vvEPA
              United States
              Environmental Protection
              Agency
              Office of Acid Deposition,
              Environmental Monitoring and
              Quality Assurance
              Washington DC 20460
EPA/600/4-86/011
August 1987
              Research and Development
Eastern  Lake Survey
Phase I
              Quality Assurance
              Report

-------
                                          EPA 600/4-86/011
                                          August 1987
Eastern  Lake  Survey

           Phase I
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 89114
           Environmental Research Laboratory - Corvallls. OR 97333

-------
                                           NOTICE


The information in this document has been funded wholly or in part by the U.S. Environmental Protection
Agency (EPA) under Contract No. 68-03-3249 and 68-03-3050 to Lockheed Engineering and Management Services
Company, Inc., No. 68-02-3889 to Radian Corporation, No. 68-03-3246 to Northrop Services, Inc., and Interagency
Agreement No. 40-1441-84 with the U.S. Department of Energy. It has been approved for publication as an EPA
document.

Mention of corporation names, trade names or commercial products does not constitute endorsement or rec-
ommendation for use.

This document has been published previously. As part of the AERP Technical Information Program, this docu-
ment has been repackaged and retitled to clearly identify its relationship to other documents produced for the
Eastern Lake Survey. The document contents and reference number have not changed. Proper citation of this
document remains:

Best, M. D., S. K. Drouse, L. W. Creelman, and D. J. Chalous. National Surface Water Survey, Eastern Lake Survey
(Phase I - Pynoptic  Chemistry) Quality Assurance Report. EPA600/4-86/011, U.S. Environmental Protection
Agency, Las Vegas, NV. 1986.

-------
                                          ABSTRACT


The National Surface Water Survey (NSWS) is a three-phase project of the National Acid Precipitation Assess-
ment Program (NAPAP). The objectives of the survey are to evaluate the current water chemistry of acid-sensi-
tive lakes and streams in the U.S., to determine the status of fisheries and other biotic resources in those
waters, and to quantify changes in a representative subset of lakes and streams through a long-term monitoring
program. Phase I of the Eastern Lake Survey is the first part of the NSWS lake study. This Quality Assurance
Report is a retrospective, comprehensive overview of the quality assurance activities and results of the Eastern
Lake Survey Phase I. The report describes the chemical  parameters measured, the sampling and analytical
methods used, and the quality assurance procedures required for field, laboratory, and data base operations.
The report also discusses the rationales and testing that led to the implementation of specific protocols. These
protocols were extensively reevaluated during and after the survey, as described in this document. The statisti-
cal testing of the analytical  and quality assurance data is explained, and the results of these tests are pre-
sented.

Overall,  Phase I of the  Eastern Lake Survey was successful in achieving its objectives. The  quality assurance
requirements proved adequate to ensure that all samples were collected and analyzed consistently, and that the
resulting data were scientifically sound and of known quality. This report was submitted in partial fulfillment of
contract numbers 68-03-3050 and 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 planning, imple-
mentation, and data review period from March 1983, to January 1986, and work was completed as of March 1986.

-------
                                           Contents

Abstract	     iii
Figures	     vi
Tables	     vii
Acknowledgement	     ix
1.  Introduction	     1
     Eastern Lake Survey	     1
     Survey Participants	     1
     Measurement Requirements	     1
     Data Quality Objectives	     1
     Development of Documents for the Eastern Lake Survey - Phase I	     6
     Pilot Studies	     6
     Sampling and Analytical Methodologies	     6
     Data Comparability Studies	     7
2  Operational Quality Assurance-Quality Control Program	     8
     Field Station Organization and Responsibility	     8
     Selection of Contract Analytical Laboratories	     8
         Statement of Work	     8
         Invitation for Bid	     9
         Analytical Laboratory Evaluation	      9
     Training	      9
     Communications	     9
         Field Communications	      9
         Daily Communications with Contract Analytical Laboratories	     10
     Sampling and Field Laboratory Quality Control Protocols	     10
         Sample Preservation	     10
         Field Laboratory Analyses	     10
     Contract Analytical Laboratory Quality Control Protocols	     13
         Contract Analytical Laboratory Sample Holding Times	     13
         Reporting Requirements	     14
     Internal Quality Control	     14
     Field Laboratory and Contract Analytical Laboratory On-Site Evaluations	     14
3. Data Base Quality Assurance	     17
     Data and Sample Transfer	     17
         Transfer of Samples and Data From Field Stations	     17
         Transfer of Contract Analytical Laboratory Data	     "17
         Raw Data Set	     22
      Data Verification	     22
         Daily Communication	     23

                                                IV

-------
         Data Receipt	     23
         Quality Assurance and Quality Control Data	     27
         Follow-up with Contract Analytical Laboratories	     30
         Preparation and Delivery of Verification Tapes	     30
     Data Validation	     31
4.  Results	     35
     Operations Evaluation	     35
         Lake and Sample Information	     35
         Field Problems and Their Resolutions	     35
         Contract Analytical Laboratory Problems and Resolutions	     37
         Data Verification Problems and Resolutions	     37
     Methods Evaluation 	     38
         Fluoride Determinations	     38
         Acid-Neutralizing Capacity and Base-Neutralizing Capacity	     38
         Total Extractable Aluminum	     38
         pH Sample Chamber	     39
         Formation of Filterable Iron and Aluminum Complexes in Synthetic Field Audit
         Samples	     39
         Nitrate Contamination	     39
     Evaluation of Quality Assurance Data	     42
         Blank Data	     42
         Duplicate Data	     44
         Audit Sample Data	     48
5.  Data Variability in the Eastern Lake Survey • Phase I	     59
     Comparisons of Precision Estimates	     59
     Expected Relationships Between Precision Estimates	     59
Summary	     61
References	     62
Appendices 	
A - Analysis of Quality Assurance Data for the Eastern Lake Survey	     65
B - Instrumental Detection Limits, System Detection Limits, and System Decision Limits by
   Laboratory, Eastern Lake Survey - Phase 1	     117
C - Overall Within-Batch Precision by Laboratory for 23 Parameters, Eastern Lake Survey -
   Phase  I	     123
D - Analytical Within-Batch Precision by Laboratory for 23 Parameters, Eastern Lake Survey -
   Phase  I	     129
E - Overall and Analytical Among-Batch Precision by Laboratory for 23 Parameters, Eastern
   Lake Survey -  Phase  I	     135

-------
                                         Figures

Number                                                                              Page

   1     Regions and subregions sampled during the Eastern Lake Survey - Phase 1	      2
   2     Flow of samples and data through field and analytical laboratories, Eastern Lake
         Survey - Phase I	     11
   3     NSWS Form 3 - Shipping	     18
   4     NSWS Form 1 - Lake Data	     19
   5     NSWS Form 2 - Field Laboratory Data	     20
   6     Eastern Lake Survey - Phase I data flow scheme	     21
   7     Data verification process for the Eastern Lake Survey - Phase I	     26
   8     NSWS Data Package Completeness Checklist	     28
   9     Data validation process for the Eastern Lake Survey - Phase I 	     34

-------
                                           Tables

Number                                                                                Page
   1      Parameters Measured, Quality Assurance Criteria, and Analytical Methods for the
          Eastern Lake Survey - Phase I	      3
   2      Sample Preservation Requirements	     12
   3      Descriptions, Applications, and Frequencies of Quality Assurance and Quality Con-
          trol Samples, Eastern  Lake Survey - Phase I	     15
   4      Descriptions and Applications of Quality Control Samples, Eastern Lake Survey -
          Phase I	     16
   5      List of Data Forms Used by the Contract Analytical Laboratory, Eastern Lake Survey
          - Phase 1	     22
   6      Eastern Lake Survey - Phase I Field and Laboratory Data Qualifiers (Tags')	     23
   7      Eastern Lake Survey - Phase I Verification	     24
   8      Exception Generating and Data Review Programs, Eastern Lake Survey - Phase I..     30
   9      NSWS Missing Value Codes	     31
   10      Data Validation for the Eastern Lake Survey - Phase I: Comparison of Variables Used
          to Check for Random and Systematic Errors	     32
   11      Physical Parameters Subject to Validation, Eastern Lake Survey - Phase  I	     33
   12      Numbers of Samples Received and Analyzed by Contract Laboratories, Eastern Lake
          Survey - Phase I	     36
   13      Numbers of Samples Delivered by Field Stations, Eastern Lake Survey - Phase I  ..     36
   14      Cleaning Procedure Used for Aliquot 3 Sample Containers, Eastern Lake Survey -
          Phase I Pilot Study	     40
   15      Filtration  Procedure Originally Used  by  Field Personnel, Eastern Lake Survey -
          Phase I	     40
   16      Field Laboratory Filtration Procedures Used in Nitrate Contamination Experiment,
          Eastern Lake Survey - Phase I	     41
   17      Description of Field Laboratory Blank Samples Collected in Nitrate Contamination
          Experiment, Eastern Lake Survey - Phase I	     42
   18      Nitrate Concentrations in Field Laboratory Blank Samples in Nitrate Contamination
          Experiment, Eastern Lake Survey - Phase I	     43
   19      Instrumental Detection Limits, System Detection Limits, and System Decision Lim-
          its for 20 Parameters, Eastern Lake Survey - Phase I	     45
   20      Overall Within-Batch Precision Estimated from Field Duplicate Data and Field Blank
          Data for Measurements of 23 Parameters, Eastern Lake Survey - Phase I	     46
   21      Overall and Analytical Within-Batch Precision  Estimated from Field Duplicate and
          Trailer Duplicate Data for Measurements of Four Parameters, Eastern Lake Survey -
          Phase I	     49
   22      Analytical Within-Batch Precision Estimated from Contract Laboratory  Duplicate
          Data and Calibration Blank Data for Measurements of 23 Parameters, Eastern Lake
          Survey - Phase I	     50
                                             vu

-------
23     Overall Among-Batch Precision Estimated from Field Natural, Lot 2 (FN2, Big Moose
       Lake) and Field Natural, Lot 3 (FN3, Lake Superior) Audit Sample Data For Measure-
       ments of 23 Parameters, Eastern Lake Survey - Phase I	      52
24     Mean Measured Values, Overall and Analytical Among-Batch Precision Estimates,
       and Theoretical Concentrations of  High Synthetic Audit Samples, Eastern Lake Sur-
       vey - Phase I	      53
25     Mean Measured Values, Overall and Analytical Among-Batch Precision Estimates,
       and Theoretical Concentrations of Low Synthetic Audit Samples, Eastern Lake Sur-
       vey - Phase I	      55
26     Contract Analytical Laboratory Performance Windows for Audit Sample Measure-
       ments, Eastern Lake Survey - Phase 1	      58

-------
                                     Acknowledgments


The completion of this report would not have been possible without the efforts of many individuals who are
associated with the National Surface Water Survey. In particular, the authors wish to acknowledge the technical
assistance, contributions, and editorial comments that were provided by the following people: R. Linthurst, E.
Meier, and R. Schonbrod (U.S. Environmental Protection Agency); D. Landers (State University of New York); P.
Kanciruk(Oak Ridge National Laboratory); C. Macleod, D. Hillman, D. Peck, T. Lewis, M. Faber, J. Engels, J. Villa,
M. Stapanian, J. Lau, C. Mericas, and S. Simon (Lockheed-EMSCO); J. Eilers (Northrop Services, Inc.); D. Brakke
(Western Washington University); T. Permutt and A. Pollack (Systems Applications Inc.); and the word process-
ing staff of Computer Sciences Corporation.
                                               IX

-------
                                           Section 1
                                         Introduction
The National Surface Water Survey (NSWS) is a three-
phase project within the National Acid Precipitation
Assessment Program (NAPAP). The NSWS was initi-
ated by the U.S. Environmental Protection Agency
(EPA) in 1983. The purpose of Phase I of the NSWS is
to document the present chemical status of lakes and
streams in areas of the U.S. that are potentially sus-
ceptible to the effects of acidic deposition (Linthurst
et al., 1986). Phase II of the survey is intended to deter-
mine the present status of  biotic resources and to
assess the chemical variability within and among sur-
face waters characterized during Phase I. Phase III, a
long-term monitoring program, will quantify changes
in the aquatic resources of a subset of Phase II sur-
face waters.

The  scientific and legislative decisions that will be
based on the data from the NSWS must be well sup-
ported.  Therefore,  an  extensive quality assurance
(QA) program has been established to ensure that the
best possible data are collected and that the quality
of the data can be defined and defended at the com-
pletion of the field surveys.


Eastern Lake Survey

This report summarizes the results of the QA program
for Phase I of the Eastern Lake Survey (ELS-i). The QA
program, including data verification and  validation
procedures, is discussed in greater detail in Drouseet
al. (1986). Analytical methods and field sampling pro-
tocols for the ELS-I are discussed in detail  in Hillman
et al. (1986) and Morris et al. (1986), respectively. In
total, 1800 lakes were sampled  in the ELS-I. The lakes
were selected from  three regions east of the  Missis-
sippi River that are potentially  susceptible to acidifi-
cation (Figure 1).


Survey Participants

Planning, conducting, and  interpreting a study of the
magnitude of the NSWS required the cooperation of
numerous government agencies and private  organi-
zations. Development of the NSWS included defini-
tion of measurement requirements and survey objec-
tives, which was accomplished through discussions
at meetings and workshops. Participants included
representatives  of several government agencies as
well as other scientists involved in acidic deposition
research (U.S. EPA 1984a and 1984b).

The U.S. EPA Environmental Monitoring Systems Lab-
oratory-Las Vegas, Nevada (EMSL-LV) had  primary
responsibility for the ELS-I QA program and sampling
operations. The Agency receives assistance in this
area from Lockheed Engineering  and Management
Services Company, Inc. (Lockheed-EMSCO) which is
the prime contractor for EMSL-LV. Sampling and qual-
ity assurance activities were performed by Lockheed-
EMSCO personnel. The relevant state agencies and
EPA regional offices were also involved in the sam-
pling activities. The data base operations were per-
formed  by Oak Ridge National Laboratory  (ORNL).
The EPA Environmental Research Laboratory-Corval-
lis, Oregon (ERL-Corvallis) had primary responsibility
for design and planning of the ELS-I, as well  as for
data validation and interpretation.


Measurement Requirements

The first step in  design of the QA program was to
define the measurement requirements for the  ELS-I.
Thirty-two parameters were selected for in situ or lab-
oratory  measurement during Phase I  (Table 1). The
EPA data users (see Linthurst, et al. 1986) reasoned
that measurement  of these chemical and physical
parameters would provide adequate information from
single lake samples for the evaluation of the present
status of lakes and thus would meet the objectives of
Phase I. A brief description of each parameter  is pre-
sented in Overton et al. (1986).

Data Quality Objectives

Data quality objectives (DQO's) were defined early in
1984 in  terms of anticipated value range, detection
limits, and precision for each measurement parame-
ter. These objectives were developed using data from
published literature and from statistical error simula-
tion. Equipment, sampling  protocols, and analytical
methodologies were selected and standardized in
order to achieve the DQO's. The pilot studies then
afforded the opportunity to evaluate and revise the
analytical  methodologies,  equipment, and DQO's.
One change to the DQO's was in the detection limit for
total phosphorus which was lowered from 3 to 2 ppb
because experience indicated the lower limit was

-------
           2C
      2A
2D
    UPPER
   MIDWEST
   REGION
                                               NORTHEAST
                                                 REGION
\        ili
 \        /   v
  1        \    >
  v>—A  /
    ^      \^~r
                               \   —

                           SOUTHEAST
                             REGION
      Figure 1. Regions and Subregions Sampled During the Eastern Lake Survey - Phase I.

-------
Observed
Parameter3 Range
IN SITU
pH 360-10.46
Conductance pS crrf1 00-1267.0
Lake Temperature. *C
Secchi disk
Transparency, m
FIELD LABORATORY
Laboratory pH 3.81 - 9.36
(closed system)
Dissolved inorganic carbon 0 16 - 48.99
(closed system), mg L'1
True color, PCU 0 - 345
Turbidity, NTU 0 - 290
CONTRACT ANALYTICAL LABORATORY
Acid-neutralizing capacity -209.1
(ANC).Meq L"1 +40466
Intra- Maximum
Required Laboratory Sample Holding
Detection Precision Time - Days Instrument
Limit Goal (%)" (Analytical Lab) or Method0
Potentiometer
(Hydrolab)
Conductivity cell
(Hydrolab)
Thermistor
(Hydrolab)
Secchi disk
0.1 g — pH meter
Orion Model 611
0.05 10 — Infrared spectroscopy
Dohrmann DC-80
carbon analyzer
0 5 g — Comparator Hach
Model CO-1
2 10 — Nephelometer Monitek
Model 21
5 10 14 Acidimetric titration,
modified Gran analysis
Reference
(Laboratory
Methods)''
Morris et al.
(1986)
Morris et al
(1986)
Morris et al.
(1986)
Morris et al.
(1986)
EPA 150.1
EPA 41 5 2
(modified)
EPA 110. 2
(modified)
EPA 180.1
Hillman et al.
(1986);
Kramer (1984)
Table 1. Parameters Measured, Quality Assurance Criteria, and Analytical Methods for the Eastern Lake Survey - Phase I.

-------
Parameter3
Total Extractable
Total
Base-neutralizing capacity
Calcium (ca),
mgr1«>
Chloride (C1"),
mg L~1(l)
Conductance, /uS cm"1
Dissolved inorganic carbon
(DIG) (air-equilibrated), mg L"1
Dissolved inorganic carbon
(DIG) (initial ANC), mg L"1
Dissolved organic carbon
(DOC).mg L"1
Fluoride (F~), total dissolved,
mg L
Iron (fe), mg L"1(e)
Magnesium (Mg), mg L"1(e)
Manganese (Mn), mg L~1(d)
Observed
Range
-0.009-3.594
-0.002-9.678
	
0.19-60.94
0.01-609
7.8-3613.3
-0.07-46.91
0.15-4983
0.0-48.2
0 001 -0.839
-0.034-2.64
0.10-39 76
-0 02 - 2 03
Required
Detection
Limit
0.005
0.005
5
0.01
0.01
-h
0.05
0.05
0 1
0.005
001
0.01
0.01
Intra-
Laboratory
Precision
Goal(%)b
10 (X). 010)
20 (<0.010)
10(>0.010)
20(<0.010)
10
5
5
1
10
10
5 (>5.0)
10 (<5 0)
5
10
5
10
Maximum
Sample Holding
Time - Days
(Analytical Lab)
7
28
14
28
28
14
14
14
14
28
28
28
28
Instrument
or Method11
Furnace AAS on MIBK
extract
Furnace AAS
Alkalimetric titration,
modified Gran analysis
Flame AAS, ICPES
1C
Conductivity cell
IR
IR
IR
ISE
Flame AAS, ICPES
Flame AAS, ICPES
Flame AAS
Reference
(Laboratory
Methods/
Hillman et al.
(1986)
EPA 202.2
Hillman et al.
(1986); Kramer
(1984)
EPA 215.1
ASTM (1984);
O'Dell etal.
(1984)
EPA 120.1
EPA 415. 2
(modified)
EPA 415. 2
(modified)
EPA 415. 2
EPA 340.2
(modified)
EPA 236
EPA 242.1
EPA 243.1
Table 1. Parameters Measured, Quality Assurance Criteria, and Analytical Methods for the Eastern Lake Survey - Phase I (Continued).

