends on the concentrations and identities of all
the analytes that comprise the audit sample matrix.
ANC, for example, is not spiked into the audit in the
same manner that Ca or other ions are, but the level
of ANC is affected directly by the anions, cations, and
other components of the audit matrix and is
calculated from the acid titrated into this matrix.
Consequently, if the volume of Ca added to the audit
sample was inaccurate, the accuracy of the ANC
measurement (and of other measurements such as
conductance) could be affected.
Initial DIG does have a theoretical concentration
(0.959 mg/L). Accuracy for initial DIG was poor
overall, by laboratory, and by lot. The matrix effect is
a factor for this analyte as well. The theoretical
concentration of the sample is based on the
assumption that the deionized water and all the
additives were pure and were mixed properly;
therefore, the theoretical concentration assumes no
matrix effect. DIG was added to the synthetic sample
in the form of HCOa", as part of the audit sample
preparation protocol (see Appendix C). Equilibration of
the sample also was expected to minimize variability
of initial DIG, but because the sample was not natural
lake water, complete equilibration was difficult to
achieve.
For two of the ten variables, Fe and extractable Al,
accuracy DQOs were not met for either lot or by
either laboratory (Table 30). This is consistent with
the results found in ELS-I (Best et al., 1987) and in
the NSS Phase I Pilot Survey (Drouse, 1987). In the
presence of oxygen, these analytes precipitate out of
solution within 24 hours. Therefore, they were not
totally soluble and were often filtered out or were
adsorbed onto filtrator walls during the aliquot
preparation process in the field laboratory. As a
result, accuracy for Fe and total Al was expected to
be poor. The accuracy has been calculated for these
two analytes, but any results should be considered
with caution.
In summary, the accuracy results exhibit the following
characteristics:
• Where calculation was applicable, overall
accuracy estimates for most analytes were
within the DQOs. Where analytes did not meet
the DQOs, the inaccuracy generally is
attributable to one laboratory's measurement
error.
• Of the 14 analytes for which accuracy can be
calculated reliably, only Ca and total Al are
outside the DQOs when laboratories and audit
sample lots are pooled. In each case, the
inaccuracy can be attributed to one laboratory or
the other, not to both. Laboratory II showed
some measurement variability with Mn and total
Al for both audit samples and with DOC for one
audit sample; Laboratory I showed some
measurement variability with Ca for both audit
samples and with DOC and Si02 for one. Each
laboratory showed a bias for total P for one audit
sample but not for the other; when the audit
sample values were pooled, this bias was within
the DQO for the analyte.
• NH4+ accuracy estimates for both laboratories
met the DQO for FL12; the laboratories had
similarly poor accuracy estimates for FL11
(about -20%). The FL11 inaccuracy may be
the result of NH4* degradation in the lot
overtime; that is, the result of inconsistency in
the sample rather than inconsistency or
inaccuracy in the analytical procedure (see
Appendix G). When the NH4+ values for the
FL11 and FL12 were pooled, however, overall
accuracy was within the ±10 percent DQO.
73
-------
• For 8 of 24 variables (ANC; BNC; air-
equilibrated and initial DIG; acidity, alkalinity, and
air-equilibrated pH; and conductance) accuracy
cannot be reliably calculated because the
analytical methods for these variables are
subject to matrix effects.
• Two variables, Fe and extractable Al, are subject
to physicochemical problems inherent in the
analytical method; thus, reliable accuracy
calculations cannot be derived from field
synthetic audit data for these variables.
Summary of Audit Sample Data for
Precision and Accuracy
Data for all audit sample lots (FN3, FN4, FN5, FN6,
FLU, and FL12) contribute to the understanding of the
precision, accuracy, and laboratory biases associated
with the WLS-I data base. When these data are
used to help interpret regional and subregional
characteristics or to characterize one lake or a subset
of lakes sampled, the data user must consider the
proposed use of the data. For example, the data user
interested in precision at a certain concentration
range must consider that concentrations and
precision estimates are specific to the audit sample
type and to each analyte in that audit sample. Thus,
the user should determine which analytes are of
interest, should evaluate the mean sample
concentration of the field audit lots, should assess the
analytical laboratory involved in a particular subregion,
and should select for review those audits that cover
the concentration range of interest. The selected
subset of field audit data then can be used in
conjunction with the accuracy data provided by the
synthetic audits (also at the concentrations of interest
only) to yield a degree of confidence for a particular
subset of the data.
The confidence assigned to subregional population
estimates can be supported by the audit sample
results. In any application of audit data, however, the
user must be aware that there are nuances
associated with precision estimates for laboratory
data, whether the data are pooled or separated by
laboratory. Often, where precision estimates are well
outside the DQOs, one or a few outlying sample
values are responsible. For example, of the 68 FN5
samples, 44 were analyzed in Laboratory II. The
%RSD for the 44 samples analyzed for NOa" is 26.5
percent at a mean concentration of 0.151 mg/L.
When four unusually high sample values are removed
from the %RSD calculation, the n of 40 yields a
%RSD of 9.6 percent. This example illustrates the
profound effect that a few unusual values may have
on the overall precision of the audit lot analyzed in
one of the analytical laboratories. In this example,
removing 9 percent (4 of 44) of the audit samples
from the population improves the precision from 26.5
percent to 9.6 percent. This illustration suggests that
the four data points that were removed did yield
unusual results. The data user may wish to identify
such data as outliers by applying appropriate
statistical tests when analytes yield high precision
estimates.
Specific examinations of data subsets, such as the
NOa" example given above, will continue to generate
questions concerning the WLS-I data base:
• What are the causes of outlier values?
• Was contamination introduced by the analytical
laboratory, by the field laboratory, by the audit
sample supplier laboratory, by the supplier of the
aliquot bottles, or by a combination of these
components?
• Does this variability affect the primary survey
goal of subregional lake characterization?
Audit data alone cannot answer these questions. Field
duplicate pairs, which are the only QA samples that
reflect all components of system variability that can
affect the routine lake samples, must be used to
estimate overall system variability. Audit sample data,
however, can isolate variability that is related to the
controlled environment of the laboratory from the
variability associated with the field sampling
component (sampling procedures and uncontrollable
lake-site environmental factors such as high winds).
Audit samples cannot measure system variability
because they are not collected at each lake site and
they are not processed through the Van Dorn
samplers as are field blanks and field duplicate pairs.
Consequently, field audits are most useful in
identifying method and daily analytical problems
related to the field laboratories and analytical
laboratories, in detecting and quantifying laboratory
bias, and in estimating the accuracy of the analytical
measurements with the aid of synthetic audits.
Thus, the WLS-I QA audit samples can be used to
make only certain inferences about data variability,
and these inferences are limited to the realm of
analytical measurements. The usefulness of audits in
the daily QA and data verification aspects of the
program, however, is certain. Initially, the field audit
samples gave the QA staff immediate feedback on
daily performance in the field laboratory and in the
analytical laboratory. Monitoring daily laboratory
performance through telephone calls and obtaining
hard-copy, raw data results for audit samples
identified trends or problems in sample analysis,
sample handling, and data reporting. Subsequent
evaluation and statistical analysis of the full suite of
audit sample results provided a basis for determining
whether or not requests for reanalysis of sample
batches were necessary. Evaluation of the FN5 audit
sample data, for example, identified a silica dilution
74
-------
error: a suspicious trend indicated by a few samples
resulted in value changes for 80 samples from one
analytical laboratory; 60 of these samples had dilution
calculation errors and 20 were reanalyzed because of
incorrect dilution procedures.
For the analytical laboratory audit sample results, 168
precision estimates could be calculated (7 audit lots
times 24 variables per audit lot). Of the 168
estimates, 90 percent (1) were reasonable in
comparison to the survey precision goals, (2)
represented a mean concentration that was too low to
provide a meaningful estimate, or (3) had high %RSD
values as a result of one or a few aberrant sample
values. In cases where the aberrant sample values
may have influenced the %RSD greatly, statistical
outlier tests should be applied. Only 17 of the 168
estimates did not fit one of these three categories; 7
of the 17 were pH, DIG, and DOC determinations for
synthetic audits and reflected sample instability or
analytical problems. The NH4+ measurement for FLU
indicated decomposition over time, which is
confirmed in poor precision and accuracy. Ca
exhibited similar precision and accuracy estimates for
FLU. Precision for K was less than desirable for FLU,
but accuracy was acceptable ( + 6.4%) at a
concentration of 0.22 mg/L.
Only six variables among the four lots of field natural
audits exhibited high variability (imprecision) that was
difficult to explain; however, accuracy cannot be
determined reliably from natural audit sample data.
For FN5, initial and air-equilibrated DIG,
conductance, and NO3" were highly variable (see
Table 26). For conductance, the 6.4 %RSD is well
above the intralaboratory precision DQO; however,
because the standard deviation is only 1.1 jiS/cm, a
high precision estimate should be of little concern to
the data user. For FN4, extractable Al values ranged
from 0.106 to 0.315 mg/L, which is a large spread in
the data for a sample size of 20. The sources of
variability are not known, but they could be
contamination, poor laboratory technique (i.e.,
extraction), problems with instruments or problems
with methods. Fortunately, 99 percent of the WLS-I
lake samples collected had extractable Al
concentrations below 0.050 mg/L; therefore,
imprecise measurements at the extractable Al
concentrations found in FN4 should not be of concern
to the WLS-I data user. Of all the analytes, SiO2
appears to be the most variable. Si02 values for FN3
and FN4 had less than desirable precision (see Table
26), and the FL12 accuracy value for Laboratory I is
slightly outside the desired range (see Appendix G,
Table G-2).
Field audit precision data for the field laboratory
determinations of pH, DIG, true color, and turbidity
indicate no systematic problems in any of the five
field laboratories. In most cases, even the pH and
DIG precision estimates for synthetic audits are
acceptable, which indicates that the variability of
these measurements for synthetic audits increases
with time; specifically, the difference can be attributed
to the time elapsed between analysis in the field
laboratory and in the analytical laboratory.
General conclusions that can be drawn regarding the
audit sample data are as follows:
1. The field audits are essential to daily QA
operations and laboratory monitoring, and they
provide evidence on which to base requests for
reanalysis.
2. Most of the precision estimates are at or near the
DQOs for precision, are too low in concentration
to allow precision to be estimated reliably as
%RSD, or had one or a few values that were
responsible for the poor precision estimates.
3. The analyte that shows the most variability in
precision and in accuracy is SiO2- Whether this
variability is attributable to method, procedure
(e.g., poor digestions), or contamination is not
known. Extractable Al also shows poor precision,
but levels of extractable Al were extremely low in
WLS-I lake samples.
4. Synthetic audits are useful in estimating accuracy
and precision, but the precision and accuracy
values for some synthetic audit variables may not
be reliable. DIG and pH seem to be unstable over
time, even after equilibration; conductance, ANC,
BNC, and DOC also have inherent problems
when theoretical or true values are determined.
Precision and accuracy for Fe and extractable Al
are in question because of the instability of these
analytes in the audit solution. The accuracy data
for the remaining 14 analytes are acceptable
overall when compared to the DQOs; however,
total Al and total P are exceptions (see Table 29).
The inaccuracy estimated for these two analytes,
however, may be a function of their low
theoretical concentrations. Ca values suggest
inaccuracy for one laboratory but not for the
other, and not for the data pooled. This evidence
may relate to a laboratory bias problem.
5. The results and conclusions regarding relative
interlaboratory bias are discussed in Permutt et
al. (Appendix I of this report). In general, these
results indicate statistically significant bias
between analytical laboratory measurements for
most analytes. An analyte-by-analyte
inspection of the audit sample data (Appendix I)
shows that it is difficult to interpret these biases in
terms of quantifying the differences over the
range of concentrations for the routine samples.
In many cases, the percentage of bias between
the analytical laboratories is large only because
the mean analyte concentration is small for the
75
-------
audit lot. In other cases, the percentage of bias is
different at different mean concentrations. In
some cases, the percentage of bias is different at
the same mean concentration; and, in some
cases, one laboratory is biased high at one audit
lot mean concentration and low at a different
mean concentration. In addition, one laboratory
may have analyzed a larger percentage of the
audit sample lot population, thus weighting the
results. Depending on the analyte, such situations
can confound the ability to quantify interlaboratory
bias. Because these biases are relative and often
are lot-specific, data calibration between
laboratories was not performed. The WLS-I QA
audit sample program was not designed to do this
a priori. This report provides the statistical data
(Appendix I) in the event that the data user
wishes to assess biases for individual laboratories
or for analytes within specific concentration
ranges or subregions.
6. Audit sample preparation appears to have added
only minimal variability to the analytical precision,
and the procedure of preparing natural audit
aliquots en masse appears to be effective;
however, within the scope of the QA program,
there is no effective way of quantifying the
contribution that audit sample preparation makes
to variability. Designers of future surveys may
want to inchde a mechanism for determining
such variability. Accuracy calculated from
synthetic audit data indicates that the stock
concentrates were prepared properly and that the
subsequent dilution procedure was performed
correctly.
7. Field laboratory performance shows negligible
variability across all audit lots. The estimated
precisions for DIG and pH were as good as or
better than the counterpart measurements in the
analytical laboratories. This observation regarding
DIG and pH indicates that time, increased
handling, and exposure to the atmosphere may
have a small but detectable effect on the amount
of CO2 in a sample, especially for the synthetic
audits. Measuring precision for true color and
turbidity is hindered by inherent problems.
Because turbidity is measured on unfiltered
samples for lake water analysis, a direct
comparison (with the use of filtered natural lake
water audits or with synthetic audits using
reagent-grade water) raises questions about the
validity of the precision estimates. Duplicate
sample pairs are a more appropriate tool for
estimating the precision of turbidity
measurements. Meaningful quantification of the
true color precision estimates was hindered by
the fact that only one audit (FN4) has a mean
value above 5 PCU, at about 20 PCD. Because
color determinations are quantified in increments
of 5 PCU, variability of 1 or 2 increments can
indicate poor precision when, in fact, the
incremental precision is good. With this in mind,
precision of the true color measurements was
reasonable in comparison to the DQO for this
determination.
A key concept is that precision is dependent on
concentration. Once the concentration levels of
interest for each variable have been established, the
composition of the audit samples must be scrutinized.
The precision for a variable from one audit sample lot
can be used in conjunction with the precision for the
same variable from a different lot if the concentration
levels carry different importance for specific
applications. Combining this information with the
precision estimates provided from duplicate sample
analysis gives a more complete picture of overall data
quality (see Appendix J).
Pooling the data from the analytical laboratories is
useful for an overview of survey analytical precision.
Looking at the precision separated by laboratory, on
the other hand, is helpful in quantifying bias. For
example, Laboratory I analyzed most of the samples
from subregions 4D and 4E, and Laboratory II
analyzed the samples from subregions 4A, 4B, and
4C. If the audit samples showed significant
interlaboratory bias, the bias would indicate biases in
the subregional population estimates as well. No
adjustment was made for bias in part because of
some limitation in the data to do so (see Permutt et
al., Appendix I). A synthetic audit program that
employs samples that are more representative of the
entire routine sample concentration range is needed
to better quantify the interlaboratory bias in terms of
accuracy (the deviation from a theoretical or true
value) so that the data can be adjusted confidently.
This process can be accomplished by varying analyte
concentration in the synthetic aduit sample lots so
that they represent the range of routine sample
concentrations. Natural audit samples cannot be used
for this purpose because they can only be used to
estimate relative biases. The more intensive programs
implemented in subsequent NSWS programs (i.e.,
ELS-II Fall Chemistry Survey) employ synthetic audit
sample lots that have different concentration
increments for each analyte to cover the expected
range of routine samples. These synthetic audits are
laboratory audit samples, which are more useful in
assessing interlaboratory bias in absolute terms than
are field audit samples.
Field audit samples include the variability introduced
in the field laboratory as a result of sample
processing. In WLS-I, the field laboratory variability
was spread among five field laboratories. Laboratory
audits, which were not used in WLS-I, were not
subject to the field laboratory variability; therefore,
they were more appropriate for determining analytical
laboratory bias. Laboratory synthetic audits also did
not exhibit instability with extractable Al and Fe, as
76
-------
field synthetic audits did, because the holding time
between sample preparation and preservation was
reduced from about 24 hours to 2 hours (Best et al.,
1987).
The final step in evaluating the success of the audit
sample programs used in ELS-I and WLS-I is to
establish DQOs specific to these QA samples. These
DQOs can be determined only when the needs of the
individual data user are known. These needs are
based on the answers t o several questions: Do the
data generated from the audit programs suit the
needs of the data user? That is, are the precision and
accuracy adequate for the intended data
interpretations? If the precision and accuracy are
adequate, are they adequate at all the concentrations
of interest, or are "sliding-scale" DQOs required for
different concentration ranges? If the precision and
accuracy are not adequate, what methodological or
procedural changes are necessary to produce data of
the necessary quality?
77
-------
-------
Section 8
Results and Discussion - Detectability
Introduction
To answer questions related to detectability and
sample contamination, the WLS-I sampling design
employed three types of blank samples: field blanks,
trailer blanks, and calibration or reagent blanks (see
Figure 3 in Section 3). Data from these blanks were
used to establish three major statistical limits: the
system detection limit, the system decision limit, and
the instrument detection limit. Together, the blank
data and these limits, calculated from that data,
provide information on detectability. A fourth statistical
limit, the quantitation limit, can be derived from blank
data as well, but this limit is used in estimating
precision (see Section 6). Plots showing blank sample
data results for each analyte are given in Appendix J.
All types of blank measurements are compared to the
required detection limits (see Table 2), which are the
DQOs for laboratory detectability. The required
detection limit is the highest instrument detection limit
allowable in the analytical laboratory contract.
Laboratory blanks, which are used to calculate
instrument detection limits, are the only blank
samples that apply directly to required detection limit
criteria. The relation of blank sample data to the
associated statistical limits and the required detection
limits is discussed below and is illustrated in Figure
10. The discussion also shows how blank sample
data relate to lake sample data.
Method of Estimating System
Detectability from Field Blank
Measurements
During WLS-I, 236 field blanks were used in
estimating cumulative (system) background noise and
contamination levels that were inherent in WLS-I
sampling and analytical methods (i.e., all the
components of variability or contamination that can
affect a routine lake sample from the time it is
collected until the final data are reported; see Figure
3). Contamination is most often caused by the
extensive sample handling that field blanks (and lake
samples) undergo.
Field blanks also can be used to detect positive and
negative bias that results from analytical drift
associated with poor instrument calibration. A
negative instrument response derived from a field
blank or routine sample directly relates to analytical
instrument calibration, not field contamination (except
for ANC and BNC). Field blanks, however, cannot
indicate degradation of an analyte in a water sample
(e.g., precipitation or oxidation-reduction); it is
necessary to rely on field duplicate pairs and on field
audit samples (see Sections 6 and 7) for this
purpose.
System Decision Limit
One method of evaluating analytical detectability and
levels of contamination in the field samples is to
calculate the system decision limit, which uses a
nonparametric statistical analysis of all the WLS-I
field blank samples. The system decision limit is
defined as the 95th percentile (Pgs) of the distribu-
tion of field blank values (see Permutt and Pollack,
1986, Appendix A in Best et al., 1987; see also Figure
10). The system decision limit provides an estimate
of the level of an analyte that potentially can be
introduced during sample collection, handling,
processing, and analysis. For measured values below
this limit, it cannot be known with certainty (i.e., 95%
confidence) whether the analyte was present in the
lake or was introduced at some stage of handling.
Thus, when the analyte concentration of a routine
sample is at or below the system decision limit, it
cannot be distinguished confidently from the system
background shown in the field blanks. Analyte at a
concentration above the system decision limit is not
system background noise. The system decision limit
for each WLS-I analyte, based on the analysis of the
236 WLS-I field blanks, is given in Table 31. Field
blank measurement data are presented by analytical
laboratory in Appendix D, Table D-1, and are
illustrated in Appendix J.
System Detection Limit
The system detection limit is the highest
concentration of an analyte that could be present in a
lake water sample in which the analyte was not
detected. Any measured concentration less than the
system decision limit should be considered "not
79
-------
Figure 10. Relation of statistical limits to data derived from blank samples. Western Lake Survey - Phase I.
«J
4-* Q)
$ C 01
> ":.?:-; Limit'.•/••/.':;
Instrumental
£:£ Drift jrjV
'Acceptable:
Unacceptable'
Quantitation
Limit
[10 SB]
Fora Given Analyte
the System Detection
Limit May be Above
or Below the System
Decision Limit
System
——— Decision
Limit
[Pad
Required
1 Detection
Limit
Negative of the
Required Detection
Limit
Any Sample Lab Blank
Field Blank
80
-------
Table 31. Required Detection Limits, System Decision Limits, and
System Detection Limits for all Variables, Western Lake
Survey - Phase I
Variable3
Al, extractable
Al, total
ANC (iieq/L)
BNC (iieq/L)
Ca
cr
Conductance
(nS/cm)
DIG, air equilibrated
DIG, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH4 +
N03"
P, total
pH, acidity
pH, alkalinity
pH, air equilibrated
Si02
so42-
Required
Detection
Limit
0.005
0.005
10.0
10.0
0.01
0.01
c
0.05
0.05
0.1
0.005
0.01
0.01
0.01
0.01
0.01
0.01
0.005
0.002
N/A
N/A
N/A
0.05
0.05
System System Detection
Decision Limit
Limit (P95)b 2 (P95 - P50)&
0.004
0.019
3.9
29.8
0.07
0.04
1.6
0.33
0.43
0.3
0.003
0.01
0.01
0.01
0.01
0.01
0.01
0.071
0.006
N/A
N/A
N/A
0.18
0.07
0.006
0.020
5.2
22.2
0.12
0.08
2.0
0.26
0.50
0.3
0.004
0.03
0.02
0.01
0.03
0.03
0.02
0.126
0.010
N/A
N/A
N/A
0.26
0.11
a All variables were measured in mg/L unless otherwise noted.
t> Pg5 is the 95th percentile of 236 field blank measurements; P&Q is the 50th
percentile of 236 field blank measurements.
c The mean of six nonconsecutive blank measurements was required to be <
0.9 pS/cm.
N/A = not applicable.
detected." The true concentration, in such a case,
could be as low as zero and as high as but no higher
than the system detection limit at the 95 percent
confidence level. The system detection limit is
calculated as 2 (Pgs-Pso), where P$Q is the 50th
percentile of the distribution of field blank
measurements. Permutt and Pollack (1986) in Best et
al. (1987) discuss further the statistical basis of the
system detection limit calculation. System detection
limits for WLS-I variables are given in Table 31. The
system decision limit (Pgs) is most useful in
estimating background contamination. The system
detection limit, however, can aid the data user in
determining whether or not background contamination
greatly affects precision (Section 6) and accuracy
estimates (Section 7) for audit lots that have low
analyte concentrations.
Detectability Results Estimated from
Field Blank Measurements
Results of the field blank analyses indicate that there
was no significant contamination for any variable that
would affect population estimates; however, each
data user must assess the contamination levels to
suit the specific purpose. Random contamination as a
result of sampling, processing, or analytical error may
have caused the system decision limit to exceed the
required detection limit for some variables. For most
variables, the system decision limit was less than or
near the required detection limit. Variables for which
the system decision limit was higher than the required
detection limit include total Al, BNC, CI", initial and
air-equilibrated DIG, and DOC. For these variables,
however, the system decision limits are comparable
81
-------
to the field blank values from ELS-I (see Best et al.,
1987).
For three other variables (Ca, NO3~, and SiO2), the
system decision limits were outside the acceptable
ranges. The system decision limit for NOa" was
0.071 mg/L. The system decision limit for Ca was
0.07 mg/L; for SiO2 it was 0.18 mg/L. Therefore,
there is at least a 5 percent chance of introducing
contamination at concentrations above these limits for
NOa', Ca, and SiO2-
Possible explanations for the high background
concentrations for these three analytes include:
• High concentrations of Ca and SiC-2 could have
been caused by (1) incomplete rinsing of the
sampling apparatus, (2) incomplete rinsing of the
filtration apparatus, or (3) instrument response
(carryover) from one sample to the next.
• Nitrate concentrations (a mean of 0.105 mg/L) in
WLS-I routine lake samples were relatively low.
Therefore, carryover was probably not a
significant factor in field blank sample
concentrations. There is an indication that
elevated NOa" concentration levels might have
been introduced in the field laboratory (see
Appendix J, Figure J-19a).
The concentrations of Ca and Si02 in the WLS-I
lakes (each with a mean of about 3.7 mg/L) indicate
that field blank sample contamination (0.07 and 0.18
mg/L, respectively) was relatively insignificant;
therefore, the effect that the background
concentrations have on population estimates should
be negligible. Nitrate, in contrast, was not abundant in
WLS-I lake samples; therefore, the high N0a~ field
blank sample concentrations could be significant in
relation to the routine lake samples. Since NOa" was
not a major contributor to the anion sum, the
background concentration levels should be
insignificant in determining population estimates. The
significance of these concentrations, however, should
be assessed by the individual data user (see
Appendix K).
Comparison of Results for Field Blank Samples
Collected by Helicopter Crews and Ground
Crews
The differences between the sampling methods used
by the helicopter crews and the ground crews were of
great interest and concern. (See Section 9 for a
discussion of calibration study results.) Also of
interest were ways in which contamination levels
introduced into the lake samples might have differed
between methods. Table 32 shows the mean and
standard deviation of the field blanks collected by the
helicopter crews and the ground crews. The data
indicate that there was no practical difference in the
level of contamination or in the variability in blank
collection between methods for any analyte
measured. The data are also consistent with the
results for system detectability for all field blanks
combined. The comparability of blank values for the
two sampling methods is excellent considering the
number of individuals who were involved in collecting
samples (60 ground crews and 7 helicopter crews).
Therefore, regardless of sampling method, the blank
collection procedure appears to be highly efficient.
Method of Estimating Detectability from
Trailer Blank Sample Measurements
Trailer blanks were not used regularly as QA samples
in WLS-I. Although useful information on analytical
performance can be obtained from the standard use
of trailer blanks, for WLS-I these samples were used
only when deviations from the normal sampling and
batch design occurred (e.g., when the ground crew
did not collect a field blank and the helicopter crew
did not sample on that day). For comparison, 236
field blanks were collected during WLS-I; only 22
trailer blanks were used. Because WLS-I employed
only 22 trailer blanks, separate statistical criteria were
not used for field blanks and trailer blanks when
trends and systematic contamination were evaluated
during data verification.
Trailer blank samples have a purpose similar to that
of field blank samples. The difference between the
two blank sample types is that contamination levels
detected from trailer blanks do not include the effects
that the entire system has on variability because they
originate at the field laboratory rather than at the lake
site. Consequently, they are not carried through any
of the steps in the sample collection procedure and
are expected to have lower background levels. Any
statistical conclusion derived from the trailer blanks
relates solely to analytical detectability, i.e., variability
contributed by field laboratory and analytical
laboratory activities combined. It follows that any
negative response is an instrument calibration
problem (bias) rather than a contamination problem
because negative analyte contamination cannot exist
(except for ANC and BNC).
Detectability Results Estimated from
Trailer Blank Sample Measurements
The results for trailer blank sample analyses are
presented in Appendix D, Table D-2 and in Appendix
J as the median (Pso) and 95th (Pgs) percentile of
the measurements. Trailer blank samples indicated
background contamination above the required
detection limit for six of the variables studied. Slight
contamination was indicated for BNC, Ca, and total P.
However, the only analytes that showed systematic
contamination at levels well above the required
detection limit were CI", NOa", and SiO2- For CI",
95 percent of the lakes sampled had analyte
concentrations above the Pgs of the trailer blanks.
82
-------
Table 32. Evaluation of Field Blank Data by Sampling Method, Western Lake Survey - Phase I
Field Blank Samples Collected
by Helicopter Crews
(n = 124)
Field Blank Samples
Collected by Ground Crews
(n = 112)
Mean
Variable3 Concentration
Al, extractable
Al, total
ANC(neq/L)
BNC(ueq/L)
Ca
cr
Conductance
(liS/cm)
DIC, air equilibrated
DIG, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH/
NOV
P, total
pH, acidity (pH units)
pH, alkalinity (pH units)
pH, air equilibrated (pH
units)
SiO2
SO42"
0.000
0.009
1.1
17.4
0.022
0.008
0.7
0.22
0.18
0.15
0.001
0.003
0.002
0.002
-0.001
0.001
-0.005
0.015
0.001
5.67
5.66
5.72
0.08
0.02
Standard
Deviation41
0.0022
0.0099
1.92
8.00
0.028
0.012
0.55
0.077
0.093
0.24
0.00097
0.0071
0.0085
0.0026
0.011
0.022
0.011
0.033
0.0040
0.058
0.053
0.151
0.125
0.040
Required
Detection
Limit
0.005
0.005
5.0
5.0
0.01
0.01
0.9
0.05
0.05
0.1
0.005
0.01
0.01
0.01
0.01
0.01
0.01
0.005
0.002
-
-
--
0.05
0.05
Mean
Concentration
0.001
0.008C
1.2c
17.7C
0.026
0.014
0.8
0.19
0.22
0.15
0.001
0.002
0.002
0.002
-0.002
-0.001
-0.007
0.018
0.001
5.66
5.63
5.70
0.076
0.030
Standard
Deviation^
0.0020
0.0067
3.02
10.47
0.032
0.028
0.67
0.058
0.161
0.14
0.0021
0.0070
0.0077
0.0042
0.0096
0.0086
0.014
0.035
0.0046
0.084
0.079
0.058
0.271
0.070
a All variables were measured in
b Although nonparameteric tests
standard deviations are useful
en = 111
mg/L unless otherwise indicated.
are useful in determining contamination effects on samples, means and
in comparing the sampling ability of one method with that of the other method.
SiC>2 was found at relatively high concentrations in
the routine lake samples. Therefore, the slight
contamination should have little effect on the routine
sample concentrations or on subsequent population
estimates for CI" and for SiO2- The contamination of
NOa" in the field and trailer blanks may have
resulted in part from field laboratory activities. This is
plausible, because HNOs is used as a standard
laboratory preservation and cleaning reagent. Some
aerosol or vapor generated from processing
laboratory procedures may have affected some blank
measurements.
Method of Estimating Detectability from
Calibration Blank and Reagent Blank
Sample Measurements
The third type of blank sample employed in WLS-I
was the calibration blank. Analytical laboratory
calibration blank analyses are useful in determining
instrument performance capability and instrumental
drift. Their use is limited to evaluating performance in
the laboratory only. Calibration blanks were analyzed
after daily instrument calibration and before analysis
of lake water samples as a check on instrumental
83
-------
drift. In the field laboratory, calibration blanks were
used in the calibration of the carbon analyzer. In the
analytical laboratory, calibration blanks were analyzed
for every variable except ANC, BNC, air-equilibrated
DIG, and all pH measurements. Digestion procedures
were performed before analysis of total Al and SiO2;
this procedure required that reagent blanks be
analyzed and used in the same capacity as calibration
blanks.
Determining Instrument Detection Limit
Calibration and reagent blanks were used in two
facets of analytical instrument detection. The first
facet, dictated by the SOW, required the analytical
laboratory to determine and report instrument
detection limits at periodic intervals during the survey.
