;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, ^-^ ^*. 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 ° 
-------

-------
                                              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

-------