-------

Observed
Parameter3 Range

Required
Detection
Limit
Intra-
Laboratory
Precision
Goal (%)"
Maximum
Sample Holding
Time - Days
(Analytical Lab)

Instrument
or Method1

Reference
(Laboratory
Methods/
                  "1(e>
 Nitrate (N03~, mg L



 pH (air-equilibrated)

 pH (initial ANC)

 pH (initial (BNC)
 Phosphorus (P), total, mg
  -'(e)
 Potassium (K), mg L~1(f)

 Silica (Si02), mg L"1(e)


 Sodium (Na), mg L"1|f)

 Sulfate (S042~), mg L'1"1
 -0.106-30 6



  3.82-893

  380-8.78

  3.81 -882

-0.006- 0.833





 000-24.98

 -1.14-4353


  0.06-323

  00-119
0.005
0.002





0.01

0.05


0.01

0.05
    10



    0.5

   0.05

   0.05

10 (>0 010)
20(<0.101)



     5

     5


     5

     5
 7           1C



 7           pH meter

 7           pH meter

 7           pH meter

28           Colorimetry
             (phosphomolybdate, or
             modification,
             automated)

28           Flame AAS

28           Colorimetry
             (automated)

 7           Flame AAS,  ICPES

28           1C
ASTM (1984),
O'Dell etal.
(1984)

EPA 150.1

EPA 150 1

EPA 150.1

USGS I-4600
EPA 258 1


USGSi-2700

EPA 273 1

ASTM (1984),
O'Dell etal.
(1984)
a Dissolved ions and metals were determined, except where noted

  Relative precision was calculated for samples at concentrations above 10 times instrumental detection limits, except where noted
c AAS  = atomic absorption spectroscopy, MIBK = methyl isobutyl ketone, ICPES =  mductilely coupled plasma emission spectroscopy, 1C  = ion chromatography, IR
  infrared spectrophotometry; ISE =  ion-selective electrode

d In situ measurements are outlined in Morris et al. (1986), EPA methods are from U.S. EPA 1983); USGS methods are from Skougstad et al (1979)
e Values converted to/*g L"1 for data analysis Required detection limits are in mg L"'.
  Values converted to Meq L~1 for data analysis Required detection limits are in mg L"'.
9 Absolute precision  goal in applicable units

  The mean of six nonconsecutive blank measurements was required to be less than 0 9*iS cm"'.
 Table 1.  Parameters Measured, Quality Assurance Criteria, and Analytical Methods for the Eastern Lake Survey - Phase I (Concluded).

-------
necessary. The final DQO's were provided in a docu-
ment delivered to the quality assurance management
staff of the EPA in October 1984. The observed ranges
of values in the verified data set, required detection
limits,  intralaboratory  precision  goals,  maximum
sample holding times,  and measurement methods
used for  each variable are listed in Table  1. The
instances where extreme values  affected the  esti-
mated precision are listed in Section 5.


Development of Documents for the
Eastern Lake Survey (Phase I)

Protocols for sampling, chemical analysis, and  data
processing  were based on the best  available  pub-
lished methods. A draft QA plan and a draft analytical
methods  manual were used during pilot studies con-
ducted to test all aspects of the ELS-I research plan,
including logistics, methods, and QA.  The evaluation
and modification of the data quality objectives, sam-
pling and analytical methodologies, and verification
and validation procedures were of particular impor-
tance to the QA program. Prior to commencement of
the ELS-I, the draft QA plan and the draft analytical
methods  manual were revised on the basis of results
obtained  from the pilot studies. The final QA plan and
the final  methods manual incorporated both those
revisions  and any changes implemented during the
ELS-I.


Pilot Studies

Two pilot studies for the ELS-I were undertaken dur-
ing the winter and spring of  1984. The winter  pilot
study was conducted in January 1984 and consisted
of sampling  50  ice-covered  lakes in  Maine,  New
Hampshire, and  Vermont. During the  spring  pilot
study in May-June, 137 lakes in Maine, New Hamp-
shire, Vermont, Massachusetts, and New York were
sampled. These studies were implemented to evalu-
ate all aspects of the National Surface Water Survey
research  plan including  lake selection, proposed
sampling protocols, the QA program, and data man-
agement. Objectives of the pilot studies are outlined
in greater detail in Drouse et al., 1986.  The process of
lake selection is discussed in Linthurst  et al. (1986).


Sampling and Analytical  Methodologies

Modifications to the original sampling and analytical
protocols were made based on the experience gained
during  the pilot studies. These modifications inclu-
ded elimination of the dissolved oxygen  measure-
ments which were deemed to be time consuming and
unnecessary, and revision of the sample-bottle wash-
ing procedure which was discovered  to introduce
nitrate contamination  (see Section 4). In addition,
four experiments were conducted prior to the ELS-I to
investigate (1) the possibility of contamination by hel-
icopter exhaust, (2) representativeness  of  single-
point sampling, (3) variability of measurements made
using the Hydrolab units, and (4) effects of increased
pressure (depth) on Hydrolab accuracy (Morris et al.
1986).

The primary goals of base site operations were to
obtain accurate physicochemical and geographical
data at each lake site, to collect representative lake
samples without introducing contamination,  to pre-
serve the integrity of samples until their analysis at
contract laboratories, and to perform selected chemi-
cal analyses. The objectives of the field  laboratory
are defined as follows:

  • receive lake and QA samples and field data from
    each sampling team and assess sample condi-
    tion upon receipt

  • review lake data forms for accuracy and com-
    pleteness

  • incorporate audit samples with lake samples to
    form a batch

  • analyze the batch samples for pH, DIG, true color,
    and turbidity

  • perform aluminum extraction

  • filter,  preserve,  and ship  samples to contract
    analytical laboratories for detailed analysis

  • coordinate  sample shipment  information with
    the Sample Management Office and EMSL-LV

  • distribute copies of NSWS Forms 1 and 2 to the
    appropriate offices

The protocols for collection of field data and water
samples during the ELS-I were implemented in three
phases: preflight  preparation,  lake site  activities
(including sampling  methodologies), and postflight
operations (Morris et al. 1986).

Standardized forms were developed to record meas-
urements made at each lake and at the field and con-
tract analytical laboratories. The multicopy field and
field laboratory forms were checked  for complete-
ness and  internal consistency at  the field station.
One  copy  of each  form was  sent to  Oak  Ridge
National  Laboratory for entry into the NSWS data
base and a second copy was sent to quality assur-
ance personnel in  Las Vegas; both copies were sent
by overnight  mail  service. Transfer of samples and
data is discussed in Section 3.

Analytical methodologies were selected for the ELS-I
on the basis of guidelines from the DQO's. For some
analytes, two or more techniques were considered to
be equivalent based on published literature (Hillman
et al. 1986). The pilot studies and field and laboratory
experiments provided opportunities to evaluate the
relative merits of  each technique. Two techniques
were judged to be equivalent for determining the con-

-------
centrations of total phosphorus  and of dissolved
metals (Ca, Fe, K, Mg, Mn, Na), and either technique
was  permitted. However,  measurements of  free
(uncomplexed) dissolved  fluoride by ion-selective
electrode and of total dissolved fluoride by ion chro-
matography were found to lack reproducibility, and
these methods were eliminated. The field measure-
ment and analytical methods used for the ELS-I are
summarized in Table 1.

Data Comparability Studies

Standardized techniques were specified for sampling
and chemical analysis of water samples from the
ELS-I. Standardization ensured that we could identify
the effect of variance (if any) of sampling and analysis
on differences identified between lakes.

A study was also undertaken to determine whether
ELS-I  data could be compared to survey data from
other countries. Subsamples were obtained from 215
ELS-I samples collected in the southern Blue Ridge
Mountains (NSWS subregion 3A, see Figure  1) and
shipped via commercial courier to Norway for chemi-
cal analysis of 14 parameters. Similarly, 105 subsam-
ples from the Adirondack Mountains (NSWS  subre-
gion 1A) were analyzed in Canada for 18 chemical
parameters. Results of this study are summarized in
Stapanian et al. (1986).

A second study utilized 2047 split samples from the
ELS-I to compare chemical analyses by flame atomic
absorption spectroscopy (AAS) and inductively coup-
led plasma emission spectrscopy (ICPES). The ICPES
analyses  were performed at ERL-Corvallis and by
ELS-I laboratories, and flame AAS analysis by ELS-I
laboratories only. These analyses will be presented in
Drouse and Best (1986).

-------
                                           Section 2
              Operational Quality Assurance - Quality Control Program
The QA plan (Drouse et al. 1986) describes the quality
assurance-quality control (QA-QC)  program of the
ELS-I, and the analytical methods manual (Hillman et
al. 1986)  documents all methodologies used in the
survey at the standard-operating-procedure  (SOP)
level of detail. The major aspects of the operational
QA-QC program are summarized in the following sec-
tions. Quality assurance aspects related to the data
base are  discussed in Section 3.  More detail is pro-
vided in the documents just referenced.


Field Station Organization and
Responsibility

The operation of a field station required the establish-
ment of a field laboratory, a calibration room, and a
local  communications center to coordinate  field
activities. Fifteen people were based at each field sta-
tion, including two helicopter pilots  and a mechanic,
an EPA base coordinator, an EPA duty officer, five lab-
oratory personnel, and five field samplers. Field labo-
ratory personnel  included a laboratory coordinator, a
laboratory supervisor, and three analysts. All labora-
tory personnel were cross-trained in sample prepara-
tion and  analysis. At some sites, individuals rotated
analyst positions on  a weekly basis. All personnel
reported  to the  EPA  base  coordinator  who was
responsible for the overall operation of the  base site.
The duties of the field station personnel are summa-
rized below and are discussed in greater detail in Mor-
ris et al. (1986).


Selection of Contract Analytical
Laboratories

The analytical requirements and QA approach were
defined during the development of the QA program.
The estimated number of samples to be analyzed and
the estimated rate of sample collection were defined
in the logistics planning.  It  was recognized early  in
the  planning that no  single analytical  laboratory
could analyze the number of expected samples at the
rate they were to be collected and still meet the QA-
QC  requirements (especially the required  holding
times) that had been established. There was not a sin-
gle EPA laboratory that had all of the analytical  capa-
bilities or resources to provide the required  analytical
support.  This meant  that  the analytical support
would have to be obtained via contracts with commer-
cial analytical  laboratories. A Contract Laboratory
Program (CLP)  had already been established to sup-
port the hazardous waste monitoring activities of the
Environmental  Protection  Agency across the United
States. Use of the CLP to  obtain contract analytical
laboratories for the NSWS was reasonably inexpen-
sive. Use of multiple contract analytical laboratories
also  meant  that  additional care  would have to  be
placed in the selection and documentation of analyti-
cal methods and QA activities to assure 9 consistent
and adequate  performance in all laboratories. The
contracting process required the following activities:

  • preparation of a statement of work (SOW) that
    defined the analytical and QA-QC requirements
    in a contractual format

  • preparation and  advertisement of an invitation
    for bids (IFB)  to  solicit contractor support

  • evaluation of the lowest bidders to assure that
    qualified laboratories were selected

Statement of Work

Control of data quality within each contract analyti-
cal laboratory was necessary in order to control data
quality among  the different laboratories and to mini-
mize the data variability. The methods manual and QA
plan had been drafted in the early phases of the plan-
ning process. However, in order to obtain analytical
support contracts,  the  methods and  the QA-QC
requirements had to be prepared in a contractual for-
mat or statement of work (SOW). This involved careful
review of the analytical and logistics requirements to
assure that they  were clearly stated and could be
enforced under the  terms of the contracts. Every
effort was made to assure that the reporting and QC
requirements were clearly stated in  the SOW.  The
major contractual requirements in the SOW were the
following:

   •  A contractor  could bid on the analysis of one or
     more bid lots (500 samples per bid lot) that would
     be delivered  to  the  analytical laboratory at a
     maximum rate of 30 samples per day per bid lot.

   •  Maximum holding times were specified. Failure
     to meet this requirement resulted in a penalty to
     the contractor of 20 percent per day per analyti-

-------
    cal subunit (7-, 14-, or 28-day holding time group)
    to a maximum of 100 percent of the bid price.

  • Delivery of the completed data package was
    required within 35 days of sample receipt by the
    contractor. An incentive fee for early delivery of
    data and a penalty for late delivery of data (1 per-
    cent per day up to a maximum of 10 percent of
    the bid price in each case) were established.

  • Failureof thecontractorto provide adequate QA-
    QC data and deliver ables as required by the
    SOW resulted in a penalty of up to 15 percent of
    the bid price.

The contractor laboratories were required to follow
the methods exactly as specified in the SOW. The QA
manager was allowed to  make interpretations for the
contract laboratory, but contractual changes  were
only made with the approval of the EPA contract offi-
cer.

Invitation for Bid

The SOW and IFB reflected the experience gained
from using contract analytical laboratories during the
pilot studies. The IFB was prepared and advertised in
Commerce Business Daily. Approximately 200 labora-
tories responded to the advertisement and were sent
copies of the SOW. Twenty-five laboratories  submit-
ted bids. The twelve lowest bidders were considered
to have submitted reasonable bids and were selected
to receive the preaward performance evaluation (PE)
samples for the second step of the selection process.

Analytical Laboratory Evaluation

The lowest bidders were  required to analyze PE sam-
ples and report the results within 15 days after sam-
ple receipt. The PE samples were prepared to repre-
sent  lake  samples  at   both  the   low  and  high
concentrations expected for the survey. Each bidder's
data and data package  was evaluated and  scored.
The scoring was based on evaluation of the quality of
the analytical data as well as the quality and com-
pleteness of the data package itself. This process
eliminated those laboratories that could not perform
the analytical and data reporting requirements. A
more  detailed description of  analytical  laboratory
performance evaluations is provided in Drouse et al.
(1986).

Each of the laboratories that passed the  PE  sample
evaluation was visited by an EPA team in order to ver-
ify the qualifications and capabilities of the laborato-
ries to meet the contractual requirements. The EPA
team determined whether the analytical  laboratory
had adequate equipment, personnel, and facMities to
analyze samples in accordance with the SOW. These
visits also provided an opportunity to clarify contract
requirements with the analytical staff and to identify
deficiencies that were observed  in the  PE  sample
data evaluation. The results of these onsite evalua-
tions are on file with the QA manager. The lowest bid-
ders who passed both the PE sample and onsite eval-
uations were awarded contracts to provide analytical
services for the ELS-I.

Training

Data quality depended on the ability of the field and
laboratory personnel to properly collect, process, and
analyze the samples. Operation of the ELS-I required
a large staff composed of Lockheed-EMSCO employ-
ees, EPA regional and EMSL personnel, and the con-
tract analytical laboratories. Training was essential
to ensure consistent application of all operational
and  quality   assurance  procedures.   Lockheed-
EMSCO field sampling personnel received six days of
intensive technical and safety  training during a 15-
day orientation program in September 1984 at the U.S.
EPA EMSL-LV. Personnel  from  the regional  EPA
offices who would  be involved  in sampling were
trained at the field stations prior to commencement
of sampling. Details of these training sessions are
contained in Morris et al. (1986).

A meeting was held in early August 1984 to review and
discuss the objectives and requirements for the ELS-
I.  Participants included representatives  from  the
NSWS management  team, EMSL-LV QA  staff, data
base management team, contract analytical labora-
tories, and the analytical  support laboratory.  The
objective of the meeting was to ensure that all parties
would implement both 11 the analytical methods and
the QA and QC procedures accurately  and consist-
ently. This meeting also provided an opportunity for
the participants to clarify the survey requirements.

Communications

Coordination of the  ELS-I operations required close
communication to ensure all  program  objectives
were met. During the actual sampling phase, the most
critical lines of communication were between  Las
Vegas and the  field stations and between Las Vegas
and the contract analytical laboratories.  Daily com-
munication was required concerning both logistical
and quality assurance topics.

Field Communications

Field  sampling activities  were closely  monitored
each day to ensure safety and logistical coordination.
Regular communication among  the field stations and
Las Vegas was also necessary. A local  communica-
tions center staffed by the EPA base coordinator and
the duty officer was established at each field station,
and a primary communications center was  estab-
lished in Las Vegas for these purposes.

The logistics communications  center in  Las  Vegas
was a clearinghouse for information about the num-

-------
her and type of lakes sampled, sample shipment
schedules,  helicopter flight  hours, and  long-range
weather forecasts.  Logistics personnel  monitored
the survey by coordinating and tracking shipment of
QA and analytical samples to contract laboratories,
and by coordinating the shipment of supplies to field
stations.

Computer software  utilized in tracking the progress
of lake sampling activities  was developed before
sampling began. Maps for the daily tracking of field
activities were  inventoried and were  displayed by
region.  Bulletin  boards  and  chalkboards  were
installed  to effectively monitor field activities. All
communications were logged on a field communica-
tion form. Sampling progress was graphically dis-
played on regional  maps with color-coded flags to
indicate lakes sampled  and  lakes remaining  to be
sampled  Progress reports were made by  phone, and
a written report  was made twice weekly to the NSWS
management team.

The establishment of communications centers and
the  implementation  of  communications   plans
enabled field operations to proceed in a coordinated
and consistent  manner although  field stations were
located over a wide  geographic area.


Daily Communications with Contract Analytical
Laboratories

Daily calls were made to each contract analytical lab-
oratory by  the  QA  staff. The primary objective of
these calls was  to ensure that QA and QC  procedures
were  being  implemented  according  to  the survey
requirements and that the samples were being han-
dled and analyzed properly. Othertechmcal and  logis-
tical issues were addressed as they arose. Examples
of issues that were quickly identified or resolved as a
result of these calls include:

   • aluminum contamination in aliquot 7 during the
    digestion step of the analysis for total aluminum

   • incorrect calculations for reporting  nitrate and
    silica data

   • sample overload at one laboratory

   • nitrate contamination in aliquot 3

   • illegible data reporting

The daily QA contact with each laboratory continued
until sample analysis was completed  Preliminary
sample data were obtained either verbally, via com-
puter, or via TELEFAX, the method of data collection
depended on the resources available to the contract
analytical laboratory. The preliminary data were eval-
uated by comparing QA sample values with prelimi-
nary acceptance criteria, calculated from pilot  study
data or early QA sample data.
Sampling and Field Laboratory Quality
Control Protocols

The ELSI-I also included  (QC) procedures  for sam-
pling and analytical activities.  Specific procedures
are outlined in the QA plan, the methods manual, and
the field operations report. The flow of samples and
data through the field and analytical laboratories is
illustrated in Figure 2.

Sample Preservation

For each chemical parameter measured during the
ELS-I, it was  necessary to identify the procedures
required for sample preservation. The objectives  of
sample preservation were to (1) inhibit chemical and
biological activity, (2) prevent changes due to volatil-
ity, and (3) prevent precipitation or adsorption effects.
These considerations led  to a sample preparation
process in the field laboratory in which seven pre-
served aliquots were prepared from each bulk (rou-
tine, field duplicate, field blank, or fie Id audit) sample.
The preservation  process  used  for each aliquot  is
listed in Table 2.

Filtration through a 0.45Mmmembrane filter was used
to remove suspended particulate and  large colloidal
material. This process provided subsamples that rep-
resented the dissolved fraction of  analytes. Sus-
pended material  was filtered from these aliquots  at
the field laboratory because such material may have
been a source of  biological activity or,  through disso-
lution, of additional analyte. It may also have pro-
vided surface area for adsorption or precipitation  of
dissolved analytes which would serve as a transport
mechanism for removing these analytes from solu-
tion. Acid was added to some aliquots to prevent loss
of dissolved analytes caused by precipitation or  by
chemical or biological reactions. Storage at 4°C was
specified for aliquot 2 to reduce volatilization of the
solvent, and for aliquots 3 through 6 to reduce biologi-
cal activity.

Field Laboratory Analyses

After the sample preparation and preservation steps
were identified, it was necessary to establish maxi-
mum holding times to assure that the samples were
analyzed  before   any   significant   degradation
occurred.

Four parameters (pH, DIG,  true color,  and  turbidity)
were identified as requiring immediate analysis. Use
of the mobile field laboratory for these analyses per-
mitted all measurements to be completed  within  16
hours of sample  collection.