This exercise consisted of analyzing 10
nonconsecutive, replicate calibration blanks, then
determining the value for three times the standard
deviation of the 10 measurements. These 10
measurements were taken on one day (i.e., on one
calibration curve). The result had to equal or be less
than the required detection limit (see Table 31 and
Figure 10). Drouse et al. (1986) provide a detailed
discussion of the instrument detection limit
calculation. The instrument detection limit is useful in
determining the lowest possible concentration at
which an instrument can detect the analyte. Attaining
instrument performance at this level indicates that
contamination introduced at the analytical laboratory
could be minimal if the conditions under which the
instrument was calibrated remained constant.
(Appendix D presents the instrument detection limit
data.)
The second facet required that the analytical
laboratory assess the laboratory blank data by
observing the distribution, the median (Psn, 50th
percentile), and the 95th percentile (Pgs) of the daily
calibration blank data. These data characterize the
distribution of calibration blanks that were analyzed
day-to-day, once per batch, during the course of
the survey. Comparison of these data to the required
detection limit differs from similar comparisons for the
instrument detection limit in two ways: (1) each
calibration blank was analyzed on a different
calibration curve; day-to-day variability is expected
to be higher than the variability of blanks analyzed on
the same curve, and (2) the concentration for each
calibration blank used in daily instrument calibration
could be as high as two times the required detection
limit according to the SOW.
Like the measurements of field and trailer blanks, the
Pgs of the calibration blank measurements can alert
the data user to contamination or instrumental drift
that could affect the routine lake sample
concentrations. The Pgs also provides detectability
data that eliminate the effects of sample collection,
processing, and shipping on the routine samples.
Because field blanks were subject to more handling
(and thus to more sources of error) than were
calibration blanks, variability in field blanks was
expected to be higher than in calibration blanks. The
data user may find it informative to refer to the Pgs of
the calibration blanks as an "analytical decision limit"
in a comparison with the Pgs of the field blanks (the
system decision limit). (See Appendix J for a
comparison of the distribution of all blank sample
types.)
Detectability Results Estimated from
Calibration and Reagent Blank Sample
Measurements
In all cases for all analytes, the DQO for instrument
detectability was met by each WLS-I analytical
laboratory. The instrument detection limit was
consistently at or below the required detection limit,
which indicates that the analytical laboratory
instrumental response did not contribute significantly
to the background levels, and, therefore, had little or
no effect on the routine samples.
The concentration levels of the daily calibration blanks
from each analytical laboratory were always within the
criteria (two times the required detection limit) set
forth in the SOW. This indicates that daily instrument
calibration resulted in negligible background
contamination.
Matrix Spike Sample Results
A final component of detectability in the QA analysis
of WLS-I data is the evaluation of matrix spike
samples. The criterion for the matrix spike QC check,
which was applied to 15 variables, was 100 ± 15
percent spike recovery, which was calculated as
follows:
concentration of spiked sample — spike concentration
original sample concentration
100
The overall results (Table 33) indicate that matrix
effects produced minimal, if any, interference with
routine sample analysis. This indicates that sample
matrix did not affect instrumental detection of
analytes.
84
-------
Table 33. Results of Matrix Spike Percent Recovery Analysis9, Western
Lake Survey • Phase I
Variable
Al, total
Ca
cr
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH4 +
N03"
P, total
Si02
SO42"
Number of
Batches for
which Criteria Not
Met (n = 149)
1
2
0
0
0
0
0
0
7
0
8
1
0
0
0
% Of
Batches
<1%
1.5%
0
0
0
0
0
0
4.6%
0
5.4%
<1%
0
0
0
Number of
Samples
Affected
(n = 1,642)
10
25
0
0
0
0
0
0
117
0
102
17
0
0
0
% of
Samples
<1%
1 .5%
0
0
0
0
0
0
7.1%
0
6.2%
1 .0%
0
0
0
a Matrix spike recovery analysis was applied only to the 15 variables listed above.
85
-------
-------
Section 9
Special Studies
Calibration Study
Introduction
As a part of the overall WLS-I sampling and
analytical strategy, a subset of WLS-I lakes was
sampled in a calibration study. The study included 50
of the 455 wilderness-area, roadless-area, and
national park lakes that were selected for study in
WLS-I and that were targeted to be sampled by
ground crews only. The 50 calibration study lakes
represented a random sample of the WLS-I
wilderness-area lakes. Legislation restricts activities
that jeopardize the pristine character of wilderness
areas, and considerable precedent has been
established to limit helicopter and other motorized
access to such areas. However, because information
obtained from WLS-I might be of great help in
long-term maintenance of wilderness characteristics,
the Forest Service approved helicopter access to the
50 lakes so that the established sampling method
(helicopter access) could be compared to the new
method (ground access). A detailed discussion of
calibration study lake selection can be found in
Landers et al. (1987).
Data derived from the chemical analyses conducted
during the calibration study were used to perform
calibration by linear regression (see Appendix A in
Landers et al., 1987). These calibration data were
intended to be applied to analytical values reported
for all WLS-I samples that were collected by the
ground crews. The calibrations by regression were
designed to eliminate value differences that resulted
from variations in sampling protocol, sample holding
time, or laboratory bias. The regression analyses and
the significance of those analyses are discussed in
Landers et al. (1987); the goals and design of the
calibration study are presented in Silverstein et al.
(1987) and are summarized below.
The calibration study was designed to meet three
goals:
1. Detect differences between two sampling
methods.
2. Evaluate the effects of holding samples for
different lengths of time before they were
processed (preserved) and analyzed.
3. Detect interlaboratory bias between the two
contract analytical laboratories that analyzed
WLS-I samples.
Sampling Design
Comparison of Sampling Methods-
For the WLS-I calibration study, an established
sampling method (helicopter-access sampling
protocols used during ELS-I) was compared to a
new sampling method (ground-access sampling
protocols not previously tested for NLS). Each
calibration lake was sampled by one helicopter crew
and by one ground crew. The two crews collected
samples from approximately the same location (the
perceived deepest spot) on the lake. The plan called
for the ground crew to sample the lake first, and the
helicopter crew to sample the lake as soon as
possible thereafter (optimally, within 1 hour). The
ground crew collected a routine sample and a
duplicate sample; the helicopter crew collected a
routine sample, a duplicate sample, and a triplicate
sample. Both types of sampling crews used sample
collection techniques standard for all WLS-I lakes.
To ensure that each ground crew's sampling
procedure would be representative of all WLS-I
lakes sampled, the ground crews were not told which
lakes were calibration lakes. Because the helicopter
sample collection procedure was tested and proven in
ELS-I, there was no need to conceal the identity of
calibration lakes from the helicopter crews. Some
minor modifications were made to the ELS-I
helicopter sampling protocol for WLS-I (Bonoff and
Groeger, 1987), but these modifications did not affect
the method used to collect lake samples (collecting
lake water through a Van Dorn sampler).
When the sampling scenario was designed, there was
an indication that the samples collected by the ground
crews might arrive at the field laboratory from 1 to 5
87
-------
days after they were collected. As a result of this
indication, the possible effects of delayed sample
preservation (or "holding time") were of interest.
Processing procedures were designed to account for
possible delays in delivering samples from the lake
site to the field laboratory and were used to observe
the effects of different holding times. Each procedure
assumed a different relationship of the sampling time
to the arrival time of the helicopter crew's samples
and the ground crew's samples at the field laboratory.
The field laboratory personnel preserved the ground
crew's samples on the date they were received and
preserved two of the helicopter crew's samples on
the date they were received. The third sample
collected by the helicopter crew was not preserved
immediately. Instead, this sample, which was selected
randomly from among the helicopter crew's three
samples, was held at the field laboratory at 4°C in the
dark for a specified length of time before it was
processed and preserved. The holding time for the
withheld sample depended on which processing
procedure applied (See Figure 11). Sample
processing and preservation procedures used for
calibration lake samples are discussed further in
Silverstein et al. (1987) and in Kerfoot and Faber
(1987).
Comparison of Analytical Laboratories-
The calibration study also was designed to provide
data that could be used to evaluate differences
between the analyses performed by the two analytical
laboratories. To meet this goal, the survey design
called for the field laboratories to randomly assign all
calibration-lake samples for shipment to the
analytical laboratories. The assignments were
designed to ensure that, for each calibration lake, one
of the two samples collected by the ground crew and
one of the three samples collected by the helicopter
crew would be sent to each analytical laboratory. The
results of analyses of potential bias are presented in
the following discussion and in Table 34 (later in this
section); a more detailed discussion concerning the
effect that these results may have on population
estimates appears in Landers et al. (1987).
Design Modifications
The calibration study originally was called the
comparability lakes study. During the survey, the
lakes to be sampled were referred to as comparability
lakes, a term that appears in many of the early
internal documents. After the survey, the official name
of the study was changed to describe correctly the
purpose of the study. Although the study did compare
two sampling methods, the primary purpose was to
determine whether or not there were systematic
differences in the sampling methods or in the
analytical laboratory performance and, if so, to
develop factors that would allow the data from the two
ground-access samples to be calibrated to the data
from the helicopter-access samples.
Modifications to the sampling design and survey
protocols limited the number of calibration lake
samples that could be used for statistical
comparisons. The design of the calibration study
called for 50 lakes (9 to 12 per subregion) to be
sampled. Of these 50 lakes, 5 were not sampled. All
five of the lakes were in subregion 4D (Bozeman,
Montana, field base). Three of these lakes were
frozen, one was too shallow, and one was
inaccessible from the ground. The inaccessible lake
(Red Rock Lake; ID 4D2-006), was sampled by a
helicopter crew, but weather and trail conditions
prevented the lake from being sampled by the ground
crew. Because no comparison could be made
between sampling methods, data for this lake were
deleted from calibration study statistics; however,
they were included in the routine statistics for WLS-I
samples collected by helicopter.
Ground crews at the Carson City field base
(subregion 4A) were told inadvertently which lakes
were part of the calibration study. As a result, some
new lakes had to be selected for the study, and
seven of those lakes were not sampled in duplicate.
However, all helicopter crews collected all three
samples as per original protocol. The loss of these
seven lakes from the statistical analysis was not
considered as crucial as ensuring that the ground
crews could not identify lakes as calibration study
lakes. Therefore, of the 45 calibration lakes sampled,
only 38 were used in the determination of calibration
for sampling method or for laboratory bias (Landers et
al., 1987). In most cases (23 of the first 28 lakes
sampled), samples collected by the ground crews
arrived at the field base on the day they were
collected, so there was no holding-time
consideration. As a result, an artificial holding time
was used for some subsequent samples so that
sufficient data would be available to allow comparison
of differences in sample concentration as a result of
different lengths of time before preservation.
All of the combinations of sample collection dates,
holding times, and different analytical laboratories
took considerable coordination among personnel at
each field base and among field bases. Scheduling
dates for helicopter crew and ground crew sampling
was a difficult and intricate task. Because the
calibration lakes were in wilderness areas, many of
the lakes were difficult for ground crews to reach.
Ensuring that helicopter and ground crews could
sample the same lake on the same day was difficult,
especially when weather conditions were poor.
Successful timing required constant radio
communication and constant rescheduling of daily
sample itineraries. Two-thirds of the lakes sampled
(30 of 45) were sampled on the same day by both
crews. In many of the remaining cases, the weather
88
-------
Figure 11. Sample flow for the calibration study. Western Lake Survey - Phase I.
Ground Samples Helicopter Samples
(Forest Service) (Lockheed-EMSCO, EPA)
Routine ^
1st
Sample
Taken
RGC
Duplicate Routine Duplicate Triplicate
2nd 1st 2nd 3rd
Sample Sample Sample Sample
Taken Taken Taken Taken
DGC J 1 RHC DHC THC 1
1
i '
Field Laboratory
\
^_A_^
» * .. t
RGC
7
Aliquots
DGC RHC DHC THC
7 777
Aliquots Aliquots Aliquots Aliquots
Randomly Selected
Sample Shipment
1
1
1
1
t
Analytical
Laboratory
1
1
1
t
^
Withheld
Helicopter
Sample
I i
1 1
I |
J
|
«. X
\
\
~ -N
1
1
1
1
t
Alternate
Analytical
Laboratory
RGC-Routine Ground Calibration
DGC-Duplicate Ground Calibration
RHC-Routine Helicopter Calibration
DHC-Duplicate Helicopter Calibration
THC-Triplicate Helicopter Calibration
was such that the helicopter could not fly, yet the
ground crew was already on its way to the lake. In
these cases, the ground crew could not be called
back without identifying the lake as a calibration lake
and thereby jeopardizing the integrity of the study.
Consequently, 14 of the lakes were sampled one or
more days earlier by the ground crews than by the
helicopter crews (see the detailed discussion on
sampling times later in this section). Close
communication between field bases also was
essential to ensure that calibration lake samples were
inserted in the proper batches and were sent to the
proper analytical laboratories. The laboratory
coordinator had to maintain daily contact with the
89
-------
communication center and the QA staff in Las Vegas
to coordinate protocol changes and shipment of the
calibration study samples to the appropriate analytical
laboratory.
Verification of Calibration Lake Data
Calibration lake samples were tracked and evaluated
as a separate data set because there was concern
that these samples could not be used to check
precision in the same way that field duplicate pairs
were used. The calibration lake study provided five
samples from each lake (routine, duplicate, and
triplicate samples collected by the helicopter crew and
routine and duplicate samples collected by the ground
crew). Duplicate samples taken for the calibration
study were not used as QA samples because
sampling methods, holding times, and batches often
differed for the five comparable samples collected
from one lake. Therefore, flags were not generated
from calibration lake samples the way flags were
generated for standard duplicate pairs. All five
samples from each lake were compared to each other
visually, however, to check for outlier values or
reporting errors. In addition, the QA staff performed
standard verification checks for ion balance,
conductance, and protolytes for each calibration lake
sample. In that regard, calibration study lake samples
were treated like any other individual sample in the
survey. These samples also received tags and flags
applicable to the batch in the same manner as routine
samples received tags and flags.
Determination of Sampling Method Bias
A primary objective of the calibration study was to
determine whether data collected by the ground
crews, in either the calibrated or the unadjusted form,
were accurate enough to be included in the WLS-I
data base. This determination was intended to show
that these data could be used in estimating
populations for wilderness-area lakes. The
management team was concerned that if variables
with large systematic error were included in the data
base, they would bias the overall survey results
significantly.
Preliminary statistical analyses showed that the data
on samples collected by the ground crews were of
suitable quantity and quality to permit calibration
analyses to be performed (Landers et al., 1987).
Sampling method, lake sampled, analytical laboratory,
and 2-way interactions of these three factors were
tested with an analysis of variance (ANOVA) with
interaction. The null hypothesis of no difference is
expected to be rejected when it is in fact true 1 time
in 20 (Snedecor and Cochran, 1967). The ANOVA
results are given in Table 34. For 1 of 24 analytes
(N03~) the data collected by ground crews were
significantly different from the data collected by
helicopter crews. Samples showed such low
concentrations for NO3~ that precision was poor for
both sampling methods. The relative error for NO3~
was large; however, the absolute error was small.
Therefore, the method of sample collection did not
have an overall significant effect on WLS-I data.
Analysis of the calibration study data is also
presented in Appendix A of Landers et al. (1987). On
the basis of that interpretation of the data, and
decision criteria outlined there, it was determined that
the ground-access data were as acceptable and
usable as the helicopter data without performing
calibrations for any analyte. Although some
differences in the two sampling methods were
detected, the practical differences between the two
methods were not significant (i.e., slope of 0.99
versus 1.00). When a large relative difference was
detected, the absolute difference was small, usually
as a result of low sample concentrations that
confounded the results. Differences that predicted
helicopter-access data from ground-access data
could not be detected when the imprecision of the
measurement of the analyte was greater than the
difference in the sampling method. Therefore,
although decision criteria indicate that some statistical
benefit may have been shown, there was no practical
benefit to calibrating the data.
Determination of Relative Bias Between Analytical
Laboratories
The calibration study also was used to determine
whether or not a bias between analytical results
reported by the two analytical laboratories existed
and, if so, to what extent this laboratory bias would
affect the WLS-I data base. The effect of
laboratories can be evaluated with the same ANOVA
with interaction (Table 34) that was used to analyze
sampling method differences.
The analytical laboratories had significantly different
values for 12 variables. Of those 12 analytes, the
interaction between lake and laboratory was
significant for all but Fe, Mn, and SO42". Overall, the
interaction between lake and laboratory was
significant for 14 analytes. Analyte concentration is
site-specific. This was reflected by the fact that
there was a significant difference among lakes for the
concentration of 24 analytes. Only Mn did not have a
significant lake effect. The low concentrations
observed for Mn in all lakes in this study may account
for this situation. These observations suggest that
laboratory bias changed with analyte concentration.
Landers et al. (1987) analyzed the calibration lake
data for relative bias between the analytical
laboratories with standard and weighted regression
techniques. Relative bias was found to be statistically
significant for some analytes. It was concluded,
however, that relative bias between laboratories was
not meaningful in the context of the survey objectives
for most variables. Because it is difficult to establish
90
-------
Table 34. Calibration Study Regression With and Without
Western Lake Survey - Phase I
2-Way Interactions of Its Components,
Variable
Al, extractable
Al, total
ANC
BNC
Ca
cr
Conductance
DIC, air
equilibrated
DIC, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH4 +
N03"
P, total
pH, acidity
pH, alkalinity
pH, air
equilibrated
SiO2
SO42'
P < 0.05
Laboratory
(F1,37)a
198.916
51.416
1.40
36.456
22.466
1.547
1.61
0.915
2.03
1.40
0.306
24.336
0.308
13.816
8.16&
0.843
11.47b
0.771
0.21
98.106
29.896
4.28C
0.177
6.597C
12
Method
(Fl,37)a
0.002
0.767
0.09
2.33
0.18
0.523
0.016
0.669
0.147
2.57
0.067
0.009
1.16
0.003
0.465
1.98
4.08d
6.57=
1.14
0.000
0.033
0.102
0.835
0.615
1
Lake
(F37,37)a
53.756
54.176
7497.776
3.516
26593.736
62.206
1408.006
844.896
1999.606
147.536
17.976
93.526
728.726
11160.916
1.31
2117.526
3.646
26.536
3.596
203.376
207.086
59.276
54.856
473.756
24
Laboratory by
Method
(F 1.37)a
0.004
1.62
0.10
0.74
0.25
0.85
2.62
0.06
0.30
1.38
2.Q7d
0.00
1.18
0.24
3.21<*
1.92
1.78
0.63
1.11
8.346
4.59C
0.68
1.16
0.05
2
Lake by
Laboratory
(F37,37)a
5.896
2.416
9.916
2.8Q6
48.406
1.31
1.62<*
8.356
22.756
1.66d
1.09
1.36
1.22
10.346
1.10
2.336
2.18C
1.36
1.00
1.90C
2.596
2.04C
3.126
0.85
14
Lake by
Method
(F37,F37)a
5.056
1.18
1.846
0.775
2.876
1.08
1.46
0.73
0.82
1.08
1.11
1.55<*
1.16
2.16C
1.09
2.15C
1.07
0.88
1.09
1.74C
1.52
0.98
0.97
1.09
6
a F-ratio is the statistical test of analysis of variance; 1,37 = degrees of freedom.
6 p < 0.01.
c p < 0.05.
d0.05 < p < 0.10.
the absolute bias (accuracy) for an analyte (because
neither laboratory is considered the standard),
Landers et al. (1987) concluded that accounting for
the observed relative bias in WLS-I requires more
information than is currently available. Permutt et al.
(Appendix I of this report) present similar conclusions
from an analysis of data from WLS-I field audit
samples.
Determination of Calibration by Linear Regression
The second objective of the calibration study was to
determine whether the data on the samples collected
by the ground crews could be entered directly into
the final data set, or if the data needed to be
calibrated before population estimates were
calculated. To address this objective, linear
regression techniques were used.
For 22 analytes, the difference in the bias
measurements is very small. Consequently, the data
were comparable, and no correction for sampling bias
was applied. This conclusion also was supported by
the results of field blank and field duplicate data
analyzed by sampling method (see Section 6 and
Section 8). For two analytes (NOs" and extractable
Al) to which regression analysis was applied, both
types of samples showed such low concentrations
that precision was poor for both sampling methods.
The relative error for these two variables was large;
however, the absolute error was small. Therefore,
there is little risk in using the ground crew sample
91
-------
data for NOa" and extractable Al. No correction for
sampling bias was applied.
Holding-Time Effects on Sample Concentration
As a part of the sampling design of the calibration
study, one of the three samples collected by the
helicopter crew was selected randomly to be stored
(in the dark at 4°C) at the field laboratory for as long
as 4 days before it was processed and preserved.
The length of time that the withheld sample was kept
at the field laboratory depended on the amount of
time required to transport the corresponding samples
collected by the ground crew from the lake site to the
field laboratory. This element of the sampling design
was intended to evaluate the impact of delayed
sample processing and preservation that could result
if the ground crews did not deliver their samples to
the field laboratory on the day that the lake was
sampled. The Forest Service ground crews, however,
were extremely efficient in providing same-day
sample delivery, even from lakes that were difficult to
reach. Consequently, the number of samples for
which delivery was delayed (and thus the number of
corresponding withheld samples) was much smaller
than had been anticipated. The number of withheld
samples assigned to each holding time (i.e., the
number of days it took to process the sample after
collection) and the analytical laboratories to which the
samples were sent are summarized in Table 35.
Table 35. Holding Times for Calibration Study Samples
Analyzed by Analytical Laboratories, Western
Lake Survey - Phase I
Holding
Time3
(days)
&
1
2
3
4
Total
Laboratory I
(No. of
Samples)
19
1
4
1
3
28
Laboratory II
(No. of
Samples)
6
2
6
3
_g
17
Total
(No. of
Samples)
25
3
10
4
_3
45
a Holding time here refers to the time between sample collection
and sample processing.
b Zero indicates that the sample was processed on the same date
that the sample was collected.
For each analyte, the effects of holding time on the
concentration were tested with standard linear
regression. The difference between the concentration
for each variable in the withheld sample and that in
the other sample collected by the helicopter crew and
analyzed by the same analytical laboratory was
calculated. The differences were analyzed as a
function of holding time by using standard linear
regression. The number of days that the withheld
sample was stored before it was processed and
preserved was the independent variable in each
regression. The dependent variable was the
difference between the concentration of the withheld
sample and the concentration of the other sample
analyzed at the same analytical laboratory (see Figure
11).
The results of the regression analyses are presented
in Table 36. A significant effect of holding time was
demonstrated in only three cases: extractable Al in
samples analyzed at Laboratory II (p < 0.032), air-
equilibrated DIG in samples analyzed at Laboratory II
(p < 0.010), and air-equilibrated pH in samples
analyzed at Laboratory I (p < 0.001). For extractable
Al, the effect probably was due to one exceptionally
low concentration from one sample held for three
days. Because values for all sample pairs in this
study were near the detection limit, no conclusions
can be drawn about the effect of holding time on
extractable Al concentrations. For air-equilibrated
DIG, the effect probably was due to one exceptionally
low value for this variable analyzed in one sample
held for three days. Removal of the results of the
analysis on this sample, and the fact that the initial
DIG regression showed no statistical significance,
indicate that holding time did not affect air-
equilibrated DIG concentrations. For air-equilibrated
pH, the effect was due to a relatively large difference
in this variable in three samples held for four days.
Two of these sample pairs had differences of 0.1 pH
unit. Because a difference of 0.1 pH unit is
acceptable even for field duplicate pairs processed on
the same day (see Table 2), it has no practical
significance. The third sample pair showed a
difference of 0.5 pH unit; of 45 samples, this was the
only one that showed a large difference. This
difference probably is attributable to random error.
Therefore, a practical difference in holding time for
this variable cannot be concluded from calibration
study data.
Relation of Calibration Study Sampling Times and
Locations
Of the 45 lakes sampled in the WLS-I calibration
study, 14 were sampled one or more days earlier by
the ground crew than by the helicopter crew, 1 was
sampled four days earlier by the helicopter crew than
by the ground crew, and 30 were sampled by both
crews on the same day. Of these 30 lakes, 23 were
sampled by the ground crew first. The ground crews
sampled these 23 lakes 1 hour 20 minutes to 4 hours
55 minutes earlier than the helicopter crews, with a
mean difference of 2 hours 57 minutes. Of the 7
times a helicopter crew sampled before a ground
crew, the range was from 25 minutes to 3 hours 45
minutes, with a mean difference of 2 hours 3 minutes.
The spatial variability cannot be assessed with
confidence. It is not always possible to determine that
a lake was sampled in precisely the same location by
both sampling methods. Nor is it certain in some
cases (i.e., where several lakes were in immediate
92
-------
Table 36. Regression Statistics for the Differences Between
Routine and Withheld Samples Versus Holding
Time by Laboratory, Western Lake Survey -
Phase I
Laboratory I
Laboratory II
Variable
Al, extractable
Al, total
ANC
BNC
Ca
cr
Conductance
DIG, air
equilibrated
DIG, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH/
N03'
P, total
pH, acidity
pH, alkalinity
pH, air
equilibrated
SiO2
S042'
r2
0.102
0.0004
0.002
0.094
0.070
0.046
0.105
0.029
0.024
0.012
0.027
0.013
0.001
0.133
0.018
0.069
0.011
0.004
0.008
0.039
0.024
0.349
0.047
0.0004
P
0.103
0.912
0.823
0.119
0.182
0.281
0.098
0.398
0.443
0.582
0.417
0.566
0.871
0.062
0.504
0.186
0.610
0.754
0.667
0.326
0.440
0.001 a
0.275
0.920
r2
0.243
0.033
0.062
0.001
0.054
0.085
0.051
0.364
0.126
0.045
0.003
0.005
0.093
0.010
0.001
0.059
0.013
0.055
0.083
0.003
0.005
0.005
0.0003
0.090
P
0.0323
0.451
0.304
0.890
0.335
0.226
0.350
0.006a
0.136
0.385
0.819
0.770
0.205
0.685
0.878
0.318
0.642
0.332
0.231
0.830
0.781
0.779
0.945
0.210
r2 = fraction of total variance explained by the linear model.
p = statistical probability of occurrence using standard linear
regression procedure.
a holding times with significance at p <0.05.
proximity to the lake targeted for sampling) that the
helicopter crew and the ground crew sampled the
same lake. The analytical data indicate, however, that
the correct lakes were sampled.
Summary
In general, analyses of the calibration study samples
showed no significant effects of sample holding time
on analyte concentration. Possible effects of holding
time, however, were not adequately tested for
samples with low analyte concentration (i.e.,
extractable -Al, NOa", and NH4 + ) because
concentrations for these samples were near the
detection limits.
Nitrate-Sulfate Stability Study
Introduction
From the onset of WLS-I, there was concern that
the samples collected by the ground crews would
arrive at the field laboratory days after the samples
were taken. The calibration study was designed to
determine the effect of the late arrival. However,
there also was concern about the possible instability
of nitrate and sulfate in these samples. For example,
the instability could be caused by biological activity in
the unpreserved samples between the time of
collection and the time of processing. Therefore, a
special study was conducted in which split samples
were collected directly from the Van Dorn sampling
apparatus to compare sample preservation methods
and to study the effects of holding samples for
different lengths of time before preserving them.
These split samples (annotated with the code "L" on
the field data forms) were analyzed for nitrate and
sulfate content at EMSL-LV. The data provided an
auxiliary check on sampling, processing, and
analytical performance for these analytes.
Sample Processing, Preservation, and Analysis
The nitrate-sulfate split sample consisted of one
125-mL aliquot taken directly from the Van Dorn
sampler after the sample syringes and Cubitainers
had been filled. The aliquot was preserved with 0.1
mL of 5 percent HgCl2 at the lak,e site. The
preservative was added to stop biological activity that
might occur within the split after sampling. These
nitrate-sulfate aliquots were prepared by ground
crews for all samples they collected (including
calibration lake samples, field blanks, and field
duplicates). These split samples were collected by
helicopter crews at calibration study lakes only. The
procedure for collection and preparation of split
samples is given in Appendix L.
When the split samples arrived at the field laboratory,
the laboratory coordinator assigned and recorded the
batch and sample ID numbers on the upper portion of
each aliquot label. The upper portion of the label was
removed and was taped into the nitrate-sulfate
logbook (by batch). The aliquots were then stored at
4°C in the dark until they were shipped to EMSL-LV
the following day.
The field laboratory also prepared aliquots of field
natural audit samples for the nitrate-sulfate study
(see Appendix L). For each field natural audit sample
(FN3, FN4, FN5, and FN6) the field laboratory
received extra 2-L samples from Radian
Corporation. For nitrate-sulfate aliquot batches, the
laboratory coordinator substituted the nitrate-sulfate
audit aliquots for the regular field natural audits being
processed that day. Because of volume limitation and
93
-------
sample instability, synthetic audit samples were not
employed in this study.
On a daily basis, the nitrate-sulfate aliquot batches
were shipped to EMSL-LV for analysis by ion
chromatography. Samples were shipped in coolers
that contained enough frozen freeze-gel packs to
maintain the samples at 4°C during shipment.
Analytical Results
Systems Applications, Inc. (SAI), in San Rafael,
California, performed statistical analyses to compare
results for the samples preserved with HgCl2 with
results for the samples that were analyzed by the
analytical laboratories, which used standard NSWS
preservation techniques.
SAI compared the pairs of sulfate measurements and
the pairs of nitrate measurements. For each
comparison, the analytical laboratory sample
concentration was compared to the split sample
concentration. For each pair, two values were
computed: (1) the signed difference between the
analytical laboratory sample value and the split
sample value and (2) the mean of the two values.
The signed difference and the mean were used to
compute the relative difference:
analytical laboratory value — split sample value
mean of the two values
The relative differences for both analytes are
summarized in Table 37.
Table 37. Summary Statistics for Relative Differences9 in
Analyte Concentrations for the Nitrate-Sulfate
Stability Study, Western Lake Survey - Phase I
Number of Sample Pairs
Mean
Standard Deviation
Signed Rank
Median
Lower Quartile
Upper Quartile
Low Extreme
High Extreme
Nitrate
918
1.002
5.315
139125
0.628
0.015
1.914
-32
91.333
Sulfate
919
-0.308
2.885
-39322
-0.007
-0.216
-0.093
-50
9.143
a Relative difference equals the analytical laboratory value minus
the HgCk" preserved EMSL-LV split sample value, divided
by the mean of the two values.
Sulfate Stability Results—
In this study, 919 pairs of sulfate measurements were
compared. The moments (e.g., mean and standard
deviation) in the first part of Table 37 are not very
useful because of the presence of extreme values.
For example, in one pair, one negative value and one
equally positive value combine to yield a small mean,
and therefore, a relative difference of 50. The
percentiles in the second part of the table, however,
are not significantly affected by the few extreme
cases.
The median relative difference is negative, about 0.7
percent, whereas the upper quartile is 9 percent and
the lower quartile is -22 percent. Thus, there are
considerable random differences in the pairs; the
analytical laboratory value exceeds the split sample
value by 9 percent or more 25 percent of the time,
and the split sample value is as much as 22 percent
greater 25 percent of the time. Nevertheless, the
systematic difference, represented by the median, is
less than 1 percent. Overall, therefore, it appears to
make little difference whether or not the sulfate.
samples are treated with HgCl2, or whether they are
analyzed by the analytical laboratory or the EMSL-
LV laboratory. For a given sample, however, the
difference can be considerable.