To assure that reliable measurements of DIG and pH
were obtained, samples for those two field laboratory
measurements were collected and were analyzed in a
closed system within the shortest possible holding
times. This procedure was followed  to avoid prob-
                                                 10

-------
                         FIELD BLANK
                          SAMPLES
                        FIELD DUPLICATE
                            SAMPLE
                             FIELD/LABORATORY
                              AUDIT SAMPLE(S)
                                                  FIELD LABORATORY
                               DAILY BATCH
                               OF SAMPLES
                                                                                1
                                              RELABELING
                                                   '
                      ANALYSIS
              (DIG, pH, Turbidity, True Color)
      QCCHECK
       SAMPLES
 BATCH
SAMPLES
 TRAILER
DUPLICATE
 SAMPLE
                       DATA

                                 ALIQUOT PREPARATION

                                 ALUMINUM EXTRACTION


                                    PRESERVATION
                                                            SHIPMENT TO
                                                       ANALYTICAL LABORATORY
                                                                 ANALYTICAL LABORATORY
                                                            ANALYSIS
                                                    INTERNAL
                                                  QC SAMPLES
                                             BATCH
                                            SAMPLES
                                        LABORATORY
                                         DUPLICATE
                                               LABORATORY BLANK,
                                                  MATRIC SPIKE,
                                                QCCHECKSAMP1E
                                                       I
                                                                   DATA
                                            1	|
Figure 2. Flow of Samples and Data Through Field and Analytical Laboratories, Eastern Lake Survey
- Phase I.
                                             ll

-------
Aliquota"
Container Size

Preservation Required
Parameters to be Measured





a Aliquots 2, 3, 4, 5, and 6 were
1 2 3
250mL 10 ml 250 ml
Filtered Filtered
ph<2 MIBK-HQ Filtered
w/HN03 Extract
Ca Total Extractable C1~
A1
Mg F"
K S042~
Na N03~
Mn Si02
Fe
stored in the dark at 4°C.
4567
125 ml 500ml 125 ml 125mL
Filtered Not Filtered Not Filtered
pH<2 Not Filtered pH<2 pH<2
w/H2S04 w/H2S04 w/HN03
DOC pH Total P Total A1
NH4+ ANC
BNC
Conductance
DIG


Table 2.  Sample Preservation Requirements.

-------
lems  with changes in dissolved  inorganic carbon
(DIG) content and thereby to represent accurately the
in situ values.

True color and turbidity were determined in the field
laboratory because the recommended  holding time
was 48 hours (U.S. EPA 1983). Turbidity had to be mea-
sured as soon as possible to avoid settling of  sus-
pended matter. Both true color and turbidity are sub-
ject to  changes  due  to  chemical  activity  and
adsorption effects.

There is  evidence that dissolved aluminum species
change rapidly after sample collection (Driscoll et al.
1983); therefore, it was recommended that the alumi-
num extraction be performed in the field as soon as
possible. It was not possible to perform the extrac-
tion in the  helicopter because  of concerns about
exposure of the  crew  to  MIBK vapors  and  the
increased possibility of sample contamination. All
samples  were extracted in a laminar-flow hood in the
field laboratory within 16 hours of collection.


Contract Analytical Laboratory Quality
Control Protocols

Analytical methods and quality control protocols
were developed for use by the contract laboratories in
accordance with the ELS-I research plan and the pilot
study results. These  methods  and protocols are
described below and are outlined in greater detail in
the analytical methods manual and the QA plan.

In general, specific models of instruments were not
required  for the contract laboratories  although the
analytical methods manual  contains particular rec-
ommendations for instruments in some methodolo-
gies. Instrumentation for all  of the methods required
some form of calibration. For all methods except con-
ductance, a series of standards were analyzed and a
calibration curve was calculated.

Several   requirements  for  sample concentration
ranges were specified for the analytical methods. The
concentration range of the standards used to calcu-
late the calibration curve was required to bracket the
range of concentrations observed in the  samples ana-
lyzed.  If the concentration of analyte was at or below
the detection limit, the concentration of  the detection
limit QC standard had to be within two to three times
the detection limit. This QC requirement assured that
the reported results were based upon interpolation 16
within the calibration curve and not on extrapolation
outside the curve which could result in  significant
error. The calibration required for conductance meas-
urements was dependent upon the type  of conductiv-
ity meter  used and was either a single point calibra-
tion or was internally set by the factory.

Calibration of laboratory equipment was initially veri-
fied prior  to sample analysis by analyzing an indepen-
dent QC sample (either commercially or internally
prepared). If the measured value for the QC sample
was not within established control limits, new cali-
bration standards were prepared, and, where applica-
ble, a new calibration curve was generated. Calibra-
tion was  also reverified  on  a  routine  basis  by
reanalyzing the QC sample after every 10 samples
analyzed and after the last sample for the batch. If the
measured value was not within established control
limits, a new calibration was required, and all sam-
ples after the previous acceptable QC check had to be
reanalyzed.

Contract Analytical Laboratory Sample Holding
Times

A maximum sample holding time (determined from
the time of  sample collection to  sample analysis,
Table 1) was established for each  parameter mea-
sured in the contract  laboratories. These  holding
times were based upon information from  the litera-
ture,  the  best  scientific judgment related to  the
defined needs, and the logistical demands and limita-
tions of the ELS-I.

A 7-day holding time was specified for the contract
analytical  laboratory measurement of  pH,  nitrate,
and  total  extractable  aluminum.  McQuaker et  al.
(1979) reported that the pH of a sample remains sta-
ble for up to 15 days if the sample is kept at 4°C and
sealed from the atmosphere; however, in the EPA
manual Methods for Chemical Analysis of Water and
Wastes (U.S. EPA 1983), it is recommended that pH be
measured  immediately after  sampling.  The same
manual  also  specifies  a 48-hour  holding time for
nitrate in  unpreserved (not acidified)  samples  of
water and wastes. Other  sources  (Williams, 1979;
Quave, 1980) indicate that nitrate is stable for 2 to 4
weeks if the sample is stored in the dark at 4°C. The 7-
day holding time for pH and nitrate was selected as a
limit that was practical with respect to the logistical
constraints of the survey and that was conservative in
relation to the available guidelines.

Barnes  (1975)  reported that  the  MIBK-aluminum
extract is stable for several weeks following extrac-
tion. However, for the ELS-I, a 7-day holding time was
specified for total extractable aluminum in order to
minimize potential  changes in  sample composition
due to volatilization of the MIBK.

A 14-day holding time  was selected for ANC, BNC,
conductance, DIG, and DOC; selection of this holding
time was based on U.S. EPA (1983) and on practical
considerations with regard to the logistical design of
the survey. In most cases, the DIG analyses were per-
formed almost simultaneously with the correspond-
ing  pH measurements because of specific require-
ments in the analytical laboratory contract.

A 28-day maximum holding time was specified for the
remaining  contract  analytical laboratory  measure-
                                                13

-------
ments. In many cases, especially for the metals, hold-
ing times up to 6 months are acceptable with proper
sample handling and preservation (U.S.  EPA 1983).
However, the analytical laboratory contract  17
required that the final data package be submitted
within 35 days after receipt of the samples. A 28-day
holding time was specified as a conservative limit
both to meet that requirement and to protect sample
quality.

Reporting Requirements

As noted above, the contract analytical laboratories
were required to analyze the samples and report  the
results within  35 days after sample receipt. The
reporting requirements  included the submission of
both analytical results and specific  information
related to the maximum sample holding times dis-
cussed above This information was used by the  QA
auditors to judge the  performance of the laboratory
and the quality of the data. The requirement for a spe-
cific delivery time assured that the data packages
were delivered in time to allow for preliminary review
by the QA auditors and for corrective action, if neces-
sary.
Internal Quality Control

The QA program utilized a variety of QA and QC sam-
ples. The numbers of QA and QC samples used in the
survey were based on the need both to keep program
costs within  reason and  to provide a  maximum
amount of information. The QC samples were used by
field samplers, field  laboratory personnel (Morris et
al. 1986), and  contract analytical laboratory person-
nel (Drouse et ai, 1986); QA samples were used by the
QA staff to evaluate data quality and to judge overall
field  and  laboratory  performance. Descriptions,
applications, and frequencies of QA and QC samples
are provided in Tables 3 and 4.
Field Laboratory and Contract Analytical
Laboratory On-site Evaluations

Onsite evaluations of the contract analytical labora-
tories and the field stations were performed by a QA
audit team to assure that the sampling and analysis
activities were being implemented as planned. These
evaluations are described in Drouse et al. (1986).
                                                14

-------
Sample Type
Field Blank


Laboratory Blank a
Field Duplicate

Trailer Duplicate a

Contract Laboratory
Duplicate3
Field Audit


Laboratory Audit


a Serves as both a QA
Description
Deionized water (ASTM
Type 1) treated as a
lake sample
Zero analyte standard
Duplicate lake sample

Lake sample, split

Sample aliquot, split

Synthetic sample or
natural lake sample

Synthetic sample


sample and a QC sample.
Application
Estimate system decision limit
and quantitation limit

Identify sample contamination
Estimate overall withm-batch
precision
Estimate analytical
withm-batch precision
Estimate analytical
within-batch precision
Estimate overall among-batch
precision, estimate laboratory
bias
Estimate analytical
among-batch precision,
estimate laboratory bias

Frequency
One per sampling crew per day


One per laboratory batch
One per field station per day

One per field batch

One per laboratory batch

A minimum of one field or
laboratory audit per field batch

A minimum of one field or
laboratory audit per field batch


Table 3. Descriptions, Applications, and Frequencies of Quality Assurance (QA) and Quality Control (QC)
Samples, Eastern Lake Survey - Phase I.
                                                 15

-------
       Sample Type
      Description
         Application
          Frequency
   Trailer Duplicate
   Contract Laboratory
   Duplicate
   Laboratory Calibration
   Blank
   Matrix Spike
   Quality control Check
   Sample
Lake sample; split
Sample aliquot, split
Zero analyte standard
Batch sample plus
known quantity of
analyte

Standard, source other
than calibration
standard
Field lab, determine analytical
within-batch precision

Contract lab, determine
analytical within-batch
precision

Field and contract lab; identify
signal drift and sample
contamination

Contract lab, determine
sample matrix  effect on
analysis

Field and contract labs,
determine accuracy and
consistency of calibration
One per field batch


One per laboratory batch



One per laboratory batch



One per laboratory batch
Before, after every 10, and after
final sample in batch
Table 4.  Descriptions, Applications, and Frequencies of Quality Control Samples, Eastern Lake Survey - Phase I.
                                                            16

-------
                                           Section 3
                                Data Base Quality Assurance
 The data  base for the ELS-I was  managed  by Oak
 Ridge National Laboratory (ORNL) which has consid-
 erate expertise in managing  large data bases, in
 manipulating data, and in restructuring data bases to
 satisfy data analysis needs. The data are stored in
 four major data sets: (1) a raw data set of field and
 analytical laboratory data, (2) a verified data set, (3) a
 validated data set, and (4) an enhanced data set.

 Development of the data base began when samples
 and data were transferred from the field stations to
 the analytical laboratories and continued with receipt
 of analytical results and verification reports. The
 process was complete after final  validation of the
 data by ERL-Corvallis. The  enhanced data set was
 generated from the validated data by using known
 relationships between physicochemical  parameters
 (Linthurst et al. 1986). The enhanced data set provided
 a representative summary of sample values for use in
 generating population estimates.


 Data and Sample Transfer

 Data from in situ, field laboratory, and contract ana-
 lytical laboratory measurements were transferred to
 ORNL and EMSL-LV initially  in hardcopy form. Docu-
 mentation related to shipment of samples from the
 field stations to the contract analytical laboratories
 was also transferred in this  manner. After the appro-
 priate information was entered into the data  base, a
 magnetic tape containing raw data was sent to the
 EPA IBM 3081 computer at  the National Computer
 Center (NCC), Research Triangle Park, North Caro-
 lina. Each tape received by the NCC tape library was
 assigned a volume serial number and  a BIN number
 that indicated the physical location of the tape. The
 EMSL-LV QA staff then remotely loaded the tape to
 disk files and reviewed the data.

 Transfer of Samples and Data From Field Stations

 Following sample processing at the field laboratory,
the aliquots were shipped in Styrofoam-lined ship-
ping cartons with frozen freeze-gel packs to maintain
a temperature of 4°C. A 4-part  carbonless shipping
form (Form 3, Figure 3) was completed for each ship-
ping container. One copy was retained at the field sta-
tion, one copy was sent to the U.S. EPA Sample Man-
agement Office (SMO), and  two copies (sealed in a
 plastic bag) were enclosed in the carton. The ship-
 ping container was then sealed and shipped by over-
 night delivery to the contract analytical  laboratory.
 Preserved splits of all samples were sent to ERL-Cor-
 vallis for chemical analysis by ICPES (Drouse and
 Best  1986). Additionally,  splits  of  samples  from
 selected regions were sent to two Canadian laborato-
 ries and to a Norwegian laboratory forchemical anal-
 ysis (Stapanian et al. 1986).

 When the samples arrived at the destination contract
 laboratory,  a  receiving  clerk recorded  the  date
 received on each shipping form, verified that the sam-
 ple contents matched those listed on the shipping
 form, and completed the "sample condition" portion
 of the shipping form. The sample  condition notation
 included such information  as  leakage, insufficient
 sample, noticeable suspended particulates, partially
 frozen samples, and internal temperable  conditions
 ordiscrepancies in the listed contents. Upon comple-
 tion of sample inspection, the receiving clerk mailed
 one copy of each shipping form to SMO and retained
 the other copy.

 In addition to shipping samples, the field laboratory
 personnel  transferred data forms to various  loca-
 tions. Copies of the lake data form (Form 1, Figure 4)
 and the  field laboratory data form (Form 2, Figure 5)
 were mailed to the QA manager and to ORNL. Upon
 receipt of Forms 1  and 2, ORNL personnel entered the
 data from these forms into the raw data set. This data
 flow scheme is summarized in Figure 6. Verification
 of the field data is discussed below.

 Transfer of Contract Analytical Laboratory Data

 After all analyses for a single batch of samples were
 completed, the contract  laboratory  personnel  pre-
 pared an analytical report called a sample data pack-
 age. All analytical results were recorded on the forms
 listed in  Table 5.  The  laboratory  manager  was
 required to sign each form signifying that he or she
 had reviewed the data and that the samples were ana-
 lyzed exactly as described in the contract. Any devia-
tions from the contract  required the authorization of
the ELS-I QA manager prior to sample analysis. The
data package also contained a narrative description
of any difficulties encountered. All original raw data
were  retained by the  laboratory manager  until
                                                17

-------
NATIONAL SURFACE WATER SURVEY
SAMPLE MANAGEMENT OFFICE
P.O. BOX 8 I 8
ALEXANDRIA, VA   22314
                  NSWS
                 FORM 3
                SHIPPING
RECEIVED BY 	
IF INCOMPLETE IMMEDIATELY NOTIFY:
   SAMPLE MANAGEMENT OFFICE
       (703)557-2490
FROM
(STATION ID):
SAMPLE
ID
0!
02
03
04
05
06
07
08
09
10
1 1
12
1 3
14
15
16
1 7
18
19
20
2 1
22
23
24
25
26
27
28
29
30
TO
(LAB):
3ATCH
ID
DATE SAMPLED

ALIQUOTS SHIPPED
(FOR STATION USE ONLY)
1






























2






























3



























1


4






























5






























6






























7






























DATE SHIPPED DATE RECEIVED
AIR — BILL N1"1,


SAMPLE CONDITION UPON LAB RECEIPT
(FOR LAB USE ONLY)






























QUALIFIERS:
         V: ALIQUOT SHIPPED
         M: ALIQUOT MISSING DUE TO DESTROYED SAMPLE
 WHITE - FIELD COPY
 PINK - LAB COPY
YELLOW - SMO COPY
GOLD - LAB COPY FOR RETURN TO SMO
                          Figure3. NSWS Form 3 - Shipping.
                                             18

-------
    NATIONAL SURFACE WATER SURVEY
                     FORM1
                   LAKE DATA
                                                                     D   D   M M  M   Y  Y
STATE
                LAKE ID
                                    LAKE NAME
                        LORAN READINGS   }
                                                 DATE

                                         SAMPLING TIME

                                          HYDROLAB ID
                                                INITIAL
                                                 FINAL

                                                INITIAL i
                                                                                i  pH
LATITUDE i 	 n 	 ii 	 i°i 	 ii 	 i.i 	 ii 	 .LONGITUDE i 	 u 	 n 	 i°i 	 11 	 i.i 	 n 	 i FINAL. — . i — n — .. — . IJS v--/
PHOTOGRAPHS^)
FRAME ID AZIMUTH
^^ LAP CARD
DISTURBANCES WITHIN 100 METERS OF SHORE
EH ROADS EH LIVESTOCK EH MINES/QUARRIES
EH DWELLINGS EH INDUSTRY EH LOGGING LH OTHER
SITE DEPTH (ft) x 0 3048 m/ft - i 	 n 	 i.i 	 i m
AIR TEMP i 	 ii 	 ii 	 i°C
SITE DEPTH
                           SECCHI DEPTH  DISAPPEAR i
                             LAKE STRATIFICATION  DATA
                 1 5mV_y
                                  T°C
BOTTOM -1 5m
 AT°C (1 5, B-1 5m) i	n—I.L


06 DEPTH            T°C
                                                       IF A > 4° C PROCEED
                                                         IF NOT, STOP HERE   |
                                                                             pH
                                                                             PH
AT°C (1 5, 06 DEPTH) i — n — i.i — i (^J
[ IF AT >4°C FILL IN j
FOLLOWING DATA BLOCK

LAKE DIAGRAM # I 1
Flpvatinn ft Ontlpts LAKE DEPTH
CHECK ONE
TN InlPK D-20m D 20m T°C pS
* 4
6
8
10
12
14
16
18
20

c ,_,,_,, 	 ,Q ,__, ,__,,_, 1_J
10 ^i^.^Q ,_,,_,,_,,_,
15 i_n_i.i 	 iQ i 	 i I—" L— > i— i
20 i — i i — i.i — i \^) i — i i — i i — i ' — i
25 i 	 , i 	 tll 	 i O i 	 i , 	 i , 	 , , 	 i
30 i 	 i i 	 i . i 	 i O i 	 i i 	 i i 	 i i 	 i
35 ,_,,_,. i 	 iQ i— i •—! i— ' i 	 i
40 ^_,^.^_,O ^^^^
45 L-JL^.I—.O L-IU-J ,_,,_,
50 __._0 ^^^^
0
O
o
o
o
o
o
o
o
o
COMMENTS  D NOT SAMPLED, SEE BELOW
                                                               DATA QUALIFIERS
                                                              ® INSTRUMENT UNSTABLE

                                                              @ REDONE FIRST READING NOT
                                                                 ACCEPTABLE

                                                              © INSTRUMENTS SAMPLING GEAR
                                                                 NOT VERTICAL IN WATER COLUMN

                                                              @ SLOW STABILIZATION

                                                              (D HYDROLAB CABLE TOO SHORT

                                                              (N) DID NOT MEET OCC

                                                              ® OTHER (explain)	
               FLOWING WATER  D INACCESSIBLE  D NO ACCESS PERMIT  DURBAN/INDUSTRIAL
            DHIGHCOND  (>1500/jS)   GNON-LAKE  DTOOSHALLOW   DOTHER-
FIELD LAB USE ONLY
TRAII FR in
RATP.H in
SAMPI F in


FIELD CREW DATA
P.HFW in
ORSERVER
SAMPLER
nRR SIGN
GROUND CREW MEMBER
SIGN

WHITE COPY— ORNL
PINK COPY— EMSL-LV


                        Figure 4. NSWS Form 1 - Lake Data.
                                           19

-------
NATIONAL SURFACE WATER SURVEY
                 FORM 2
       BATCH/QC FIELD DATA
DATE RECEIVED
BY DATA MGT	
ENTERED	

RE-ENTERED	
BATCH ID
NO SAMPLES
IN BATCH
BASE SITE ID

SAMPLE ID
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
DUP
FIELD
CREW
ID
































LAKE
ID
(XXX-XXX)































LAB TO WHICH
BATCH SENT
DATE SHIPPED
LAB CREW ID

SAMPLE
CODE






























TD

DIG (mg/L)
QCCS LIMITS
UCL - 2.2
LCL — 1.8
VALUE































QCCS































STATION pH
QCCS LIMITS
UCL - 4.1
LCL - 3.9
VALUE































QCCS































DATE PRO(
AIR-BILL N
FIELD LAB
SUPERVIS
MESSED
0
ORATORY
->R

TURBIDITY (NTU)
OCCS LIMITS
UCL - 5.5
LCL — 4.5
VALUE































OCCS































COLOR
(APHA
UNITS)
VALUE































SPLIT
CODES
(E,L)
































COMMENTS
                WHITE — ORNL COPY   YELLOW — FIELD COPY  PINK — EMSL LV COPY
            Figure 5. NSWS Form 2 - Field Laboratory Data.
                                20

-------
 CONTRACT
ANALYTICAL
LABORATORY
                             2 copies
    Form 3
       FIELD STATION
         (One copy
       each of Forms
         1, 2, and 3
          retained)
           1 copy
           Form 3
   SAMPLE
MANAGEMENT
   OFFICE
1 copy
Form 3
                             ELS-I QA
                            MANAGER
                    1 copy each
                                             Forms 1 and 2
                              ORNL
                           DATA BASE
                                               1 copy each
                                               Forms 1 and 2
             Figure 6. Eastern Lake Survey - Phase I Data Flow Scheme.
                                21

-------
Data Form
11
13
14"
15b
16b
17
18
19
20
21
22
23
a These forms are shown in Drouse et al.
b Form not required to be submitted with
Description
Summary of Sample Results
ANCand BNC Results
QC Data for ANC and BNC Analyses
Conductance (Measured and Calculated)
Anion-Cation Balance Calculations
Ion Chromatography Resolution Test
Instrumental Detection Limits
Sample Holding Time Summary
Results of Blank Sample and QCCS Analyses
Results of Matrix Spike Analyses
Results of Duplicate Sample Analyses
Results of Standard Additions Analyses
(1986).
data package but recommended for internal QC requirements.
    Table 5. List of Data Forms Used by the Contract Analytical Laboratory, Eastern Lake Survey (Phase l)a.
otherwise notified. Raw Data included data system
printouts, chromatograms,  notebooks,  QC charts,
standard preparation data, and any other information
pertinent to sample analysis. The original data pack-
age was retained by the laboratory, and copies were
mailed to EMSL-LV, SMO, and  ORNL The analytical
results were double-entered into the raw data set by
ORNL

Raw Data Set

At ORNL, the field and laboratory data and data quali-
fiers ('tags',  Table 6) reported on Forms 1, 2, 11, 13,
and 18 through 23 were directly entered into the data
base using Statistical System software (SAS  Insti-
tute, Inc. 1985).