The systematic difference is statistically significant
(signed-rank test, p < 0.0001) even though it is
small. There is either a sample-handling effect or an
interlaboratory bias of a fraction of a percent.
Furthermore, the direction of the effect is that the split
sample results are systematically higher. The effect
probably has no practical significance, however, given
its size and the much larger random variation.
Therefore, it should not affect calculation of
population estimates.
Nitrate Stability Results-
For nitrate, 918 sample pairs were compared (see
Table 37). The median relative difference for nitrate is
about 63 percent. Thus, 50 percent of the time,
analytical laboratory measurements of nitrate
concentration exceeded the split sample
measurements by nearly or more than a factor of two.
Many of the measured concentrations of nitrate were
near the detection limits, and even a large relative
difference may not be of practical significance at very
low concentrations. It is therefore desirable to explore
the difference in sample concentration between the
sample analyzed by the analytical laboratory and the
split sample analyzed by EMSL-LV. The pairs are
divided into 10 groups, or deciles. These are deciles
of the distribution of the pair means. The median pair
difference and the median pair mean for each decile
are plotted with a "d" in Figure 12. The horizontal
scale is logarithmic for convenience; therefore, the
first decile, which consists mainly of blanks and has a
very slight negative median pair mean, is eliminated
from the graph.
In every decile except the first and the last, the nitrate
concentrations measured by the analytical
laboratories were significantly higher than those
94
-------
Figure 12.
0.04-
_ 0.03-
8 0.02-
to
1
? 0.01-
'ra
o.
c
ra
"S 0.00-
-0.01-
Relative differences in nitrate concentrations,
nitrate-sulfate stability study. Western Lake
Survey - Phase I. Note that two observations
were out of range.
-0.02-
d = Deciles
n = Natural Audits
d d
d d
0.001 0.010 0.100 1.000 10.000
Median Pair Mean Concentration (mg/L)
measured by the EMSL-LV laboratory (p < 0.0001,
signed-rank test). The median differences are
largest, about 0.02 mg/L, at concentrations between
0.01 mg/L and 0.1 mg/L, where differences this large
represent a substantial fraction of the concentration.
It is not easy to distinguish an effect of the HgCl2
preservation from a possible relative bias between the
analytical laboratories and the EMSL-LV laboratory
because neither laboratory analyzed both types of
samples. The natural audit samples, however, may
help somewhat. These samples were stored for
several weeks before they were processed and
preserved. As a result, it is likely that microbiological
activity had reached a steady state. Thus, the effect
of preserving these audit samples with HgCl2 a day
or two later is of little concern when the audit sample
may have been collected months earlier. Therefore,
these audit samples were used as a tool to assess
interlaboratory bias; their primary purpose did not
include the study of the stability of nitrate over time.
A systematic difference between the analytical
laboratory values and the EMSL-LV values for
natural audit samples might be ascribed to bias rather
than to actual change in nitrate concentrations.
Furthermore, if the difference for routine lake samples
were about the same as for natural audit samples, the
difference in the routine samples might be ascribed to
bias also.
The median pair means and median pair differences
for natural audit lots FN3, FN5, and FN6 are
represented in Figure 11 by the letter n. (Lot FN4 is
not shown because the median difference, -0.11
mg/L, is so large that it would compress the graph
severely. The median pair for lot FN4, however, is
2.35 mg/L; the apparent bias of -0.11 mg/L is an
acceptably small fraction of the concentration.)
Interlaboratory bias does appear to account for some
of the difference between the analytical laboratory
and split sample measurements. For example, the
measurements in the eighth and ninth deciles, where
most of the FN5 audit values fall, were not
significantly different for natural audit samples than for
other samples (p = 0.11, rank-sum test). In the
tenth decile, the FN3 and FN4 audits were
significantly different from the other samples, but only
by a few percent of sample concentration. In the
second and third deciles, the FN6 audits were
significantly different from the other samples (p =
0.012, rank-sum test), but these concentrations may
be too low (i.e., near the instrument detection limit) to
be of much interest.
None of the natural audit samples had concentrations
between 0.01 mg/L and 0.10 mg/L, where the
differences between the analytical and EMSL-LV
laboratory measurements were greatest. The
possibility that the differences were entirely due to
bias cannot be ruled out on the basis of the data.
It seems likely that concentrations of nitrate near 0.02
mg/L can be produced in lake samples during an
extended storage before preservation. If it is important
to measure concentrations in this range accurately,
special precautions such as preservation with HgCl2
are indicated. In future studies, if a 0.02-mg/L
difference is of concern, an additional study
accounting for laboratory bias at the concentration
levels of interest must be employed. The results of
the WLS-I stability study indicate that sample
holding time before preservation had minimal effect
on population estimates.
95
-------
-------
Section 10
References
American Society for Testing and Materials, 1984.
Annual Book of ASTM Standards, Vol. 11.01,
Standard Specification for Reagent Water,
D1193-77 (reapproved 1983). ASTM,
Philadelphia, Pennsylvania.
Best, M. D., S. K. Drouse, L. W. Creelman, and D. J.
Chaloud, 1987. National Surface Water Survey
Eastern Lake Survey (Phase I - Synoptic
Chemistry) Quality Assurance Report.
EPA/600/4-86/011. U.S. Environmental
Protection Agency, Environmental Monitoring
Systems Laboratory, Las Vegas, Nevada.
Bonoff, M. B., and A. W. Groeger, 1987. National
Surface Water Survey, Western Lake Survey -
Phase I (Synoptic Chemistry) Field Operations
Report. EPA 600/8-87/018. U.S. Environmental
Protection Agency, Environmental Monitoring
Systems Laboratory, Las Vegas, Nevada.
Drouse, S. K., 1987. National Surface Water Survey,
National Stream Survey Phase I - Pilot Survey,
Summary of Quality Assurance Data Results. In
press. U.S. Environmental Protection Agency,
Environmental Monitoring Systems Laboratory,
Las Vegas, Nevada.
Drouse, S. K., D. C. Hillman, L. W. Creelman, and S.
J. Simon, 1986. National Surface Water Survey,
Eastern Lake Survey (Phase I - Synoptic
Chemistry) Quality Assurance Plan. EPA/600/4-
86/008. U.S. Environmental Protection Agency,
Environmental Monitoring Systems Laboratory,
Las Vegas, Nevada.
Eilers, J. M., P. Kanciruk, R. A. McCord, W. S.
Overton, L. Hook, D. J. Blick, D. F. Brakke, P. E.
Kellar, M. D. DeHaan, M. E. Silverstein, and D. H.
Landers, 1987. Characteristics of Lakes in the
Western United States. Volume II. Data
Compendium for Selected Physical and Chemical
Variables. EPA/600/3-86/054b. U.S. Environ-
mental Protection Agency, Washington, D. C.
Hillman, D. C., J. F. Potter, and S. J. Simon, 1986.
National Surface Water Survey, Eastern Lake
Survey (Phase I - Synoptic Chemistry) Analytical
Methods Manual. EPA/600/4-86/009. U.S.
Environmental Protection Agency, Environmental
Monitoring Systems Laboratory, Las Vegas,
Nevada.
International Science Technology, Inc, 1986. Data
Base Audit - National Surface Water Survey,
Phase I - Western Lake Survey (IS&T 10067.)
Internal Report for the U.S. Environmental
Protection Agency, Environmental Research
Laboratory, Corvallis, Oregon.
Kanciruk, P., M. Gentry, R. McCord, L. Hook, J.
Eilers, and M. D. Best, 1987. National Surface
Water Survey: Western Lake Survey - Phase I,
Data Base Dictionary. Oak Ridge National
Laboratory, Environmental Sciences Division
Publication No 2838. Oak Ridge National
Laboratory, Oak Ridge, Tennessee.
Kanciruk, P., R. J. Olson, and R. A. McCord, 1986.
Quality Control in Research Data Bases: The U.S.
Environmental Protection Agency National
Surface Water Survey Experience, pp. 193-207
in W. K. Michener (ed.) Research Data
Management in the Ecological Sciences. Belle W.
Baruck Library in Marine Science, Number 16.
University of South Carolina Press, Columbia,
South Carolina.
Kerfoot, H. B., and M. L. Faber, 1987. National
Surface Water Survey, Western Lake Survey
(Phase I - Synoptic Chemistry) Analytical
Methods Manual. EPA/600/8-87/038. U.S.
Environmental Protection Agency, Environmental
Monitoring Systems Laboratory, Las Vegas,
Nevada.
Knapp, C. M., C. L. Mayer, D. V. Peck, J. R. Baker,
and G. J. Filbin, 1987. National Surface Water
Survey: National Stream Survey - Phase I Pilot
Survey Field Operations Report. EPA/600/8-
87/019. U.S. Environmental Protection Agency,
Environmental Monitoring Systems Laboratory,
Las Vegas, Nevada.
Landers, D. H., J. M. Eilers, D. F. Brakke, W. S.
Overton, P. E. Kellar, M. E. Silverstein, R. D.
97
-------
Schonbrod, R. E. Crowe, R. A. Linthurst, J. M.
Omernik, S. A, league, and E. P. Meier, 1987.
Characteristics of Lakes in the Western United
States. Volume I. Population Descriptions and
Physico-Chemical Relationships. EPA/600/3-
86/054a. U.S. Environmental Protection Agency,
Washington, D.C.
Linthurst, R. A., D. H. Landers, J. M. Eilers, D. F.
Brakke, W. S. Overton, E. P. Meier, and R. E.
Crowe, 1986. Characteristics of Lakes in the
Eastern United States. Volume I: Population
Descriptions and Physico- Chemical
Relationships. EPA/600/4-86/007, U.S.
Environmental Protection Agency, Washington,
D.C.
Morris, F. A., D. V. Peck, M. B. Bonoff, K. J. Cabbie,
and S. L. Pierett, 1986. National Surface Water
Survey, Eastern Lake Survey (Phase I -
Synoptic Chemistry) Field Operations Report.
EPA/600/4-86/010. U.S. Environmental
Protection Agency, Environmental Monitoring
Systems Laboratory, Las Vegas, Nevada.
Permutt T. J., and A. K. Pollack, 1986. Analysis of
Quality Assurance Data for the Eastern Lake
Survey. Prepared by Systems Applications, Inc.,
for Lockheed Engineering and Management
Services Company, Inc., Las Vegas, Nevada.
Appendix A in Best, M.D., S. K. Drouse, L. W.
Creelman, and D. J. Chaloud. 1986. National
Surface Water Survey, Eastern Lake Survey
(Phase I - Synoptic Chemistry) Quality
Assurance Report (1987). EPA/600/4-86/011.
U.S. Environmental Protection Agency,
Environmental Monitoring Systems Laboratory,
Las Vegas, Nevada.
SAS Institute, 1982. The Statistical Analysis System,
SAS Institute, Gary, North Carolina.
Silverstein, M. E., S. K. Drouse, J. L. Engels, M. L
Faber, and T. E. Mitchell-Hall, 1987. National
Surface Water Survey, Western Lake Survey
(Phase I - Synoptic Chemistry) Quality
Assurance Plan. EPA/600/8-87/026. U.S.
Environmental Protection Agency, Environmental
Monitoring Systems Laboratory, Las Vegas,
Nevada.
Snedecor, G. W., and W. G. Cochran, 1973.
Statistical Methods. The Iowa State University
Press, Ames, Iowa.
U.S. Environmental Protection Agency, 1979.
Handbook for Analytical Quality Control in Water
and Wastewater Laboratories. EPA/600/4-
79/019. U.S. Environmental Protection Agency,
Environmental Monitoring and Support Laboratory,
Cincinnati, Ohio.
98
-------
Appendix A
National Surface Water Survey Form 26
Data Confirmation/ Reanalysis Request Form
The NSWS Form 26 was created to document data analytical laboratory results from the raw data set to
changes. The form was used to track reported the verified data set.
99
-------
National Surface Water Survey Form 26
Data Confirmation/Reanalysis Request Form
Date Sent
Date Received
Batch #_
The following values require:
National Surface Water Survey
Form 26
Data Confirmation/Reanalysis Request Form
Contract Analytical Laboratory Laboratory Supervisor
Confirmation (See I)
Reanalysis (See II)
Variable
NSWS
Form*
Sample I.D.
Suspect
Original
Value
Reconfirmed/
New
Value
Explanation
Contract
Analytical
Laboratory
LEMSCO
Yes
No
Confirmation Request: Did ANY values change:
If yes, reason (note above in explanation column):
(A) Reporting Error (C) Original reported value did not change
(B) Calculation Error (D) Data Previously Omitted
(E) Other - Explain
If values changed, submit supporting raw data AS REQUIRED.
Additional Comments Regarding Confirmation:
II. Reanalysis Requested Due to:*
External QA Data
Internal QC Data Indicated Below:
1C Resolution
IDL > CRDL
Blank > 2 x CRDL (Reagent: Calibration)
QCCS Outside Criteria (DL; Low; High)
Sample Concentration Outside Calibration Range
QCCS Not in Mid-Range of Calibration Range
Duplicate Precision (% RSD) Outside Criteria; Insufficient Number of Duplicates Analyzed
Additional Comments Regarding Reanalysis:
* An abbreviated version of NSWS Forms 11, 18, 19, and 20 must be submitted for all reanalyzed data.
NSWS Forms 13, 17, and 22 must be submitted when applicable.
FOR LEMSCO USE ONLY: INITIAL REVIEW NUMBER OF VALUES SUBMITTED
VERIFICATION
NUMBER OF VALUES CHANGED
100
-------
Appendix B
Calculation of Field Blank Sample Control Limits
Criteria for determining contamination were needed in
order to check for systematic contamination problems
during sample collection and analysis and before
preparing a verified data set. These criteria, termed
control limits, were determined by a variety of
nonstatistical methods during ELS-I. Some control
limits were established on the basis of specifications
provided by the instrument manufacturer; others
reflected DQOs (i.e., the level of detectability needed
to meet the goals of the survey). Control limits for
some analytes could be defined only in terms of
analytical experience and intuitive assumptions based
on that experience, because there were not any
acceptable precedents.
Upper control limits for WLS-I blank samples were
determined statistically, on the basis of ELS-I
experience and the analytical results obtained for
ELS-I field blanks. The 95th percentile (Pgs)
nonparametric test used to calculate the system
decision limit in ELS-I (Best et al., 1987) was
selected for the WLS-I control limits for the following
reasons:
1. Although negative response values can be valid
and were required to be reported, one of the
ELS-I analytical laboratories, which analyzed
almost 50 percent of all ELS-I samples, reported
all negative values as zero. This biased any
negative values and further skewed the field blank
distribution toward positive values.
2. A field blank is subjected to handling, shipping,
and preservation effects at the lake site, in the
field laboratory, and in the analytical laboratory.
Contamination may result at any or all of these
locations. A field blank can yield a positive value
as a result of contamination; however, it cannot
yield a negative value as a result of
contamination. Consequently, the distribution of
values may be skewed toward the contaminated
levels, and the associated curve will represent a
nonnormal distribution about 0. Contamination
could reduce a negative bias that resulted from
calibration error.
3. Negative values can be reported for an analysis,
but they will not result from contamination or from
the presence of analyte. Negative values are
caused by instrumental drift, analytical error, and
standard regression curves with negative y-
intercepts. Therefore, negative values are created
in the analytical laboratory and do not result from
field activities.
Two methods of calculating blank windows were
considered in the WLS-I survey design. One
calculation was the prediction interval:
X ± (t) sJl + Ifn
which is the standard 95 percent confidence interval
about the mean and assumes a normal distribution.
This calculation was rejected because the distribution
of ELS-I blanks was not normal; most of the ELS-I
blanks showed a skewed distribution to positive
values, and one ELS-I laboratory had adjusted all
negative values to zero. To accommodate the
nonnormal distribution, the non-parameteric Pgs
statistic was used in determining the field blank
acceptance criteria. As long as at least 5 percent of
the blank values were above zero, this calculation
was not affected by distribution or by the number of
negative values set to zero.
The Pgs statistic was used to calculate the upper limit
at which blank values would be flagged. The lower
limit, however, was designated as the negative value
of the required detection limit. Anything less than this
negative value was unacceptable and was attributed
to excessive instrumental drift or to inaccurate
calibration of the instrument.
Table B-1 presents the field blank control limits for
ELS-I and WLS-I. Field blank concentrations that
were outside these limits were considered suspect
and were flagged. Establishing these limits prior to a
full-scale statistical analysis was essential to
identifying contamination trends as they occurred.
The detailed statistical analysis of the WLS-I field
blank values was performed after data verification
was completed.
101
-------
Table B-1. Comparison of Field Blank Control Limits, Eastern Lake
Survey - Phase I and Western Lake Survey - Phase I
Variable^
Al, extractable
Al, total
ANC (jieq/L)
BNC (peq/L)
Ca
cr
Conductance
(uS/cm)
DIC, air
equilibrated
DIC, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH4 +
N03'
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air equilibrated
(pH units)
P, total
SiO2
SO42'
Low
ELS-I
-0.005
-0.005
-10.0
0.00
-0.005
-0.010
-0.01
0.10
0.20
-0.20
-0.005
-0.005
-0.005
-0.005
-0.005
-0.005
-0.010
-0.0106
5.40
5.40
5.40
-0.005
-0.050
-0.01
Limit
WLS-I
-0.005
-0.005
-10.0
-10.0
-0.010
-0.010
-0.9
-0.05
-0.05
-0.1
-0.005
-0.010
-0.010
-0.010
-0.010
-0.010
-0.010
-0.005
5.5QC
5.50C
5.41
-0.002
-0.050
-0.050
High
ELS-I
0.009
0.009
10.0
40.0
0.050
0.050
2.0
0.30
0.40
0.6
0.009
0.05
0.05
0.05
0.010
0.050
0.030
0.020b
5.90
5.90
5.90
0.005
0.050
0.10
Limit
WLS-I
0.008
0.033
7.18
22.45
0.034
0.094
1.31
0.294
0.426
0.45
0.005
0.023
0.018
0.008
0.012
0.031
0.039
0.023
5.87C
5.87C
5.90
0.008
0.117
0.094
a Units are in mg/L unless otherwise noted.
b NOs" limits were calculated by using only the last 99 blanks processed
during ELS-I; earlier ELS-I blanks were contaminated.
0 ELS-I pH (acidity) and pH (alkalinity) values were pooled to calculate
WLS-I limits.
102
-------
Appendix C
Preparation of Audit Samples
1.0 Preparation of Field Natural Audit
Samples
To ensure that all field natural audit samples of a
particular lot were uniform, EMSL-LV instructed the
preparation laboratory (Radian Corporation in Austin,
Texas) to follow the protocol specified below.
1. Clearly label the field natural (FN) stock barrels
with the lot number.
2. Label the 2-L bottles to be filled.
3. Operating in a clean environment, flush the Tygon
tubing lines with lake water. Discard the water.
4. Pump 20 to 25 ml_ lake water into the audit bottle,
cap the bottle, rinse the bottle to get complete
coverage, and discard the rinse.
NOTE: The Tygon tubing must not touch the
sidewalls of the bottle.
5. Perform step 4 two more times. Discard the rinse
water each time.
6. Fill the bottle to the top (no head space) and cap
the bottle.
NOTE: The bottle must be capped immediately
after it is filled to minimize the possibility
of contamination.
7. Secure the cap to the bottle with tape.
8. Log in the total number of samples prepared, the
date prepared, and the name of the analyst or
technician.
9. Place samples in storage at 4°C by lot and ID
number to await shipment.
10. Discard any water remaining in Tygon tubing. Do
not drain residual lake water into the stock barrel.
2.0 Preparation
Samples
of Synthetic Audit
To prepare the field synthetic audit samples of the
desired concentrations, Radian technicians diluted the
lot stock concentrates with ASTM Type I reagent-
grade water. Each diluted 2-L synthetic audit
samples were prepared for shipment to the field
laboratory as follows:
1. Fill a 2-L volumetric flask with 1.5 L deionized
water.
2. Add a predetermined volume of each of the four
stock concentrates (see Table C-1) to the flask.
3. Fill the flask to volume and mix the solution
thoroughly.
4. When the dilution is complete, transfer the 2-L
sample to a carboy. (If 10 samples were prepared
in one day, the carboy would eventually contain
20-L of diluted stock, prepared 2-L at a time.)
5. When these dilution and transfer steps are
completed, sparge the audit sample solution in
the carboy with 300 ppm CO2 and equilibrate.
(The equilibration raises the acidity of the sample,
thereby counteracting the effect of adding the
strong base Na2SiO3- It also restores any DIG
lost during sample preparation steps and, by
stabilizing the sample, it minimizes day-to-day
sample variation caused by shipping and
handling.)
103
-------
Table C-1. Composition of the Field Synthetic Audit Sample
Concentrates, Western Lake Survey - Phase I
Stock
Concentrate
Chemical Formula
Analytes to be Measured
1 AI2(SO4)3-(NH4)2SO4-24H2O
2 FeNH4(SO4)2-12H2O
3 Na2SiO3
4 CaCI2
NaHCO3
C6H4(COOH)2
MgSO4
NaF
MnSO4-H2O
NH4NO3
Na2HPO4
KHC8H4O4
Extractable Al, total Al,
NH/,SO4S'
Fe, NH4 + ,SO4'
Na, SiO2
Ca, Cf
DIC, Na
DOC
Mg, SO42'
Total dissolved F", Na
Mn, SO42'
NH/,NO3'
Na, Total P
DOC, K
104
-------
Appendix D
Distribution of Data for Field, Trailer, and Calibration Blank
Samples Analyzed in the Analytical Laboratories
Table D-1. Distribution of Data for Field Blank Samples Analyzed in the Analytical Laboratories
... !-> V I— aUVJi ClUJi v *-*» VJCM i IMKI ivj ivit^u \\j\j
Laboratories
and Sampling
Methods Pooled
(n = 236)
Variable3
Al, extractable
Al, total
ANC (ueq/L)
BNC (peq/L)
Ca
Cf
Conductance
(liS/cm)
DIG, air
equilibrated
DIG, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH/
P50
0.001
0.009
1.3
18.7
0.014
0.005
0.7
0.200
0.180
0.14
0.001
0.001
0.001
0.001
0.000
-0.001
-0.003
P95
0.004
0.019
3.9
29.8
0.072
0.043
1.6
0.330
0.430
0.30
0.003
0.014
0.012
0.006
0.014
0.012
0.010
Labi
(n = 97)
PSO
0.000
0.002
-0.4
8.3
0.042
0.013
1.0
0.160
0.110
0.22
0.000
0.000
0.003
0.002
0.000
-0.001
-0.007
P95
0.003
0.014
5.4
17.5
0.077
0.047
2.0
0.240
0.340
0.35
0.006
0.005
0.014
0.007
0.001
0.012
0.012
Lab II
(n = 139)
P50
0.001
0.011
1.5
22.5
0.007
0.004
0.4
0.220
0.210
0.10
0.002
0.005
0.000
0.001
-0.001
-0.001
0.002
P95
0.004
0.019
3.2
30.3
0.037
0.024
1.6
0.370
0.440
0.23
0.003
0.020
0.008
0.005
0.021
0.014
0.009
Ground
(n = l12)
P5o
0.001
0.008
1.1
18.0
0.018
0.005
0.6
0.189
0.200
0.15
0.001
0.001
0.001
0.001
0.000
-0.001
-0.003
P95
0.004
0.019
4.0
37.1
0.075
0.052
1.7
0.313
0.490
0.31
0.005
0.014
0.014
0.004
0.009
0.008
0.008
Helicopter
(n = l24)
PBO
0.001
0.010
1.3
19.8
0.012
0.005
0.6
0.210
0.170
0.13
0.001
0.001
0.001
0.001
0.000
-0.001
-0.003
P95
0.004
0.019
4.4
28.3
0.071
0.033
1.8
0.370
0.390
0.32
0.003
0.015
0.012
0.008
0.016
0.014
0.012
(con-
tinued)
105
-------
Table D-1. Continued
Variables
N03"
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
SiO2
SO,2"
Laborator
Sampling
Pool
(n = 2
P50
0.008
0.001
5.66
5.64
5.70
0.047
0.012
ies and
Methods
ed
36)
P95
0.071
0.006
5.78
5.74
5.83
0.179
0.065
oy uauL
Labi
(n = 97)
PSO
0.010
0.000
5.70
5.67
5.68
0.046
0.028
P95
0.045
0.017
5.79
5.77
5.75
0.271
0.123
naiury
Lab II
(n = 139)
P5Q
0.006
0.001
5.64
5.63
5.71
0.047
0.006
P95
0.078
0.005
5.73
5.71
5.88
0.116
0.021
t
ay aampiin
Ground
(n = 112)
PBO
0.009
0.001
5.65
5.64
5.69
0.040
0.013
P95
0.061
0.008
5.78
5.75
5.80
0.150
0.129
g meinoa
Helicopter
(n = 124)
PBO
0.006
0.001
5.66
5.65
5.70
0.052
0.011
P95
0.072
0.006
5.78
5.74
5.86
0.248
0.047
a units in mg/L unless otherwise noted; units shown at the number of significant figures reported by the analytical laboratories.
PSO = 50th percentile.
Pg5 = 95th percentile.
106
-------
Table D-2. Distribution of Data for Trailer Blank
Samples Analyzed in the Analytical
Laboratories
Trailer Blank Samples
(n = 22)
Variable3
Al, extractable
Al, total
ANC (iieq/L)
BNC (neq/L)
Ca
CI"
Conductance (nS/cm)
DIG, air equilibrated
DIG, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH4 +
N03-
P, total
pH, acidity (pH units)
pH, alkalinity (pH units)
pH, air equilibrated
(pH units)
SiO2
S042-
P5o
0.001
0.008
0.9
16.9
0.001
0.007
0.6
0.195
0.122
0.09
0.001
0.001
-0.001
0.000
0.000
-0.003
-0.004
0.007
0.001
5.67
5.64
5.69
0.057
0.013
P95
0.003
0.013
2.0
30.4
0.029
0.050
1.3
0.300
0.243
0.27
0.002
0.009
0.002
0.003
0.014
0.002
0.004
0.074
0.012
5.96
5.90
5.81
0.152
0.057
a units in mg/L unless otherwise noted; units shown at the
number of significant figures reported by the analytical
laboratories.
P50 = 50th percentile.
P95 = 95th percentile.
107
-------
Table D-3. Distribution of Data for Anal'
DC
Variable^
Al,
extractable
Al, totald
ANC (ueq/L)
BNC (ueq/L)
Ca
Cf
Conductance
(uS/cm)
DIC, air
equilibrated
DIC, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH4*
N03"
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
SiO2d
SO42'
Required
Detection
Limit
0.005
0.005
10.0
10.0
0.01
0.01
0.9
0.05
0.05
0.1
0.005
0.01
0.01
0.01
0.010
0.01
0.01
0.005
0.002
N/A
N/A
N/A
0.05
0.05
Labs
Pooled
0.002
0.001
N/A
N/A
0.007
0.004
0.2
N/A
0.020
0.08
0.001
0.007
0.004
0.002
0.005
0.003
0.005
0.004
0.001
N/A
N/A
N/A
0.021
0.012
ytical Labora
Instrument
election Limita
Labi
0.001
0.001
N/A
N/A
0.007
0.010
0.3
N/A
0.030
0.09
0.002
0.006
0.007
0.001
0.002
0.005
0.007
0.005
0.001
N/A
N/A
N/A
0.038
0.016
itory Calibration and Reagent Blank Samples
Calibration Blankb Concentrations
Lab II
0.002
0.001
N/A
N/A
0.007
0.003
0.2
N/A
0.020
0.07
0.001
0.009
0.004
0.002
0.009
0.003
0.004
0.004
0.001
N/A
N/A
N/A
0.016
0.011
LabsP
/„ j
(n 1
PSO
0.001
0.005
N/A
N/A
N/A
0.003
0.0
0.001
0.002
0.10
0.000
N/A
N/A
N/A
N/A
N/A
-0.002
0.005
0.000
N/A
N/A
N/A
0.008
0.005
'ooled
49)
P95
0.004
0.008
N/A
N/A
N/A
0.013
0.6
0.046
0.060
0.17
0.002
N/A
N/A
N/A
N/A
N/A
0.006
0.013
0.002
N/A
N/A
N/A
0.065
0.033
Lat
(n =
P50
0.000
0.000
N/A
N/A
0.004
0.004
0.5
0.020
0.020
0.12
0.000
0.003
-0.007
0.001
0.000
-0.008
-0.001
0.008
0.000
N/A
N/A
N/A
0.024
0.022
) I
58)
P95
0.005
0.009
N/A
N/A
0.009
0.016
0.7
0.090
0.090
0.19
0.000
0.008
0.008
0.004
0.005
0.000
0.010
0.018
0.001
N/A
N/A
N/A
0.093
0.039
Lab II
(n = 91)
PSO
0.001
0.005
N/A
N/A
N/A
0.003
0.0
-0.010
-0.010
0.04
0.001
N/A
N/A
N/A
N/A
N/A
-0.002
0.004
0.000
N/A
N/A
N/A
0.008
0.003
P95
0.002
0.008
N/A
N/A
N/A
0.007
0.1
0.007
0.007
0.11
0.003
N/A
N/A
N/A
N/A
N/A
0.002
0.009
0.002
N/A
N/A
N/A
0.024
0.010
a Calculated at three times the standard deviation of 10 nonconsecutively analyzed calibration blanks; required to be performed weekly.
b Calibration blanks analyzed as part of daily instrument calibration.
c Units in mg/L unless otherwise noted; units shown at the number of significant figures reported by the analytical laboratories.
d Reagent blanks used in instrument calibration.
N/A = not applicable P5o = 50th percentile P95 = 95th percentile
108
-------
Appendix E
Field Laboratory Precision Data for Audit Sample Measurements of Dissolved
Inorganic Carbon, pH, Turbidity, and True Color
Table E-1. Comparison of Field Laboratory and Analytical Laboratory Measurements for Dissolved Inorganic Carbon,
Western Lake Survey - Phase I
Analytical Laboratory
Field Base (Subregion) Measurement for Measurement for
Closed-System DIC Open-System DIG
Audit Sample
FN3
(Lake
Superior)
FN4
(Big
Moose
Lake)
FN5
(Bagley
Lake)
FN6
(Bagley
Lake)
FL11
(Synthe-
tic)
FL12
(Synthe-
tic)
FL11 and
FL12
(Pooled
Synthetics)
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
Cali-
fornia
(4A)
5
10.34
2.6
4
0.62
3.6
19
2.16
4.1
22
1.65
3.9
2
1.64
1.0
8
1.41
5.3
10
1.45
8.1
Pacific
NW
(4B)
5
10.35
1.1
4
0.64
11.5
12
2.12
3.9
11
1.63
3.3
2
1.70
0.1
8
1.36
4.7
10
1.43
10.6
No.
Rockies
(4C)
10
10.28
4.7
4
0.63
8.3
15
2.21
4.4
-
-
-
6
1.57
11.6
4
1.43
4.2
10
1.52
10.3
Cent.
Rockies
(4D)
9
10.23
1.8
4
0.57
9.8
12
2.14
3.5
—
-
-
5
1.56
13.5
2
1.46
0.5
7
1.53
11.7
So.
Rockies
(4E)
9
10.31
1.40
4
0.65
2.0
10
2.18
3.4
4
1.67
1.9
6
1.52
6.5
4
1.43
6.7
10
1.48
6.9
Sub-
regions
Pooled
38
10.29
2.7
20
0.62
8.5
68
2.16
4.1
37
1.65
3.6
21
1.57
9.5
26
1.41
5.2
47
1.48
9.5
Labi
19
10.34
12.8
9
0.49
14.4
24
2.04
17.9
8
1.63
5.1
11
1.49
26.4
6
1.31
5.8
17
1.42
22.8
Lab II
19
9.39
3.0
11
0.53
38.7
44
1.83
6.1
29
1.42
3.6
10
1.97
2.7
20
1.41
9.9
30
1.60
18.3
Labs
I and II
Pooled
38
9.86
10.7
20
0.51
30.4
68
1.91
13.3
37
1.47
7.0
21
1.72
21.9
26
1.39
9.7
47
1.54
20.4
X = mean in mg/L.
109
-------
Table E-2. Comparison of Mean pH Values for Field Audit Samples, Western Lake Survey - Phase I
Field Base (Subregion) Measurement for
Closed-System pH
Analytical Laboratory3
Measurement for
Open-System pH
Audit Sample
FN3
(Lake
Superior)
FN4
(Big
Moose
Lake)
FN5
(Bagley
Lake)
FN6
(Bagley
Lake)
FL11
(Synthe-
tic)
FL12
(Synthe-
tic)
FL11 and
FL12
(Pooled
Synthetics)
n
X
SD
n
X
SD
n
X
SD
n
X
SD
n
X
SD
n
X
SD
n
X
SD
Cali-
fornia
(4A)
5
7.76
0.07
4
4.72
0.02
19
6.98
0.04
21
7.14
0.05
2
6.87
0.01
8
7.00
0.07
10
6.98
0.08
Pacific
NW
(4B)
5
7.66
0.17
4
4.75
0.04
12
7.01
0.05
11
7.08
0.09
2
6.86
0.06
8
7.06
0.07
10
7.02
0.11
No.
Rockies
(4C)
10
7.81
0.06
4
4.86
0.06
15
6.95
0.06
—
-
-
6
6.97
0.06
4
7.04
0.10
10
7.00
0.08
Cent.
Rockies
(4D)
9
7.86
0.04
4
4.79
0.03
12
7.04
0.06
—
--
-
5
7.01
0.02
2
7.12
0.02
7
7.04
0.06
So.
Rockies
(4E)
8
7.83
0.05
4
4.79
0.06
10
7.05
0.07
4
7.18
0.07
6
6.91
0.11
4
7.06
0.02
10
6.97
0.12
Sub-
regions
Pooled
37
7.80
0.10
20
4.78
0.06
68
7.00
0.07
36
7.13
0.07
21
6.94
0.08
26
7.04
0.07
47
7.00
0.09
Labi
19
7.88
0.05
g
4.68
0.03
24
7.08
0.08
8
7.11
0.06
11
6.97
0.15
6
6.99
0.07
17
6.98
0.13
Lab II
19
7.84
0.09
11
4.68
0.02
44
6.97
0.07
29
7.07
0.07
10
6.92
0.12
20
6.93
0.11
30
6.93
0.11
Labs
I and II
Pooled
38
7.86
0.08
20
4.68
0.03
68
7.01
0.09
37
7.08
0.07
21
6.94
0.14
26
6.95
0.10
47
6.95
0.12
a All analytical laboratory pH precision estimates are calculated from the pooled pH (acidity and alkalinity) determinations.
X = mean in pH units.
110
-------
Table E-3. Comparison of Mean Turbidity Values for Field Audit Samples, Western Lake Survey - Phase I
Field Base (Subregion)
Audit Sample
FN3
(Lake
Superior)
FN4
(Big Moose
Lake)
FN5
(Bagley
Lake)
FN6
(Bagley
Lake)
FL11
(Synthetic)
FL12
(Synthetic)
FL11 and
FL12 (Pooled
Synthetics)
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
California
(4A)
4
0.1
66.7
3
0.2
50.0
7
0.3
51.3
20
0.1
110.0
1
0.2
6
0.1
122.5
7
0.1
105.0
Pacific
NW
(4B)
5
0.1
36.2
4
0.3
12.4
g
0.2
37.1
11
0.1
27.6
2
0.3
0.0
7
0.1
38.0
g
0.2
52.0
No.
Rockies
(4C)
g
0.1
50.0
4
0.7
73.7
15
0.1
83.5
—
--
--
6
0.2
31.6
4
0.1
66.7
10
0.2
56.7
Cent.
Rockies
(4D)
g
0.2
101.7
4
0.3
62.1
12
0.3
63.1
--
--
--
3
0.1
43.3
2
0.3
84.8
5
0.2
72.4
So.
Rockies
(4E)
6
0.2
ig2.2
1
0.1
-•
5
0.1
104.6
4
0
0.0
4
0.4
82.6
4
0.3
148.0
8
0.3
103.2
Sub-regions
Pooled
33
0.1
123.0
16
0.4
go.4
48
0.2
74.6
35
0.1
90.0
16
0.2
70.2
23
0.1
125.8
sg
0.2
gs.s
Note: Audit samples were filtered before shipment to field laboratories. Field natural audit samples were filtered in the audit preparation
laboratory (see Appendix C); the synthetic audits were deionized water with added analytes spiked into them. Routine samples and
audit samples were prepared differently; therefore, for turbidity no inferences should be drawn from the precision estimates
calculated for audit samples. The data are presented here for illustrative purposes. To estimate precision for the turbidity
measurement with confidence, the data user should employ field duplicate pairs.
X = mean in NTU.
111
-------
Table E-4. Comparison of Mean True Color Values for Field Audit Samples, Western Lake Survey - Phase I
Field Base (Subregion)
Audit Sample
FN3
(Lake
Superior)
FN4
(Big Moose
Lake)
FN5
(Bagley
Lake)
FN6
(Bagley
Lake)
FL11
(Synthetic)
FL12
(Synthetic)
FL11 and
FL12 (Pooled
Synthetics)
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
n
X
%RSD
California
(4A)
3
1.7
173.2
3
21.7
13.3
7
0.7
264.6
17
0.3
412.3
2
0
--
6
0.8
244.9
8
0.7
282.8
Pacific
NW
(4B)
5
6
69.7
4
13.8
86.0
9
1.7
150.0
11
1.8
139.0
2
2.5
141.4
7
2.9
93.5
9
2.8
93.0
No.
Rockies
(4C)
9
3.9
56.7
4
17.5
16.5
15
4.7
63.6
..
..
--
5
6
37.3
4
3.8
66.7
9
5
50.0
Cent.
Rockies
(4D)
9
3.9
107.1
4
28.8
8.7
12
2.5
104.4
..
--
4
2.5
115.5
2
7.5
47.1
6
4.2
90.3
So.
Rockies
(4E)
3
5
100.0
1
25
2
2.5
141.4
4
3.8
66.7
1
5
3
1.7
173.2
4
2.5
115.5
Sub-regions
Pooled
29
4.1
86.0
16
20.6
39.5
45
2.8
105.5
32
1.3
176.0
•\A
I f
3.6
85.6
22
2.7
109.2
36
3.1
98.0
X = mean in PCU.
112
-------
Appendix F
Estimated Precision for Audit Sample by Lot
The following tables, figures, and discussions are
useful for the data user who is interested in
components of variability in each audit sample.
FN3 (Lake Superior) - Table F-1, Figures J-1
through J-26. Of the 38 samples, 19 were analyzed
by each laboratory. The %RSDs for SiC>2 (18.3%),
total dissolved F' (21.5%), and DOC (20.6%) are
higher than is desirable for pooled values or for
single-laboratory values. (The mean concentration
for DOC, however, is only 1.4 mg/L and for total
dissolved F" is 0.035 mg/L). This factor indicates
that for FN3 the ability to make consistent DOC and
SiO2 measurements was difficult, regardless of
laboratory. Much of the imprecision appears to result
from one unusually low measurement in Laboratory II
and one unusually high measurement in Laboratory I.
The precision estimates for Cl" were higher for
Laboratory I (9.6%) than for Laboratory II (2.1%).
There were not sufficient levels of extractable Al, total
Al, BNC, Fe, Mn, NH4 + , or total P in this Lake
Superior sample to determine confidence in precision.
Precision estimates for all other analytes are
considered reasonable relative to the DQOs.
FN4 (Big Moose Lake) - Table F-2, Figures J-1
through J-26. Laboratory II analyzed 11 samples and
Laboratory I analyzed 9 samples. Of all six audit lots,
only FN4 contained levels of total Al and extractable
Al high enough to allow meaningful precision
estimates to be determined. Both laboratories had
difficulty with extractable Al precision; however, only
one routine sample had a concentration above 0.100
mg/L. That sample was collected from a hot spring
that had a pH of about 5.70. Only Laboratory II had
difficulty with total Al precision (probably as the result
of two unusually low measurements). Laboratory I had
much better precision than did Laboratory II for Mn
(1.3% as compared to 13.6%); however, the pooled
%RSD was 9.9 percent. The mean concentrations for
Mn are the same, which indicates that the
concentrations are variable about the same mean for
the two laboratories. Mn, like extractable Al, generally
was found in extremely low concentrations in WLS-I
lakes; consequently, it may be of little concern. The
same trend was observed for total dissolved F" as
for Mn. Laboratory I's %RSD for SiO2 (16.2%) is
higher than desired for FN4, but Laboratory II had a
low %RSD (3.4%). Precision estimates for DIG (initial
and air equilibrated), NH4 + , and total P could not be
determined confidently because of their low
concentrations. All other analytes for this field audit
were close to the DQO for precision.
FN5 (Bagley Lake, sampled in January 1985) -
Table F-3, Figures J-1 through J-26. Because
Laboratory II analyzed almost twice as many samples
(44) as Laboratory I (24), Laboratory ll's results have
a greater effect on overall precision. That is, overall
precision estimates are weighted toward Laboratory
ll's results. Cl" was measured with a %RSD of 17.1
percent by Laboratory I, whereas Laboratory ll's
%RSD for Cl" was 6.2 percent; the pooled estimate
was 11.8 percent. For conductance, Laboratory II had
a %RSD of 6.9 percent; Laboratory I's %RSD for
conductance was below 4.8 percent; the pooled
%RSD was 6.4 percent. Laboratory I's %RSDs were
above 17 percent for air-equilibrated and initial DIG;
Laboratory ll's %RSDs were near 6 percent. Each
laboratory's mean values, however, were identical in
concentration, which indicates greater variability
among Laboratory I's measurements. The %RSDs for
NOa" were much higher for FN5 than for any other
field audit sample that contained NOa" levels high
enough to permit relevant precision estimates to be
calculated (FN3, FN4, FL11, and FL12). The pooled
%RSD of 23.1 percent represents Laboratory ll's
variability of 26.5 percent and Laboratory I's variability
of 11.7 percent and shows the effect of the weighting
factor. Air-equilibrated pH for both laboratories is
about 0.13 pH unit. Although the pooled value (8.6%)
for Si02 in this field audit was only slightly higher
than the DQO, Laboratory I's %RSD of 12.3 percent
is twice that of Laboratory ll's %RSD of 5.9 percent.
Analytes with concentrations below which reliable
precision estimates are questionable include
extractable Al, total Al, BNC, DOC, total dissolved F"
, Fe, Mn, NH4 + , and total P.
FN6 (Bagley Lake, sampled in September 1986) -
Table F-4, Figures J-1 through J-26. Of the 37
samples analyzed, 29 were analyzed by Laboratory II
and 8 were analyzed by Laboratory I; consequently,
the pooled precision is weighted toward Laboratory II.
Although the means and the %RSDs for BNC are
very different for Laboratory II (mean = 31.3, %RSD
= 6.1) and Laboratory I (mean = 18.5, %RSD =
113
-------
Table F-1. Precision Estimated from Audit Samples Analyzed Among Batches (Field Natural Audit
Lot 3 [FN3, Lake Superior]), Western Lake Survey - Phase I
Laboratories Pooled
(n = 38)
Laboratory I
Laboratory II
Variable3
Al, extractable
Al, total
ANC(neq/L)
BNC(iieq/L)
Ca
cr
Conductance
(liS/cm)
DIG, air
equilibrated
DIC, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH4 +
NO3"
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
SiO2
S042'
Mean
Concentration
0.002
0.012
846.1
21.9
13.84
1.43
95.5
9.90
9.86
1.4
0.035
0.005
0.52
2.90
-0.002
1.36
-0.010
1.418
0.001
7.86
7.85
8.13
2.51
3.24
Estimated
Precision
as %RSD
116.1
51.8
5.0
59.5
4.8
6.9
1.9
9.7
10.7
20.6
21.5
155.8
4.4
2.6
439.2*
2.5
221.2*
4.7
296.7
0.08C
0.07C
0.09C
18.3
6.4
Mean
Concentration
0.001
0.008
867.3
16.9
14.40
1.43
95.2
10.36
10.34
1.5
0.037
0.001
0.51
2.95
0.000
1.35
-0.019
1.393
-0.000
7.89
7.86
8.11
2.47
3.13
Estimated
Precision
as %RSD
233.0
66.8
4.3
63.6
1.8
9.6
1.7
11.5
12.8
15.1
16.6
649.7
4.4
2.6
668.6
2.8
1 54.6*
5.3
1011.2*
0.05C
0.06C
0.04C
21.1
7.2
Mean
Concentration
0.003
0.017
824.9
27.0
13.28
1.43
95.9
9.45
9.39
1.3
0.033
0.009
0.53
2.86
-0.004
1.37
-0.001
1.443
0.002
7.83
7.84
8.15
2.55
3.35
Estimated
Precision
as %RSD
59.6
20.1
4.5
49.8
3.2
2.1
2.0
2.1
3.0
23.2
25.3
87.8
3.9
1.5
259.9*
2.0
358.6*
3.3
81.6
0.1 Oc
0.08C
0.1 2c
15.5
3.6
a All variables are measured in mg/L unless otherwise noted.
* The absolute value of the %RSD.
c Standard deviation values were calculated for pH measurements.
18.1), these values are close enough to zero that
they should not be of great concern to the data user.
The overall precision for pH (air equilibrated) for
Laboratory II is 0.16 pH unit, which is slightly higher
than Laboratory I's 0.10 pH unit. For Si02, the
differences in the mean concentration between
laboratories (9.65 mg/L for Laboratory II and 8.33
mg/L for Laboratory I) may be of practical
significance. Extractable Al, total Al, BNC, DOC, total
dissolved F", Fe, Mn, NH4 + , NOa" and total P all
had mean concentrations that were too low to allow
precision to be estimated confidently for FN6. Of
added note is the NH4+ precision estimate of
20,069.1 percent for Laboratory II where, of the 29
values, 28 were near or at 0.000 mg/L and 1 was
0.034 mg/L. This shows how variable the %RSD can
be at extremely low concentrations.
There are differences between FN5 and FN6, the two
Bagley Lake samples that were collected during
different seasons of the same year. Of the 24
variables measured, 13 showed a significant change
between FN5 and FN6, 8 were near the detection
limit for both field audits, and 3 had no significant
change. All the measurable anions and cations
decreased from FN5 to FN6 (most notably NOa", the
mean concentrations for which decreased from 0.147
mg/L to 0.016 mg/L). Air-equilibrated and initial DIG,
conductance, ANC, BNC, and SiO2 mean
concentration also decreased. Only the initial and
air-equilibrated pH values showed no significant
difference between the two field audits. Although
there may be many factors that contribute to these
decreases in analyte concentrations over time,
seasonal effects are likely to be the primary factor.
114
-------
Table F-2. Precision Estimated from Audit Samples Analyzed Among Batches (Field Natural Audit
Lot 4 [FN4, Big Moose Lake]). Western Lake Survey - Phase I
Laboratories Pooled
(n = 20)
Variable*
Al, extractable
Al, total
ANC(ueq/L)
BNC(neq/L)
Ca
cr
Conductance
(yS/cm)
DIC, air
equilibrated
DIC, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH4*
NO3~
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
SiO2
SO42'
Mean
Concentration
0.195
0.352
-24.1
119.8
2.10
0.54
32.2
0.32
0.51
8.1
0.074
0.07
0.68
0.36
0.078
0.74
-0.001
2.351
0.002
4.68
4.68
4.70
4.45
6.68
Estimated
Precision
as %RSD
31.1
13.2
10.26
10.9
3.6
6.1
3.6
80.9
30.4
2.1
11.7
10.2
2.7
1.3
9.9
3.7
1770.46
4.8
141.4
0.03C
0.02C
0.03C
11.0
5.5
Laboratory I
(n = 9)
Mean
Concentration
0.236
0.368
-22.9
107.1
2.16
0.56
32.0
0.15
0.49
8.0
0.079
0.07
0.67
0.36
0.078
0.74
-0.007
2.330
0.001
4.68
4.68
4.69
4.29
6.81
Estimated
Precision
as %RSD
18.4
5.5
12.76
6.9
1.8
8.0
2.4
20.2
14.4
1.8
7.5
5.3
2.3
1.1
1.3
5.1
215.36
6.0
345.1
0.05C
0.02C
0.02C
16.2
7.3
Laboratory II
(n = 1 1 )
Mean
Concentration
0.162
0.339
-25.0
130.2
2.06
0.53
32.4
0.47
0.53
8.2
0.069
0.08
0.69
0.36
0.078
0.74
0.004
2.368
0.003
4.68
4.68
4.71
4.58
6.58
Estimated
Precision
as %RSD
32.4
17.2
6.26
2.9
3.4
2.5
4.4
60.3
38.7
1.0
10.8
9.3
2.4
1.3
13.6
2.3
182.8
3.8
87.4
0.02C
0.02C
0.04C
3.4
2.5
a All variables are measured in mg/L unless otherwise noted.
b The absolute value of the %RSD.
c Standard deviation values were used for pH measurements.
FL11 (Field Low Synthetic, Lot 11) - Table F-5,
Figures J-1 through J-26. Of the 21 samples
analyzed, Laboratory II analyzed 10 and Laboratory I
analyzed 11. For Ca, the mean concentrations for the
two laboratories differ; this difference, coupled with
Laboratory I's precision estimate at almost three
times that of Laboratory ll's precision estimate, results
in a %RSD of 17.3 percent for the laboratories
pooled. This is consistent with the field natural audit
precision results. Both DIG determinations
(laboratories pooled and by laboratory) show a %RSD
that is higher than desired (between 16% and 26%)
except for the initial DIG precision for Laboratory II
(2.7%). For initial DIG, however, the mean
concentrations for Laboratory I (1.49 mg/L) and
Laboratory II (1.97 mg/L) are far enough apart that the
precision for the laboratories pooled results in a
%RSD of 21.9 percent. For DOC, there are laboratory
(mean concentration) differences and there is
variability for the laboratories pooled and separated,
but the concentration (0.9 mg/L) may be too low to be
of concern. Precision for Na was much better for
Laboratory II (5.3%) than for Laboratory I (17.0%),
with a precision of 12.7 percent for the laboratories
pooled. In fact, there is one sample concentration that
makes the estimate so high; it was identified as a
dilution error at Laboratory I. When the dilution error
is corrected, the precision for Na is 4.8 percent.
Similarly, the precision estimates for K were much
better for Laboratory II (6.3%) than for Laboratory I
(11.6%). The NH4+ means and %RSDs were almost
identical, with a %RSD of 18.8 percent at a mean
concentration of 0.13 mg/L for the laboratories
pooled. Concentrations for extractable Al, total Al,
BNC, and Fe were too low for confident statistical
comparisons to be obtained.
FL12 (Field Low Synthetic, Lot 12) - Table F-6,
Figures J-1 through J-26. Of the 26 samples
115
-------
Table F-3.
Precision Estimated from Audit Samples Analyzed Among Batches (Field Natural
Audit Lot 5 [FN5, Bagley Lake, First Sampling]), Western Lake Survey - Phase I
Laboratories Pooled
(n = 68)
Variable^
Al, extractable
Al, total
ANC(neq/L)
BNC(neq/L)
Ca
cr
Conductance
(uS/cm)
DIG, air
equilibrated
DIG, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH4*
N03-
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
SiO2
so42-
Mean
Concentration
0.002
0.010
146.7
37.1
1.99
0.24
17.8
1.83
1.91
0.4
0.025
0.004
0.36
0.24
0.003
1.06
0.011
0.147
0.004
7.00
7.02
7.29
11.37
0.97
Estimated
Precision
as %RSD
120.4
43.1
3.5
16.1
3.2
11.8
6.4
11.3
13.3
79.2
8.0
177.8
5.5
2.0
351.7
5.2
106.1
23.1
197.2
0.106
0.096
0.136
8.6
7.6
Laboratory I
(n = 24)
Mean
Concentration
0.001
0.007
149.4
34.7
2.05
0.25
17.5
1.83
2.04
0.5
0.025
0.001
0.36
0.24
0.001
1.08
0.007
0.139
0.003
7.10
7.07
7.29
11.17
0.93
Estimated
Precision
as %RSD
193.5
37.4
5.0
16.1
2.3
17.1
4.8
17.6
17.9
25.6
11.0
744.4
4.9
2.1
359.8
7.1
160.8
11.7
140.9
0.076
0.096
0.136
12.3
9.2
Laboratory II
(n = 44)
Mean
Concentration
0.002
0.012
145.3
38.3
1.95
0.23
18.0
1.83
1.83
0.4
0.025
0.006
0.36
0.24
0.004
1.04
0.013
0.151
0.004
6.95
7.00
7.29
11.47
0.98
Estimated
Precision
as %RSD
89.4
34.8
1.6
15.2
2.2
6.2
6.9
5.9
6.1
102.7
5.3
131.7
5.7
1.9
299.0
3.2
86.7
26.5
202.7
0.076
0.076
0.136
5.9
5.8
a All variables are measured in mg/L unless otherwise noted.
6 Standard deviation values were used for pH measurements.
analyzed, Laboratory II analyzed 20 and Laboratory I
analyzed 6. For FL11, the Ca mean concentrations
differ considerably between laboratories. Laboratory
I's %RSD of 16.0 percent for Ca contributes
significantly to the %RSD of 12.2 percent for the
laboratories pooled. For the laboratories pooled, the
conductance precision (4.5%) was affected most by
the performance indicated from Laboratory II, which
had a %RSD of 5.0 percent for conductance. As with
FL11, the %RSDs for air-equilibrated DIG are less
than desirable for each laboratory and for the
laboratories pooled (about 15%). The initial %RSD for
DIG, however, was reasonable for all subsets. The
fact that Laboratory I analyzed only 6 FL12 samples
may help to account for the high variability for total
dissolved F" (11.5%) as compared to Laboratory II
(4.9%). Although the laboratories had very similar
mean concentrations for SiO2 (1.19 mg/L and 1.14
mg/L), Laboratory I's %RSD was 16.3 percent as
compared to Laboratory ll's %RSD of 7.2 percent.
The %RSD of 9.9 percent for the laboratories pooled
also was acceptable. Precision for total P showed a
large degree of variability for Laboratory II and for the
laboratories pooled; all %RSDs were 19 percent or
greater. Laboratory ll's imprecision for all pH precision
greatly contributed to the estimates for the
laboratories pooled. Extractable Al, total Al, BNC, and
Fe all had levels too low to allow consideration of
their precision estimates.
116
-------
Table F-4. Precision Estimated from Audit Samples Analyzed Among Batches (Field Natural
Audit Lot 6 [FN6, Bagley Lake, Second Sampling]), Western Lake Survey - Phase 1
Laboratories Pooled
Variable^
Al, extractable
Al, total
ANC(neq/L)
BNC(neq/L)
Ca
Cf
Conductance
(liS/cm)
DIG, air
equilibrated
DIG, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH/
N03-
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
SiO2
SO42'
-------
Table F-5. Precision Estimated from Audit Samples Analyzed Among Batches (Field Low Synthetic
Audit Lot 11 [FL11]), Western Lake Survey -Phase I
Laboratories Pooled Laboratory I Laboratory II
(n = 2l) (n = 11) (n = 10)
Variable^
Al, extractable
Al, total
ANC(neq/L)
BNC(ueq/L)
Ca
cr
Conductance
(US/cm)
DIG, air
equilibrated
DIG, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH/
N03-
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
Si02
so42-
Mean
Concentration
0.004
0.027
114.2
28.1
0.23
0.37
19.6
1.35
1.72
0.9
0.044
0.005
0.22
0.46
0.087
2.79
0.13
0.490
0.024
6.95
6.93
7.29
1.04
2.35
Estimated
Precision
as %RSD
39.0
37.5
7.4
31.2
17.3
5.3
4.4
19.7
21.9
38.9
5.6
153.7
9.9
4.1
12.9
12.5
18.8
5.3
10.2
0.1 6&
0.12*>
0.1 2t>
8.0
6.2
Mean
Concentration
0.005
0.019
112.6
22.6
0.26
0.37
19.5
1.22
1.49
1.1
0.044
0.003
0.22
0.46
0.097
2.75
0.13
0.474
0.024
6.98
6.95
7.21
1.00
2.27
Estimated
Precision
as %RSD
35.5
28.1
6.4
19.6
15.0
6.6
4.0
18.9
26.4
28.0
5.8
222.6
11.6
3.5
3.2
17.0
19.8
5.2
10.8
0.1 6<>
0.1 5*
O.Q8b
7.9
5.2
Mean
Concentration
0.003
0.035
116.0
34.1
0.20
0.38
19.6
1.48
1.97
0.8
0.044
0.008
0.21
0.46
0.077
2.83
0.13
0.507
0.025
6.93
6.91
7.37
1.07
2.44
Estimated
Precision
as %RSD
35.3
20.3
8.5
24.7
5.7
3.4
5.1
15.9
2.7
45.9
5.6
115.9
6.3
4.8
9.5
5.3
18.9
2.9
9.5
0.166
O.Q8b
0.116
6.6
5.0
a All variables are measured in mg/L unless otherwise noted.
b Standard deviation values were used for pH measurements.
118
-------
Table F-6. Precision Estimated from Audit Samples Analyzed Among Batches (Field Synthetic
Audit Lot 12 [FL12]), Western Lake Survey - Phase I
Laboratories Pooled
Variable3
Al, extractable
Al, total
ANC(neq/L)
BNC(ueq/L)
Ca
cr
Conductance
(nS/cm)
DIG, air
equilibrated
DIG, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH4 +
N03"
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
SiO2
S042-
(n =
Mean
Concentration
0.005
0.028
108.4
31.6
0.20
0.35
19.7
1.51
1.39
1.1
0.042
0.006
0.21
0.44
0.105
2.76
0.17
0.478
0.025
6.93
6.96
7.20
1.15
2.26
26)
Estimated
Precision
as %RSD
50.9
21.8
3.0
28.0
12.2
4.1
4.5
16.1
9.7
8.8
7.6
154.2
7.1
1.8
9.4
2.5
5.2
7.4
28.0
0.1 1C
0.1 QC
0.1 4C
9.9
3.8
Laboratory I
(n-
Mean
Concentration
0.007
0.022
106.8
19.0
0.24
0.34
19.6
1.33
1.31
1.18
0.044
-0.001
0.22
0.45
0.092
2.84
0.18
0.456
0.029
7.02
6.97
7.15
1.19
2.16
6)
Estimated
Precision
as %RSD
19.2
24.2
5.4
7.5
16.0
6.4
1.6
15.6
5.8
2.2
11.5
248.66
5.0
2.0
1.1
2.4
4.7
6.3
19.6
0.05C
0.08C
0.06C
16.3
2.6
Laboratory II
(n =
Mean
Concentration
0.005
0.030
108.9
35.4
0.19
0.35
19.8
1.56
1.41
1.1
0.041
0.009
0.20
0.44
0.109
2.74
0.17
0.485
0.024
6.91
6.96
7.22
1.14
2.29
20)
Estimated
Precision
as %RSD
58.8
17.4
1.9
17.2
3.3
2.8
5.0
14.7
9.9
9.5
4.9
113.2
6.9
1.4
7.3
1.9
2.1
7.1
29.4
0.1 DC
0.11C
0.1 5C
7.2
2.9
a All variables are measured in mg/L unless otherwise noted.
b The absolute value of the %RSD.
c Standard deviation values were used for pH measurements.
119
-------
-------
Appendix G
Estimated Analytical Accuracy for Field Synthetic Audit Samples by Lot
Table G-1 presents accuracy calculations for FL11
only, with the laboratories pooled and separate. Ca
was biased high as a result of Laboratory I's
inaccuracy of +32.6 percent. Accuracy for total Al
was poor solely due to Laboratory It's value of + 73.0
percent. Mn was biased marginally low at -10.8%
also as a result of Laboratory ll's values; total P was
biased high as a result of Laboratory I's values. The
only specific trend is for NhU , where both
laboratories were biased low. When NH4 +
concentration is plotted over time (over a 3-week
period), the concentration tends to drop, producing
high inaccuracy (-13.1% in week 1 and -33.0% in
week 3). DOC also was biased low only for
Laboratory II.
Table G-2 presents accuracy calculations for FL12
(laboratories pooled and separate). Two analytes
showed high inaccuracy: total Al reflects the high
value for Laboratory II ( + 46.5%) and DOC reflects
the high value for Laboratory I ( + 18.0%). Laboratory
II showed slight inaccuracy for Mn ( + 11.2%) and
total P (-12.2%), but these values are not high
enough to affect overall accuracy. Laboratory ll's
values for Ca ( + 21.6%) and SiO2 ( + 11.2%) also
were not high enough to affect overall accuracy.
121
-------
Table G-1. Estimated Analytical Accuracy for Field Synthetic Audit Lot 11 (FL11), Western Lake Survey - Phase I
Laboratories Pooled
Laboratory I
Laboratory II
Variable*
Al, extractable
Al, total
ANC(neq/L)
BNC(neq/L)
Ca
Cl~
Conductance
(nS/cm)
DIG, air
equilibrated
DIG, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH/
NO3'
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
SiO2
so4*-
Concentration
0.020
0.020
— -
0.194
0.343
—
0.959
1.0
0.042
0.059
0.203
0.447
0.098
2.74
0.168
0.464
0.027
—
—
—
1.07
2.28
(n = 21)
0.0041
0.0265
114.2
28.1
0.231
0.371
19.6
1.35
1.720
0.942
0.0441
0.0053
0.216
0.460
0.087
2.791
0.134
0.490
0.0241
6.95
6.93
7.29
1.04
2.35
(%)
-79.5
+ 32.5
— -
....
+ 19.1
+ 8.2
—
+ 79.4
-5.8
+ 4.8
-91.0
+ 6.4
+ 2.8
-11.2
+ 1.9
-20.2
+ 5.6
-10.7
—
—
—
-2.9
+ 3.1
(n = 11)
0.0047
0.0192
112.6
22.6
0.257
0.366
19.5
1.22
1.49
1.10
0.0443
0.0026
0.222
0.459
0.097
2.751
0.134
0.474
0.0235
6.98
6.95
7.21
1.00
2.27
(%)
-76.5
-4.0
....