All data were entered independently by two different
operators. A computer  program (COMPARE) was
developed to identify any inconsistencies  between
the two data sets (Rosen and Kanciruk 1985). The
inconsistencies were then corrected using the SAS
full-screen editing  procedure. The purpose of this
double entry and comparison process was to mini-
mize data entry errors.

DATA VERIFICATION

Verification  procedures  for the  raw data set were
developed and implemented by the EMSL-LV QA staff.
The objective of data verification was to identify and
correct, qualify, or eliminate data of unacceptable
quality. Data qualifiers added during the verification
process ('flags') are listed in Table 7. This objective
was  accomplished  using the  following organized
process to examine the data: (1) establish daily com-
munication with the field and analytical laborato-
ries;^) verify completeness  and consistency of the
data package on receipt and  review any comments or
questions associated with the batch or sample under
evaluation (i.e., tags and narrative comments); (3)
evaluate preliminary QA sample data and  routine
sample data; (4) obtain confirmation, correction, or
reanalysis data from the laboratories as needed to
address atypical values; and (5) provide correcting
entries to ORNL for establishing the verified data set.
A  computer software package  (AQUARIUS)  was
developed to automate this procedure as much as
possible (Fountain and Hoff 1985). The AQUARIUS
package was tested during the ELS-I  and is being
modified to provide additional verification  proce-
dures for other phases of the NSWS.

Data verification procedures are summarized below
and are illustrated in Figure 7. Additional details are
provided in Drouse et al. (1986).
                                                22

-------
        Qualifier
                               Indicates
           D
           E
           F

           G
           H
           J

           K
           L
           M

           N
           P
           Q

           R
           S
           T
           U

           V
           W
           X

           Y
           Z
Instrument unstable
Redone, first reading not acceptable
Instruments, sampling gear not vertical in water column

Slow Stabilization
Hydrolab cable too short
Result outside QA criteria (with consent of QA manager)

Result obtained from method of standard additions
Holding time exceeded criteria (Form 19 only)
Result not available, insufficient sample volume shipped to laboratory from the field

Result not available; entire aliquot not shipped
Not analyzed due to interference
Result not available; sample lost or destroyed by lab

Not required
Result outside QA criteria, but insufficient volume for reanalysis
Result outside QA criteria

Result from reanalysis
Contamination suspected
Leaking container
Result not required by procedure; unnecessary

Anion-cation balance outside criteria due to high DOC
Percent Difference (%D) calculation (Form 14) outside criteria due to high DOC
Available for miscellaneous comments in the field only

Available for miscellaneous comments in the field only
Available for miscellaneous comments in the field only
Measurements taken at <1.5 m
              Table 6. Eastern Lake Survey (Phase I) Field and Laboratory Data Qualifiers ('Tags').
Daily Communication

Daily communication was  maintained between the
EMSL-LV QA staff and each contract analytical labo-
ratory during the periods when samples were being
analyzed. The objectives of daily  communication
were to assure that the laboratory was implementing
the QC requirements and to obtain a preliminary eval-
uation of data quality and  laboratory performance.
Performance was judged by reviewing the analytical
results for QA samples (e.g., field blanks, field dupli-
cates, and audit samples). Daily contact enabled the
QA auditors to be familiar with analytical problems
and some results so that data analysis was already
partially underway prior to  receipt of the data pack-
age. The bulk of the data evaluation occurred concur-
rently with input of the raw  data by ORNL.


Data Receipt

As  noted  above, the QA staff at EMSL-LV received
copies of the field data  forms (Forms 1 and 2) from
each field laboratory coordinator.  Upon receipt, the
auditor checked for the following items:
                         Lake ID. The lake data form (Form 1) was com-
                         pared with the field laboratory data form (Form 2)
                         for transcription errors.

                         Trailer Duplicate. The duplicate lake sample ID
                         recorded on Form 2 had to match a routine lake
                         sample ID, and the precision criteria for pH, DIG,
                         true color, and turbidity between those two sam-
                         ples had to be achieved.

                         Hydrolab Calibration Data. The Hydrolab pH and
                         conductance QCCS values from  Form 1 were
                         compared with the data from the Hydrolab cali-
                         bration forms to assure that  initial calibration
                         criteria were met  or that data qualifiers were
                         added.

                         Hydrolab pH. The Form 1 pH value at 1.5 meters
                         was compared with the field laboratory (Form 2)
                         pH values. Values were expected to agree within
                         0.5 pH unit.

                         Field laboratory pH and DIG. Form 2 values for
                         field audit samples were compared with accept-
                         ance criteria. Routine/field duplicate pairs and
                                                   23

-------
AO       Anion-Cation Percent Ion Balance Difference (%IBD) is outside criteria due to unknown cause
A1       Anion-Cation Percent Ion Balance Difference (%IBD) is outside criteria due to Nitrate contamination.
A2       Anion-Cation Percent Ion Balance Difference (%IBD) is outside criteria due to anion (other than nitrate) contamination.
A3       Anion-Cation Percent Ion Balance Difference (%IBD) is outside criteria due to cation contamination.
A4       Anion-Cation Percent Ion Balance Difference (%IBD) is outside criteria due to unmeasured organic protolytes (fits Oliver
         Model).
A5       Anion-Cation Percent Ion Balance Difference (%IBD) is outside criteria due to possible analytical error - anion
         concentration too high (list suspect anion}.
A6       Anion-Cation Percent Ion Balance Difference (%IBD) is outside criteria due to possible analytical error - cation
         concentration too low (list suspect cation).
A7       Anion-Cation Percent Ion Balance Difference (%IBD) is outside criteria due to possible analytical error - anion
         concentration too low (list suspect anion).
A8       Anion-Cation Percent Ion Balance Difference (%IBD) is outside criteria due to possible analytical error - cation
         concentration too high (list suspect cation).
BO       External (field) blank is  above expected criteria for pH, DIG, DOC, conductance, ANC, and BNC determinations
B1       Internal (lab) blank is >2 x  RDL for PH, DIG, DOC, conductance, AJ\IC, and BNC determinations.
B2       External (field) blank is  above expected criteria and contributed  >20% to sample value. (This flag is not used for pH,
         DIG, DOC, ANC, or BNC determination.)
83       Internal (lab) blank is >2 x  RDL and contributes >10% to the sample concentrations. (This flag is not used for pH,
         DIG, DOC, ANC, or BNC determinations.)
B4       Potential negative sample bias based on internal (lab) blank data.
B5       Potential negative sample bias on external (field) blank data.
CO       Percent Conductance Difference (%CD) is outside criteria due to unknown cause (possible analytical error - ion
         concentration too high).
C1        Percent Conductance Difference (%CD) is outside criteria due to possible analytical error - anion concentration  too high
         (list suspect anion).
C2       Percent Conductance Difference (%CD) is outside criteria due to anion contamination.
C3       Percent Conductance Difference (%CD) is outside criteria due to cation contamination.
C4       Percent Conductance Difference (%CD) is outside criteria due to unmeasured organic anion (fits Oliver Model).
C5       Percent Conductance Difference (%CD) is outside criteria due to possible analytical error in conductance measurement
C6       Percent Conductance Difference (%CD) is outside criteria due to possible analytical error - anion concentration too low
          (list suspect anion).
 C7       Percent Conductance Difference (%CD) is outside criteria due to unmeasured protolyle anions (does not fit Oliver
          Model).
 C8       Percent Conductance Difference (%CD) is outside criteria due to possible analytical error - cation concentration too low
          (list suspect cation)
 C9       Percent Conductance Difference (%CD) is outside criteria due to possible analytical error - cation concentration too high
          (list suspect cation).
 DO       External (field) duplicate precision exceeded the maximum expected percent relative standard deviation (%RSD), but
          either the routine or the duplicate concentrations were  >10 x  RDL.
 D2       External (field) duplicate precision exceeded the maximum expected percent relative standard deviation (%RSD), and
          both the  routine and duplicate sample concentrations were  >10  x RDL
                  Table 7.  Eastern Lake Survey - Phase I Verification Data Qualifiers ('Flags').
                                                           24

-------
03       Internal (lab) duplicate precision exceeded the maximum allowable percent relative standard deviation (%RSD), and
         both the routine and duplicate sample concentrations were >10  x RDL
FO       Percent Conductance Difference (%CD) exceeded criteria when Hydrolab conductance value was substituted.
F1       Hillman-Kramer protolyte analysis program indicated field pH problem when Hydrolab pH value was substituted.
F2       Hillman-Kramer protolyte analysis program indicated unexplained field pH or PIC problem when Hydrolab pH value was
         substituted.
HO       The maximum holding time criteria were not met.
H1       No "Date Analyzed" data were submitted for reanalysis data.
L1       Instrumental Detection  Limit (IDL) exceeded RDL and sample concentration was  <10 x  IDL.
MO      Value was obtained using a method which is unacceptable by the contract.
NO       Audit sample value exceeded upper control  limit.
N1       Audit sample value was below control limit.
N2       Audit sample value exceeded control limits due to suspect audit sample preparation.
N5       N03~ data obtained from analysis of aliquot 5.
PO       Field problem - station pH.
P1       Field problem - station DIG.
P2       Field problem - unexplained pH or DIG.
P3       Lab problem -  initial ANC pH.
P4       Lab problem-initial BNC pH.
P5       Lab problem -  unexplained - initial pH (ANC or BNC).
P6       Lab problem -  initial DIG.
P7       Lab problem -  air-equilibrated pH or DIG.
P8       Lab problem -  unexplained - initial pH or DIG.
P9       Lab problem - ANC determination.
Q1       Quality control check sample (QCCS) was above contractual criteria.
Q2       Quality control check sample (QCCS) was below contractual criteria.
Q3       Number of quality control check samples (QCCS) measured was  insufficient.
Q4       No quality control check sample (QCCS) analysis was performed).
SO       Matrix spike percent recovery (%REC) was above contractual criteria.
S1       Matrix spike percent recovery (%REC) was below contractual criteria
V        Data verified.
         Table 7.  Eastern Lake Survey - Phase I Verification Data Qualifiers ('Flags') (Concluded).
                                                        25

-------
                       Field Data
                     Analytical Data
 Preliminary
 Acceptance
 Criteria in
 QA Samples
        LEMSCO
      Preliminary
        Review
        Fix Data
        Initiate
      Corrections
                                     Yes
Generation of
Raw Data Set
                     Daily Communications
                       With the Field and
                     Analytical Laboratory
                    Generate Performances
                     Based on Acceptance
                           Criteria
                                                     Magnetic Tape
                                                     Mailed From
                                                     ORNLtoNCC-IBM
                                   Automated QA Review
                                   1.  Identify Exceptions
                                      to QA Criteria
                                   2.  Preliminary Statistics
                                      to Identify Outliers
                                                         ±.
                                      Fix Mistakes
                                     and Set Flags
                                 Using SAS Full-Screen
                                         Editor

Apply
Correcting
Entries to Raw


Data Set

                                                  Magnetic Tape Mailed
                                                    From NCC-IBM to
                                                         ORNL
Generation of
Verified Data
     Set
No
Yes
       Figure 7. Data Verification Process for the Eastern Lake Survey - Phase I.
                                      26


-------
    routine/trailer duplicate pairs were evaluated for
    precision.

  • pH and DIG QCCS Data. Form 2 QCCS data were
    compared with acceptance criteria.

Data anomalies were reported to the field laboratory
coordinator for review. Continual review by the field
laboratory staff minimized the number of data tran-
scription errors  encountered by the EMSL-LV QA
staff. Data reporting errors were reported to ORNL for
correction before values were entered into the data
base.  All  telephone communications  were docu-
mented in a  bound notebook. Data changes were
annotated on the appropriate form.

The contract  laboratory data packages were deliv-
ered to the EMSL-LV QA staff upon completion of
analysis or within 35 days following sample receipt.
As they were received, data packages were reviewed
for  completeness,  internal  QC compliance,  and
appropriate use  of data qualifiers.  A  data  package
completeness checklist was used by each EMSL-LV
QA auditor to assure consistency in the review of all
data packages (Figure 8).  Problems were reported to
the appropriate contract laboratory manager for cor-
rective action. Comments provided by  the laboratory
with the data package were also reviewed  to deter-
mine their impact on data quality and the need for any
followup action by the laboratory. Completion of this
checklist was important in verifying that the labora-
tory had met all contractual requirements for the pur-
pose of payment.


Quality Assurance and Quality Control Data

ORNL personnel entered  the data into the raw data
set as the data packages were received (see Section
3). During the data entry period, QA sample data were
evaluated by the EMSL-LV QA staff  to establish per-
formance-based  acceptance  criteria. The review
process utilized  the computer  programs listed in
Table 8 to identify or flag results that were exceptions
to the expected QA-QC limits. These programs auto-
mated  much of the QA review process and enabled
the auditor to concentrate more effort on the correc-
tion or flagging of questionable data.

The QA auditor used the output from these programs
(along with the original data and field  notebooks) to
evaluate the data and to complete the NSWS verifica-
tion report contained in the QA plan. The verification
report was actually a worksheet designed to system-
atically guide  the auditor through  the  verification
process. It listed procedures for qualifying ('flagging')
data and for tracking  both requests for confirmation
reanalysis and for data resubmissions. It also listed
the steps used to help explain the QA exceptions and
to summarize all modifications to the raw data set.
These auditing procedures are described in detail in
Drouse et al. (1986).
After being entered into the raw data set, the data
were reviewed on a batch basis by making use of both
computer programs and manual review. These efforts
checked internal consistency (e.g., ion balance differ-
ence and conductance difference) for each sample
and the acceptability of both the QA sample data and
the laboratory QC data for each batch..  Examination
of univariate distributions was also performed on QA
sample data to identify statistical outliers  and to
establish or update performance acceptance criteria.
Samples  which met acceptance criteria for these
checks were transferred into the verified data set.

When exceptions could be explained by the presence
of organic compounds based on the Oliver et al. (1983)
model  or on a correctable reporting error, these val-
ues associated with these samples were qualified or
corrected and entered into the verified data set. This
was accomplished through the Hillman-Kramer  Pro-
tolyte Analysis Program as described in Drouse et al.
(1986).  When exceptions to the ioin balance differ-
ence or conductance difference criteria could not be
explained by calculated organic ion concentrations,
that sample was reanalyzed to determine if the result
was due to reporting or analytical error.

Suspected analytical errors were referred to the ana-
lytical  laboratory  for reanalysis. Acceptable values
from reanalysis were qualified and substituted for the
original values  in the  verified data set. For each
parameter, samples that were  not  analyzed within
maximum allowable holding times or that were asso-
ciated  with unacceptable QC or QA sample results
were flagged before entry into the verified data set.

When the QA sample for a given parameter  did not
meet the acceptance criteria,  that parameter  was
flagged for all samples in the batch. A parameter was
also flagged when internal QC checks were not met.
Those checks included matrix spike recovery, calibra-
tion  and  reagent blank analysis, internal duplicate
precision, required   instrumental detection limit,
QCCS  percent  recovery, and maximum allowable
holding times. In all cases, each flag generated by the
computer was reviewed by the QA auditor for reason-
ableness and consistency before it was added to the
data base.

Less than 3 percent of the raw data reported for  lake
samples was classified as reporting errors and  was
corrected before transfer to the verified data  set.
Sample reanalysis was requested for less than 4 per-
cent of the originally reported raw data values. Less
than 1  percent of the reported data required correc-
tion because of transcription or data entry errors. The
overall  error rate for data entry and updating of the
raw data set to the verified data set was estimated to
be less than 0.03 percent.
                                                27

-------
Lab Name:

Date:
   NATIONAL SURFACE WATER SURVEY

Data Package Completeness Checklist

                 Batch ID:
                                                                    Page 1 of 2
                 Auditor's Initials:
   All major difficulties during analyses have
   been discussed with the QA manager or designee.
   Anion-cation balance and conductance balance
   checks exceeding criteria are reported on cover
   letter.
   a. Required forms (11, 13, 17, 18-23) submitted.
   b. Lab name, batch ID, and lab manager's
      signature submitted on all forms.
   c. Sample ID reported on Forms 13, 21, 22, and 23.
   d. Analyst's signature on Form 13.
   e. Correct units indicated on all forms.
   Form 11:
   a. Correct number of samples analyzed and
      results for each parameter tabulated.
   b. Data qualifiers (J, K, M, or U) reported when
      results are missing.
   c. Data qualifier R is reported when a sample
      is reanalyzed for QC purposes.
   d. F is reported as a data qualifier when a
      result is outside criteria (with consent
      of QA manager).
   e. G is reported as a data qualifier when the
      method of standard additions is used and
      Form 23 is submitted.
   f. ANC initial pH and BNC initial pH are within
      ±0.1 pH unit.
   Percent 1C resolution reported as greater than 60%
   on Form 17.
   Form 18:
   a. Instrumental detection limits and associated
      dates of determination tabulated.
   b. Instrumental detection limits less than or
      equal to the required detection limits
      (Table 1-1).
   Form 19:
   a. Date sampled, date received, holding time
      plus date sampled, and dates of analyses for
      the correct number of samples are tab-
      ulated.
                                 Yes
Par-
tial
No
Comments
                                                                     (continued)
                  Figure 8. NSWS Data Package Completeness Checklist.
                                      28

-------
                                                                     Page 2 of 2
10.
11.
b. Date analyzed is less than or equal to the
   reported holding time plus date sampled.
c. The data qualifier H is reported for dates
   of analyses which exceed the holding time
   plus date sampled with consent of QA manager.
Form 20:
a. Calibration blanks, reagent blanks, and DL
   QCCS are reported where required.
b. Calibration blanks and reagent blanks are
   less than 2 times the required DL.
c. DL QCCS is approximately 2 times the required
   DL and the measured values are within 20% of
   the theoretical  values.
d. QCCS true values are in the midrange of
   sample values.
e. If high QCCS true values are reported, the
   samples analyzed on high range are discussed in
   the cover letter.
f. Diluted samples and their dilution factors
   are discussed in the cover letter.
Percent recovery of matrix spikes is reported
on Form 21 for each required analysis, and
the values are within the range of 85% to 115%.
Duplicate precision results are reported for
each parameter and are less than or equal
to the maximum %RSD (Table 1-1).
Standard additions are performed and Form 23 is
submitted when the matrix spike analyses do
not meet contract requirements.
Yes











Par-
tial











No











                                                                        Comments
 Note:  Checklist is not required in data package but is recommended
        to be included in the raw data.
                             Figure 8.   (continued).
               Figure 8. NSWS Data Package Completeness Checklist (Concluded).
                                       29

-------
                             Title
                                  Sample Type
 Audit Sample Summary

 Lab and File Blank Summary

 Field Duplicate Precision Summary

 Instrumental Detection Limit Summary

 Holding Time Summary

 DIG Check Calculations

 ANC Check Calculations

 Conductivity 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 - (DIG, DOC, pH, ANC, and BMC Data Evaluation)

 Audit Sample Window Generation

 Raw Data Listing - Format for QA Manager

 Complete Raw Data Listing - Format for Audit Staff

 Reagent-Calibration Blanks and QCCS

 Calculation of Laboratory Penalties

 Matrix Spike Summary
                             (LH, LL, FH, FL, FN)

                             (B, LB, FB)

                             (R, D Pairs)

                             (All)

                             (All)

                             (All)

                             (All)

                             (All)

                             (All)

                             (All)

                             (pH and DIG)

                             (pH and DIG)
           Table 8.  Exception Generating and Data Review Programs, Eastern Lake Survey (Phase I).
Follow-up with Contract Analytical Laboratories

Completion of Step 2 (data receipt) and Step 3 (QA-QC
review of data) included communications with  the
appropriate contract analytical laboratory. This  fol-
low-up (Step 4) was needed to obtain corrections to
the data package, confirmation  or correction  of
reported data, and sample reanalysis when required.
Step 4 was the most difficult and time-consuming
process in data verification, especially if the requests
to the laboratory were not clearly supported by con-
tract requirements.  Typically, responses to requests
for confirmation or  correction of reported data were
completed within 2 to  4 weeks.  Reanalyses were
either completed within 2 to 3 months or were not per-
formed because of negotiations with the contractors
for additional payment.  Although every effort was
made to identify  samples  for reanalysis within  the
maximum allowable holding times, the contract labo-
ratory often had to  choose between samples in pro-
gress and  samples for reanalysis.  Therefore, many
reanalyses, especially for nitrate in aliquot 5, were
performed outside the maximum allowable holding
time.

Preparation and Delivery of Verification Tapes

After the first four verification steps were completed
by the EMSL-LV QA staff, the data were adequate for
establishing the verified data set. In order to translate
the raw data set into the verified data set, a method
forquickly transferring the information from EMSL-LV
to ORNL was required. Changes to the raw data set
were made using computer entries called  tuples.  A
tuple is an ordered group of elements (e.g., batch ID,
sample ID, variable, old flag, new flag, old value, new
value). For the ELS-I, tuples identified changes into a
data set such as adding, changing, or deleting sam-
ple values or data qualifiers. A tuple was also created
when a computer program generated a flag for a spe-
cific parameter.

The system  that was used for modifying  the  ORNL
raw data set was designed to  minimize data entry
errors. This system used separate areas for each  pro-
                                                  30

-------
gram-generated tuple and each manual tuple (value
change or deletion), which  facilitated  searching,
modifying, and checking of tuple listings. When a
tuple listing was ready to send to ORNL, a computer
program (Database Administrator Program) was used
to combine all of the tuple areas including flags, tags
(data qualifiers added to a value  in the field), and
value changes, and to append the listing to the data
base. This listing included only those tuples for which
the batch ID, sample  ID, variable name, and old value
from EMSL-LV matched those from ORNL. The com-
bined tuble was then written on magnetic tape and
mailed to ORNL from the NCC, RTP, North Carolina.
ORNL  processed  the combined tuble listing  and
returned a magnetic tape  to RTP with a listing of
unmatched tubles which were mistakes  that could
not be applied to the data  base  until they were cor-
rected. This cycle took about 10 days.

The overall outcome of the five steps described above
was a verified data set in which all values which did to
meet criteria  were flagged, replaced with either cor-
rected  or reanalyzed data or replaced with missing
value codes (Table 9). Confirmation (C) and reanalysis
(R) codes were removed from the data set when the
correct values were received from the contract labo-
ratory.


Data Validation

The  data validation process  identified  potential
errors  in  chemical  analyses  that  could  not  be
revealed by verification  procedures.  These values
included potential  outliers  and systematic  errors
because data were evaluated on  a regional basis
(Table 10). The quality of non-chemical variables was
also evaluated during the validation process (Table
11).

Data validation was a joint activity performed by ERL-
Corvallis and EMSL-LV. While interpreting the results,
the EPA data  users  identified  questionable data
ponts  (Figure 9).  The  atypical  data  points  were
reported to the EMSL-LV QA staff who then reviewed
the data packages and QA information to reverify the
quality of that particular value. The final decisions
regarding data quality  were made  by the  EPA data
users at ERL-Corvallis  on the  basis of all available
information.

The data validation process included the identifica-
tion of possible outliers and the evaluation of possi-
ble systematic error in the  measurement process.
Both of these aspects were exploratory (as opposed
to  test-oriented).  Thus,  the  validation  methods
stressed  visual  presentation    and  subjective,
although conservative,  data selection procedures.
The objective was to identify data  values or sets of
data values that warranted special  attention or cau-
tion when used for analysis of survey results when
used for model-building based upon survey data. The
methods selected for detection of  outliers and  sys-
tematic errors were chosen for simplicity of computa-
tion by  using pre-existing software whenever possi-
ble (see Figure 9).  Data validation is discussed in
greater detail in Drouse et al. (1986) and the Eilers et
al. (1986).
   A   - Carbonate alkalinity, C02-Acidity and mineral acidity data are eliminated from data base due to method inconsistencies.

   C   - Temporary flag indicating raw data incomplete pending CONFIRMATION by analytical laboratory.

   N   - Eliminate from data base pending review of aliquot 5 nitrate data.

   R   - Temporary flag indicating raw data incomplete pending REANALYSIS.

   X   - Permanent flag indicating IMVALID data based on QA review.

       - Value never reported.

         (NOTE: These codes appear in [nuJNUMERIC fields only.)
                                  Table 9.  NSWS Missing Value Codes.
                                                 31

-------
 Conductance (contract analytical laboratory
 AND
 pH  (field laboratory)
 DOC



 True color



 Turbidity

 Ca


  Na

  DIG (equilibrated)

 Al (total)
vs.    Conductance (field)
       Calculated conductance
       Sum of cations
       Sum of anions
       Ca
       Ca plus Na
       pH (field laboratory)
       ANC
       DIG

vs.    Ca
       pH (all measures)
       Sum of cations
       Sum of anions
       Calculated ANC [sum of base cations minus (S042~ and Cl

vs.    pH (all measures)
       Sum of cations
       Sum of anions
       Al (extractable)
       S042~ (expressed as percent of total anions)

vs.    True color
       Anion deficit
       Secchi disk transparency
       Turbidity

vs.    Anion deficit
       Secchi disk transparency
       Turbidity

 vs.    Secchi disk transparency

 vs.    Wig
       Si02

 vs.    Cf

 vs.     DIC (initial)

 vs.    Al (total extractable)
Table 10.  Data Validation for the Eastern Lake Survey - Phase I: Comparison of Variables Used to Check for
Random and Systematic Errors.
                                                           32

-------
          Variable                                        General Description of Validation Checks


Latitude,                          Lake location as measured by LORAN was compared against location on U.S.G.S. maps.
Longitude

Lake Elevation,                    Lake and watershed characteristics were checked against state records, where available, to
Lake Area,                        confirm lake identification.
Watershed Area,
Site Depth,
Stream Inlets and Outlets,
Lake Hydrologic Type

Shoreline Land Use                Compared against aerial photographs

Water Temperature                 Compared against range of appropriate temperature

Secchi Disk Transparency,          Compared with each other for internal consistency
True Color,
Turbidity


              Table 11.  Physical Parameters Subject to Validation,  Eastern Lake Survey - Phase I.
                                                         33

-------
                        c
       Data Set No. 2
         (Verified)
   UNIVARIATE
• Box Plots
• Probability Plots
       MULTIVARIATE
     • PCA
     • Cluster Analysis
     • Trilinear Plots
     • MLR
                  Outliers
                                                            BIVARIATE
                  • Scatter Plots
                  • Regression
                                                     Relational
                           Comparative
 Modify or
Delete Value
                 Yes
 /ValueV
\   in   /
 \Error/
                                                                        Systematic
                                                                        Differences
(
U
                                            Hag or
                                         flodify Values
                            No
                                                   No
                                                       Data Set
                                                         No. 3
                                                      (Validated)
            Figure 9. Data Validation Process for the Eastern Lake Survey - Phase I.
                                        34

-------
                                            Section 4
                                             Results
 Evaluation of the quality assurance and quality con-
 trol data was an ongoing process during and follow-
 ing the ELS-I. A substantial part of this evaluation
 process  involved the statistical analysis of the QA
 and  QC sample results which  are presented below
 and in Appendix A. Evaluations were also made of the
 field and laboratory operations, the verification pro-
 cedures, and the analytical methods. Results of the
 chemical  analyses of samples  split between  ELS-I
 analytical laboratories and ERL-Corvallis, Canadian,
 or Norwegian laboratories are summarized in Drouse
 and Best (1986), Stapanian et al. (1986), and Yfantis et
 al. (1986).

 Operations Evaluation

 Overall, operation of the ELS-I  proceeded smoothly.
 The QA-QC program was strictly adhered to through-
 out the period of operations. Some specific problems
 were encountered that were expected for a study of
 this magnitude, and these were usually detected and
 resolved quickly. Some protocol changes were imple-
 mented during the survey; others were made after the
 survey as a result of debriefing and evaluation recom-
 mendations. All changes were incorporated into the
 final QA plan  (Drouse et al. 1986) and  were  imple-
 mented during subsequent NSWS operations.

The ELS-I was completed in a timely manner, and data
of known quality were collected consistently through-
out the period of operations. The statistical require-
 ments for adequate sample size were achieved at all
field stations (Linthurst et al. 1986). There were-no
 major interruptions in field operations due to acci-
dents, weather, or equipment failure. The sampling
and  laboratory protocols were  very successful for
most procedures and should serve as a model for
future field studies of a similar nature.

Lake and Sample Information

 Field operations were a successful means for obtain-
ing samples  and field data consistent with the  ELS-I
research plan. Approximately 90 percent of the lakes
initially selected for sampling were visited by sam-
pling crews,  and of those  lakes visited, 96 percent
(1,612) were  actually  sampled.  In addition to those
lakes, 188 special interest lakes were sampled for a
total sample of 1,800 lakes. The numbers of regular
lakes selected, visited, and sampled during the ELS-I
are listed by subregion in Morris et al. (1986).

Less than 20 percent of  the lakes visited were sam-
pled at a depth other than the one originally specified.
Only 5 percent of the lakes sampled were thermally
stratified; thus,  95 percent of all  samples  were
acceptable in terms of the research plan requirement
that a single water sample was to be collected from
each lake during a period when the  lakes were iso-
thermal.

In total, 2,389 routine, field duplicate, and field blank
samples were delivered from the field laboratories to
the contract analytical laboratories (Table 12). This
resulted in 2,639 sets of analyses because the con-
tract required that one matrix spike and one labora-
tory dupicate be analyzed per batch of samples. The
distribution of samples by field station and analytical
laboratory is provided in Table 13.

Field Problems and Their Resolutions

In general, the sampling teams performed as planned
during the ELS-I with an average yield of 20 processed
samples per operating day from each field laboratory.
Three problems were identified and corrected during
field operations. These concerned an inaccurate pH
meter, contamination of blanks from one field labora-
tory, and destruction of one set of split samples. The
identification of  the  problems  and  the corrective
actions taken are discussed below.

Prior to the beginning of routine lake sampling from
the Duluth and Rhinelander field stations, a practice
sampling excursion was undertaken. Samples col-
lected during that excursion were split between the
Duluth and Rhinelander field laboratories for proc-
essing and chemical analyses. When the data were
evaluated, it was found that the Duluth pH measure-
ments differed from the  Rhinelander measurements
by one pH unit at pH values greater than 6.0. Measure-
ments  of the buffers used for calibration and the
QCCS were found to be accurate. After changing the
pH  electrodes, checking the  buffers, and reviewing
the measurement technique, the problem was eventu-
ally traced to a faulty electronic display on the Duluth
pH meter. A replacement meter was obtained and was
used throughout the  survey.  The problem with the
Duluth pH meter led to implementation of an addi-
                                                35

-------
Laboratory
Versar
EMS!
Global
USGS-CO
Total
a Includes analysis of matrix spikes
Samples Received
872
983
332
202
2,389
and duplicates (11% lab QC)
Samples Analyzed3
970
1,081
366
222
2,639

 Table 12.  Number of Samples Received and Analyzed by Contract Laboratories, Eastern Lake Survey - Phase I
tional, mid-level QC check at  all  field stations. For
each batch of samples, the laboratory supervisor was
required to compare the Hydrolab pH measurements
with the field laboratory measurement and to take
corrective action if the two values differed by  more
than 0.5 pH unit.

Some field blanks associated with the Lexington field
station were found to have  high levels  of several
major cations and anions.   Further investigation
revealed that, because of failure of the reverse osmo-
sis-deionization (RO-DI) water system in the field lab-
oratory,  the  field personnel had been instructed to
take distilled water for two batches from the EPA
regional laboratory still. Each field station was sub-
sequently informed of the problem and cautioned to
use only field laboratory RO-DI water for field blanks.
No  routine  samples were  lost  during  the ELS-I
although one batch of samples was temporarily mis-
routed in shipment to the contract analytical labora-
tory. These samples were located using the sample
tracking procedures forthe survey and were analyzed
within the  required sample holding time. The Norwe-
gian split  samples collected from Region  1E were
inadvertently destroyed by the commercial courier
service.  A  second set of split samples was collected
from Region 3A for shipment to Norway (see Section
1).

Several  general recommendations to improve field
operations of future NSWS activities were obtained
from summaries provided by each EPA base coordina-
tor.  Many of the temporary employees hired  by Lock-
heed-EMSCO as  field sampling and field laboratory
Field Station
Bangor
Lexington
Lake Placid
Mt Pocono
Duluth
Rhinelander
Asheville
Lakeland
Total Samples

Versar
—
77
—
99
343
353
—
—
872
Analytical Laboratories
EM SI Global
213
332
332
106
54
116
162
—
983 332

USGS-CO Total
213
409
332
205
397
469
162
202 202
202 2,389
           Table 13.  Number of Samples Delivered by Field Stations, Eastern Lake Survey - Phase I.

                                                36

-------
 personnel also provided debriefing letters before or
 shortly after the ELS-I was completed. In addition, all
 field laboratory notebooks were turned over to the QA
 manager. A debriefing was held for all EPA base coor-
 dinators and duty officers and members of the NSWS
 management team in Plant City, Florida, in December
 1984. Topics from that meeting related to the quality
 assurance program included communications, sam-
 ple shipment, and  training. These topics are dis-
 cussed below.

 Efficient and complete information transfer between
 the NSWS management team and the field stations
 was necessary to ensure that all new developments
 or modifications in  operational  protocols were con-
 sistently disseminated and understood.

 Shipment  of samples to the contract analytical labo-
 ratories was a problem on weekends especially when
 there was no service by overnight courier. Weekend
 shipments using commercial airlines required  close
 coordination  between the field  laboratory and con-
 tract laboratory personnel to assure that the samples
 were received at the contract laboratories within  the
 appropriate time. The contract laboratories later sug-
 gested  holding weekend shipments until  Monday,
 since it was easier for them to meet the holding time
 constraints than to pick up samples at the airport.

 More comprehensive instruction to field samplers in
 the completion of field data forms was recommended
 to ensure  clarity and consistency. It was suggested
 that training at each field  station be  lengthened,
 since some field sampling personnel lacked experi-
 ence and an understanding of limnology and certain
 types of sampling equipment. At some field stations,
 state or  regional  EPA  sampling  personnel  were
 rotated and were replaced by new people on  a fre-
 quent basis.  It was suggested that this practice be
 discouraged if possible because it added to inconsis-
 tencies in  data reporting and required that additional
 time and effort be spent on training activities.

 Contract Analytical Laboratory Problems and
 Resolutions

 During the ELS-I, there was one occurrence of a sam-
 ple volume overload at one contract laboratory. This
 problem was immediately resolved by  distributing
 some samples to another contract laboratory.

 Several analytical problems were identified and cor-
 rected during contract laboratory operations. Two
 significant difficulties centered on the  calibration
 method used  with silica measurement at one labora-
tory and aluminum contamination problems at  three
 laboratories. These are discussed below.

The daily QA checks and audit sample data compari-
sons by the QA staff indicated that the silica results
from one  contract  laboratory had a negative  bias.
 Field blank data with highly negative values were also
being reported by the same laboratory. The bias was
eventually traced to a calibration problem. The other
contract laboratories did not experience the same dif-
ficulty. The laboratory in error was calibrating for sil-
ica in the 10 to 60 ppm range which was specified for
the method. However, the silica method also speci-
fied that  the  calibration standards bracket the
expected sample concentrations  which were well
below the 10 to 60 ppm range. The laboratory involved
calibrated its instruments for silica analyses from 0
to 10 ppm thereafter and reanalyzed the affected sam-
ples. Data from reanalysis for silica were thoroughly
reviewed for technical merit before substitution into
the raw data set.

During the  pilot study, laboratory blank values were
reported for total aluminum that were  equal to or
greater than the values found  in routine samples. The
high aluminum concentrations, believed to be caused
by either airborne dust contamination, reagent and
glassware contamination,  or both,  occurred  at two
laboratories. The major source of contamination was
traced to the use of borosilicate  glassware which
containsaluminumoxide. Duringthedigestion proce-
dure aluminum was being leached from the glass.
Teflon beakers were substituted for glass in each lab-
oratory.

The initial sample batches analyzed at one laboratory
also exhibited  sporadic aluminum  contamination.
Through the daily QA contact with this laboratory, the
problem was identified  and was quickly traced to a
building maintenance operation. The floors  in an
adjacent room had been sanded during the weekend
while several sample batches were being digested for
measurement of total aluminum. As a result, all labo-
ratories were instructed to improve contamination
control  procedures  by  performing  digestions in  a
laminar-flow hood or in an isolated station within a
standard fume hood to  prevent airborne contamina-
tion. This procedure was implemented at all partici-
pating laboratories.

Data Verification Problems and Resolutions

Several instances of misreported data were  uncov-
ered through daily QA checks and the data verifica-
tion process. Although uncommon, errors were iden-
tified   and  traced  to  data  transcription   errors,
switched aliquots in the field laboratory, mislabeled
samples, and data entry errors at the ORNL data cen-
ter. Through careful QA evaluation and follow-up with
all participants, the incidents were identified and cor-
rected.

As discussed in Section 3, general problems existed
with the system used for modifying the ORNL raw
data base. The system was slow; the entire cycle took
about 10 days. A second  system was implemented for
a short  time which was faster (3 days) but required
manual  entry error checking and, thus, was labor-
                                                37

-------
intensive. The modification procedure has since been
redesigned to facilitate error detection  and data
tracking while providing an efficient turnaround time.

Methods Evaluation

Several analytical questions arose during the plan-
ning for the ELS-I; others arose as problems during
the pilot studies. Several methods studies were con-
ducted at EMSL-LV to address  and  resolve these
issues. Questions  that arose  during the planning
stages included  the practicality  of measuring free
dissolved fluoride, the relative merits of ion chroma-
tography (1C) versus ion-selective  electrodes (ISE) for
measurement of total  dissolved  fluoride, and the
accuracy of Gran analysis for determination of ANC
and BNC. These issues were evaluated during and fol-
lowing the pilot study. Other issues that were identi-
fied during the pilot study concerned the procedure
for measuring total extractable aluminum, the design
of the pH sample chamber, the effects of filtration on
iron and aluminum  in the audit samples, and nitrate
contamination in blank samples. These methods
evaluations are discussed below.
eluteatthesametimeasfluoride(Hillmanet al. 1986).
It was concluded that despite its advantages 1C is not
a suitable method for measuring total dissolved fluo-
ride in natural waters under standard  1C conditions.
Total dissolved fluoride was therefore measured with
ISE during the ELS-I.