—
+ 32.6
+ 6.7
....
+ 55.4
+ 10.0
+ 5.5
-95.6
+ 9.4
+ 2.7
-1.0
+ 0.4
-20.2
+ 2.2
-13.0
+ 7.0
-0.4
(n = 10)
0.0034
0.0346
116.1
34.12
0.202
0.375
19.6
1.48
1.97
0.80
0.0438
0.0082
0.209
0.461
0.077
2.828
0.134
0.507
0.0247
6.93
6.91
7.37
1.07
2.44
(%)
-83.0
+ 73.0
—
—
+ 4.2
+ 9.4
+ 105.4
-20.0
+ 4.3
-86.1
+ 3.0
+ 3.1
-21.4
+ 3.2
-20.2
+ 9.3
-8.5
+ 0.3
+ 7.0
a All variables are measured in mg/L unless otherwise noted. Mean concentrations are presented at the number of significant
figures useful in estimating accuracy.
b A plus sign ( + ) indicates that the mean concentration was higher than the theoretical concentration; a minus sign (-) indicates
that the mean concentration was lower than the theoretical concentration.
122
-------
Table G-2. Estimated Analytical Accuracy for Field Synthetic Audit Lot 12 (FL12), Western Lake Survey - Phase I
Variable*
Laboratories Pooled
Laboratory I
Laboratory II
Mean Mean Mean
Theoretical Concentration Accuracy*1 Concentration Accuracy'3 Concentration Accuracy^
Concentration (n = 26) (%) (n = 6) (%) (n = 20) (%)
Al, extractable
Al, total
ANC(peq/L)
BNC(neq/L)
Ca
cr
Conductance
(nS/cm)
DIG, air
equilibrated
DIG, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH/
NO3'
P, total
pH, acidity
(pH units)
pH, alkalinity
(pH units)
pH, air
equilibrated
(pH units)
SiO2
SO42'
0.020
0.020
....
0.194
0.343
—
....
0.959
1.0
0.042
0.059
0.203
0.447
0.098
2.74
0.168
0.464
0.027
—
....
....
1.07
2.28
0.0050
0.0275
108.4
31.6
0.204
0.348
19.7
1.51
1.39
1.12
0.0419
0.0063
0.207
0.440
0.105
2.76
0.170
0.478
0.0249
6.93
6.96
7.20
1.15
2.26
-75.0
+ 37.5
....
....
+ 5.2
+ 1.5
....
....
+ 44.9
+ 12.0
-0.2
-89.3
+ 2.0
-1.6
+ 7.1
+ 0.7
+ 1.2
+ 3.0
-7.8
....
....
....
+ 7.5
-0.9
0.0066
0.0216
106.8
19.0
0.236
0.338
19.6
1.33
1.31
1.18
0.0443
-0.001
0.219
0.448
0.092
2.84
0.184
0.456
0.0290
7.02
6.97
7.15
1.19
2.16
-67.0
+ 8.0
....
....
+ 21.6
-1.4
....
....
+ 36.6
+ 18.0
+ 5.5
-101.7
+ 7.9
+ 0.2
-6.1
+ 3.6
+ 9.5
-1.8
+ 7.4
— -
— -
+ 11.2
-5.2
0.0046
0.0293
108.9
35.4
0.195
0.351
19.8
1.56
1.41
1.10
0.0412
0.009
0.204
0.438
0.109
2.74
0.166
0.485
0.0237
6.91
6.96
7.22
1.14
2.29
-77.0
+ 46.5
....
....
+ 0.5
+ 2.3
....
....
+ 47.0
+ 10.0
-1.9
-84.7
+ 0.5
-2.0
+ 11.2
0.0
-1.2
+ 4.5
-12.2
....
....
+ 6.5
+ 0.4
a All variables are measured in mg/L unless otherwise noted. Mean concentrations are presented at the number of significant
figures useful in estimating accuracy.
b A plus sign ( + ) indicates that the mean concentration was higher than the theoretical concentration; a minus sign (-) indicates
that the mean concentration was lower than the theoretical concentration.
123
-------
-------
Appendix H
Field Audit Sample Control Limits and Summary of Field Audit Samples Outside
Control Limits
Final audit sample control limits were generated after outside' the control limits were considered suspect
all analytical laboratory data (149 batches) had been and were the basis for requesting confirmation of the
entered into the raw data set. Values that were values reported by the analytical laboratories.
125
-------
Table H-1. Field Audit Sample Control Limits,
FN3
(Lake Superior;
Variable3
Al, extractable
Al, total
ANC (neq/L)
BNC (jieq/L)
Ca
cr
Conductance
(liS/cm)
DIC, air equilibrated
DIC, initial
DOC
F ", total dissolved
Fe
K
Mg
Mn
Na
NH/
N03-
P, total
pH, acidity (pH units)
pH, alkalinity (pH
units)
pH, air-equlibrated
(pH units)
Si02
S042'
Control
Lower
Limit
-0.0029
-0.001
807.36
-4.81
12.48
1.354
93.67
9.034
7.696
1.13
0.0307
-0.010
0.472
2.748
-0.0120
1.300
-0.015
1.283
-0.0030
7.700
7.724
8.018
2.196
2.814
Limits
Upper
Limit
0.0071
0.0250
882.89
48.64
15.21
1.510
97.78
10.138
12.033
1.58
0.0405
0.019
0.567
3.058
0.0075
1.423
0.010
1.554
0.0041
8.028
7.988
8.224
3.002
3.668
Western Lake Survey - Phase 1
FN4
n = 38) (Big Moose Lake; n = 20)
Number of Samples
Outside Control Limits
Below
2
-
1
1
-
4
2
-
-
2
3
2
-
1
1
5
-
-
2
2
1
4
1
Above
-
-
5
1
-
2
2
5
4
3
3
2
1
1
3
1
1
1
2
-
1
2
2
2
Control
Lower
Limit
0.0655
0.2488
-29.32
91.96
1.941
0.491
30.18
-0.235
0.338
7.74
0.0552
0.058
0.638
0.351
0.063
0.679
-0.027
2.116
-0.0056
4.607
4.636
4.654
3.561
6.089
Limits Number
Outside <
1 Ippor
Limit Below
0.3244
0.4517 1
-18.81
147.59
2.263
0.588 1
33.89
0.881
0.624
8.46
0.0920 1
0.088
0.716
0.371
0.094
0.796 1
0.026 1
2.614
0.0107
4.756 1
4.717
4.759 1
5.134 1
7.085
• of Samples
Control Limits
Above
-
1
-
1
-
.
1
2
1
-
1
1
1
-
2
1
-
-
-
-
1
1
1
3
(continued)
mg/L unless otherwise indicated.
126
-------
Table H-1. (continued)
FN5
(Bagley Lake; o = 68)
FN6
(Bagley Lake; n = 37)
Control Limits
Variable3
Al, extractable
Al, total
AIMC (peq/L)
BNC (neq/L)
Ca
cr
Conductance
(nS/crn)
DIG, air equilibrated
DIG, initial
DOC
F ", total dissolved
Fe
K
Mg
Mn
Na
NH/
N03-
P, total
pH, acidity (pH units)
pH, alkalinity (pH
units)
pH, air-equlibrated
(pH units)
SiO2
SO/'
Lower
Limit
-0.0024
-0.0032
137.86
27.61
1.860
0.213
15.68
1.409
1.397
0.09
0.0214
-0.008
0.325
0.228
-0.013
0.980
-0.013
0.116
-0.0021
6.800
6.852
7.096
9.157
0.827
Upper
Limit
0.0059
0.0161
154.87
45.53
2.116
0.248
19.89
2.257
2.407
0.66
0.0284
0.016
0.394
0.246
0.019
1.124
0.035
0.160
0.0057
7.208
7.195
7.510
13.725
1.095
Number of Samples
Outside Control Limits
Below
2
-
1
2
1
-
2
3
1
2
1
1
1
3
1
2
-
-
2
2
6
1
-
Above
2
3
6
4
3
9
4
3
7
4
2
4
4
1
4
2
1
7
8
1
1
1
3
3
Control
Lower
Limit
0.0012
0.0095
1 19.82
166.67
1.505
0.147
13.21
1.305
1.285
-0.01
0.0169
-0.006
0.274
0.167
-0.015
0.779
-0.017
-0.005
-0.0017
6.909
6.952
7.051
8.182
0.576
Limits
Upper
Limit
0.0115
0.0212
122.01
40.41
1.673
0.178
15.24
1.662
1.632
0.49
0.0255
0.029
0.309
0.180
0.021
0.850
0.013
0.021
0.0036
7.223
7.237
7.503
10.654
0.692
Number of Samples
Outside Control Limits
Below
-
1
1
2
-
1
1
2
-
-
2
-
1
1
-
-
3
-
-
1
2
3
2
-
Above
1
1
5
-
-
1
_
-
3
1
1
2
-
2
-
2
5
3
-
_
_
-
2
(continued)
mg/L unless otherwise indicated.
127
-------
Table H-1. (continued)
FL11
(Synthetic, Lot 11; n = 21)
Variable3
Al, extractable
Al, total
ANC (iieq/L)
BNC (neq/L)
Ca
cr
Conductance
(liS/cm))
DIG, air equilibrated
DIG, initial
DOC
F ", total dissolved
Fe
K
Mg
Mn
Na
NH4 +
N03~
P, total
pH, acidity (pH units)
pH, alkalinity (pH
units)
pH, air-equlibrated
(pH units)
Si02
so/-
Control
Lower
Limit
0.0007
0.0054
99.14
9.45
0.158
0.337
17.72
0.628
0.902
0.16
0.0388
-0.012
0.170
0.420
0.064
2.565
0.078
0.440
0.0188
6.619
6.673
7.026
0.860
2.039
Limits Number of Samples
Outside Control Limits
Limit Below Above
0.0075
0.0477
126.87 - 1
46.73
0.293 - 1
0.399 - 1
21.40
2.223
2.377
1.83
0.0494
0.023 - 1
0.262
0.500
0.111
3.149 1
0.184
0.544 1
0.0293
7.295 - 1
7.189
7.547 - 1
1.211
2.663
FL12
(Synthetic, Lot 12- n = 26)
Control
Lower
Limit
-0.0003
0.0149
102.46
13.08
0.181
0.318
18.09
0.999
1.108
0.909
0.0370
-0.014
0.176
0.423
0.084
2.616
0.154
0.447
0.0202
6.714
6.749
6.912
1.053
2.081
Limits
Upper
Limit
0.0104
0.0401
113.68
50.14
0.208
0.378
21.19
2.016
1.673
1.321
0.0454
0.027
0.237
0.457
0.125
2.905
0.184
0.521
0.0290
7.156
7.173
7.496
1.238
2.440
Number of Samples
Outside Control Limits
Below Above
-
1 1
2 1
_
4
1
2
_
_
-
2
2
1
1
_
2
2
2 1
2 2
1
1
2
1 1
1 1
mg/L unless otherwise indicated.
128
-------
Table H-2. Cumulative Number of Field Audit Samples
Outside Control Limits, Western Lake Survey -
Phase I
Variable
Al, extractable
Al, total
ANC
BNC
Ca
CL"
Conductance
DIC, air
equilibrated
DIC, initial
DOC
F", total
dissolved
Fe
K
Mg
Mn
Na
NH/
N03'
P, total
pH, acidity
pH, alkalinity
pH, air
equilibrated
SiO2
SO/'
No. of
Samples
Below Limit
4
3
5
5
1
7
5
5
-
3
8
1
4
2
4
4
11
3
2
6
7
11
g
2
No. Of
Samples
Above Limit
1
6
18
6
8
13
g
10
15
8
g
12
7
5
g
6
6
14
15
3
3
7
7
11
Total
(n = 210)a
5
g
23
11
g
20
14
15
15
11
17
13
11
7
13
10
17
17
17
g
10
18
16
12
Total 320
' n = the number of audit samples. There are 24 variables per n;
therefore, 5,040 analyses were performed.
129
-------
-------
Appendix I
Relative Interlaboratory Bias in the Western Lake Survey - Phase I
Prepared by
Thomas Permutt, Mithra Moezzi and Stella C. Grosser
Systems Applications, Inc.
101 Lucas Valley Road
San Rafael, CA 94903
Introduction
In the Western Lake Survey, water from
approximately 700 lakes in the western United States
was sampled to study its chemical composition. After
preservative treatment at field laboratories, the
samples were shipped to one of two contract
laboratories for analysis. Water from each lake was
thus analyzed by one laboratory. Futhermore, water
from all lakes in an area was analyzed by the same
laboratory.
Some water was analyzed by both laboratories. For
example, 50 wilderness lakes were visited both by
helicopter and ground access, and duplicate samples
from these lakes were sent to both laboratories. This
report focuses on audit samples, another example of
water analyzed by both laboratories.
Two types of audit samples were included. Natural
audits were made of water collected one time in large
quantities from Lake Superior and Big Moose Lake
and two times from Bagley Lake. Small samples of
this water were shipped to the field laboratories,
treated as usual, and reshipped to the contract
laboratories. Synthetic audits were made up from
stock solutions according to a recipe designed to give
concentrations close to the limits of quantitation for
most analytes. These were shipped to the field
laboratories also, and treated and reshipped for
analysis along with routine samples.
Since they are repeated measurements of the same
water, the audit determinations provide information
about the precision of the measurement process.
Also, for the synthetic audits, the measurements can
be compared with the theoretically known
composition to provide information about absolute
bias. The subject of this work, however, is relative
interlaboratory bias. That is, how do measurements of
the same water by the two contract laboratories differ
on average?
The question is especially important in view of the
design of the Western Lake Survey. Because
samples from different regions went to different
laboratories, what is really a difference between
laboratories could appear as a difference between
regions or vice versa. The audit samples, however,
should allow any differences that are only due to
laboratories to be distinguished.
Preliminary Analysis
We will illustrate the analysis of relative bias by
looking at one parameter, calcium, in detail. Figure
1-1 shows the concentrations of calcium measured
by the two contract laboratories in audit samples of
water from Lake Superior (sample code FN3). The
measurements by Versar (V) are consistently
somewhat higher than those by EMSI (E). The
difference in means is 1.11 mg/L, or about 8 percent
of the concentration. The standard deviations are
only 0.43 and 0.26 mg/L, so the measurements by
the two laboratories overlap very little. The standard
errors are 0.10 and 0.06 mg/L, so the standard error
of the difference is 0.12 mg/L. That is, we can be 95
percent certain that if there .were a very large number
of audit samples from Lake Superior, the difference in
means would be within 0.24 mg/L of 1.11 mg/L.
Whatever the practical significance of a relative bias
of this magnitude, there is clearly no doubt as to its
statistical significance.
Figure I-2 shows the measurements of calcium in
audit samples from Bagley Lake (sample code FN5).
The pattern is similar, on a different scale. The
131
-------
Figure 1-1. Measurements of calcium in Lake Superior audit samples (FN3). (V = VERSAR, E = EMSI).
O!
as
O
14.75
14.50
14.25
14.00
13.75
13.50
13.25
13.00
12.75
12.50
V
V V
V
V
V
V
E
E
13Sep1985 19Sep1985 25Sep1985 01Oct1985 07Oct1985 13Oct1985
Note: 1 DBS Hidden Date Sampled
difference in means is 0.10 mg/L, or about 5 percent
of the concentration. Apparently neither the amount of
the bias (in mg/L) nor the bias as a percentage of the
concentration stays the same from lake to lake.
This variation complicates the application of audit data
to routinely sampled lakes. If, for example, the
estimated bias for all the lots of audit samples were
0.10 mg/L, we might reasonably suppose that a
similar bias applied to routine samples as well. If,
instead, the bias appeared to be 8 percent of the
concentration in each lot of audit samples, we might
suppose the bias in routine samples to be 8 percent.
With different biases in different lots, however, it is
difficult to say what bias should be expected in a
given routine sample.
Perhaps bias in the audit samples can be explained
as a slightly more complex function of concentration,
and this function can be assumed to apply to routine
samples. A plausible model is that bias is a linear
function of concentration with nonzero slope and
intercept. To examine its applicability, we have
summarized the data from Figures 1-1 and I-2
along with the four other lots of natural and synthetic
audits in Figure I-3. Because Lake Superior is so
different from the rest, we have drawn the same
figure on two different scales, one including and one
excluding Lake Superior. Each of the six stars
represents the two mean measurements of calcium in
one of the six lots of audit samples. Error bars show
the standard deviations, and ellipses show 95 percent
confidence regions for the means. Thus most of the
spread of the individual measurements is within the
error bars; and we can be confident that the means of
a large number of measurements would fall inside the
ellipse. The line of identity is shown; if there were no
bias the stars would lie close (within an ellipse or so)
to this line. We also show the straight line that fits
best, in a sense that we will make precise later.
The intercept of the line is 0.02 mg/L, and the slope
is 1.04. The points on this line therefore represent a
relative bias of 0.02 mg/L plus 4 percent of the
concentration. At low concentrations the intercept
dominates, so the line indicates a bias of about 0.02
mg/L, whereas at high concentrations the slope
dominates, so the line indicates a bias of about 4
percent of the concentration. The estimated bias at a
concentration of 14 mg/L (Lake Superior) would be
0.02 +• (.04) (14) = 0.58 mg/L. This is only about
132
-------
Figure 1-2. Measurements of calcium in Bagley Lake audit samples (FN5). (V = VERSAR, E = EMSI).
\
en
g
|
'3
CO
O
2.125
2.100
2.075
2.050
2.025
2.000
1.975
1.950
1.925
1.900
1.875
1.050;
V
V
V V
V
V V
V
V V
V
E
E
E E
V VE
E E
E E
E E
E E
E
11Sep1985 20Sep1985 29Sep1985 09Oct1985 17Oct1985
Date Sampled
Note: 2 DBS had missing values 3 OBS Hidden
260ct1985
half the observed difference in laboratory means for
Lake Superior.
Thus, the fit is not very good. The scale on which
deviations of the stars from the line should be
regarded is indicated by the ellipses.These show the
amount of uncertainty in the position of the stars due
to random errors in measurement. It is impossible to
draw a line through all the ellipses. Thus, relative bias
cannot be described simply as a linear function of
concentration.
Several explanations of the variation in bias can be
posited. For example, the deviation of the extreme
upper-right star (Lake Superior) could be supposed
to result from a nonlinear effect. That is, the line
could be bent between this star and the next one.
Obviously there is little information about the shape of
the curve between these two stars, so that the bias at
concentrations between about 2 and 13 mg/L would
be hard to estimate.
Even below 2 mg/L, the fit of the line is not as good
as might be expected if bias were really a linear
function of concentration. Lots FL11 and FL12, for
example, have about the same concentration of
calcium, but the differences in laboratory means are
0.055 mg/L for lot FL11 and 0.041 mg/L for lot FL12.
The difference here is not statistically significant, but
the pattern persists for some other parameters,
especially total aluminum and dissolved inorganic
carbon. Different lots have different biases, even at
the same concentration; therefore, the bias for a
given lot may also depend on properties other than
the concentration of the analyte in question. For
calcium for the synthetic audits FL11 and FL12, the
important other property (if there is one) may be time.
Lot FL12 was analyzed later, and the bias may have
changed a little. It seems likely that the bias for a
given analyte for a given lot depends not only on the
concentration of the analyte but also on the
concentrations of other species.
Unfortunately, the data do not allow statistical
evaluation of the many possible explanations for
variation in bias. After all, there are only six lots of
audit samples; they come from only three lakes; and
the three lakes were not chosen randomly. The
replication of measurements within lots permits fairly
precise estimation of the bias for each lot as well as
the testing of hypotheses concerning linear
relationships. The limited number of lots, however,
133
-------
Figure 1-3.
20.0
18.0
16.0
Means by laboratory of natural and synthetic audit measurements for CA11. Each point represents one lot
Uncertainty shown by bars (standard deviation) and elipses (95% confidence region).
3.0
0
'0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.016.0 18.0 20.0
EMSI Measurement
1.0 2.0
EMSI Measurement
3.0
prevents authoritative statistical treatment of the
variation from lot to lot.
The audit data appear to be sufficient for the intended
purpose of assuring that interlaboratory biases are
within quality assurance guidelines. A precise
knowledge of the bias that would apply to replicated
measurements from any given lake, such as would be
required to correct individual observations for bias,
appears to be out of reach for calcium and some
other parameters because of the variability of the
bias. In these cases, the best estimate of the bias for
a given lake would depend on a judgment as to which
of the audit lots the lake is most like. This judgment
may be based on factors other than concentration
and cannot be considered a purely statistical problem.
Statistical Methods
In this section we propose five alternative statistical
models of the relative bias and random variation in
the measurements by the two laboratories on the six
lots of natural and synthetic audit samples. We
describe a method of estimating the parameters of
these models. We also discuss statistical hypothesis
tests that can be used to choose an appropriate
model for each parameter.
When we speak of relative bias, we mean the
difference between the long-run means of
measurements by Versar and EMSI. We would like to
be able to conclude that there is no relative bias at
all. Failing this, it would be good to be able to identify
a simple relationship among the biases for the six
lots. For example, if the bias were the same for each
of the six lots, it would lend credence to the argument
that the bias is also the same for routine lake
samples. If this bias could be estimated with
satisfactory precision, it could then be used to adjust
the routine measurements for relative bias. Similarly,
if the bias were the same fraction of the measured
concentration for each lot, this relationship might
apply to routine samples. Alternatively, the relative
bias might be a linear function of concentration with
both slope and intercept nonzero.
Obviously many other, more complicated relationships
could exist between bias and concentration. Indeed,
there is certain to be some fairly simple function
whose graph passes through all six stars, as with any
six points. Given the wide choice of functions,
however, it would be impossible to obtain statistical
confirmation that the chosen function was the correct
one, and difficult to argue its applicability to routine
samples.
We therefore confine our attention to four simple,
useful models of bias; for the purpose of hypothesis
testing we also include a completely general model.
The four simple models are no bias, constant bias,
constant fraction bias, and bias as a linear function of
concentration. The general model is that the bias is
different for each lot; that is, six parameters are
required to describe the bias in six lots rather than the
one or two parameters of the linear model.
134
-------
The general model has much in common with
random-effects models. Besides the within-lot
variability, there is assumed to be an additional
source of variation that affects different lots differently
but is consistent within lots. Formal random-effects
models are not appropriate to this problem, however.
In the first place, the lots are not a random sample
from any population of lots, and no a priori
assumption about the distribution of lot effects seems
reasonable. Second, with only six lots no empirical
model of the lot-to-lot variation could be adequately
tested.
With respect to Figure 1-1, the five models can be
summarized as follows:
1 . The true, long-run positions of the stars are on a
line through the origin with slope 1. Deviations of
the actual locations from the line result from
random error.
2. The true positions are on a line with slope 1 , but
not through the origin.
3. The line goes through the origin, but the slope is
not 1 .
4. The true positions of the stars are .on a line not
through the origin and with slope different from 1 .
5. The true positions of the stars are not on a line.
The observed deviations are too large to have
resulted from random error.
More formally, we may write
EJJ = pi + 8jj
and one of the following:
(1) Vjj = pj + eij
(2) Vjj = pi + a + ejj
(3) Vjj = pi + Ppi + eij
(4) Vjj = W + a + fe + cjj
(5)
= pi + Yi + eij;
with
8jj - N (O, q2)
N (O,
where EJJ is the jth measurement by EMSI on the ith
lot and Vij is the same by Versar. The terms 8 and e
are independent, normally distributed errors; their
variances, oj2 and Xi2, may vary from lot to lot as well
as from laboratory to laboratory. The term pi is the
long-run mean concentration that would be
measured by EMSI; the terms in a, p, or yi, depending
on the model, represent the interlaboratory bias
(Versar relative to EMSI). Of course, there is no
implication that it is Versar's measurements that are
biased; the model could be rewritten, switching the
roles of the two laboratories, without affecting the
results.
The problem is conceptually similar to what has been
called structural or orthogonal regression. In
ordinary.regression a line is chosen to minimize the
sum of squared vertical distances from points to the
line. This imposes an asymmetry on the problem in
that vertical and horizontal; variables, which we have
assigned arbitrarily to Versar and EMSI, are treated
differently. The effect would be to underestimate the
slope of the underlying relationship, and to
overestimate the intercept. Thus a slope significantly
less than 1 and an intercept significantly different
from zero might be expected to result from random
error alone even if there were no interlaboratory bias.
In structural regression the two variables are treated
symmetrically. In the common situation of paired
observations the structural problem is difficult. Not all
the parameters can be estimated, and it is therefore
necessary to make assumptions about relationships
among them. The present problem is different
because of the multiple observations on each lot. All
the parameters a, p, yi. Pi» °i2. and ^i2 can be
estimated by the method of maxium likelihood.
The computations can be done by iteratively
reweighted least squares using a nonlinear least-
squares program like SAS PROC NLIN or BMDP3R.
Given estimates of the oj2 and Tj2 can be estimated
from the deviations of the data from the fitted values.
The weighted least-squares estimates and the
weights are alternately recalculated until convergence
is achieved.
The method of maximum likelihood also provides
statistical tests of various relevant hypotheses. If the
statistic 1 = -2 log L is calculated for each model,
where L is the maximum value of the likelihood
function, differences in 1 from model to model can be
referred to the chi-square distribution. For example,
the difference between the no-bias model and the
two-parameter linear-bias model was approximately
a chi-square distribution with two degrees of
freedom when the hypothesis of no bias is true.
Similarly, either of the one-parameter models can be
tested against the two-parameter linear-bias model;
and the linear-bias model can be tested against the
six-parameter general model.
Results
The results of our study of interlaboratory bias in
natural and synthetic audit samples are best seen in
the figures collected at the end of this section. Each
figure shows the following for a single parameter.
135
-------
The line of identity. If there were no bias, mean
measurements would fall near this line.
A star for each lot of audit samples, representing
the mean measurements by EMSI and by
Versar. The distance of a star from the line of
identity is the apparent relative bias for that lot.
Error bars showing the spread of measurements
by EMSI and Versar around the means.
A 95 percent confidence ellipse for each pair of
means. It is very likely that the bar would lie
somewhere in this region if there were many
samples in each tot. Therefore, deviations from
the line of identity or the calibration line that are
larger than the ellipses cannot be supported to
result from random error.
Our best estimate of the calibration line,
assuming bias is a linear function of
concentration. It has intercept a and slope 1 +
P, where d and 0 are the maximum-likelihood
estimates of the slope and intercept.
Table 1-1 contains likelihood-ratio statistics for
testing several hypotheses. The first column is used
to test the hypothesis that the bias is a linear function
of concentration against the general alternative that
the bias is different for each lot. The number reported
is the difference between the values of 1 = -2 log L
for the two models, where L is the maximum value of
the likelihood function. Values above 9.5 lead to
rejection of the hypothesis at the 5 percent level of
significance. This happens for 10 parameters, which
we designate Group 1 and list in Table I-2. For
these parameters bias cannot be considered to be a
linear function of concentration. Table I-3 contains
estimates of the bias for these (and other) parameters
by lot. The mean for each laboratory is given along
with the estimated relative bias, the standard error of
this estimate, and the bias as a percentage of the
EMSI mean.
For the other 14 cases, where the hypothesis of
linear bias can be accepted, the next three columns
test simpler models. The second column tests the
hypothesis of no intercept against the two-parameter
alternative. If the number here is less than 3.8, bias
can be considered to be proportional to
concentration. We have designated the six
parameters for which this is so as Group 3.
Similarly, the third column tests the hypothesis that
the slope of the bias is zero; i.e., that bias in absolute
terms is constant across lots. We catl the four
parameters for which this hypothesis is accepted
Group 4. Conductivity and air-equilibrated pH are in
both Groups 3 and 4. The bias for these parameters
is approximately constant, but the measurements are
so far from zero that the fit is about equally good
whether the bias is forced to have intercept zero or
slope zero. Certainly, in the case of pH and perhaps
in the case of conductivity, a measurement of zero
has no special importance. The constant-bias modeJ
therefore seems preferable on grounds of simplicity.
Statistically, however, it is not possible to discern
whether the bias in conductivity or air-equilibrated
pH varies with the measured value.
For two parameters, designated Group 5, the line of
identity fits the data acceptably. This is indicated by a
number befow 4.6 in the last column. In these two
cases the hypothesis that there is no bias at all can
be accepted.
The four remaining parameters are called Group 2. In
these cases a straight calibration line fits acceptably,
but both a nonzero slope and a nonzero intercept are
required.
Conclusions and Recommendations
Most of the estimated biases are well below 10
percent, which is the objective for most parameters.
There are numerous exceptions at low
concentrations, but this is not surprising. To take an
extreme example, at zero concentration any bias at all
is an infinite percentage bias. It is implicit that the
data quality objective is not meant to apply at such
low concentrations, but perhaps the lower limit of
applicability should be an explicit part of the data
quality objective.
Even though the measurements appear to meet data
quality objectives in respect of interlaboratory bias,
the question arises whether data quality can be
improved by adjusting for the apparent bias. We do
not believe this can be achieved. For almost half the
parameters no acceptable model of the variation of
bias with concentration was found. For the others,
hypotheses that simple models apply could not be
rejected, but this failure to reject should not be
interpreted as strong evidence that the simple models
are correct. There is evidence that interlaboratory
bias varies from lake to lake. Since the sample of
lakes used for audit samples is neither large nor
random, not much information about the nature of this
variation is available. Correcting for the estimated bias
therefore seems almost as likely to do harm as good
for a particular lake, and to improve overall measures
of data quality very little.
We believe that most users of the data will find the
relative interlaboratory bias in the Western Lake
Survey to be within acceptable limits. Furthermore,
the bias is unusually well documented because of the
design of the quality assurance program. This
documentation should provide the most sophisticated
users with the means of adjusting data to suit their
purposes.
136
-------
Table 1-1. Likelihood-Ratio Test Statistics for Testing Linear Models of Bias
Parameter*
Acidity (neq/L)
Aluminum
(extractable)
Alkalinity (neq/L)
Aluminum (total)
Calcium
Chloride
Conductivity (nS/cm)
DIC (equilibrated)
DIC (initial)
DOC
Iron
Fluoride (total)
Postassium
Magnesium
Manganese
Sodium
Ammonium
Nitrate
pH (acidity)
pH (alkalinity)
pH (equilibrated)
Phosphorus (total)
Silica
Sulfate
Linear Bias?
(4 d.f.)
37.4
8.2
13.8
23.1
23.5
8.5
1.3
19.0
28.1
7.7
7.3
7.2
3.1
4.6
61.3
13.9
2.5
5.3
15.7
6.3
8.4
8.2
9.6
9.2
X42 0.95 = 9-5
Zero
Intercept?
(1 d.f.)
8.6
1.1
1.0
36.6
16.5
1.3
13.9
3.3
15.1
1.5
2.4
0.7
5.7
0.4
*t2 0.95 = 3-8
Zero
Slope?
(1 d.U
23.8
0.3
0.5
20.4
1.6
3.9
19.1
13.1
13.7
4.0
3.5
2.0
0.4
15.1
No Bias?
(2 d.f.)
24.7
1.4
7.1
39.3
67.2
10.0
19.8
26.9
22.7
17.4
5.3
7.5
6.4
33.5
X22 0.95 = 6.0
'All measurements are in mg/L unless otherwise noted.
137
-------
Table 1-2. Linear Models of Bias
GROUP 1. Bias not a linear function of concentration.
Acidity
DIG (equilibrated)
pH (acidity)
GROUP 2. Bias linear
Aluminum
(extractable)
DOC
Potassium
Ammonium
Alkalinity
DIC (initial)
Silica
in concentration
Intercept
-0.0019
0.17
0.026
-0.010
Aluminum (total)
Manganese
with nonzero slope and
Standard
Error of
Intercept
0.0007
0.02
0.006
0.002
intercept.
Slope
0.68
-0.052
-0.075
0.16
Calcium
Sodium
Standard
Error of
Slope
0.17
0.008
0.015
0.03
GROUP 3. Zero intercept.
Fluoride (total)
Magnesium
Nitrate
Sulfate
Conductivity*
pH (equilibrated)"
GROUP 4. Zero slope.
Iron
Phosphorus (total)
Conductivity*
pH (equilibrated)*
-0.0083
-0.001 1
-0.32
-0.026
0.0009
0.0004
0.12
0.011
0.048
0.016
-0.041
-0.055
-0.01 1
-0.0048
0.016
0.003
0.009
0.007
0.004
0.0018
GROUP 5. No statistically significant bias.
Chloride
pH (alkalinity)
'Conductivity and pH (equilibrated) can be fit with a line of slope zero or with a different line of
intercept zero, but not by a line with slope and intercept both zero.
138
-------
Table 1-3. Estimated Relative Bias by Lot
PARK
ACCO
ACCO
ACCO
ACCO
ACCO
ACCO
ALEX
ALEX
ALEX
ALEX
ALEX
ALEX
ALKA
ALKA
ALKA
ALKA
ALKA
ALKA
ALTL
ALTL
ALTL
ALTL
ALTL
ALTL
CA11
CA11
CA11
CA11
CA11
CA11
CL11
CL11
CL11
CL11
CL1!
CL11
COND
COND
COND
COND
COND
COND
DICE
DICE
DICE
DICE
DICE
DICE
DICI
DICI
DICI
DICI
LOT
3
4
5
6
11
12
3
4
5
6
11
12
3
4
5
6
11
12
3
4
. 5
6
11
12
3
4
S
6
11
12
3
4
5
6
11
12
3
4
5
6
11
12
3
4
5
6
11
12
3
4
5
6
EMS I
26.958
130.173
36.355
31 .303
34.120
35.395
.003
.162
.002
.006
.003
.005
824.868
-25.036
145.268
120.934
116.050
108.940
0.017
0.307
0.012
0.016
0.035
0.029
13.2B5
2.057
1.954
1.574
0.202
0.195
•1.428
0.532
0.234
0.162
0.375
0.351
95.937
32.355
18.014
14.290
19.620
19.770
9.447
0.466
.835
.472
.651
.562
.393
0.528
.828
.424
VERSAR
16.879
IJT7.07B
34.679
18.537
22.609
18.983
BIAS »t»nd«rd
•rror of
bias
-10.079 3.9439
-23.095 2.7072
-3.671 1.4404
-12.766 1.2371
-11.511 2.9E74
-16.412 1.4785
0.001
1
r.236
0.001
0.008
0.005
0.007
867.253
4
-22.878
149.408
122.450
112.609
106.817
1
.008
.368
.007
.012
.019
.022
.402
.156
.051
.643
.257
.235
.430
.559
.248
.169
.367
.338
~
-
-
-
-
-
-
95.153
32.044
17.525
13.975
19.509
19.633
10.351
0.148
1.829
1.524
1.220
1.328
-
10.337
0.494
2.039
1.626
.002
.074
.001
.002
.001
.002
.384 1
.159
.140
.516
.441
.123
.009
.060
.005
.004
.015
.008
.117
.099
.096
.069
.055
.041
.002
.027
.014
.008
.009
.013
.784
.310
.489
.315
.111
.137
.904
.318
.006
.052
.431
.234
.944
.034
.212
.202
.0007
.0214
.0006
.0009
.0006
.0008
.0174
.0755
.5718
.4437
.7897
.3846
.0014
.0325
.0008
.0010
.0028
.0024
.1146
.0245
.0115
.0093
.0122
.0154
.0323
.0145
.0089
.0054
.0083
.0091
.5613
.5008
.2549
.1964
.3911
.2570
.2781
.0652
.0681
.0243
.1389
.0990
.3105
.0660
.0762
.0310
PCS IAS*
-37.39
-17.74
-9.57
-40.78
-33.74
-46.37
-63.49
45.84
-35.03
40.05
40.80
44.74
5.14
.
2.85
1.25
-2.97
-1 .95
-54.95
19.63
-42.02
-23.59
-44.51
-26.23
8.41
4.62
4.93
4.37
27.19
20.99
0.12
5.04
6.06
4.70
-2.29
-3.63
-0.62
-0.96
-2.71
-2.20
-0.57
-0.69
9.57
-68.27
-0.33
3.50
-26.10
-14.98
10.05
-6.46
11.57
14.20
Continued
139
-------
Table 1-3. Continued
FARM
DICI
DICI
DOC1
DOC1
DOC I
OOC1
DOC1
DOC1
mi
mi
FEl 1
FE1 1
FEU
FEU
FTL1
FTL1
FTL1
FTL1
FTL1
FTL1
Kll
Kll
Kll
Kll
Kll
Kl 1
MCI!
MG11
MG11
HG1 1
MCI 1
MC11
MN11
MN11
MN11
MN1 1
MN11
MN11
HA) I
NA11
NA11
NA11
NA11
NA11
NH41
NH41
NH41
NH41
NH41
NH41
N031
N031
LOT EMSI VERSAR BIAS «t.nd»rd
•rror of
bias
11 1.80630 1.46527
12 1.41430 1.31083
3 1.25053 1.47947
4
D
6
11
12
3
4
5
6
11
12
3
4
5
6
11
12
3
4
5
6
11
12
3
4
5
6
11
12
3
4
S
6
11
12
3
4
S
6
11
12
3
4
5
6
11
12
.21454 7.96444
.36545
.19566
.90100
.09550
.00863
.07664
.00611
.01379
.00820
.00870
.03316
.06860
.02461
.02105
.04381
.04122
.52879
.66616
.36295
.29086
.20880
.20355
.65789
.35918
.23591
.17324
.46110
.43765
.00395
.07754
.00407
.00421
.07730
.10855
.36789
.73527
.04475
.81479
.82600
.73765
.00147
.00427
.01341
.00003
.12120
.16625
.46125
.40750
. 10444
. 1 6000
.00058
.06211
.00054
.00237
.00264
.00150
.03736
.07949
.02551
.02185
.04433
.04433
.51056
.66644
.35658
.29367
.22245
.21650
.94768
.36276
.23892
.17575
45909
44600
00042
07800
02019
00025
09654
09233
34947
74000
07946
61237
74909
63717
01916
00678
00721
00700
10944
16367
3 1.44347 1.39295
4 2.36754 2.36109
.323
.103
.229
.250
.076
.212
.203
.085
.006
.015
.006
.011
.006
.010
.004
.011
.001
.001
.001
.003
.016
.020
.006
.003
.014
.015
.090
.004
003
003
002
010
004
000
016
004
019
016
018
005
035
002
079
100
016
011
006
007
012
017
051
006
.132035
.044265
.084028
.054223
.064330
.032059
.174192
.025566
.001941
.006360
.001467
.001635
.003489
.002678
.002395
.002996
.000613
.001477
.001105
.002132
.006971
.007025
.004743
.004117
.008812
.005478
.019892
.001983
.001138
.001765
.008530
.003680
.002441
.003204
.013635
.001625
.002508
001630
010682
.013566
016528
007096
148927
029784
006904
005404
002945
00533JS
023882
003573
019997
056506
rceiAS*
-17.86
-7.32
18.31
-3.04
19.66
108.05
22. SB
7.71
-93.29
-18.95
-91.14
-82.78
-67.85
-117.24
12.75
15.54
3.64
3.79
1.18
7.55
-3.44
-2.88
-1.76
1.04
6.54
7.34
3. 14
1.00
1.28
1 .45
-0.44
2.36
0.59
396.36
-94.06
24.90
-14.94
-1.35
0.64
3.32
-0.30
-2.79
3.64
-258.62
-46.25
-20688.24
-9.70
10.46
-3.50
-0.27
Continued
140
-------
Table 1-3. Continued
FARM
N031
N031
N031
N031
PHAC
PHAC
PHAC
PHAC
PHAC
PHAC
PHAL
PHAL
PHAL
PHAL
PHAL
PHAL
PHEO
PHEO
PHEO
PHEO
PHEQ
PHEO
PTL1
PTL1
PTL1
PTL1
PTL1
PTL1
SI 02
SI02
SI02
SI02
SI02
SI02
S041
S041
S041
S041
S041
S041
LOT
5
6
11
12
3
4
5
6
11
12
3
4
5
6
11
12
3
4
5
6
11
12
3
4
S
6
11
12
3
4
S
6
11
12
3
4
5
6
11
12
EMSI
0.1512
0.0068
0.5068
0.4848
7.8347
4.6845
6.9523
7.0455
6.9300
6.9080
7.8432
4.6764
6.9980
7.0948
6.9060
6.9575
8.1479
4.7073
7.2898
7.2634
7.3690
7.2200
0.0017
0.0238
0.0045
0.0016
0.0247
0.0237
2.5461
4.5378
11.6770
9.6497
1.0740
1 . 1003
3.3479
6.5772
0.9845
0.6394
2.4430
2.2910
VCRSAR 1
0.1395
0.0501
0.4786
0.4557
7.B937
4.6767
7.1004
7.1412
6.9773
7.0233
7.8579
4.6767
7.0700
7.0937
6.9536
6.9733
8.1132
4.6944
7.2871
7.2162
7.2118
7.1500
-0.0002
0.0008
0.0028
0.0014
0.0235
0.0290
2.4726
4.2871
11.3816
8.3289
1.0005
1.1883
3.1337
6.8114
0.9293
0.6155
2.2680
2.1565
HAS i
<
.012
.043
.028
.029
.059
.008
.148
.096
.047
.115
.015
.000
.072
.001
.048
.016
.035
.013
.003
.047
.157
.070
.002
.023
.002
.000
.001
.005
.074
.251
.295
.321
.074
.088
.214
.234
.055
.024
.175
.135
itandard
• rror of
bias
.006912
.026413
.008730
.014094
.024382
.017018
.018561
.022500
.070815
.031018
.023677
.008542
.021076
.029358
.051914
.041919
.029296
.013488
.032944
.047074
.042069
.041335
.000634
.019842
.001591
.000757
.001068
.002791
.150276
.241476
.642900
.171586
.032665
.084956
.058478
.173785
.019544
.012842
.052086
.027485
PCBIAS*
-7.74
638.07
-5.56
-6.00
0.75
-0.17
2.13
1.36
0.68
1.67
0.19
0.01
1.03
-0.02
0.69
0.23
-0.43
-0.27
-0.04
-0.65
-2.13
-0.97
-113.98
-96.51
-36.99
-10.41
-4.56
22.46
-2.89
-5.52
-2.53
-13.69
-6.85
8.00
-6.40
3.56
-5.61
-3.75
-7.16
-5.87
Bias as percentage of EMSI mean.
141
-------
VERSAR Measurement
o
(D Q>
ig
10 S
01 "
1
I?
3 -»
(Q -»
O
3-
O
3
(B
-------
Figure 1-6. Means by laboratory of natural and synthetic audit measurements for K11. Each point represents one lot. Uncertainty
shown by bars (standard deviation) and ellipses (95% confidence region).
0.80
0.60
a) 0.40
-------
Figure 1-8.
0.12
Means by laboratory of natural and synthetic audit measurements for MN11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
0.02
0.01
c
ID
£
3
g o.oo
-0.01 -
-0.02
-O.02 0.00
0.02 0.04 0.06 0.08
EMSl Measurement
-0.02
0.10 0.12 -0.02 -0.01 0.00
EMSl Measurement
0.01
0.02
Figure 1-9. Means by laboratory of natural and synthetic audit measurements for FE11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
0.10
0.08
0.06
0.04
£ 0.02
0.00
-0.02
-0.02 0.00
I
0.03
0.02
£ 0.01
3
(/]
DC
0.00
-0.01
0.02 0.04 0.06
EMSl Measurement
0.08
0.10
-0.02
-0.02 -0.01
0.00 0.01
EMSl Measurement
0.02
0.03
144
-------
Figure 1-10. Means by laboratory of natural and synthetic audit measurements for ALEX11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
0.30
0.20
c
0)
E
to
0.10
W
DC
0.00
-0.10
-0.10 0.00 0.10 0.20
EMSI Measurement
0.02
§ 0.01
CD
OT
£ o.oo
-O.01
0.30 -0.01
0.00 0.01
EMSI Measurement
0.02
Figure 1-11. Means by laboratory of natural and synthetic audit measurements for CL11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
2.0
0.2 -
0.7
0.6
0.5
W
DC
0.0
0.0 0.2
0.6 0.8 1.0 1.2 1.4
EMSI Measurement
1.6 1.8 2.0
0.4
0.3
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5
EMSI Measurement
0.6
0.7
145
-------
Figure 1-12. Means by laboratory of natural and synthetic audit measurements for SO411. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
2
DC
DC
2
0.0 1.0 2.0
3.0 4.0 5.0 6.0 7.0
EMSI Measurement
8.0 9.0 10.0
4.0
3.0
2.0
1.0
0.0
0.0 1.0 2.0
EMSI Measurement
3.0
4.0
Figure 1-13. Means by laboratory of natural and synthetic audit measurements for NO311. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
3.0
2.5
*i 2'°
1
| ,5
=! 1-5
(D
oc
g!
5!
CE 1.0
0.5
0.0
-0.5
J_
I
I
I
0.6
0.5
„ 0.4
c
o>
E
w
oc
-0.5 0.0 0.5 1.0 1.5 2.0
EMSI Measurement
0.3
0.2
0.1
0.0
-0.1
l
2.5 3.0 -0.1 0.0 0.1 0.2 0.3 0.4
EMSI Measurement
0.5 0.6
146
-------
Figure 1-14. Means by laboratory of natural and synthetic audit measurements for SIO211. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
14.0
12.0
10.0
8.0
CO
cc
6.0
4.0
2.0
0.0
_L
_L
J_
5.0
0.0 2.0 4.0 6.0 8.0 10.0
EMSI Measurement
12.0 14.0
0.0
1.0
2.0 3.0
EMSI Measurement
4.0
5.0
Figure 1-15. Means by laboratory of natural and synthetic audit measurements for FTL11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
0.09
0.08
^ 0.07
c
0)
I
0.06
1
0.04
CO
0.03
0.02
0.01
0.00
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
EMSI Measurement
0.04
c
-------
Figure 1-16. Means by laboratory of natural and synthetic audit measurements for DOC11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
10.0
-2.0
-2.0
2.0 4.0 6.0
EMSI Measurement
0.5 1.0
EMSI Measurement
1.5
2.0
Figure 1-17. Means by laboratory of natural and synthetic audit measurements for NH411. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
0.20
0.03
0.02
0.01
0.00
-0.03
-0.04
-0.05
-0.05
0.00
0.05 0.10
EMSI Measurement
0.15
0.20
-0.05
-0.05-0.04 -0.03 -0.02 -0.01 0.00 0.01
EMSI Measurement
0.02 0.03
148
-------
Figure 1-18. Means by laboratory of natural and synthetic audit measurements for PHEQ11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
c
01
9.0
8.0
7.0
6.0
5.0
4.0
8.0
CD
0>
7.0
4.0 5.0 6.0 7.0
EMSI Measurement
8.0
9.0
6.0
6.0
7.0
EMSI Measurement
8.0
Figure 1-19.
9.0
8.0
Means by laboratory of natural and synthetic audit measurements for PHAL11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
£
D
ID
10
0>
W
DC
7.0
6.0
5.0
4.0
I
CO
DC
'4.0 5.0 6.0 7.0
EMSI Measurement
8.0
9.0
7.0
EMSI Measurement
8.0
149
-------
Figure 1-20
9.0
Means by laboratory of natural and synthetic audit measurements for PHAC11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
8.0
m
£ 7.0
at
cu
-------
Figure 1-22.
1000
Means by laboratory of natural and synthetic audit measurements for ALKA11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
-100
-100 0 100 200 300 400 500 600 700 800 9001000
EMSI Measurement
-100 -50
0 50 100
EMSI Measurement
150 200
Figure 1-23.
100.0
90.0
80.0
Means by laboratory of natural and synthetic audit measurements for COIMD11. Each point represents one
lot. Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
40.0
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.090.0100.0
EMSI Measurement
0.0
10.0 20.0
EMSI Measurement
30.0
40.0
151
-------
Figure 1-24.
12.0
10.0 -
5
oc
M
DC
8.0 -
(Means by laboratory of natural and synthetic audit measurements for DICE11. Each point represents one lot
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
4.0 6.0 8.0
EMSI Measurement
0.5
1.0 1.5
EMSI Measurement
2.0
2.5
Figure 1-25
12.0
Un. n*hK f na'uralandj svnthetic audit measurements for DICI11. Each point represents one lot
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
4.0 6.0 8.0
EMSI Measurement
1.0 1.5
EMSI Measurement
2.5
152
-------
Figure 1-26. Means by laboratory of natural and synthetic audit measurements for PTL11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
0.02
-0.01
0.00
0.01 0.02 0.03
EMSI Measurement
0.04
0.00 0.01
EMSI Measurement
0.02
Figure 1-27
0.50
Means by laboratory of natural and synthetic audit measurements for ALTL11. Each point represents one lot.
Uncertainty shown by bars (standard deviation) and ellipses (95% confidence region).
0.05
0.04 -
0.03 -
0.02 -
0.10
0.20 0.30
EMSI Measurement
0.40
0.50
O.01 -
0.00
0.00
0.01 0.02 0.03
EMSI Measurement
0.04
0.05
153
-------
-------
Appendix J
Figures Depicting Detectability Data and the Relationship Between
Precision and Mean Concentration by Analyte
The figures presented in Appendix J should be useful
in assessing the quality of data in WLS-I based on
QA and QC data.
• For each variable, where applicable, the calibration
(or reagent) blank data are plotted as (1) the
instrument detection limit and (2) the daily
calibration blank (and reagent blank) distribution
(P50, Pgs). These are presented and pooled by the
analytical laboratory and can be compared to the
required detection limits for a DQO comparison.
The sample size of the calibration blanks and
reagent blanks is noted in Appendix D, Table D-3.
• For each variable, where applicable, the distribution
of trailer blank analyses are presented. The sample
size for the trailer blanks is noted in Appendix D,
Table D-2.
• For each variable, where applicable, the distribution
of field blank data are presented, which is an
estimation of system contamination levels. The
required detection limit can be c ompared as a
gauge, but should not be used as a direct
comparison to assess data quality. The sample
size for the field blanks is noted in Appendix D,
Table D-1.
• For each variable, all of the field duplicate pair
sample mean concentrations were plotted against
the precision (%RSD or SD) of the pair. This
shows all the field duplicate pair data above and
below the quantitation limit, so precision at varying
concentrations can be observed. Refer to Tables
15 and 21 in Section 6 for companion data.
• For each variable, all six of the field audit sample
lot mean concentrations are plotted against the
precision (%RSD, SD) and can be used in
conjunction with the field duplicate pairs to observe
the relationship to precision at different
concentrations. Refer to Table 26 in Section 6 and
Tables F-5 and F-6 in Appendix F for companion
data.
155
-------
Figure J-la.
Extractable Aluminum: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution
(Pso and P95) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled
and separated by the major components. The required detection limit is shown
0.006
0.005 -
0.004 -
Extractable Aluminum
Legend
Mean of JDL
Concentrations
50thPercentile(P50).
95th Percentile (P95)
Required Detection
•5 0.003-
c
-------
Figure J-1 b.
Extractable Aluminum: Relationship between precision (percent relative standard deviation; %RSD) and mean
concentrations of field duplicate pairs and field audit samples. Western Lake Survey - Phase I. The quantitation
limit is shown; 26 field duplicate pairs were omitted because their mean concentrations were less than or
equal to 0 mg/L.
o>
D
•o
0)
EC
300 -
200 -
100-
o-
O
o
00
0
0
e®
If ° 0 °
S^^Q o
J^Q oQ°
IBSlP'lFe0 o
Extractable Aluminum
Legend
A - Field Natural Audit, Lot
O - Field Natural Audit, Lot
x - Field Natural Audit, Lot
o - Field Natural Audit, Lot
#3
#4
#5
#6
V - Field Synthetic Audit, Lot #1 1
+ - Field Synthetic Audit, Lot #1 2
o - Field Duplicate Pair
Quantitation Limit
0
0
• ' s/
D
(f .
0.00 0.01 0.02 0.03 0.04 0.05
Mean Concentration (mg/L)
0.06
0.07
0.20
157
-------
Figure J-2a. Total Aluminum: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution
(Pso and Pas) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled
and separated by the major components. The required detection limit is shown.
0.020-
0.019-
0.018-
0.017-
0.016 -
0.015;-
0.014-
— 0013 -
ra 0.012-
— 0.01H-
| 0.010.-
I 0.009-
g 0.008 -
o 0.007-
° 0.006-
0.004 -
0.003 -
0.002 -
0.001 -
0-
Legend
• Mean of IDL
Concentrations
Q 50th Percentile (P5o)
^ 95th Percentile (Pas)
Required Detection
Limit
1
To
I
i
1
al
Pooled Lab I Lab II Pooled Lab I
V J V.
Instrument
Detection
Limits
Laboratory
Calibration
Blanks
Aluminum
ir
v\S
\\\
SX\
Lab II
i
1
\N\
X\\
\
\
Pooled Pooled
Trailer
Blanks
1
I
i
—
\
1
I
i
Labi Lab II Ground Helicopter
j
Field
Blanks
158
-------
Figure J-2b. Total Aluminum: Relationship between precision (percent relative standard deviation; %RSD) and mean
concentrations of field duplicate pairs and field audit samples. Western Lake Survey - Phase I. The quantitation
limit is shown; 3 field duplicate pairs were omitted for purposes of resolution.
c
o
.2
CD
Q
•o
CO
"g
CO
CD
CO
CD
DC
120 -
110-
100-
90 -
O\J
80 -
70-
60-
50-
40-
30-
20-
10-
o-
0
0
0 0
0 0
G
°A ®
o TT
ox ®
__
§ 0 *
jQl9 (•! Q ^"*
(9^ ^*4%>l8)00Qo
^^j^^^mV) GT^JEQ
^^^"ggTisnSyo g GO ®^ 0
1 ' 1 ' 1 ' 1 ' 1
Total Aluminum
Legend
A - Field Natural Audit, Lot #3
Q - Field Natural Audit, Lot #4
x - Field Natural Audit, Lot #5
O - Field Natural Audit, Lot #6
V - Field Synthetic Audit, Lot #1 1
+ - Field Synthetic Audit, Lot #1 2
o - Field Duplicate Pair
Quantitation Limit
3
G
0 ®
° 0 0 0
0 ((
1 1 • 1 • l ' 1 • 1 •))
C
'
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
Mean Concentration (mg/L)
0.16
0.17
0.35
159
-------
Figure J-3a.
Acid Neutralizing Capacity: Distribution (P50 and P9S) of the trailer blanks and field blanks. The data are presented
pooled and separated by the major components. The required detection limit is shown. N/A denotes that blank
sample data are not available for comparison.
11
10-
9-
8-
7-
x 6-
CT
O
Concentrat
CJ A
i i
2-
1 -
0-
-1 -
Legend
[] 50th Percentile (P50)
^ 95th Percentile (P85)
Limit
Acid Neutralizing Capacity
N/A N/A N/A N/A N/A N/A |
I
1
Pooled Labi Lab II Pooled Labi Lab II Pooled
v. / v. i i
Instrument
Detection
Limits
Laboratory Trailer
Calibration Blanks
Blanks
[
i
N\\^
I
i
U
8
i
x\\
I
I
• 1
ill
\w W w
\\v \w \w
r-1 1 1
^ $$s i — ^
n i n
Pooled Lab 1 Lab II Ground Helicopter
Field
Blanks
160
-------
Figure J-3b. Acid Neutralizing Capacity: Relationship between precision (percent relative standard deviation; %RSD) and
mean concentrations of field duplicate pairs and field audit samples. Western Lake Survey - Phase I. The
quantitation limit is shown; 17 field duplicate pairs were omitted for purposes of resolution; the field natural
audit no. 4 was omitted with a mean concentration of -24.1 /veq/L.
0)
Q
"S
CD
?
CO
tfi
_to
CD
DC
30-
20-
10-
o-
Acid Neutralizing Capacity
Legend
A - Field Natural Audit, Lot #3
O - Field Natural Audit, Lot #4
x - Field Natural Audit, Lot #5
o - Field Natural Audit, Lot #6
V - Field Synthetic Audit, Lot #11
+ - Field Synthetic Audit, Lot #12
O - Field Duplicate Pair
Quantitation Limit
O
O
O
0
100 200 300 400
Mean Concentration
500
600
r-T-T-r-r
700
V/' " ' '
850
161
-------
.Figure J-4a. Base Neutralizing Capacity: Distribution (P50 and P95) of the trailer blanks and field blanks. The data are presented
pooled and separated by the major components. The required detection limit is shown. N/A denotes that blank
sample data are not available for comparison.
Concentration (/ueq/L)
-» -• NJ ro GO oo ±k
D Ol O Ol O Ol O
5 -
0-
Legend >
Q] 50th Percent! le(Pso)
^ 95th Percentile (P95)
Limit
N/A N/A N/A N/A N/A N/A
I
i
XvN
!xs
s^;
Sx^
^^
^s
-------
Figure J-4b. Base Neutralizing Capacity: Relationship between precision (percent relative standard deviation; %RSD) and
mean concentrations of field duplicate pairs and field audit samples. Western Lake Survey - Phase I. The
quantitation limit is shown; 3 field duplicate pairs were omitted for purposes of resolution; 12 field duplicate
pairs were omitted with mean concentrations less than or equal to 0 /ueq/L.
c
o
CO
8
Q
1
CO
CD
CO
CD
^
130 -
.
120 -
1 1 n
1 I U -
100 -
90 -
80-
7O -
/ \j
60-
50-
40-
30-
20-
10-
0-
Base Neutralizing Capacity
O
O
o
A
o ®
0 ^>-> o
^-O
OG n
o Q fjr^7°
fVQO f"l
^wQ^O^ ar' ^P _
*yjJ£&&g!» /
^^P^^^^^^T ^ao Q ©
f r j • • i | i | i | i | i | . ( • | i '(
0 10 20 30 40 50 60 70 80 90
O
Legend
A - Field Natural Audit,
Q - Field Natural Audit,
x - Field Natural Audit,
o - Field Natural Audit,
V - Field Synthetic Aud
+ - Field Synthetic Aud
o - Field Duplicate Pair
Quantitation Limit
0
D
*-i i ' i ' i ' r
100 110 120 13C
Lot #3
Lot #4
Lot #5
Lot #6
t, Lot #1 1
t, Lot #1 2
)
Mean Concentration (/jeq/L)
163
-------
Figure J-5a.
0.08-
0.07-
0.06-
— 0.05 -
0>
^
o
<3
0.04 -
£ 0.03 -
c
o
o
0.02-
0.01
Calcium: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution (P50 and
Pas) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled and
separated by the major components. The required detection limit is shown. N/A denotes that blank sample
data are not available for comparison.
Calcium
Legend
•
n
Mean of IDL
Concentrations
50th Percentile (P50)
95thPercentile(P95)
Required Detection
Limit
I
I
I
0-1
N/A
1
N/A
Pooled
v
Lab I Lab II
Instrument
Detection
Limits
Pooled Lab I Lab II
Laboratory
Calibration
Blanks
Pooled Pooled Labi Lab II Ground Helicopter
Trailer
Blanks
Field
Blanks
Figure J-5b.
17
"O
C
(D
>
£
o>
CC
Calcium: Relationship between precision
(percent relative standard deviation; %RSD)
and mean concentrations of field duplicate
pairs and field audit samples. Western Lake
Survey - Phase I. The quantitation limit is
shown; 11 field duplicate pairs were omitted
for purposes of resolution.
Calcium
12;
10
9
8
7
6
5
4
3
2
1
0
1+
0
O
O
0
O
°
^ s
o 0 0
JlPSlil*'**
Legend
A - Field Natural Audit, Lot #3
O - Field Natural Audit, Lot #4
x - Field Natural Audit, Lot #5
o - Field Natural Audit, Lot #6
v - Field Synthetic Audit, Lot #1 1
+ - Field Synthetic Audit, Lot #1 2
O - Field Duplicate Pair
Quantitation Limit
A
O O
0
0 o
^a0Go 0 so
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Mean Concentration (mg/L)
164
-------
Figure J-6a. Chloride: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution (Pso and
P95) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled and
separated by the major components/The required detection limit is shown.
0.06-
0.05-
0.04-
§ 0.03
c
o
o
0.02-
0.01
O-l
Chloride
Legend
Mean of IDL
Concentrations
50th Percentile (P60)
95th Percentile (P95)
Required Detection
Limit
1
I
I
1
I
r-l
I
I
1
•I
Pooled Lab I Lab II
Instrument
Detection
Limits
Pooled Labi Lab II Pooled
Laboratory Trailer
Calibration Blanks
Blanks
Pooled Labi Lab II Ground Helicopter
v ^ ;
Field
Blanks
Figure J-6b. Chloride: Relationship between precision
(percent relative standard deviation; %RSD)
and mean concentrations of field duplicate
pairs and field audit samples. Western Lake
Survey - Phase I. The quantitation limit is
shown; 4 field duplicate pairs were omitted
for purposes of resolution.
Chloride
Legend
- Field Natural Audit, Lot #3
- Field Natural Audit, Lot #4
- Field Natural Audit, Lot #5
- Field Natural Audit, Lot #6
- Field Synthetic Audit, Lot #11
- Field Synthetic Audit, Lot #12
o - Field Duplicate Pair
--- - Quantitation Limit
A.
Mean Concentration (mg/L)
165
-------
Figure J-7a. Conductance: Comparison of the mean instrument detection limit (IDL) of conductance to the distribution (P50
and P95) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled
and separated by the major components. The required detection limit is shown.
2.1-
2.0-
1.9-
1.8-
1.7-
1.6-
o 1.5-
K 1-4-
4- 10.
ro '••*
E 1.2-
(J
| 1>
3 1.0-
<£
a 09-
§ 0.8-
| 0,7-
o 0.6-
0.5-
*0.4-
0.3-
0.2-
0.1-
o-
Legend
• Mean of IDL
Concentrations
D 50thPercentile(P50)
^ 95th Percentile (P95)
Required Detection
Limit
ill
II
Conduc
tan
1
1
1
^
\s^
1
ce
Pooled Labi Lab II Pooled Labi Lab II Pooled
V 1 V 1 i J \
Instrument
Detection
Limits
1
ll
1
!
i
1
Pooled Lab 1
Laboratory Trailer
Calibration Blanks
Blanks
i
i
i
i
^
V-VS
1
i
i
1
1
Lab II Ground Helicopter
j
Field
Blanks
Figure J-7b. Conductance: Relationship between
precision (percent relative standard
deviation; %RSD) and mean conductance of
field duplicate pairs and field audit samples.