Acid-Neutralizing Capacity and Base-Neutralizing
Capacity

In order to maximize the accuracy and precision of
ANC and BNC determinations, a full  Gran analysis
was specified for interpreting titration  data. Because
Gran analysis is not a standard technique for measur-
ing ANC and BNC and because complex calculations
were involved, several samples were analyzed prior to
the ELS-I to ensure that the calculations were correct
and to provide detailed examples to the contract labo-
ratories. These  examples were included in both the
IFB and the analytical methods manual. Technical dif-
ficulties were encountered with the interpretation of
BNC titration data which are  presently unresolved
This parameter is not discussed further in the  present
report.
Fluoride Determinations

Prior to the start of the ELS-I, free dissolved fluoride
was chosen as one of the parameters of interest. It is
an important parameter to consider in modeling alu-
minum chemistry, but it is difficult to measure. No
satisfactory method could be found in the literature,
and  none  was  recommended by  the  analytical
experts at a workshop which was held in Denver, Col-
orado, during the planning stage of the ELS-I.

On  a trial  basis, a simple ion-selective electrode
method was used during the pilot  study. Pilot study
results for free dissolved fluoride tended to be varia-
ble  and often were  higher than those for total dis-
solved fluoride.  Also,  the measurement was  very
time-consuming. Based on  the pilot study  results,
free dissolved  fluoride was dropped from the list of
parameters for the NSWS until an acceptable method
for the determination could be developed.

At the Denver Analytical Workshop, two methods
were  suggested  for the determination  of total dis-
solved fluoride: ion-selective electrode (ISE) and ion
chromatography (1C). The ISE method is a standard
analytical technique, but the electrode is character-
ized by low sensitivity and slow response especially
when the fluoride concentration is below 0.1 mg L  .
The 1C method  is not commonly  used for  fluoride
determinations,  but it is very  sensitive (detection
limit <0.005 mg L~1), rapid (analysis time less than 8
minutes), and easily automated. It is also possible to
simultaneously measure anions in addition to  fluo-
ride using 1C. However, the determination of fluoride
by 1C is subject  to interferences such as the "water
dip" and measurement of small organic  species that
Total Extractable Aluminum

Although the method specified for the determination
of total extractable aluminum (Al) is based on meth-
ods in the literature (Barnes 197979), questions arose
during the pilot study regarding the effects of filtra-
tion and filter type (brand and composition), sample
temperature, and extraction technique. Estimates of
precision based on both single- and multi-analyst
measurements were desired. In order to obtain this
information, a series of experiments was performed
using the natural audit sample from Big Moose Lake.

The effect  of filter  type  on  analytical  results  was
examined by  determining total extractable Al  in an
unfiltered sample and in three filtered samples, each
obtained using  a different type of membrane filter.
The analytical results indicated that neither filtration
(0.45-fj.m membrane) nor filter type had an effect on
total extractable Al concentrations.  The membrane
filter was selected for use during the ELS-I.

The question of sample temperature arose because
an effort was not made during the  pilot  study to
ensure that samples were at the same temperature
prior to Al extraction.  If  sample temperature were
important,  then the comparability of  results among
lakes would be  affected. To determine whether tem-
perature had an effect on total extractable Al concen-
tration, portions of a sample were equilibrated at two
different temperatures (4°C and 20°C)  and  then
extracted. The results of this experiment indicated
that sample  temperature had  no effect  on  total
extractable Al concentration.

The experimental protocol for  total  extractable Al
called for a rapid  extraction with 8-hydroxyquinoline
                                                38

-------
 into methyl isobutyl ketone (MIBK). Because the com-
 plexation time and the method for this  extraction
 were not specified, it was necesary to estimate the
 effect of small variations in the extraction technique.
 Portions of the natural audit sample were extracted
 using two  mixing  styles (vigorous shaking and  no
 deliberate shaking after hydroxyquinoline addition)
 and two complexation times (5 seconds and 20 sec-
 onds). The results indicated that equivalent results
 were obtained for each of the experiments, i.e., that
 small variations in extraction technique did not affect
 measured sample concentrations.

 The lack of effect from variations in extraction tech-
 nique was seen when determining single- and  multi-
 analyst precision. The mean concentrations and
 standard deviations of total extractable Al  measure-
 ments  by a single analyst (0.192  ±  0.011 mg L~1,
 n = 24) and by several analysts (0,193 ± 0.010 mg L~1,
 11 analysts, 2 extractions each) are essentially identi-
 cal. This experiment demonstrated that, with ade-
 quate personnel training, total  extractable Al results
 were not affected by any among-analyst variations in
 extraction technique.

 pH Sample Chamber

 During  the  ELS-I, sealed  syringe samples  were
 obtained at each lake for pH determinations. Sealing
 samples in syringes was expected to preserve sample
 pH by minimizing exchange of dissolved C02   with
 the atmosphere. A sample chamber was also  devel-
 oped for the field laboratory pH measurements to pre-
 vent exposure of  the sample to air (Hillman  et  al.
 1986).

 Formation of Filterable Iron and Aluminum
 Complexes in Synthetic Field Audit Samples

 During the pilot study, concentrations of dissolved  Fe
 and total extractable Al concentrations in field syn-
 thetic audit samples were  consistently lower than
 those in laboratory synthetic audit samples. The pri-
 mary difference between laboratory  synthetic and
 field synthetic  audit samples  was that laboratory
 audits were processed (acidified for Fe analysis and
 extracted into MIBK for Al analysis) immediately after
 preparation without filtering, and field audits were
 processed (with filtration) 24 to 72 hours after prepa-
 ration. The probable cause for lower concentrations
of  Fe and total extractable Al in the field audit sam-
 ples was that, during the time delay, hydroxides  or
other  complexes  were formed that  did  not  pass
through the 0.45-Mm membrane filter.

An experiment was performed to determine the cause
of the low concentrations of Fe and total extractable
Al  in field audit samples. Three aliquots each from
both the low and high field synthetic audit samples
were acidified to pH   2 with analytical-grade  nitric
acid and analyzed for Fe and  total extractable Al.
Three  additional  aliquots from each  audit sample
were similarly prepared, except that they were ana-
lyzed after being filtered through a 0.45-fxm membrane.

The analytical results suggested that total extract-
able Al was lost from the field audit samples during
filtration.  The purpose of sample  filtration was to
remove suspended materials which at pH 6 to 8 may
have included  hydroxyaluminum complexes.  This
explanation is consistent with the observation that
laboratory audit values for total extractable Al were
higher than field audit values. By extracting the labo-
ratory  audit sample immediately after preparation
(within 1 hour), the formation of large polymeric Al
species may have been avoided, and, hence, extract-
able Al was not lost from the samples.

Furthermore, Fe was not detected in either filtered or
unfiltered field audit samples. In those samples, Fe
probably formed a precipitate which was adsorbed to
the sample bottle between the time of preparation at
the analytical support laboratory  and the time of
processing (filtration and acidification) in the  field
laboratory. Because the samples were filtered prior to
being acidified, the  adsorbed Fe  precipitate would
have been lost from these samples.  In laboratory
audit samples which were processed by the analyti-
cal support laboratory immediately after preparation,
the precipitate may not have had time to form. Field
audit results for  Fe and total extractable  Al were
therefore not included in statistical evaluations.

Nitrate Contamination

To meet contract analytical laboratory quality control
specifications, nitrate concentrations  in field blank
samples were required to be less than 0.01 mg L~1.
However, during the ELS-I pilot study, up to 18 mg L~1
nitrate was  detected in field blanks, and this  sug-
gested that a serious contamination problem existed.
This contamination was not present in contract ana-
lytical  laboratory  blank samples. After discussions
with the managers of the contract  laboratories per-
forming the nitrate analyses, it  was concluded that
the source of the contamination was not in the con-
tract laboratory but was in the cleaning procedure for
aliquot 3 sample containers. That cleaning procedure
which included a nitric acid  rinse is outlined in Table
14. As  a first attempt at eliminating the contamina-
tion, the nitric acid rinse was omitted from the clean-
ing procedure for aliquot 3 containers during the ELS-
I.

Another potential source of contamination was the
sample filtration procedure at the field laboratory
which included a nitric  acid rinse of the filter holder
and membrane followed by copious rinsing with RO-
Dl water. This step was not considered to be a prob-
lem at the time that the bottle cleaning procedure was
changed, and during the ELS-I field laboratory train-
ing sessions, the need for copious rinsing of the filtra-
tion apparatus was stressed.
                                                39

-------
   1.  Rinse container three times with deionized water.

   2.  Rinse container three times with 3 N HN03.

   3.  Rinse container six times with deionized water.

   4.  Fill container with deionized water and allow to stand for 48 hours.

   5.  Empty container, air dry in a laminar-flow hood (class 100 air).

   6.  Cap containers and place in clean plastic bags.

   NOTE:  Deionized water must meet ASTM specifications for Type I Reagent Water. Nitric acid is Baker Instra-Analyzed or
         equivalent
             Table 14.  Cleaning Procedure Used for Aliquot 3 Sample Containers, ELS-I Pilot Study.
The change in the cleaning procedure for the sample
containers and the rinsing step in the filtration proc-
ess reduced but did not eliminate the nitrate contami-
nation problem. Field blank samples obtained during
the early part of the ELS-I still contained up to 3.5 mg
L~1 nitrate. Overall, the mean nitrate concentration in
field blank samples during the early part of the ELS-I
was 0.3 ± 0.6 mg L~1 (n = 146). It became apparent
that the filtration procedure  was a source of more
contamination than was previously expected.

In the original filtration procedure (Table 15), the filter
holder and membrane  were  rinsed with 5  percent
HNC>3 between samples. In a further attempt to elimi-
nate the nitrate contamination, the nitric acid rinse
was eliminated for aliquot 3. However, the same filtra-
tor was still used for all aliquots, and the filtration pro-
cedure  during the  processing of aliquots 1 and  4
included a  nitric acid rinse of the filter holders and
membranes.  Blanks processed  in this manner still
contained about 0.05 mg L~1 nitrate. Further investi-
gation revealed that the design of the filter holder per-
mitted intermittent nitrate contamination (i.e., nitrate
was not completely removed by successive deionized
water rinses). To avoid this occurrence, the filtration
procedure for the ELS-I was modified to include the
use  of separate  filter  holders and membranes to
obtain the  unacidified  aliquots. New filter  holders
were dedicated to the aliquot 3 filtration and were
never allowed to contact nitric acid.

To determine whether the  new filtration procedure
eliminated nitrate contamination of aliquot 3, a series
of experiments was  performed at the field laborato-
ries in Duluth, Lexington, and Rhinelander. At each
laboratory,  10 4-liter deionized water  samples (two
samples every hour for 5 hours) were  collected and
processed,  each in three different ways (Table 16) at
the times and dates of sample preparation given in
Table 17. The processed samples were then analyzed
   Filtered Aliquots - Filtered sample is obtained by vacuum filtration through a 0.45-jum membrane filter into a clean 500-mL
   container The filtered sample is then transferred into the containers for aliquots 1, 3, and 4. the filtration is performed as
   follows1

      a.  Assemble the filtration apparatus with a waste container as a collection vessel. Thoroughly rinse the filter holder and
         membrane filter in succession with 20 to 40 ml of deionized water, 20 ml of 5% HN03 (Baker Instra-Analyzed grade),
         and 40 to 50 ml of deionized water (it is CRUCIAL that all traces of the HN03 rinse be removed)

      b.   Rinse the filter holder and membrane with 10 to 15 mL of the sample to be filtered.

      c.  Replace the waste container with a clean 500-mL plastic  bottle  Reapply vacuum (vacuum pressure must not exceed
         12 inches Hg), and filter 10 to 15 mL of sample into the bottle.  Rinse the container by slowly rotating the bottle so that
         the  sample touches all surfaces. Discard the rinse and place the 500-mL bottle back under the filter holder.

      d.   Filter the sample  into the 500-mL bottle until the bottle is full.

      e.  Transfer  filtered sample into aliquot containers 1,  3,  and  4 (previously labeled) after first rinsing containers with 10 to
          15 mL of filtered sample (as described in "c").

      f   If necessary return the 500-mL bottle to the filtration apparatus  and collect additional filtered sample [about 700 mL of
         filtered sample is required for aliquots 1 (250 mL), 3 (250 ml),  and 4 (125 mL)].
        Table 15. Filtration Procedure Originally Used by Field Personnel, Eastern Lake Survey - Phase I.

                                                     40

-------
   Process 1.  (Raw, unfiltered sample - no treatment )

             a.  Rinse two aliquot 3 bottles with sample (4-lrter deionized water) and discard rise water

             b.  Fill with fresh sample and seal.

             c.  Label samples as A and B

   Process 2.a (Filtered sample, filter holder rinsed with acid and then with deionized water.)

             a.  Assemble filtration apparatus, rinse once with 5% nitric acid, followed by three deionized water rinses.

             b.  Insert filter membrane, rinse once with nitric acid, followed by three deionized water rinses.

             c  Filter 300 to 400 ml deionized water sample into waste container

             d.  Replace filter  membrane with new membrane Rinse three times with deionized water and once with 20 ml
                sample.

             e.  Place clean aliquot 5 bottle (500 ml) under filter holder Filter 20 ml sample into bottle. Rinse bottle with this
                initial portion, then discard rinse. Filter and collect 500 ml of sample.

             f.  Transfer filtered sample to two aliquot 3 bottles.  Label as C and D

   Process 3.  (Filtered sample, filter holder rinsed with deionized, water only.)

             a  Assemble filtration apparatus using a new filter  holder.  Rinse three times with deionized water

             b.  Insert filter membrane. Rinse three times with deionized water, followed by one 20-mL rinse with sample.

             c.  Place a clean aliquot 5 bottle (500 mL) under filter holder. Filter 20 mL sample into bottle. Rinse bottle with
                this initial portion, then discard rinse. Filter and collect 500 mL sample.

             d.  Transfer filtered sample to two aliquot 3  bottles. Label as E and F.
    Step d of process 2 performed at Duluth field station only
Table 16. Field Laboratory Filtration Procedure Used in Nitrate Contamination Experiment, Eastern Lake Survey
- Phase I.
for nitrate at  EMSL-LV. The Lexington samples were
also split between EMSL-LV and Western Washington
University (WWU) for analysis. Detection limits for
nitrate at EMSL-LV and WWU were 0.001 and 0.002 mg
L~1, respectively.

A comparison of  the nitrate data from EMSL-LV and
WWU is given in Table 18. The data from EMSL-LV indi-
cate that the nitrate  concentrations in  deionized
water used to prepare the field laboratory blank sam-
ples were 0.005 mg L'1 or less in 28of 30 samples. The
two exceptions (both at Duluth) had concentrations
less than or equal to 0.01 mg L~1  nitrate. With one
exception, the nitrate concentration in samples pre-
pared  by  process 3 (filtered sample, filter holder
rinsed with deionized water only) was 0.005 mg L"1 or
less. That exception (Duluth 5) contained less than
0.02 mg L"1 nitrate.

Both the WWU and EMSL-LV data indicated that proc-
ess 2 (filtered sample, filter holder rinsed with HNO3,
then with deionized  water) produced field laboratory
blank  samples with  high concentrations of nitrate
(values ranging from 0.004 to 1.65 mg L"1 with an aver-
age of 0.2 = 0.5 mg L"1). Again, field blank samples
filtered using  process 3  contained  low concentra-
tions of  nitrate.  WWU reported  no nitrate values
above their instrumental detection limit (0.002 mg L~
11 nitrate) using process 3.

By making a comparison of the results using process
2 with those using process 3, it was concluded that
the nitric acid rinse of filter holders and filter mem-
branes was responsible for the nitrate contamination
seen  in field blanks (and, by  implication, in  routine
and duplicate  samples). By using  process 3 to pre-
pare aliquot 3, nitrate concentrations in field blanks
were  reduced to levels less than 0.01 mg L~1 for the
remainder of the ELS-I.

It should be noted that the sources of contamination
can be traced to the deionized water, sample contain-
ers, and filtration apparatus, which contributed up to
1.65 mg L~1 of nitrate to field laboratory blank sam-
ples in this experiment. Two raw unfiltered samples
(Duluth 8-1 and 9-1) contained 0.009 and 0.010 mg L~1
nitrate, respectively, and the two  corresponding fil-
tered  samples (deionized water only, Duluth 8-3 and 9-
                                                    41

-------
Dates3 and Times'1 of Sample Processing
Sample
1
2
3
4
5
6
7
8
9
10
11e
12f
Lexington0
10/28/84,1000
10/28/84,1000
10/28/84,1100
10/28/84,1100
10/28/84,1200
10/28/84,1200
10/28/84,1300
10/28/84,1300
10/28/84,1400
10/28/84,1400
10/28/84,1400
10/28/84,1400
Rhinelander0
10/29/84,1325
10/29/84,1330
10/29/84,1420
10/29/84,1420
10/29/84,1530
10/29/84,1530
10/29/84,1630
10/29/84,1630
10/29/84,0900
10/29/84,0900
—
—
Duluth"
10/28/84
10/28/84
10/28/84
10/28/84
10/28/84
10/28/84
10/28/84
10/28/84
10/28/84
10/28/84
—
—
3 Day the sample was collected and processed (filtered).
b Time the water sample was collected from the Millipore
0 Times and dates
d Time unknown.
e This sample was
' This sample was
of sample collection. All samples were

taken from the still in the EPA regional
system.
processed on 10/30/84.

lab in Lexington, Massachussetts.




taken from the deionized water spigot in the EPA regional lab in Lexington, Massachussetts.
Table 17. Description of Field Laboratory Blank Samples Collected in Nitrate Contamination Experiment, Eastern
Lake Survey - Phase I.
3) contained 0.003 and 0.005 mg Lr1 nitrate. The ali-
quot bottles (rather than the deionized water) were
the most likely source of the slight contamination in
Duluth samples 8-1  and 9-1. Similarly, a raw unfiltered
sample (Duluth 5-1) contained 0.005 mg L~1 nitrate
while a  corresponding  filtered aliquot (deionized
water only, Duluth 5-3) contained 0.018 mg Lr1 nitrate.
In  this case, the contamination was probably from
either the aliquot bottle or the filtration apparatus.
Both of these examples indicate the extreme  care
that was necessary to reduce nitrate contamination
to  less than 0.005 mg L~1. The ELS was about halfway
completed (November 1, 1984) before  the cause  of
nitrate contamination was isolated  and corrected.
The QA manager immediately asked the contract lab-
oratories to  reanalyze nitrate in aliquot 5, the unfil-
tered aliquot.  Unfortunately, most aliquot  5 nitrate
determinations were performed well outside the max-
imum allowable  holding time of seven days. After
extensive statistical review, it was decided to replace
all nitrate data processed on or before November 1,
1984, with the aliquot 5 data.