Western Lake Survey - Phase I. The
quantitation limit is shown; 7 field duplicate
pairs were omitted for purposes of
resolution.
c
o
(0
1
Q
T3
<5
c
as
(/)
CD
ED
O
DC
0
19
18
17-
16-
15
14
13
1 9
1 £. •
1 1-
10
9.
8
7-
6
5:
*r "
3
2
1
0-
j
i
Cjp
'ft
60 0
100
L*t>
Conductance
Legend
A
D
X
0
V
+
0
....
cF
% x
wprfa ° o
^Sf ™
j^Mp^ ®|— |
Q^&i
^^^?<*
^ S»,o ° ° Qre c
*^^*
-------
Figure J-8a. Dissolved Inorganic Carbon (air equilibrated): Distribution (Pso and Pas) of the trailer blanks and field blanks.
The data are presented pooled and separated by the major components. The required detection limit is shown.
N/A denotes that blank sample data are not available for comparison.
0.40 -
0.35-
0.30-
^ 0.25 -
D>
I 0.20 H
I 0-16 H
u
c
0 0.10H
0.05
0
-0.05
Diossolved Inorganic Carbon (Air Equilibrated)
Legend
50th Percent! le (Pso)
95th Percentile(P95)
Required Detection
Limit
N/A N/A
N/A
I
1
I
1
I
I
1
Pooled Lab I Lab II
v >
Instrument
Detection
Limits
Pooled Lab I Lab I
v
Laboratory
Calibration
Blanks
Pooled Pooled Labi Lab II Ground Helicopter
Trailer
Blanks
Field
Blanks
Figure J -8b. Dissolved Inorganic Carbon (air equilibrated):
Relationship between precision (percent
relative standard deviation; %RSD) and mean
concentrations of field duplicate pairs and
field audit samples. Western Lake Survey -
Phase I. The quantitation limit is shown; 6
field duplicate pairs were omitted for purposes
of resolution.
80 Oj
c.
o
0)
Q
C
CO
u>
tr
30
20
£ 10
Dissolved Inorganic Carbon (Air Equilibrated)
Legend
A - Field Natural Audit, Lot #3
Q - Field Natural Audit, Lot #4
x - Field Natural Audit, Lot #5
o - Field Natural Audit, Lot #6
V - Fielo*"Synthetic Audit, Lot #11
+ - Field Synthetic Audit, Lot #12
o - Field Duplicate Pair
Quantitation Limit
o 1 2345678 910111213141516
Mean Concentration (mg/L)
167
-------
Figure J-9. Dissolved Inorganic Carbon (closed system):
Relationship between precision (percent rela-
tive standard deviation; %RSD) and mean
concentrations of field duplicate pairs and field
audit samples. Western Lake Survey - Phase I.
One field duplicate pair was omitted for
purposes of resolution.
Dissolved Inorganic Carbon (Field Laboratory; Closed System)
40
c
o
|30
o
a
T3
CD
"g 20
CO
£ 10
CD
DC
10
Legend
A - Field Natural Audit, Lot #3
D - Field Natural Audit, Lot #4
x -Field Natural Audit, Lot #5
o - Field Natural Audit, Lot #6
v - Field Synthetic Audit, Lot #11
+ - Field Synthetic Audit, Lot #11
o - Field Duplicate Pair
20
30
40
Mean Concentration (mg/L)
168
-------
Figure J-IOa. Dissolved Inorganic Carbon (initial: open system): Comparison of the mean instrument detection limit (IDL)
concentrations to the distribution (Pso and P9s) of the laboratory calibration blanks, trailer blanks, and field
blanks. The data are presented pooled and separated by the major components. The required detection limit
is shown.
Dissolved Inorganic Carbon (Initial; Open System)
u>
_E
c
o
c
CD
U
C
o
o
0.5-
0.4-
0.3-
0.2-
0.1-
Legend
Mean of IDL
Concentrations
50th Percentile (P50)
95th Percentile (P95)
Required Detection
Limit
I
I
I
-T- F^1
-0.1
Pooled Lab I Lab II
i ;
Instrument
Detection
Limits
Pooled Labi Lab II
Laboratory
Calibration
Blanks
Pooled Pooled Lab I Lab II Ground Helicopter
Trailer
Blanks
Field
Blanks
Figure J-10b. Dissolved Inorganic Carbon (initial; open
system): Relationship between precision
(percent relative standard deviation; %RSD)
and mean concentrations of field duplicate
pairs and field audit samples. Western Lake
Survey - Phase I. The quantitation limit is
shown; 7 field duplicate pairs were omitted
for purposes of resolution.
Dissolved Inorganic Carbon (Initial; Open System)
-------
% Relative Standard Deviation
CO
c
Concentration (mg/L)
--
2 3 4 5 6
Mean Concentration (
3
"x >J-
00 -
Q -
-'toco^cncn-joocoo"'
oooooooooooo e_
0 « o f*Q fi Q
IfS^o "g -° g 2. a °" o
nSI: O 0> e» <»" :r S' 8 5" r~ CD <£
4" « «»<"a§3» = CD I J
o a £ i'i^c^'* 1-3
'°:D2. ? 33
o Ulfotf
« cQo*S"r~(Doi— »S'
§ »s*»g'5'oc»
D -Q 2J. ?r^j tt * r\
Q 0'COII£30?
r-i .3 < *° ^ 3 m =S
o !o + g" 8 ^ 5 g 3 3 ^
i i i i i i i i ^ J ii*^ Q. o 5T
O c0)a)a)5'(D S=^i?SS m * ™
5o.Q.Q.aaaQ. gc"SaSa^. ^CTO
^*r-JC/)r/)^-^^^ ° TI CD ^ tt ft ij-n-^j
oJc^^OjajtoD) E. — • ~ o> r+ ^ 3 ?r S p*
^^^^cccc f~ S" 8 H c o" & * wo'?
0® igifS'SStttto 3»cD"SS3.€' ^^
L^fDOO^^^^ D
i-mcc^0-0-0-
o o o o
S S ***» CDH
3,. ^ cnen 4^co 5T S
S- l«
2-n
111
0)
'•D
§
Q.
1-
01
cr
ET
cr
'•o
o
5"
Q.
1—
0)
a-
r~
0)
r°
L£
*T3
o
eo"
O,
r~
CO
|-
0)
3
a.
CD_
9
k.CD
o c
b -
2 en c
•
—
r
rifflffiwfflffffifP
1
3OOOQOOO ffl
•A^-'ioiococo*»ji *-
3cnOoioenocn ^
a
•o a Q
I - ^~
r- 3J eo en o? a o' §.
5'cDcnoo^ „§
If 3- 5g8 ?^.o
^ -D -o CD= r- 0. -0
Sf, ^-^ ^*.
-------
Figure J-12a. Fluoride, Total Dissolved: Comparison of the mean instrument detection limit (IDL) concentrations to the
distribution (P5o and Pas) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are
presented pooled and separated by the major components. The required detection limit is shown.
0.007 -
0.006 -
0.005
0.004 -
c
o
Fluoride, Total Dissolved
0.003
0 0.002-
0.001 -
Legend
•
D
Mean of IDL
Concentrations
50th Percentile(P5o)
95th Percentile(P95)
Required Detection
Limit
Pooled Labi Lab 11 Pooled Labi Lab II Pooled Pooled Labi
Instrument
Detection
Limits
Laboratory
Calibration
Blanks
Trailer
Blanks
Lab II Ground Helicopter
j
Field
Blanks
Figure J-12b. Fluoride, Total Dissolved: Relationship
between precision (percent relative
standard deviation; %RSD) and mean
concentrations of field duplicate pairs and
field audit samples. Western Lake Survey -
Phase I. The quantitation limit is shown; 6
field duplicate pairs were omitted for
purposes of resolution; 2 field duplicate
pairs were omitted because their mean
concentrations were less than or equal to
0 mg/L.
Fluoride, Total Dissolved
c
o
<0
Q
CO
?
(0
W
I
JO
<»
DC
601
50
40
30
20
10-
Legend
A -
D -
x -
o -
v -
+ -
0 -
Field Natural Audit, Lot #3
Field Natural Audit, Lot #4
Field Natural Audit, Lot #5
Field Natural Audit, Lot #6
Field Synthetic Audit, Lot #1 1
Field Synthetic Audit, Lot #1 2
Field Duplicate Pair
Quantitation Limit
0.2 0.3 0.4 0.5
Mean Concentration (mg/L)
171
-------
% Relative Standard Deviation
2
CD
a>
3
O
0
3
fj
CD
D
3
6'
D
?
a° s J> l1^
0°
; O + < O X O >
1 1 1 1 1 1 1 1
Q-n-n-n-n-n-n"n
CCDCDCDCDCDCDCD
^ Q.Q.Q.Q.Q.Q.Q.
^. Ot/)COZ22Z
S-§ 'S^"'*"'*
I'ffcfslss
(D 0 0 ££££
— •m'O-Q.Q.Q.
^' CL Q. ^* ^* r* r*
o o o" o"
o" o" ^* "* "* "*
J Jcnai^co
jo 3*
**
\ —
0)
D
Q_
§
X
i
(O
3
Concentration (mg/L)
pPPPpppoppoopppppppoo
J—i—"—i—i—i—i—i i i T
i:fi
a 3
r; to 5' (D
E :
u m I o~
= = < O
•ff
Is
to
3
w
01
r- 3j co en
« 32.
8 26
§.§-
»
T3 "D
a M TJ ;
fi) CD (o i
3 Q. Sr S
- S
I'lH
•lias
» 2 0 3
=• S 3 -.
;ks
A =: Q-
coo
^' - S
S. 2 S
Q. 3 O'
Is I
—• 2 °
"* 0) 3
S-S-S
o y> §
S,, »
sll
3^§ 2.
^ (D C
2 a —
je-o —
tss
o a a
-------
CD
O
o
01
o'
to
Relative Standard Deviation
Concentration (mg/L)
6
-------
Figure J-15a. Magnesium: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution (Pso
and P9s) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled
and separated by the major components. The required detection limit is shown. IM/A denotes that blank sample
data are not available for comparison.
0.011
0.010
0.009
0.008
0.007
o.ooe
0.005'
0)
o
o 0.004
0.003
0.002
0.001
0
Magnesium
Legend
Mean of IDL
Concentrations
50th Percentile (P50)
95th Percentile (P95)
Required Detection
Limit
N/A
N/A
Pooled Lab I Lab II
V ^ ;
Instrument
Detection
Limits
Pooled Lab I Lab I
i
Laboratory
Calibration
Blanks
Pooled Pooled Labi Lab II Ground Helicopter
Trailer
Blanks
Field
Blanks
Figure J-15b. Magnesium: Relationship between precision
(percent relative standard deviation; %RSD)
and mean concentrations of field duplicate
pairs and field audit samples. Western Lake
Survey - Phase I. The quantitation limit is
shown; 9 field duplicate pairs were omitted
for purposes of resolution.
Magnesium
c
o
O
•o
CO
c
to
CD
£t
11
10
9
8
7
6
5
4
3
2
1
0
Legend
£ - Field Natural Audit, Lot #3
Q - Field Natural Audit, Lot #4
x - Field Natural Audit, Lot #5
o - Field Natural Audit, Lot #6
V - Field Synthetic Audit, Lot #11
+ - Field Synthetic Audit, Lot #12
O - Field Duplicate Pair
Quantitation Limit
Mean Concentration (mg/L)
174
-------
Figure J-16a. Manganese: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution (P5o
and P95) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled
and separated by the major components. The required detection limit is shown. N/A denotes that blank sample
data are not available for comparison.
CD
0.022-
0.020-
0.018-
0.016-
0.014-
0.012-
0.010-
Manganese
Legend
•
D
Mean of IDL
Concentrations
50thPercentile(P5o)
95th Percentile (P95)
Required Detection
Limit
-0.002
N/A
I
I
N/A
I
I
Pooled Lab I Lab II
V j
Instrument
Detection
Limits
Pooled Lab I Lab II Pooled Pooled Lab I Lab II Ground Helicopter
Laboratory
Calibration
Blanks
Trailer
Blanks
Field
Blanks
Figure J-16b.
c
o
o>
o
•a
c
s
(/)
I
53
HI
IT
440
350 Jx
9; Relationship between precision
(percent relative standard deviation; %RSD)
and mean concentrations of field duplicate
pairs and field audit samples. Western Lake
Survey - Phase I. The quantitation limit is
shown; 12 field duplicate pairs were omitted
for purposes of resolution; 78 field duplicate
pairs were omitted because their mean
concentrations were less than or equal to
0 mg/L.
Manganese
260-
190
180
170
160
150
140
130
m
100
90
80
70
60
50
&
20
10
0
[°
0
O
** ° o
I' 9
O
00
-------
°-£
> Relative Standard Deviation
^pf""--—
94*4**
o X
O G
(D
0)
D
O
o
D
O
(D
3
CO
X
'
•fl
s
sj
IT
6 6
8f
O)
i o + < o x a t>
i t i i i i i i
D3^31J!3!2!3!21
Sj ci a a 5 S a S
|f.||c c | c
C CD n o' > J> J> J>
rt =;• a aFFFr1-'
r* r* r~ r~ r~ r~
_ ,_ o o o o
OD — C7
2-n
Q)
3 Q) I o
Is II
<
b
Concentration (mg/L)
p
8
o P P p p
b o b b b
r- 33 CD en 0?
a » TJ
"• 3
w
a
1
S 5=?
5 1 § S
8 So-3
p 5> 5T 5'
}*\
all
"= » «
—. -t Q.
Q. 3
1=1
«|s
2^8
"•oo
- 3
• °>
^.33
(B Q. Q.
-------
Figure J-18a. Ammonium: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution (P5o
and Pas) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled
and separated by the major components. The required detection limit is shown.
Legend
•
D
Mean of IDL
Concentrations
50th Percentile (P5o)
95th Percentile (P95)
Required Detection
Limit
Ammonium
-0.001-
-0.002-
-0.003-
-0.004-
-0.005-
-0.006-
-0.007-
l\
I
Pooled Lab I Lab II Pooled Lab I Lab II Pooled Pooled Lab I Lab II Ground Helicopter
Instrument
Detection
Limits
Laboratory
Calibration
Blanks
Trailer
Blanks
Field
Blanks
Figure J-18b.
1770 JD
Ammonium: Relationship between precision
(percent relative standard deviation; %RSD)
and mean concentrations of field duplicate
pairs and field audit samples. Western Lake
Survey - Phase I. The quantitation limit is
shown; 12 field duplicate pairs were omitted
for purposes of resolution; 132 field dupli-
cate pairs were omitted because their mean
concentrations were less than or equal to
0 mg/L; field natural audit lot no. 3 was
omitted because its mean concentration was
less than 0 mg/L.
i Ammonium
% Relative Standard Deviation
630
200
190
180
170
140
130
120
110
100
80
70
60-
50
40
30
20
10-
0
lb
9
8
o
X
I
Legend
A - Field Natural Audit, Lot #3
a - Field Natural Audit, Lot #4
x - Field Natural Audit, Lot #5
o - Field Natural Audit, Lot #6
V - Field Synthetic Audit, Lot #1 1
+ - Field Synthetic Audit, Lot #1 2
o - Field Duplicate Pair
Quantitation Limit
OD
03
0 )
0
O 0
0
^o 0 V
nilL4 ; "t" ^
0.00 0.05 0.10 0.15 0.20 0.25
Mean Concentration (mg/L)
0.30
177
-------
Figure J-19a. Nitrate: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution (Pso and
Pas) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled and
separated by the major components. The required detection limit is shown.
0.08-
0.07-
0.06-
0.05-
c
I 0.04 H
c
u
S 0.03 -
0.02-
0.01-
0-
Nitrate
Legend
Mean of IDL
Concentrations
50th Percent! le (P5o)
95th Percentile (P9S)
Required Detection
Limit
I
fl
fi
1
1
I
1
Pooled Lab I Lab II
V ^ ;
Instrument
Detection
Limits
Pooled Lab I Lab II Pooled Pooled Lab I Lab II Ground Helicopter
Laboratory
Calibration
Blanks
Trailer
Blanks
Field
Blanks
178
-------
Figure J-19b. Nitrate: Relationship between precision (percent relative standard deviation; %RSD) and mean concentrations
of field duplicate pairs and field audit samples. Western Lake Survey - Phase I. The quantitation limit is shown;
3 field duplicate pairs were omitted for purposes of resolution.
Nitrate
c
o
a
•^
0)
O
eg
T3
C
2
OT
§
»
Q)
DC
CT*
230-
s
V.
-
190
180-
170
160 -
15O-
140-
130-
120-
110-
100-
90-
80-
70-
60-
50-
40-
30-
20-
1 n -
I U
0 -
p
0
0
0
0 GO
Q
0 0
£j
3^0
£
*V 0
9J
^ 00
W£WQ
£rr 0
*%
3ft0 0 ®
^rt,
^^ « X
S^fcSrf'
^ Sl^f'c^ « 0
Legend
A - Field Natural Audit, Lot #3
D - Field Natural Audit, Lot #4
x - Field Natural Audit. Lot #5
0 - Field Natural Audit, Lot #6
V - Field Synthetic Audit, Lot #1 1
+ - Field Synthetic Audit, Lot #1 2
o - Field Duplicate Pair
Quantitation Limit
0 0^D 0 00 o ,f A f, D
I • I • I • I • I • I • I • I ' I • .1 // I // I
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.4
2.4
Mean Concentration (mg/L)
179
-------
Figure J-20a. Phosphorus. Total: Comparison of the mean instrument detection limit (IOL) concentrations to the distribution
(P5o and P95) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled
and separated by the major components. The required detection limit is shown.
0.018-
f\ m "7
U.U1 / —
0.016-
0.015-
0.014-
0.013 -
3- 0.012-
01 0.011-
E
~ 0.010-
o
«= 0.009-
E 0.008-
o>
c 0.007-
o
o 0.006-
0.005-
0.004-
0.003 -
0.002
0.001 -
0-
Legend
• Mean of IDL
Concentrations
D 50th Percentile (P50)
^ 95th Percentile (P95)
Reouired Detection
Limit
m m m +.
i
riiuapiiu
j_
1 US
1
i uiai
ssi
r
^
SN^S
§
svSN
I
1
\S»
1
vS
1
i
^
r
i
i
§
— • ^
1
1
H
1
1
§§
1
1
^
1
w
1
1
Xsx
1
^
1
1
I
1
o>
Q
CD
DC
Phosphorus, Total: Relationship between
precision (percent relative standard devia-
tion; %RSD) and mean concentrations of
field duplicate pairs and field audit samples.
Western Lake Survey - Phase I. The quan-
titation limit is shown; 7 field duplicate pairs
were omitted for purposes of resolution; 5
field duplicate pairs were omitted because
their mean concentrations were less than or
equal to 0 mg/L.
190^
V
150
140.
130.
120
110-
100
70.
fin
50.
40
30
20
10-
0
?
80
0 e
O
39 a
(3D
0 0 O
IHTllSFX^ ' p ®
losphorus. Total
Legend
A -
a -
x -
o -
v -
o -
Field Natural Audit, Lot #3
Field Natural Audit, Lot #4
Field Natural Audit, Lot #5
Field Natural Audit, Lot #6
Field Synthetic Audit, Lot #1 1
Field Synthetic Audit, Lot #1 2
Field Duplicate Pair
Quantitation Limit
0
o
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
Mean Concentration (mg/L)
180
-------
Figure J-21a. pH (acidity; open system): Distribution (P50 and P95) of the trailer blanks and field blanks. The data are presented
pooled and separated by the major components. The theoretical pH of deionized water is shown. N/A denotes
that blank sample data are not available for comparison.
Q.
6.0-
5.8-
5.6-
5.4-
5.2-
5.0-
48-
Legend
Q] 50th Percentile (P6o)
^ 95th Percentile (P9S)
Theoretical pH of
Deionized Water
pH (Acidity; Open System)
N/A N/A N/A N/A N/A N/A
I
1
1
1
I
I
1
1
1
\
I
\
m
Pooled Lab \ Lab \\
)
Instrument
Detection
Limits
Pooled Labi Lab II Pooled
Laboratory Trailer
Calibration Blanks
Blanks
Pooled Lab I Lab II Ground Helicopter
Field
Blanks
181
-------
Figure J-21b.
0.7
0.6
0.5 -
c
o
JO
I 0.4
£ 0.3 -I
0.2 -
0.1
0.0 -
pH (acidity; open system): Relationship between precision (standard deviation) and mean pH of field duplicate
pairs and field audit samples. Western Lake Survey - Phase I.
pH (Acidity, Open System)
Legend
A - Field Natural Audit, Lot #3
a - Field Natural Audit, Lot #4
x - Field Natural Audit, Lot #5
o - Field Natural Audit, Lot #6
v - Field Synthetic Audit, Lot #11
+ - Field Synthetic Audit, Lot #12
o - Field Duplicate Pair
O
00
D
o
10
pH Units
182
-------
Figure J-22a. pH (alkalinity, open system): Distribution
-------
Figure J-22b.
0.7 -
0.6 -
pH(alkalinity; open system): Relationship between precision (standard deviation) and mean pH of field duplicate
pairs and field audit samples. Western Lake Survey - Phase I.
pH (Alkalinity; Open System)
I
Q
a
•a
c
a
3)
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
0.0 -
Legend
A - Field Natural Audit, Lot #3
a - Field Natural Audit, Lot #4
x - Field Natural Audit, Lot #5
o - Field Natural Audit, Lot #6
v - Field Synthetic Audit, Lot #11
+• - Field Synthetic Audit, Lot #12
o - Field Duplicate Pair
pH Units
184
-------
Figure J-23a. pH fair equitiblrated): Distribution (P50 and P9s) of the trailer blanks and field blanks. The data are presented
pooled and separated by the major components. The theoretical pH of deionized water is shown. N/A denotes
that blank sample data are not available for comparison.
pH (Air Equilibrated)
6.0-
X
Q.
5.8-
5.6-
5.4-
5.2-
5.0-
4.8-
Legend
Q 50th Percentile (Pso)
ffl 95th Percentile (P95)
Theoretical pH of
Deionized Water
N/A N/A N/A N/A N/A N/A
Pooled Lab 1 Lab II Pooled Lab 1 Lab II
v j v -*
Instrument
Detection
Limits
P —
I
1
Pooled
Laboratory Trailer
Calibration Blanks
Blanks
1
•• -
s
1
i
i
Pooled Lab Lab II Ground Helicopter
i t
Field
Blanks
185
-------
Figure J-23b.
0.6
0.5
c
o
S
1
-------
Figure J-24.
1.0 .
0.9-
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
c
o
Q
"S
(0
?
%- - A
o ° <*T^^0
-------
Figure J-25a. Silica: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution (P50 and
P9s) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled and
separated by the major components. The required detection limit is shown.
c
o
I
c
ID
O
C
o
o
0.28-
0.26-
0.24-
0.22-
0.20-
0.18 -
0.16 -
0.14 -
0.12-
0.10-
0.08 -
0.06-
0.04-
0.02 -
0 -
Silica
Legend
•
D
Mean of IDL
Concentrations
50th Percentile (P50)
95th Percentile (P95)
Required Detection
Limit
I
Pooled Lab I Lab I
v
Instrument
Detection
Limits
„ Pooled Lab I Lab II Pooled
Laboratory Trailer
Calibration Blanks
Blanks
Pooled Labi Lab II Ground
Helicopter
i
Field
Blanks
Figure J-25b. Silica: Relationship between precision
(percent relative standard deviation; %RSD)
and mean concentrations of field duplicate
pairs and field audit samples. Western Lake
Survey - Phase I. The quantitation limit is
shown; 3 field duplicate pairs were omitted
for purposes of resolution.
Silica
c
o
<5
'>
o
Q
•o
CD
?
CD
w
-------
Figure J-26a. Sulfate: Comparison of the mean instrument detection limit (IDL) concentrations to the distribution (Pso and
Pas) of the laboratory calibration blanks, trailer blanks, and field blanks. The data are presented pooled and
separated by the major components. The required detection limit is shown.
_l
\
o>
c
0
cS
Concenti
0.14-
0.13-
0.12 -
0.11 -
0.10-
0.09-
0.08-
0.07-
0.06-
0.05-
0.04 -
0.03 -
0.02 -
0.01 -
0 -
Sulfate
Legend
•
D
Mean of IDL
Concentrations
50th Percentile (P50)
95th Percentile (Pas)
Required Detection
Limit
1
ri
HI
Pooled Lab I Lab I
^
Instrument
Detection
Limits
Pooled Labi Lab II Pooled
Laboratory Trailer
Calibration Blanks
Blanks
Pooled Lab I Lab II Ground Helicopter
.. '. /
Field
Blanks
Figure J-26b. Sulfate: Relationship between precision
(percent relative standard deviation; %RSD)
and mean concentrations of field duplicate
pairs and field audit samples. Western Lake
Survey - Phase I. The quantitation limit is
shown; 7 field duplicate pairs were omitted
for purposes of resolution.
50
-S 40
0)
O
T3
(0
T3
C
ID
CC
30
.> 20
10
Sulfate
9
O
0 0
3 0
o oe
o Q gfS
idv-**^
iiBwHN^j^
D 1 2
A -
Q -
x -
o -
v -
+ -
O -
A
o
o
SD o
3
Legend
Field Natural Audit, Lot #3
Field Natural Audit, Lot #4
Field Natural Audit, Lot #5
Field Natural Audit, Lot #6
Field Synthetic Audit, Lot #1 1
Field Synthetic Audit, Lot #1 2
Field Duplicate Pair
Quantitation Limit
0 e O
G
e ° • o
456789
Mean Concentration (mg/L)
189
-------
-------
Appendix K
Distribution of Analyte Concentrations for Routine Lake Samples
Table K-1 shows the distribution of the analytical
measurements for each variable, for all routine lake
samples. These data can be useful in interpreting the
importance of QA limits and of trends in QA sam-
ple data. The lowest and highest values are the
endpoints of the range of concentrations measured
in WLS-I. The highest value often was associated
with a rare high conductance, high ionic strength
sample. For comparison, the table also presents
median and mean sample concentrations.
Table K-1. Distribution of Analyte Concentrations for Routine Lake Samples, Western Lake Survey - Phase I
Variable3
Low Value
Median Value
Mean Value
High Value
Al, extractable
Al, total
ANC (neq/L)
BNC dieq/L)
Ca
cr
Conductance (pS/cm)
DIG, air equilibrated
DIC, initial
DOC
F", total dissolved
Fe
K
Mg
Mn
Na
NH4"
N03
P, total
pH, acidity (pH units)
pH, alkalinity (pH
units)
pH, air equilibrated
(pH units)
SiO2
SO42'
-o.ooe"
-0.002"
-24.0
-798.5
0.09
0.011
1.6
0.14
0.16
0.05
0.000
-0.009"
0.00
0.02
-0.043"
0.02
-0.083"
-0.01 3b
-0.003"
4.55
4.60
4.65
-0.05"
0.00
0.004
0.024
105.6
27.6
1.67
0.14
14.6
1.36
1.44
1.3
0.015
0.015
0.21
0.28
0.001
0.54
-0.002"
0.022
0.005
6.94
6.92
7.21
2.27
0.82
0.006
0.037
270.9
23.7
3.77
0.87
44.5
3.74
3.82
1.9
0.062
0.034
0.85
1.06
0.004
3.79
0.000
0.105
0.008
7.03
7.03
7.29
3.76
3.87
0.594
1.154
14,140.0
310.9
95.38
187.50
6,601.0
485.4
462.60
32.0
6.233
0.974
269.40
126.00
0.212
1 ,205.00
0.240
2.669
0.188
9.93
9.93
9.92
98.74
1 ,726.00
' Concentrations are in mg/L unless otherwise indicated. Each variable includes 811 routine sample analyses, except for ANC,
BNC, total dissolved F', SiC>2,and SO/i , each of which includes 810 routine sample analyses.
' Negative values are a result of analytical laboratory instrument calibration.
191
-------
-------
Appendix L
Collection and Preparation of Nitrate-Sulfate Split Samples
Collection Procedure (Lake Site)
1. Complete an aliquot label and affix it to a 125-mL
Nalgene bottle.
2. Fill the bottle to the shoulder with sample that has
been processed through the Van Dorn sampler.
3. Use a dropper bottle to add 2 drops (0.1 ml) of 5
percent HgCl2 to the aliquot. Note the amount of
preservative used on the aliquot label.
4. Cap the aliquot bottle tightly. Invert it several times
to mix the contents. Tape the cap with electrician's
tape, then place it in a plastic bag for transport with
the Cubitainers and syringes.
Field Natural Audit Preparation
Procedure (Field Laboratory)
1. Prepare nitrate-sulfate aliquot labels. Enter the
audit sample code in the lake ID field on the label.
Do not record a crew ID. Check the line indicating
that the sample is an audit.
2. Rinse the 125-mL Nalgene bottles (not acid-
washed) three times with 5 to 10 mL of audit
sample.
3. Fill each bottle with field natural audit sample.
Refrigerate it at 4°C until needed (1 to 10 days
later).
193
-------
-------
Appendix M
Proposed Procedure for Use of Low Ionic Strength, Circumneutral, Mid-Range pH
and DIC Quality Control Check Samples
During ELS-I and NSS Phase I Pilot, precision and
accuracy data showed that pH values in the range pH
6 to 8 were the most difficult to determine. The
readings in this range were more variable and took
more time. Also, when an electrode is malfunctioning
or is incorrectly calibrated, readings for near-neutral,
low-ionic-strength samples are affected most (see
Best et al., 1987). It is not always possible, however,
to detect instrument problems with the quality control
check sample (QCCS) used (10-4 N H2S04, pH =
4) or with commercial pH buffers (high-ionic-
strength). For this reason, the use of a low-ionic-
strength QCCS with a pH in the range 6 to 8 was
investigated.
In August 1985, experiments were performed at
EMSL-LV with a low-ionic-strength pH QCCS
(ionic strength = 4.5 x 10"5, conductance < 5 pS).
The sample also may be used as a low-level DIC
QCCS. Preparation instructions are given in Table
M-1. Sample results of the experiment are given in
Table M-2. The average ApH (difference from
theoretical) was 0.002 ± 0.069 and average ADIC
(relative difference from theoretical) was -1.3 ± 7
percent.
As a result of this experiment, during WLS-I, two
new QCCS solutions for pH and DIC were proposed,
in addition to those used previously in NSWS. The
use of these solutions was to provide the following
benefits:
• independent checks of the neutral calibration
point and mid-range linearity in pH analysis
• independent analysis of low-range sensitivity
at two levels in DIC analysis
• a field-determined cross-check between
pH and DIC analyses which would enable
early detection of suspect measurements
The protocol for measurement of pH was to be the
standard closed method used for routine lake
samples in the field laboratory, as described in the
Table M-1. Preparation of the Experimental, Cir-
cumneutral, Mid-Range, Low Ionic Strength
pH/DIC Quality Control Check Sample (pH 7,
DIC 0.7 ppm)
Temp (°C)
PH
DIC (mg/L)
10
15
20
25
30
35
40
6.91
6.93
6.97
7.00
7.03
7.08
7.11
0.734
0.715
0.685
0.665
0.643
0.640
0.618
a. Dilute 0.270 mL of the 1,000-ppm DIO stock QC solution to
1.000L.
b. Sparge with 300 ppm CC>2 for 20 to 30 minutes. The exact pH
and DIC values are given above
c. Store in sealed syringes at 4°C.
methods manuals (Hillman et al., 1986; Kerfoot and
Faber, 1987).
The theoretical pH values of these solutions are 7.00
and 5.67 at 25°C, and the theoretical DIC values are
0.665 mg/L and 0.148 mg/L at 25°C. These solutions
were to be made weekly and held under refrigeration
until used.
The procotol for the QCCS measurements is as
follows.
Weekly:
1. Prepare all reagents for 4.00, 5.67, and 7.00
QC check solutions.
2. Draw 18 syringes each of the 5.67 and 7.00
solutions, seal with syringe valves, date and
label 6 of the 5.67 pH solution syringes
"0.148 ppm DIC," date and label the
remaining 12 5.67 pH syringes "pH 5.67,"
date and label 6 of the 7.00 pH syringes
"0.665 ppm DIC," date and label the
remaining 12 7.00 pH syringes "pH 7.00.
195
-------
Table M-2. Results from Analysis of Low Ionic Strength Quality Control Sample
Day
1
2
3
6
7
8
8
8
9
10
10
13
13
14
15
15
15
15
15
15
15
Theoretical
7.09
7.09
7.04
7.08
7.06
7.00
7.01
7.01
7.02
6.99
7.00
7.03
7.03
7.02
7.01
7.01
7.02
7.02
7.02
7.02
7.02
PH
Measured
6.97
7.13
7.00
7.03
6.99
7.07
7.03
7.23
7.01
7.01
7.06
6.76
6.72
7.05
6.94
6.91
7.07
7.12
7.09
7.12
6.96
A
-0.12
+ 0.04
-0.04
-0.05
-0.07
+ 0.07
+ 0.02
+ 0.22*
-0.01
+ 0.02
+ 0.06
-0.27^
-0.31b
+ 0.03
-0.07
-0.10
+ 0.05
+ 0.10
+ 0.07
+ 0.10
-0.06
DIG (mg/L)
Theoretical
-
0.634
--
0.640
0.641
--
--
0.658
0.646
0.671
0.665
0.643
0.643
0.651
0.657
0.653
--
--
--
--
--
Measured
--
0.604
--
0.623
0.669
--
--
0.571
0.655
0.657
0.654
0.625
0.633
0.661
0.630
0.637
--
--
--
--
--
A(%>
--
-4.7
--
-2.7
+ 4.4
--
--
-13a
+ 1.4
-2.1
-1.7
-2.8
-1.5
+ 1.5
-4.1
-2.4
--
--
--
--
-
a Possible incomplete equilibration (DIG low, pH high) or preparation error.
b Possible pH error (possibly due to calibration; pH low, DIG acceptable).
3. Store all syringes in the refrigerator and use them
in determinations throughout the daily sample
analysis.
Daily (by batch):
1. Calibrate the pH meter with the pH 7.00 buffer
and the pH 4.00 buffer.
2. Analyze the 7.00, 5.67, and 4.00 pH QCCS
solutions, in that order. NOTE: If initial QCCS
values are out of range for any solutions,
reanalyze the solutions. If the measurement is still
out of range, recalibrate the instrument and
reanalyze the QCCS solutions.
3. Analyze 5 samples in the batch.
4. Analyze the 4.00 pH QCCS.
5. Repeat steps 3 and 4 until all samples have been
analyzed.
6. Analyze the final 4.00 pH QCCS.
7. Analyze the 7.00 and 5.6 pH QCCS in that order.
8. Enter the 4.00 pH QCCS value on the batch form.
Enter the 5.67 and 7.00 values in the logbook
only.
If at any time the value of a 4.00 pH QCCS solution
falls outside the acceptance criteria of ±0.10 pH unit,
analyze a fresh QCCS solution. If the QCCS value
still fails to meet the criteria, recalibrate the
instrument and reanalyze the affected samples,
available volume permitting. If the value of the 5.67 or
the 7.00 pH QCCS solutions falls outside the
acceptance window, analyze a fresh sample of the
QCCS solution. If the value still falls outside criteria,
enter a note in the pH logbook documenting the
circumstances, including the DIG values associated
with the variant QCCS solutions. Because the 7.00
and 5.67 pH QCCS checks are only conducted
before the first sample in the batch is analyzed and
after the last sample in the batch is analyzed, do not
reanalyze samples based solely on failure to meet
acceptance criteria for one of these solutions.
Reanalysis of an entire batch of samples would be
prohibitively time consuming, because pH is usually
the slowest procedure conducted in the field
laboratory.
The protocol for analysis of DIG was to be identical to
that described in the methods manual (Hillman et al.,
1986; Kerfoot and Faber, 1987), with the exception
that two new QCCS analyses were to be added
before the first sample in the batch was analyzed and
196
-------
after the last sample in the batch was analyzed. The
protocol for DIG analysis is as follows.
Weekly:
1. Prepare DIG stock solutions.
2. Prepare 7.00 and 5.67 pH QCCS solutions, as
described above.
Daily:
1. Prepare calibration and QCCS solutions as
specified in the methods manual.
2. Calibrate the DIC at 10.00 mg/L.
3. Using 0.148 mg/L DIC and 0.665 mg/L DIC
syringes from the refrigerator, analyze samples of
the DIC QCCS solutions.
4. Check linear dynamic range of the calibration
curve by analyzing a 20.00-ppm DIC calibration
solution.
5. Analyze a 2.00-mg/L DIC QCCS.
NOTE: If any of the QCCS or linearity check
concentrations vary from the
theoretical value by more than 10%,
reanalyze the solution. If the
measurement still fails to meet the
"QC criteria, recalibrate the
instrument.
6. Analyze a calibration blank sample.
7. Analyze batch samples and QC check samples
as described in the field laboratory methods
manual (Morris F. A., D. V. Peck, D. C. Hillman,
K. J. Cabbie, S. L. Pierett, and W. L. Kinney,
1985. National Surface Water Survey, Western
Lake Survey - Phase I, Field Training and
Operations Manual [internal report], U.S.
Environmental Monitoring Systems Laboratory,
Las Vegas, Nevada).
8. Analyze a 2.00-mg/L QCCS.
9. Analyze a 0.665-mg/L DIC QCCS.
10. Analyze a 0.148-mg/L DIC QCCS.
If at any time any QCCS value differs from the
expected value for that QCCS solution by more than
10 percent, a fresh QCCS sample will be analyzed. If
the QC check is still outside the acceptance criteria
for a 2.00 QCCS, the samples associated with the
acceptance QCCS measurements will be reanalyzed.
If the QC check sample for a 0.148 or 0.665 mg/L QC
solution is still outside acceptance criteria, the values
will be noted in the logbook, along with the pH values
based solely on failure to meet criteria for these two
solutions.
197
-------
Glossary
Abbreviations
ANOVA
AQUARIUS
ASTM
%CD
CLP
DBMS
DIG
DOC
DQO
ELS-I
EMSI
EMSL-LV
EPA
ERL-C
Forest Service
%IBD
ICPAES
Lockheed-EMSCO
MIBK
NAPAP
NBS
NCC
NLS
NSS
NSWS
ORNL
QA
QC
QCCS
RMS
%RSD
SAI
SAS
SMO
SOW
USGS
WLS-I
analysis of variance
Aquatics Quality Assurance Review, Interactive Users' System
American Society for Testing and Materials
percent conductance balance difference
Contract Laboratory Program
data base management system
dissolved inorganic carbon
dissolved organic carbon
data quality objective
Eastern Lake Survey - Phase I
Environmental Monitoring and Services, Inc.
U.S. Environmental Protection Agency, Environmental Monitoring
Systems Laboratory, Las Vegas, Nevada
U.S. Environmental Protection Agency
U.S. Environmental Protection Agency, Environmental Research Laboratory,
Corvallis, Oregon
U.S. Department of Agriculture, Forest Service
percent ion balance difference
inductively coupled plasma atomic emission spectroscopy
Lockheed Engineering and Management Services Company, Inc.
methyl isobutyl ketone
National Acid Precipitation Assessment Program
National Bureau of Standards
National Computer Center
National Lake Survey
National Stream Survey
National Surface Water Survey
Oak Ridge National Laboratory
quality assurance
quality control
quality control check sample
root-mean-square
percent relative standard deviation
Systems Applications, Inc.
Statistical Analysis System
Sample Management Office
Statement of Work
U.S. Geological Survey
Western Lake Survey - Phase I
199
-------
Definitions
Absolute Bias
The difference between a measured value and the true value. (See Accuracy.)
Acceptance
Criteria
The range in which the analytical measurement of a quality assurance or quality
control sample is expected to be; measurements outside that range (also referred to
as control limits) are considered suspect.
Accuracy
The closeness of a measured value to the true value of an analyte. For this report,
accuracy is calculated as:
X -T
100
where:
X = the mean of all measured values, T = the true value.
Acid Neutralizing
Capacity
Total acid-combining capacity of a water sample determined by titration with a
strong acid. Acid neutralizing capacity includes alkalinity (carbonate species) as well
as other basic species (e.g., berates, dissociated organic acids, alumino-hydroxy
complexes.
Air Equilibration
The process of bringing a sample aliquot to equilibrium with standard air (300 ppm
CO2) before analysis; used with some pH and dissolved inorganic carbon
measurements.
Aliquot
Alkalinity Class
Fraction of a sample prepared for the analysis of particular constituents, sent in a
separate container to the analytical laboratory.
One of three categories to which each lake in the survey was designated before
sampling activities began. The alkalinity class estimated the acid neutralizing capacity
of the lake. The three classes are < 100 ueq/L, 100 peq/L < 200 peq/L, and 200
peq/L < 400 ueq/L
Among-Batch
Precision
The estimate of precision that includes effects of different laboratories and day-to-
day difference within a single laboratory, calculated from field audit sample data (as
percent relative standard deviation).
Analyte
A chemical species that is measured in a water sample.
201
-------
Analytical
Laboratory
Analytical
Laboratory Duplicates
In this report, a laboratory under contract with the U.S. Environmental Protection
Agency to analyze water samples shipped from the field laboratories.
Aliquots of a sample that is split in the analytical laboratory. The aliquots are analyzed
in the same batch.
Anion
A negatively charged ion.
Anion-Cation
Balance
In an electrically neutral solution such as water, the total charge of positive ions
(cations) equals the total charge of negative ions (anions). In this report, anion-
cation balance is expressed as percent ion balance difference (% IBD) and is
calculated as follows:
S anions — 2 cations + ANC
S anions - 2 cations + ANC + 2[H + ]
100
where:
2 anions = [Cf] + [F~] + [NOs"] + [SC>42']
Scations= [Na + ] + [K + ] + [Ca2 + ] + [Mg2 + ]
ANC = Alkalinity (the ANC value is included in the calculation to
account for the presence of unmeasured ions such as organic ions)
[H + ] = (10 -PH) x 106 ueq/L
Anion Deficit
ASTM Type I
Reagent-Grade
Water
Audit Sample
Base Cation
Batch
Batch ID
Bias
The concentration (in microequivalents per liter) of measured anions less the
measured cations.
Deionized water (which meets American Society for Testing and Materials [ASTM]
specifications for Type I reagent-grade water) that has a measured conductance of
less than 1 pS/cm at 25°C. This water is used in the preparation of blank samples
and reagents.
A standardized water sample submitted to an analytical laboratory for the purpose of
checking overall performance in sample analysis. Natural audit samples were lake
water; synthetic audit samples were prepared by diluting concentrates of known
chemical composition in ASTM Type I reagent-grade water.
A nonprotolytic cation that does not affect acid neutralizing capacity; usually calcium
or magnesium.
A group of samples processed and analyzed together. A field batch of samples is
defined as all samples (including quality assurance and quality control samples)
processed at one field laboratory in one day. A laboratory batch is defined as all
samples processed and analyzed at one analytical laboratory, associated with one set
of laboratory quality control samples.
The numeric identifier for each batch.
The systematic difference between values or sets of values.
202
-------
Blank Sample
Calculated
Conductance
Calibration Blank
Calibration Curve
Calibration
(Lake) Study
Carryover
Cation
Circumneutral
Closed System
Comparability
Completeness
Component
(of a system)
Conductance
Conductance
Balance
A sample of ASTM Type I reagent-grade water analyzed as a quality assurance or
quality control sample in WLS-I. (See calibration, reagent, trailer, and field blanks.)
The sum (as pS/cm) of the theoretical specific conductances of all measured ions in
a sample.
A solution used in standardizing or checking the calibration of analytical instruments;
also used to determine instrument detection limits.
The linear regression of the analytical instrument response to a set of calibration
standards (varying in concentrations) from which the linear dynamic range is
determined.
Study conducted during WLS-I to determine whether or not the methods of sample
collection (helicopter crew versus ground crew) affected the chemistry of the water
samples; samples collected during this study were also used to evaluate analytical
laboratory bias.
An artifact of the analyte carried from a sample of high concentration to a sub-
sequent sample or samples as a result of incomplete rinsing of an instrument or
apparatus.
A positively charged ion.
Close to neutrality in pH (near pH 7).
Method of measurement in which a water sample is collected and analyzed for pH
and dissolved inorganic carbon without exposure to the atmosphere. These samples
were collected in syringes directly from the Van Dorn sampling apparatus and were
analyzed in the field laboratory.
A measure of data quality that allows the similarity within and among data sets to be
established confidently.
A measure of data quality that is the quantity of acceptable data actually collected
relative to the total quantity that was attempted to be collected.
For this report, any of the sets of procedures used to get a sample from the lake to
analysis. Major components include sample collection, sample processing, and
sample analysis. Other components include sample transport, sample shipment, and
data reporting. Together, these components are the system.
A measure of the electrical conductance (the reciprocal of the electrical resistance)
or total ionic strength of a water sample expressed as uS/cm at 25°C.
A comparison of the measured conductance of a water sample (in pS/cm) to the
equivalent conductances (in pS/cm) of each ion measured in that water sample at
infinite dilution. In this report, conductance balance is expressed as percent
conductance difference (%CD) and is calculated as follows:
Confidence
Limit (95%)
calculated conductance — measured conductance
measured conductance
100
The ions used to calculate conductance are Ca2 +, Cf, COa2"
K + , Mg2 + , Na + , N03", OH", and °^-2"
H
A value that, in association with statistics, has a 95 percent chance of being above
the true value of the population of interest.
203
-------
Cubitainer
Data Base
Data Base Audit
Data Package
Data Qualifier
Data Quality
Objectives
Data Set 1
Data Set 2
Data Set 3
Data Set 4
Detectability
Detection Limit
Quality Control
Check Sample
Dilute Lake
Dissolved
Inorganic Carbon
Dissolved Organic
Carbon
Equivalent
Exception
Exception
Program
A 3.8-L container made of semirigid polyethylene used to transport field samples
(routine, duplicate, blank) from the lake site to the field laboratory.
All computerized results of the survey, which include the raw, verified, validated, and
final data sets as well as back-up and historical data sets.
An accounting of the data and of the data changes in the data base; includes
changes made within a data set and among all data sets.
A report, generated by an analytical laboratory for each batch of samples analyzed,
that includes analytical results, acid neutralizing capacity titration data, ion
chromatography specifications, analysis dates, calibration and reagent blank data,
quality control check sample results, matrix spike recovery results, and analytical
laboratory duplicate results, and standard addition results.
Annotation applied to a field or analytical measurement related to possible effects of
the quality of the datum. (See flags and tags).
Accuracy, detectability, and precision limits established before a sampling effort. Also
includes comparability, completeness, and representativeness.
Set of files containing raw data.
Set of files containing verified data.
Set of files containing validated data.
Set of files containing final, enhanced lake data: missing values or errors in the
validated data set were replaced by substitution values; duplicate values were
averaged; negative values (except for acid neutralizing capacity) were set equal to
zero.
The capacity of an instrument or method to determine a measured value for an
analyte above background levels.
A quality control check sample with a theoretical concentration designed to check
instrument calibration at the low end of the linear dynamic range.
For this report, a lake with a conductance of less than 10 pS/cm.
A measure of the dissolved carbon dioxide, carbonic acid, bicarbonate and carbonate
anions that constitute the major part of acid neutralizing capacity in a lake.
The organic fraction of carbon in a water sample that is dissolved or unalterable (for
this report, 0.45-ym pore size).
Unit of ionic charge; the quantity of a substance that either gains or loses one mole
of protons or electrons.
An analytical result that does not meet the expected quality assurance or quality
control criteria for which a data flag is generated.
A computer program in AQUARIUS that identifies or flags analytical results classified
as exceptions.
204
-------
Extractable
Aluminum
Field Audit Sample
Field Base
Field Blank Sample
Field Duplicate
Sample
Field Duplicate Pair
Field Inter-
laboratory Bias
Field Laboratory
Field Laboratory
Among-Batch
Precision
Field Natural
Audit Sample
Field Synthetic
Audit Sample
Final Data Set
Flag
Gran Analysis
Ground Crew
Ground Sample
Operationally defined aluminum fraction that is extracted by the procedure used in
WLS-I; this measurement is intended to provide an indication of the concentration of
the aluminum species that may be available in a form toxic to fish.
A standardized water sample submitted to field laboratories to check overall
performance in sample analysis by field laboratories and by analytical laboratories.
Natural field audit samples were lake water; synthetic field audit samples were
prepared by diluting concentrates of known chemical composition into ASTM type I
reagent-grade water.
A location providing support for helicopters, sampling personnel, and field laboratories
during field sampling operations.
A sample of ASTM Type I reagent-grade water prepared at the field laboratory and
transported to the lake site by the field sampling crews. At the lake site, the blank
was processed through the Van Dorn sampling apparatus. These samples were
analyzed at field laboratories (except for pH and DIG) and at analytical laboratories
and were employed in the calculation of system decision and system detection limits
and quantitation limits.
Second sample of lake water collected by the sampling crew at the same location
and depth at the lake site immediately after the routine sample, in accordance with
standardized protocols.
A routine lake water sample and a second sample (field duplicate sample) collected
from the same lake, by the same sampling crew, during the same visit, and according
to the same procedure.
The systematic difference in measurement of an analyte between two or more field
laboratories.
Mobile laboratory (trailers) in which sample processing and measurement of selected
variables were performed. One field laboratory was located at each field base.
The estimate of day-to-day variability of the analytical measurements performed in
the field laboratory for a particular audit sample lot, calculated as percent relative
standard deviation (for turbidity, true color and closed-system DIG) and standard
deviation (for closed-system pH).
See field audit sample.
See field audit sample.
Data Set 4. (See definition for Data Set 4.)
Qualifier of a data point that did not meet established acceptance criteria that were
assigned during the verification and validation procedures.
A mathematical procedure used to identify the equivalence point or points of the
titration of a carbonate system and subsequently for acid and base neutralizing
capacities of that system.
A team of lake sampling personnel who gained access to the lake site on foot or with
pack animals and who sampled the lake from an inflatable boat.
A lake sample (routine or duplicate) or a field blank sample collected by the ground
crew.
205
-------
Helicopter Crew
Helicopter Sample
Holding Time
Hydrolab
Imprecision
In Situ
Initial DIG
Instrumental
Detection Limit
Interlaboratory
Bias
Intralaboratory Bias
Intralaboratory
Precision Goal
Ionic Strength
Laboratory Bias
Laboratory Blank
Sample
Laboratory
Duplicate Sample
Lake ID
Linear Dynamic
Range
Loran-C
A team of lake sampling personnel who gained access to and sampled the lake from
a pontoon-equipped helicopter.
A lake sample (routine, duplicate, or triplicate) or a field blank sample collected by the
helicopter crew.
(1) In the field laboratory,the time elapsed between sample collection and sample
preservation. (2) In the analytical laboratory, the elapsed time between sample
processing in the field laboratory and final sample analysis or reanalysis.
In situ water quality analytical instrument for the measurement of pH, conductance,
and temperature.
The degree of irreproductibility or deviation of a measurement from the expected or
average of a set of measurements for a particular analyte; the variation about the
mean.
Referring to measurements taken within the water column of a lake.
A measurement of dissolved inorganic carbon made on an aliquot immediately before
it is titrated for acid neutralizing capacity.
For each chemical variable, value calculated from laboratory calibration or reagent
blank samples that indicates the minimum concentration reliably detectable by the
instrument(s) used; calculated as three times the standard deviation of 10
nonconsecutive blank analyses (on the same calibration curve).
Systematic differences in performance between laboratories estimated from analysis
of the same type of samples.
The degree of imprecision or uncertainty of measurement in the analysis of an
analyte in the laboratory.
A precision goal based on the data quality objectives for the analysis of laboratory
duplicate pairs within a single laboratory.
A measure of the interionic effect resulting from the electrical attraction and repulsion
between different ions. In very dilute solutions, ions behave independently of each
other, and the ionic strength can be recalculated from the measured concentrations
of anions and cations present in the solution.
The degree of uncertainty of the measurement of an analyte within a laboratory or
between laboratories; see intralaboratory bias and interlaboratory bias.
A sample of ASTM Type I reagent-grade water prepared and analyzed by analytical
laboratories. (See calibration blank, reagent blank.)
Sample aliquot that is split and prepared at the analytical laboratories and that is
analyzed in a batch.
An identification code assigned to each lake in the survey which indicates subregion,
alkalinity characteristics, and map coordinates.
The range of analyte concentration for which the calibration curve is a straight line.
A system of long-range navigation that uses paired radio signals to determine the
geographic position of a target lake.
206
-------
Management Team
Matrix
Matrix Spike
Nitrate-Sulfate Split
On-Site Evaluation
Open System
Outlier
PSO
P95
Percent Ion
Balance
Difference
Percent Recovery
Percent Relative
Standard
Deviation (% RSD)
pH
pH, acidity
pH, alkalinity
Platinum Cobalt
Unit
Population Estimate
Practical Difference
EPA personnel responsible for overseeing the WLS-I sampling and quality
assurance design and the subsequent interpretation of lake data results.
The physical and chemical composition of a sample being analyzed.
A quality control sample, analyzed at an analytical laboratory, that was prepared by
adding a known concentration of analyte to a sample. Matrix spike samples were
used to determine possible chemical interferences within a sample that might affect
the analytical result.
A 125-mL fraction of the sample taken directly from the Van Dorn sampling
apparatus, immediately preserved with HgCIa, and subsequently analyzed at EMSL-
LV for NO3" and SO4 -
A formal on-site review of field sampling, field laboratory, or analytical laboratory
activities to verify that standardized protocols are being followed.
A measurement of pH or dissolved inorganic carbon obtained from a sample that was
exposed to the atmosphere during collection, processing, and preparation before
measurement.
Observation not typical of the population from which the sample is drawn.
The median value of blank sample analyses.
The 95th percentile of the blank sample analysis.
A quality assurance procedure used to check that the sum of the anion equivalents
equals the sum of the cation equivalents (see anion-cation balance).
A calculation of the matrix spike sample which indicates the effect of the sample
matrix on the analytical measurement (also termed matrix interference).
The standard deviation divided by the mean, multiplied by 100, expressed as percent.
Also known as the coefficient of variation.
The negative logarithm of the hydrogen-ion activity. The pH scale runs from 1 (most
acidic) to 14 (most alkaline); the difference of 1 pH unit indicates a 10-fold change
in hydrogen-ion activity.
A measurement of pH made in the analytical laboratory immediately before the BNC
titration procedure and before the KCI spike has been added.
A measurement of pH made in the analytical laboratory immediately before the ANC
titration procedure and before the KCI spike has been added.
Measure of the color of a water sample defined by a potassium hexachloroplatinate
and cobalt chloride standard color series.
A statistical estimate of the number of lakes (target lakes) with a particular
characteristic (i.e., alkalinity class of a subregion) extrapolated from the number of
lakes sampled (probability sample).
Judgemental difference between a measurement result and an expected result
(usually expressed in absolute terms as units of measure).
207
-------
Precision
Primary Variables
Protolyte
Protolyte Analysis
Program
Quality Assurance
Quality Assurance
Sample
Quality Control
Quality Control
Check Sample
Quality Control
Sample
Quantitation Limit
Raw Data Set
Reagent
Reagent Blank
Relative Bias
Remote Base Site
Representativeness
Required
Detection Limit
A measure of the capacity of a method to provide reproducible measurements of a
particular analyte.
Variables of foremost concern in the survey (pH, acid neutralizing capacity,
extractable aluminum, sulfate, calcium, dissolved organic carbon).
That portion of a molecule that reacts with either H + or OH" in solution.
An exception-generating computer program of AQUARIUS that evaluates in situ,
field laboratory, and analytical laboratory measurements of pH, DIG, ANC, BNC, and
DOC in light of known characteristics of carbonate equilibria.
Steps taken to ensure that a study is adequately planned and implemented to provide
data of the highest quality, and that adequate information is provided to determine the
quality of the data base resulting from the study.
A sample (other than the routine lake sample) that is analyzed in the analytical
laboratory and that has an origin and composition unknown to the analyst.
Steps taken during sample collection and analysis to ensure that the data quality
meets the minimum standards established by the quality assurance plan.
A sample of known concentration used to verify continued calibration of an instrment.
Any sample used by analysts to check immediate instrument calibration or response;
the measurement obtained from a quality control sample is expected to fall within
specific acceptance criteria or control limits.
For each chemical variable (except pH), a value (calculated from blank samples) that
represents the lowest concentration that can be measured with reasonable precision;
determined as 10 times the standard deviation of a type of blank sample.
The initial data set (Data Set 1) that has received a cursory review to confirm that
data are provided in proper format and are complete and legible.
A substance added to water (because of its chemical reactivity) to determine the
concentration of a specific analyte.
A laboratory blank sample that contained all the reagents required to prepare a
sample for analysis of silica and total aluminum.
The expected difference between a measured value and the true value, expressed as
a percentage of the true value.
Location serving as a base of operations for sampling crews working more than 150
miles from the field laboratory; samples collected by these crews had to be flown to
the field laboratory daily.
A measure of data quality; the degree to which sample data accurately and precisely
reflect the characteristics of a population.
For each chemical variable, the highest instrument detection limit allowable in the
analytical laboratory contract.
208
-------
Root-mean-square
A summary statistic of the relative or absolute standard deviation (SD); a pooled
standard deviation of the percent relative standard deviation (%RSD), calculated by
the formula:
Routine Sample
Sample ID
Sampling Method
Bias
SAS
Secondary
Variables
Sparging
Spike
Split Sample
Standard
Additions
Standard Deviation
SDo/oRSD (
The first lake sample collected at a site in accordance with standardized protocols.
The numeric identifier given to each lake sample and quality assurance sample in
each batch.
Systematic difference between analytical results of samples collected by helicopter
access and samples collected by ground access.
Statistical Analysis System, Inc. (Gary, NC). A statistical data file manipulation
package that has data management, statistical, and graphical analysis abilities. The
WLS-I data base was developed and analyzed primarily using SAS software and is
distributed in SAS format.
Chemical variables measured during WLS-I considered to be important in providing
additional data in quantifying the chemical status of lakes, e.g., sodium, magnesium,
potassium, nitrate, chloride, and total aluminum.
A sample preparation procedure that involves bubbling a gas into an aliquot.
A known concentration of an analyte introduced into a sample or aliquot.
A subsample (aliquot) of a field batch sample that was sent for analysis to a
laboratory other than an analytical laboratory; also a procedure of separating one
aliquot (or sample) into two.
An analytical procedure in which equal volumes of a sample are added to a series of
known and varied concentrations of the analyte. This procedure is utilized only when
there is a suspected matrix interference indicated with the matrix spike sample.
The square root of the variance of a given statistic, calculated by the equation:
( X - X)2/(n - 1)
Statistical
(significant)
Difference
Stratified Lake
Synoptic
System Decision
Limit
A high probability that two sets of measurements did not come from the same
population of measurements.
In this report, a lake with a temperature difference greater than 4°C between the
water layers at 1.5 m below the surface and 1.5 m above the lake bottom. If the
temperature difference is also greater than 4°C between the water layers at 1.5 m
below the surface and 60 percent of site depth, then the lake is strongly stratified; if
not, it is weakly stratified.
Relating to or displaying conditions as they exist simultaneously over a broad area.
For each chemical variable except pH, a value that reliably indicates a concentration
above background, estimated as the 95th percentile (Pgs) of the field blank sample
concentration.
209
-------
System Detection
Limit
System Precision
Systematic Error
Systematic Random
Sampling
Tag
Titration Data
Trailer Blank
Trailer Duplicate
Triplicate Lake
Sample
True Color
Tuple
Turbidity
Van Dorn Sampler
Validation
Verification
Withheld Sample
For each chemical variable, except pH, a value indicating the highest concentration of
analyte that could be present in a routine lake sample in which the analyte was not
detected, estimated as 2(Pg5 - P50) where Pgs is the 95th percentile and P5u is the
50th percentile (median) of the field blank sample concentration.
Cumulative variability associated with sample collection, transport, processing,
preservation, shipment, analysis, and data reporting. An estimate of data certainty for
each analyte and the amount of variability associated with analyte concentration; the
estimate is based on the statistical evaluation of field duplicate pairs.
A consistent error introduced in the measuring process. Such error commonly results
in biased estimations.
The technique used in the survey to select the lakes to be sampled.
Code on a data point that is added at the time of collection or analysis to qualify the
datum.
Individual data points from the Gran analysis of acid neutralizing capacity and base
neutralizing capacity.
An ASTM Type I reagent-grade water sample prepared and processed at the field
laboratory but analyzed at an analytical laboratory.
Split sample prepared and analyzed at the field laboratory.
The third sample of lake water collected by the helicopter crew at a lake immediately
after the routine and duplicate samples are collected in accordance with standardized
protocols; this third sample was used only as part of the calibration study.
The color of water that has been filtered or centrifuged to remove particles that may
impart an apparent color; true color ranges from clear blue to blackish-brown.
A SAS observation generated by an exception program or by a QA auditor. Used to
record changes to existing data sets or to qualify a data point.
A measure of light scattering by suspended particles in an unfiltered water sample.
A water collection apparatus with a volume of 6.2 L used to sample a water column
in the lake.
Process by which data are evaluated for quality with reference to the intended data
use; includes identification of outliers and evaluation of potential systematic error after
data verification.
Process of ascertaining the quality of the data in accordance with the minimum
standards established by the quality assurance plan.
One of the three samples collected from a lake by the helicopter crew during the
calibration study. As part of holding time experiment, this sample was held in the dark
at 4°C for a specified period prior to processing and preservation.
210
-------
O k-
o o
CL 2
U S Environmental Protection Agency
Region 5, Library (5PL-16)
23Q S Dearborn Street, Room 1670
Chicago, It. 60604
------- |