Evaluation of Quality Assurance Data

The QA program used a combination of blanks, dupli-
cates, and audit samples to provide an external check
on the quality of the ELS-I data and to allow early
detection of problems in sample collection and analy-
sis. Analytical data and associated QA and QC infor-
mation were collected to define the overall and ana-
lytical  within-batch precision and  the  overall and
analytical among-batch precision for each parameter,
the normal  background contamination that occurred
during the  sampling and analytical process, and,
where possible, the bias among the  laboratories that
performed the analyses. This section presents a sum-
mary of the QA results.  Permutt and Pollack (1986)
provided additional statistical evaluation  of  ELS-I
data which  are included in this report as Appendix A.
Blank Data

Both field blanks and analytical laboratory (calibra-
tion or reagent) blanks were analyzed during the ELS-
I. Field blanks were analyzed for turbidity at the field
laboratory and for 21 physicochemical parameters at
the contract analytical laboratory. The contract labo-
ratories used calibration and reagent blanks to con-
trol their background levels and to calculate their
instrumental detection limits. These were reported
weekly for each parameter except pH, conductance,
ANC, and BNC.
                                                42

-------
Sample
1
2
3
4
5
6
7
8

9
10
11
12

1
2
3
4

5
6
7
8

9
10
1
2
3
4
5
6
7
8
9
10

Process
1
1
1
1
1
1
1
1

1
1
1
1

2
2
2
2

2
2
2
2

2
2
3
3
3
3
3
3
3
3
3
3
a All samples were analyzed within 7 days of
b ND = not detected (less than 0.002 mg L"1

Duluth
EMSL-LV
0003
0.005
0005
0.005
0.005
0.003
0.003
0009
0.009
0.010
0.003
—
—

0.033
0.033
0.033
0036

0059
0.224
0.051
0.033
0.035
0.020
0.171
0.005
0.005
0003
0.003
0018
0.005
0.003
0.003
0.005
0.002
preparation.
N03-).
N03- (mg I/1)3
Rhinelander
EMSL-LV
0.003
0.003
0.003
0003
0.003
0.003
0.003
0.003

0.003
0.003
—
—

0.006
0.014
0.006
0.008
0.007
0.013
0.023
0.015
0.186

0.006
0.006
0.003
0.003
0.003
0003
0.003
0.003
0.002
0 003
0.003
0.003


Lexington
EMSL-LV
0001
0.001
0.001
0.001
0.001
0.001
0.001
0.005

0.001
0.005
0.044
0.001

0.042
1.130
0.146
0.046

0.-44
0.088
0.104
0.025

0023
0.008
0.001
0.001
0001
0.001
0.001
0.001
0001
0.001
0.001
0.001


Lexington
WWU
NDb
ND
ND
ND
ND
ND
—
ND

ND
ND
—
0.046
0.046
0.077
1.65
0.128
0.062

0.046
0.126
0.108
0.032

0.030
0.004
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND

Table 18. Nitrate Concentrations in Field Laboratory Blank Samples in Nitrate Contamination Experiment,
Eastern Lake Survey - Phase I.
                                                  43

-------
Internal laboratory blanks were not  routinely  ana-
lyzed  at the field laboratories because the limited
number of analyses performed there would not have
provided sufficient QA/QC information to justify the
extra  effort and expense. There were few problems
achieving detection limits for measurements made in
the field laboratories. In addition, pH and DIG meas-
urements  are highly  variable  in blank  samples
because of the tendency of blanks to absorb CO2 from
the air.

In total, 245 field blanks  were collected during the
ELS-I. The data were analyzed to provide  an overall
estimate of  the normal background contamination
that occurred during sampling and analysis and to
identify and correct any  significant contamination
problems as they occurred. A statistical evaluation of
the verified  data yielded  a system decision limit,  a
system detection  limit, and an estimated instrumen-
tal detection limit for each variable (Appendix A). The
system decision limit represents the lowest  instru-
ment  signal that  can be distinguished from  back-
ground at a =  0.05. The system detection limit is
based on the reproducibility  of the field blank analy-
sis; it  represents the lowest concentration that can be
measured above the system decision limit.

The estimated instrumental  detection limit was cal-
culated from the  reported calibration  blank data.  It
was used  to provide a preliminary QA check on the
analytical results provided by the contract laboratory.
The system  detection  limit should be comparable to
the instrumental detection limit reported by the con-
tract  laboratory if there is no variability added as a
result of sample  collection,  processing,  and ship-
ment.

Table 19 provides a summary of field and  laboratory
blank data for 20  parameters from the ELS-I.  Values
for individual laboratories are listed in Appendix B.
The  required  instrumental  detection limits (IDL's)
were  achieved  in the  contract laboratories  for  all
parameters except Fe and NO3~.  However, for some of
the parameters the system decision limit is consider-
ably higher than the required IDL; this is possibly due
to background sources of contamination.

The  evaluation of blank data demonstrated that
detection  limit goals were generally achieved. How-
ever, to interpret the data, results from the field blanks
must  be   taken  into  consideration.  Although
extremely low detection limits can be achieved in the
laboratory, they are of  little value in defining usable
data when they are lower than the background from
sample collection and handling. The system decision
limit  and  system detection  limit must therefore be
considered  as the real  limits for data interpretation.

Duplicate Data

Data  for  estimating overall  within-batch precision
were  obtained using field duplicate samples. These
were processed by the field laboratories and were
analyzed with the routine samples and field blanks at
the contract  laboratories. The identity of the field
duplicates was not disclosed to the contract labora-
tories. The field and contract laboratories also per-
formed duplicate chemical analyses on one sample
per batch as a QC check on analytical within-batch
precision.

In total, 125 field duplicates were collected during the
ELS-I. The estimates of overall within-batch precision
obtained from field duplicate data include the effects
of sample collection, processing, shipping, and anal-
ysis, but do not include the effects of among-batch
variation that may have been caused by  day-to-day
differences such as different calibration curves.

Within-batch  precision was  calculated as the root
mean square of percent relative standard deviation
(%RSD) to estimate the 'average' %RSD over  the
range of values for routine/duplicate pairs. The esti-
mated precision values are directly comparable to the
intralaboratory precision goals  which  are  also
expressed as %RSD.

A quantitation limit was calculated for each parame-
ter to distinguish values which are expected a priori
to have greater and more variable relative  errors (i.e.,
values close to the detection limit) from values which
are expected to have smaller and more consistent rel-
ative precision (values much greater than the detec-
tion limit). The quantitation limit  should not be con-
fused with  instrumental detection limits, system
detection  limits,  or system decision limits, all of
which serve to constrain the actual size of the usable
data set. All of the verified data, including values
below the quantitation limits, were considered in the
validation process;  quantitation  limits were calcu-
lated solely for the purpose of estimating the charac-
teristic relative precision. A more detailed discussion
of the statistical approach for evaluating  within-
batch and among-batch  precision is provided in
Appendix A.

Overall  within-batch  precision  estimates  for  23
parameters from the ELS-I are presented in Table 20.
The three types of pH measurements were evaluated
in terms of absolute  rather than relative standard
deviation. The estimated overall  within-batch preci-
sion for pH in all four laboratories was within a range
of 0.05 to 0.09 pH unit.  No single laboratory had uni-
formly high or low precision; however, air-equilibrated
pH data fromVersar and initial ANC and BNCpH data
from USGShad higher variability than the other meas-
urements (see Appendix C).

The measured concentrations of some parameters
(Ca, Mg, Na, SO42~) were far above both the reported
and the required  instrumental detection  limits. The
results  indicate that overall within-batch precision
was better than the intralaboratory precision goal for
each of these parameters. Most or all of these pairs
                                                 44

-------
Parameter
Al, mgL~1
total extractable
total
ANC./^eq L~1
Ca, mg ~1
C1"1, mgL"1
Conductance, /*S cm'1
DIG, mg L~1
air-equilibrated
initial
DOC, mg L"1
F~, total dissolved,
mg L~ '
Fe, mg L"1
K, mg L"1
Mn, mg L'1
Mg, mg L"1
Na, mg L"1
NH4+,mgL"1
N03", mg L"1
all batches
aliquot 3
aliquot 5
P, total, mg L"1
Si02, mg L~1
S042", mg L"1
a See Drouse et al. (1986).
Instrumental Detection
Required3 Reported3
0.005 0.003
0.005 0.003
_e 	 e
0.01 0.01
0.01 0.01
--' 0.0
0.05 0.03
	 e 	 e
' 0.1 0.1
0.005 0.002
0.01 0.02
0.01 0.00
0.01 0.01
0.01 0.00
0.01 0.00
0.01 0.01
0.005 0.006
0.002 0.001
0.05 0.02
0.05 0.02

11 Estimated instrumental detection limit (IOL) = (2{P95-P50)), where P95
(median), of laboratory calibration blank samples (see Appendix A).
Limit
Estimated11
0.013
0.004
e
0.00
0.05
1.6
0.14
e
0.3
0.008
0.02
0.00
0.01
0.00
0.10
0.03
0.037
0.003
0.06
0.09

= 95th percentile, and P5
System
Detection
Limit0
0.014
0.050
11
0.06
0.13
1.4
0.20
0.38
0.4
0.008
0.04
0.03
0.02
0.02
0.06
006
0.760
0.046
1.782
0.014
0.21
0.13

o = 50th percentile
System
Decision
Limit"
0.008
0.030
6.9
0.03
0.08
1.3
0.28
0.42
0.4
0.005
0.02
0.02
0.01
0.01
0.03
0.04
0.389
0.023
0.919
0.008
0.11
0.09


0 System detection limit = (2(P95-P50)), where P95 = 95th percentlle, and P60 = 50th percentlle (median), of field blank
samples (see Appendix A).
d System decision limit = (P95) =
95th percentile of field blank samples
(see Appendix A).


e Measurements of laboratory calibration blanks not required.
1 For conductance, the mean of six
0.9 /uS cm"1.
nonconsecutive laboratory calibration
blank measurements was
required to be less
than
Table 19. Instrumental Detection Limits, System Detection Limits, and System Decision Limits for 20
Parameters, Eastern Lake Survey - Phase I.
                                                  45

-------
Pairs With Mean>0


Parameter
A1, total extractable mg L~1
all values
J<= 0 010
Y>0 010
A1, total, mg L~1
all values
x<= 0010
x"> 0 010
ANC,Meq L"1
Ca, mg L"1
C1~, mg L"1
Conductance, /*S cm'1
DIG, mg L"1
air-equilibrated
initial
DOC, mg L"1
all values
7<= 5
x">5
F"
Fe, mg L~1
K, mg L"1
Mg, mg L~1


RDLa nb

0.005 113
69
44

0005 125
9
116
5 121
0.01 125
001 125
— e 125

0.05 124
0.05 125

0 1 125
66
59
0005 125
0.01 118
0.01 125
001 125
Root
Mean
Square
of %RSDC

46
59
21

35
55
33
20
23
20
1.9

12
69

9.8
6.7
12
8.1
97
4 7
2.3
Quanti-
tation
Limit
(10sB)d

0 039



0.277


56.6
0.14
0 38
90

0.56
1 09

2 5


0.034
0.12
0.37
0 04
Pairs With Mean >10sBc


nb

9
0
9

1
0
1
90
125
85
125

94
85

105
46
59
62
32
82
125
Root
Mean
Square
of %RSDC

11
	
11

24
	
24
10
2.3
17
1 9

5.0
3 7

10
5.6
12
8 9
10
3 7
2 3
Overall
Intra-
Laboralory
Precision
Goal
(%RSD)C


20
10


20
10
10
5
5
3

10
10


10
5
5
10
10
10
Table 20. Overall Within-Batch Precision Estimated From Field Duplicate Data and Field Blank Data for Measurements of 23 Parameters, Eastern Lake
Survey - Phase I.

-------


Pairs With Mean>3

Root
Mean
Square
Parameter
Mn.mg L"1
Na, mg L"1
NH4 + , mg L"1
N03~, mg L"1
all batches
aliquot 3
aliquot 5
PH
air-equilibrated
initial ANC
initial BNC
P, total, mg L"1
all values
x<= 0.010
x>0.010
Si02, mg L"1
S042~, mg L"1
RDL3
001
0.10
0.01

0.005



::|

0.002


0.05
0.05
n" of%RSD°
82
125
113

113
45
68

125

122
57
65
120
125
57
4.2
34

443
213
544

51

37
44
30
44
6.6
Pairs With Mean>10sB°
Quanti- Root
tation Mean
Limit Square
(lOsu)" nb of%RSDc
0.093 6 11
0.20 121 43
0.56 0

0 894 5 60
0.163 9 45
1.143 0

"I ::f ::|
-t -f -f

0.037 4 9.7
0
4 97
2.34
1.68 115 6.5

Overall
Intra-
Laboratory
Precision
Goal
(%RSD)C
10
5
10

10


u
0.010
0.010
0 01n


20
10

5
aRDL = required detection limit (in applicable units).
bn = number of routine/field duplicate
pairs.




°%RSD = percent relative standard deviation.
dsB = standard deviation of field blank
aliquot 5, n = 145.
measurements

(see Appendix A) with n = 245

except for total

e For conductance, the mean of six nonconsecutive calibration blank measurements was required to be
1 Not applicable.

9 Root mean square of absolute standard deviation (pH
h Absolute precision goal (pH unit).


unit).




A1, n = 244; NH4 + , n = 243; N03~, aliquot 3, n = 99,

less than 0.9/uS CM"1.



and N03~,





Table 20. Overall Within-Batch Precision Estimated From Field Duplicate Data and Field Blank Data for Measurements of 23 Parameters, Eastern Lake
Survey - Phase I (Continued).

-------
had a mean concentration greater than the quantita-
tion limit (see Table 20.

Other chemical parameters (Mn, Fe, total extractable
Al, total Al, Cf, SiO2, and total P) were characterized
by low concentrations (at or below the quantitation
limit) in many of the sample pairs. The relative preci-
sion of the data is biased high due to the increase in
absolute variability of  the  measurement near  the
instrumental detection limit (Appendix A). Even if the
absolute variability in the chemical analysis was uni-
form at all concentrations, an estimate of the true pre-
cision would be difficult to ascertain over the entire
range of values when it Is expressed as a percentage
of the mean concentration.

The estimated precision for a!! duplicate data with
concentrations greater than zero is also provided in
Table 20. For these parameters, all pairs with concen-
trations  above the quantitation  limit  had  overall
within-batch precision that was better than the over-
all intralaboratory precision goal, except for_tptal
extractable Al, total Al, Cl~, DOC (for pairs with X >5
mg L~1), Mn, NO3~2,  and SO4~2.  The data user should
again consider the estimated quantitation limit when
interpreting the precision of data.

Duplicate analyses, designated as trailer duplicates,
were also performed for all measurements made in
the field laboratory. One trailer duplicate was ana-
lyzed per sample batch. This duplicate analysis was a
QC step to assure consistent measurements within
the field laboratory. A QC limit of ±0.1 pH unit was
established for pH measurements, and a QC limit of
±10 percent was established for DIG, true color, and
DIG measurements. If the QC limits were  exceeded,
the laboratory was required to  analyze  a  second
duplicate. If the limits were still exceeded after rean-
alysis, the laboratory  supervisor  made  additional
efforts to identify and correct the problem until it was
resolved.

Precision estimates from field duplicate and trailer
duplicate data are presented in Table 21.  For trailer
duplicates, the observed analytical within-batch pre-
cision  for each parameter was within the analytical
intralaboratory precision goal established  for  the
ELS-I.  The overall within-batch  precision for field
duplicates was within the overall intralaboratory pre-
cision  goal for pH, but was not for DIG, true color, or
turbidity. There was little bias apparent among field
stations for these analyses, based on gross observa-
tion of the data. Bias is discussed in greater detail in
Appendix A.

The  contract laboratories were required to  analyze
one sample per batch in duplicate. For each parame-
ter, the analytical within-batch precision was required
to be within the analytical intralaboratory precision
goal. Analytical  within-batch  precision  estimates
from contract analytical laboratory duplicate pH data
are presented in Table 22. The results for pH  indicate
that precision was within a range of ±0.04 to 0.08 pH
unit. These data may have been biased because the
laboratory analyst knew which samples were dupli-
cates and that the QC process required that the preci-
sion  goal be achieved. However,  the results are an
indication of the precision that was achieved within
each laboratory with the method QC requirements.

Precision estimates  for the other 20 parameters from
contract laboratory duplicate data are also shown in
Table 22. The estimated analytical within-batch preci-
sion  for pairs with mean >10sB was better than the
analytical  intralaboratory precision goal for each of
the parameters except total extractable Al>0.010 mg
L~1, CI~,Mn, SO42", and conductance.

In general, the overall and analytical intralaboratory
precision goals were achieved between values for
routine/field duplicate pairs as well as between con-
tract laboratory duplicate pairs.  The  exceptions  to
these results are discussed in  Section 5. Thus, esti-
mated overall within-batch precision was considered
to be adequate to meet the DQO's established at the
beginning of the ELS-I.

Audit Sample Data

Audit sample data can be used to obtain estimates of
among-batch precision and interlaboratory bias (see
Table 3). Six types of audit samples (two field natu-
rals, two field synthetics, and two laboratory synthet-
ics) were analyzed for 23 parameters during the ELS-I.
Field naturals and field synthetics were handled in
the same manner as routine samples (see Section 2)
to estimate the overall among-batch precision includ-
ing the effects of sample processing. Laboratory syn-
thetics were prepared as processed aliquots by the
analytical support laboratory; these were relabeled at
the field labratory and incorporated with the sample
batch to estimate analytical among-batch precision.

The natural audit samples were also used to estimate
relative interlaboratory bias by comparing measured
values from the contract laboratories with theoretical
values and with referee  laboratory measurements.
The  synthetic audits were used to provide  informa-
tion  on absolute interlaboratory bias which  required
that  they be prepared from solutions of known com-
position. Examination of both the theoretical values
and the measured values for synthetic audit  samples
indicated  that the actual  sample composition may
have differed on different days,  and  that measure-
ment imprecision and interlaboratory bias were small
by comparison  (see  Appendix A).

One type of field  natural sample (FN2, from  Big
Moose Lake in the Adirondack Mountains, New York)
was representative  of lakes which are sensitive to
acidic deposition; the other type  (FN3, from  Lake
Superior near Duluth, Minnesota)  represented sys-
tems with high ANC. In total, 41  FN2 samples and 7
FN3 samples were used during the survey. Two con-

-------









Parameter
PH
DIG, mg I/1
True color, PCU
Turbidity, NTU




Overall and
Analytical
Intralaboratory
Precision
Goal
(%RSD)a
0.1°
10
10
10
Overall
Within-Batch
Precision
(Field Duplicates)



Root Mean
Square of
nb %RSD"
124 0.04d
123 16
125 22
125 19
Analytical
Within-Batch
Precision
(Trailer Duplicates)



Root Mean
Square of
nb %RSDa
93 0.01°
116 4.6
118 1 5
117 8.4
aRSD = relative standard deviation.
bn = number of duplicate pairs.
Absolute standard deviation (pH

unit).




dRoot mean square of absolute standard deviation (pH unit).
Table 21. Overall and Analytical Within-Batch Precision Estimated From Field Duplicate and Trailer Duplicate
Data for Measurements of Four Parameters, Eastern Lake Survey - Phase 1.
centrations of synthetic audit samples were also pre-
pared including 36 field highs, 21 laboratory highs, 42
field lows, and 43 laboratory lows.

Table 23 summarizes the mean sample composition
and the overall amongbatch precision for 23 parame-
ters estimated from FN2 and FN3 data. Mean values
were less than quantitation limits  for 7 parameters
measured in FN2 samples and for 7 parameters mea-
sured in FN3 samples. For ANC, air-equilibrated DIG,
and initial DIG, estimated precision was better (i.e.,
%RSD was lower) for FN3 samples which had means
greater than quantitation limits than for FN2 samples
which had means less than quantitation limits. This is
consistent with our expectations.  The precision of
total extractable Al, total Al, and DOC measurements
was slightly better with FN2 samples than with FN3
samples; for those  parameters, mean  FN3 values
were less than quantitation limits. A test of signifi-
cance of the differences was not performed because
of the small number of FN3 samples used.

For both FN2 and FN3 samples, means were less than
quantitation limits for Fe, Mn,  NH4+, and total  P, and
means were greater than quantitation limits for the
remaining 10 parameters.  Quantitation limits were
not applicable for the 3 pH measurements. For the
field natural audit samples with means greater than
quantitation limits, the overall interlaboratory preci-
sion goal (estimated as twice the overall intralabora-
tory precision goal, see Table 20) was achieved for all
parameters except total extractable Al, total Al, and
Cl~ in FN2 samples.

Tables 24 and 25 summarize the mean values for sam-
ple composition  and the  overall and  analytical
among-batch percision  estimated from  measure-
ments of synthetic audit samples.  For the field high
synthetics, the mean value for total Al was less than
the quantitation limit; all other parameters (except
pH) had  means above the quantitation limits. The
overall interlaboratory precision goal was not met  for
total Al, as was expected, and was also not met for Cl~
,  air-equilibrated and  initial DIG,  DOC, Mn, NH/,
NO3~, and SiO2. Means for laboratory high synthetics
were greater than quantitation limits for all parame-
ters (not applicable for pH). Analytical interlaboratory
precision goals were met for all parameters except
total  Al, Ca (Lot 4), Cl- (Lots 5 and 6), conductance, air-
equilibrated  and initial DIG, Mn,  NH4 + ,  NO3~, and
SiO2.

For the field low synthetic audit samples, the mean
values for total Al, Cl", DOC, K, Mn, NH4', NO", total P,
and SiO2 were less than the quantitation limit; and the
mean of Ca (Lot 4) values was equal to  the quantita-
tion limit. The overall interlaboratory precision goals
were not met for any of these 10 parameters, with the
exception of K and Mn. For the other 13 parameters,
the overall interlaboratory precision goals were met
                                                49

-------
Pairs With Mean<0


Parameter
A1, total extractable mg L"1
all values
x"<= 0 010
x">0 010
A1, total, mg L~1
all values
'x<= 0 010
I> 0.010
ANC, Meq L"1
Ca, mg IT1
Cr, mg L"1
Conductance, /uS cm'1
DIG, mg L"1
air-equilibrated
initial
DOC, mg L"1
all values
7<= 5
>f>5
F~, total dissolved,
mg L"1
Fe, mg L"1
K, mg L~1
Mg, mg L"1


RDLa

0.005



0.005


5
0.01
0.01
-e

0.05
0.05

0.1



0.005
0.01
0.01
0.01


nb

123
22
101

126
8
118
116
123
127
125

126
127

126
85
41

127
120
125
121
Root Quanti-
Mean tation
Square Limit
of%RSD° (10sB)d

13 0.038
14
12

19 0.012
57
12
30 56. 6f
0.88 004
23 0.11
11

7.1 0.60
3.7 0 54

68 10
8.1
2.4

2.5 0015
14 0.05
18 0.04
0.64 0 01
Pairs With


nb

46
0
46

54
0
9
86
123
124
5.4

94
113

114
73
41

123
101
121
121
Mean <10sBd
Root
Mean
Square
of %RSD°

18
—
18


—
5.4
2.1
0 88
7.1
123

2.2
3 5

2.4
2.5
2 4

2.5
4.3
1 5
0 64

Analytical
Intra-
Laboratory
Precision
Goal
(%RSD)C


20
10


20
10
10
5
5
101

10
10


10
5

5
10
5
5
Table 22. Analytical Within-Batch Precision Estimated From Contract Laboratory duplicate Data and Calibration Blank Data for Measurements of 23
Parameters, Eastern Lake Survey - Phase I.

-------
Pairs With Mean>0 Pairs With Mean >10sBc

Parameter RDLa
Mn, mg L"1 0.01
Na, mg L~1 0.10
NH4 + ,mgL"1 0.01
N03~, mg L"1
all batches 0.005
aliquot 3
aliquot 5
pH
air-equilibrated -g
initia ANC -g
initia BNC --g
P, total, mg L~1
all values ' 0.002
7 = 0.010
x 0.010
Si02, mg L"1 0,05
S042", mg L"1 0.05
aRDL = required detection limit (in applicable units).
bn = number of contract laboratory duplicate pairs.
C%RSD = percent relative standard deviation.
dsB = standard deviation of measurements of contract laboratory
N03~, aliquot 5, n = 72.

Analytical
Intra-
Root Quanti- Root Laboratory
Mean tation Mean Precision
Square Limit Square Goal
n" of%RSD° (10sB)° n° of%RSDc (%RSD)°
102
123
121

122
51
71

127
127
127

125
30
95
127
127



calibration blanks

e For conductance, the mean of six nonconsecutive calibration blank measurements
1 Quantitation limit calibrated using field blank data (see text).
9 Not applicable.
h Root mean square of absolute standard deviation (pH unit).
' Absolute standard deviation (pH unit).




21 0.030 73 17 10
13 0.10 121 0.96 5
7.1 0.15 54 2.3 5

13 0.157 86 3.6 10
20 0.110 39 3.7
3.9 0.185 46 3.3

0.08h --g --g --g 0.01J
0.04h "8 -Q --Q 0.05
o.07h --g --§ -g o.os1

13 0.008 104 9.9
20 9 8.9 20
10 95 10 10
2.5 2.37 64 2.2 5
11 1.18 124 11 5



with n = 127 except for DIC, n = 125; Mn, n = 126; N03", aliquot 3, n = 53; and

was required to be less than 0.9/iS cm"1.




Table 22. Analytical Within-Batch Precision Estimated From Contract Laboratory duplicate Data and Calibration Blank Data for Measurements of 23
Parameters, Eastern Lake Survey - Phase I (Continued).

-------

Parameter
A1, mg I/1
total extractable
total
ANC.neq L~1
Ca, mg L"1
C1", mg L"1
Conductance, /iS cm'1
DIG, mg L"1
air-equilibrated
initial
DOC, mg L"1
F", total
dissolved, mg L"'
Fe, mg L"1
K, mg L"1
Mg, mg L"1
Mn, mg L~'
Na, mg L"1
NH4' , mg L"'
N03~, mg L"1
all batches
aliquot 3C
aliquot 5d
PH
air-equilibrated
initial ANC
initial BNC
P, total, mg L"1
Si02, mg L"1
SO/", mg L"1
FN2 (n =
Mean
0 182
0 305
2.4e
1 91
0 61
26.7
0 19e
0.42e
33
0 077
0 02e
0 49
0 35
0.07e
0 67
0 063
1 425
1 473
1.403
5 18
507
5.08
0.0026
4.33
6.95
41)
%RSDa
34
28
258
9.2
51
3.6
GO
27
9 3
4.0
66
4.3
2 7
26
5 1
54
5 5
3.0
5 8
0.271
0041
0.051
144
8.1
7 3
FN3 (n
Mean
0 002"
0 021e
8482
13 28
1.39
96.4
9.68
9.93
1.4e
0 035
OOO3
0 49
2 79
0 00e
1.30
0.01e
1.401
1.422
1.385
8.23
776
7.79
0 001e
2.72
3.26
= 7)
%RSDa
87
123
2 3
55
2 0
1.7
3 8
4 2
9 7
5 2
137
12
3 6
362
6 4
130
58
3 6
8 5
0 081
O1
0'
143
6.0
3.3
Quantitation
Limit
(10sB)b
0 039
0 277
56 6
0 14
0 38
9.0
0 56
1 09
2 5
0 034
0 12
0 37
0 04
0093
0 20
0 56
0984
0 163
1 143
—9
"9
"9
0 037
2 34
1 68
a Percent relative standard deviation.
bsB = standard deviation of measurements of field
0 Aliquot 3 samples after filtration protocol change
d Aliquot 5 samples (FN2, n = 28; FN3, n = 3)
e Mean less than quantitation limit
1 Absolute standard deviation.
g Quantitation limits not applicable


blank samples
(FN2, n = 13;


(see Appendix A).
FN3n = 4).










for pH measurements.
Table 23. Overall Among-Batch Precision Estimated From Field Natural, Lot 2 (FN2, Big Moose Lake) and Field
Natural, Lot 3 (FN3, Lake Superior) Audit Sample Data for Measurements of 23 Parameters, Eastern Lake Survey
- Phase I.
                                                 52

-------
Parameter
A1,mgL"1
total extractable10
total
ANC,/ueq L"1
Ca, mg L"1
Lot 4
Lot 5 and 6
Cr, mg L"1
Lot 4
Lot 5 and 6
Conductance, ^S cm'1
DIG, mg L"1
air-equilibrated
initial
DOC. mg L"1
F", total dissolved, mg L"1
Fe, mg L~lc
K, mg L"1
Mg. mg L~1
Mn, mg L~1
Na. mg L'1
NH^, mg L"'


X
0 199
476.7
1.66
1 95
366
4.00
104 5
4.72
594
100
0445
—
3 03
2 37
1 19
11 84
1 26
Overall Among-Batch
(Field Highs)
%RSDa
d 30
17
66
8 2
40
18
4 0
27
33
20
94
—
63
38
13
8 7
13
Precision
Quantitation.
Limit
(10sB)b
0.039
0.277
57.6
0.14
038
90
056
1 09
2 5
0034
0.12
0 37
0 04
0 093
020
0 56
Analytical Among
(Laboratory Highs)
-Batch

Y %RSDa
0.161
0.194
485.9
1 55
1.96
3.39
581
104.3
4.79
5.94
103
0.435
0.18
3 02
236
1.39
11 87
1 34
19
34
20
12
5.1
7.1
92
39
27
32
16
2.8
11
5 0
34
29
8 8
13
Precision
Quantitation
Limit
(10sB)b
0.038
0012
56. 69
0 04
0.11
5.4
0.60
0.54
1 0
0015
0 05
0 04
0 01
0 030
0 10
0 15
Theoretical
Value
-h
0.19
-h
1.54
2.39
2.72
4 22
-h
--h
3.10
10.0
0452
0 15
2.97
2 43
1 50
12.41
1 25
Table 24. Mean Measured Values, Overall and Analytical Among-Batch Precision Estimates, and Theoretical Concentrations of High Synthetic Audit
Samples, Eastern Lake Survey - Phase I.

-------



Parameter
N03", mg L~1
(all batches)
PH
air-equilibrated
initial ANC
initial BNC
P, total, mg I/1
Si02, mg L"1
S042", mg LM
a Percent relative standard deviation.
Overall Among-Batch Precision Analytical Among-Batch Precision
Quantitation Quantitation
(Field Highs) Limit (Laboratory Highs) Limit
Y %RSDa (10sB)b X" %RSDa (10sB)b

1 816 37 0 894 1 431 47 0 157

7.72 0 54e — g 7.87 0 15e -g
7.08 0.25e -g 7.05 0 26e --g
7.12 027e -g -h
0057 23 0037 0061 16 0008
9.23 14 2.34 9.42 15 237
14.39 5.6 1.68 14.47 4.8 0.18



Theoretical
Value

1.707

--h
-h

0.075
10.70
14 09

bsB = standard deviation of measurements of field blank or laboratory blank samples (see Appendix A).
0 Laboratory synthetic audit samples
d Mean less than quantitation limit.
only (see Section 4).



e Absolute standard deviation (pH unit).
1 Quantitation limits not applicable for pH measurements.
9 Quantitation limit calculated using field blank data.
h Theoretical concentration not available.
Table 24. Mean Measured Values, Overall and Analytical Among-Batch Precision Estimates, and Theoretical Concentrations of High Synthetic Audit
Samples, Eastern Lake Survey - Phase I (Continued).

-------
Parameter
A1, mg I/1
total extraclable10
total
ANC,/ieq L"1
Ca, mg L"1
Lot 4
Lot 5 and 6
CT, mg L"1
Lot 4
Lot 5 and 6
Conductance, ^S cm'1
DIG, mg L~1
air-equilibrated
initial
DOC, mg L'1
F~. total dissolved, mg L~1
Fe, mg L"'°
K. mg L"1
Mg, mg L"1
Mn, mg L"1
Ma, mg L"1
NH4 + , mg L"1


X

—
0 023
112

0.14d
1 17

033d
037d
19.0

1.35
1.65
10d
0.042
~
023d
042
1.091
2.71
1.16d
Overall Among-Batch
(Field Highs)
%RSDa

—
d 22
7.4

19
7.1

71
22
68

8.9
13
48
21
—
19
73
10
5.2
27
Precision
Ouantitation
Limit
(10sB)b

0.039
0.277
56.6
0.14


038


9.0

0.56
1.09
2.5
0.034
0.12
0.37
0.04
0093
0.20
0.56
Analytical Among-Batch
(Laboratory
X"

0.018d
0.031
110

0.14
0.16

0.30
0 35
18.7

1.35
1.65
1.2
0.040
0 07
0.23
0.43
0.092
2.69
0.19
Highs)
%RSDa

33
90
9 4

11
8 7

14
31
6.4

11
15
61
50
23
10
4.5
16
9.3
22
Precision
Quantitation
Limit
(10sB)b

0.038
0 012
56 69
004


0.11


5.4

060
0.54
1 0
0 015
005
0.04
001
0.030
0.10
0 15
Theoretical
Value

-f
002
-f

0.13
0 19

0.22
0 34
--f

--f
0.96
1 0
0042
0 06
0 20
0.45
0 10
2 79
0.17
Table 25. Mean Measured Values, Overall and Analytical Among-Batch Precision Estimates, and Theoretical Concentrations of High Synthetic Audit
Samples, Eastern Lake Survey - Phase I.

-------



Parameter
N03", mg I/1
pH
air-equilibrated
initial ANC
initial BNC
P, total, mg L"1
Si02, mg L~1
SO/", mg L"1
Overall Among-Batch Precision Analytical Among-Batch Precision
Quantitation Quantitation
(Field Highs) Limit (Laboratory Highs) Limit
Y %RSDa (10sB)b 7 %RSD" (10sB)b
0 547d 66 0894 0471 37 0.157

7.34 0 14e -f 7.29 0.13e -f
6.87 0.11e --f 686 0.12e -f
6.96 Oe --f 6.93 0.15e -f
0.021" 29 0.037 0022 41 0.008
1.00d 31 2 34 1 02d 20 2.37
227 10 1.68 2.28 4.4 0.18


Theoretical
Value
0.466

-f
--f
-f
0 027
1.07
2 28
3 Percent relative standard deviation.
bsB = standard deviation of
c Laboratory synthetic audit
d Mean equal to or less than
measurements of field blank or laboratory blank samples (see Appendix A).
samples only (see Section 4).
quantitation limit.



e Absolute standard deviation (pH unit).
1 Not applicable


9 Quantitation limit calculated using field blank data.
Table 25. Mean Measured Values, Overall and Analytical Among-Batch Precision Estimates, and Theoretical Concentrations of High Synthetic Audit
Samples, Eastern Lake Survey - Phase I (Continued).

-------
by all  parameters  except conductance, total  dis-
solved  F~, air-equilibrated pH, initial ANC pH, and ini-
tial BNC pH. Means for  laboratory  low  synthetics
were greater than quantitation limits for all parame-
ters except total extractable Al and SiO2 (not applica-
ble for pH)  The analytical interlaboratory precision
goals were met for ANC, Ca (Lots 5 and 6), air-equili-
brated  and initial DIG, total dissolved F", K, Mn, Na,
NH4  ,  and SO4~2; they were not met  by the other 13
parameters.

It is evident from the results for synthetic audit sam-
ples  that the levels  of among-batch  precision
expected for these measurements were not attained
for all  parameters,  despite the high  sample values
The  reasons for these discrepancies  presumably
include mixing error or sample instability or  both;
they are discussed  in greater detail in Appendix A.

The  results  for field  and laboratory audit  samples
show that quantitation limits were a useful means of
classifying the data, which was necessary to objec-
tively evaluate among-batch precision estimates for
the 23  parameters. It  was possible to determine
whether the observed precision estimates  for each
parameter were reasonable in relation to the DQO's
established  for the survey. High  precision was not
expected for measurements  close to the detection
limits, and it was not  achieved for any parameters
with means less than quantitation limits except DOC.
Conversely, high precision was expected and was
generally achieved for measurements with  higher
mean values (i e., for 4 parameters with FN2 samples
and for  1 parameter with FN3 samples). Estimated
overall and analytical within batch precision of DOC
measurements showed a pattern which was opposite
to that shown by all other parameters; measurements
at higher concentrations exhibited greater variability
than those at lower concentrations.

Natural  and synthetic audit samples were also used
to judge the performance of contract analytical labo-
ratories on a batch basis (Drouse et al., 1986). Table 26
provides exam pies of the 95 percent confidence inter-
vals (performance windows) used for data evaluation.
Actual laboratory performance was judged according
to 95 percent confidence intervals for each lot of each
type of audit sample. Any audit sample outside the
interval  was flagged for further verification.  These
flags were used to concentrate verification efforts on
potential analytical  problems. Field  synthetic audit
samples were found to contain precipitated iron and
aluminum which were removed by filtering at the field
laboratory (see above); those samples were not inclu-
ded in the statistical evaluation.
                                                 57

-------
Field and
FN2a
Parameter
A1, mg L"1
total extractable0
total
ANC,Meq L"1
Ca, mg L"1
CT, mg L'1
Conductance's cm"1
DIG, mg L"1
air-equilibrated
initial
DOC, mg L"1
F~ total dissolved, mg L"1
Fe, mg L"1c
K, mg L"1
Mg, mg L"1
Mn, mg L"1
Na, mg L"1
NH4+, mg L"1
N03~, mg L"1
pH
air-equilibrated
initial ANC
initial BNC
P total, mg L"1
Si02, mg L"1
SO/", mg L"1
a FN2 = Field natural, lot
b Ft\!3 = Field natural, lot
0 Laboratory synthetic aud
High
0.308
0.443
14.8
2.03
0.61
28.0
0.21
0 61
3 5
0.083
0.04
0.53
0.36
0.08
0.74
0.12
1.557
520
5.15
5.18
0.006
5.05
7.52
2 (Big Moose Lake)
3 (Lake Superior).
samples only (see Section
Low
0056
0.195
-10.1
1.75
0.45
25.2
0.10
021
29
0.071
-0.00
0.45
0.33
0.06
0.60
0.00
1.316
5.07
4.98
4.98
-0.003
3.62
6.20


4).
FN3b
High
0 007
0.023
8680
1395
1.46
100 5
10.61
10 99
1.7
0 039
-0.01
0.57
3.05
0.00
1.51
0.03
1.607
8 28
8.07
8.08
0 007
3 14
354



Low
-0.003
-0 006
815.1
12.12
1 32
92 3
8.76
888
1 0
0.030
0.21
0.46
2 54
-0.01
1.09
-0.02
1.205
8.19
7.44
7.49
-0.004
231
2.99



Laboratory
High Synthetics
High
0.22
0.27
654 4
2.21
4.35
112.9
7 29
9.88
14 5
0.471
0.15
330
2 54
1.58
13.87
1.58
3.090
8.14
757
7.64
0.078
11.90
15.67



Low
0.10
0.14
305.9
1.30
2.67
96.1
2.14
1.99
6.1
0.404
009
2 83
2.20
0.94
9.76
0.97
0.264
7.58
6.56
6.58
0.038
6.74
13 05



Field and
Laboratory
Low Synthetics
High
0 03
0.01
128 3
0.19
0.38
21.1
1.61
2 07
1.9
0045
0 05
0 27
4.58
0.10
2.98
0 25
0927
7.59
7.09
7 21
0 031
1.43
2 47



Low
0.01
94.4
0.11
0 23
16.5
1.11
1.24
0.1
0 036

0.18
0.39
0 08
2 45
0.10
0.037
7 04
6 63
6.65
0.010
0 57
2 09



Table 26.  Contract Analytical Laboratory Performance Windows for Audit Sample Measurements, Eastern Lake Survey - Phase I.

-------
                                           Section 5
                                Data Variability in the ELS-/
Four types of precision estimates (overall  within-
batch, overall amongbatch, analytical within-batch,
and analytical among-batch) identify the amounts of
data variability that can be attributed to sample col-
lection,  processing, storage, and analysis. For the
ELS-I, each type of precision estimate was used to
estimate a different aspect of data variability.

Overall within-batch precision (the  total amount of
data variability for samples collected and processed
on a given day) was estimated using field duplicate
data. Analytical within-batch precision (the portion of
the total data variability that occurred during chemi-
cal analysis of the samples collected and processed
on a  given day) was estimated using  laboratory dupli-
cate  data. The differences between the amounts of
overall (field  duplicate)  and analytical (laboratory
duplicate)  within-batch  precision  indicate  the
amount of data variability that occurred during sam-
ple collection, processing, and storage.

Overall among-batch precision (total data variability
among all sample batches) was estimated using data
from field natural audit samples and field synthetic
audit samples. Analytical among-batch precision (the
portion of the total among-batch variability that could
be attributed to measurement imprecision including
temporal effects) was estimated using data from lab-
oratory synthetic audit samples. The  field natural
audit samples were also used to provide an estimate
of relative interlaboratory bias, and the field synthetic
and laboratory synthetic audit samples were used to
determine absolute  interlaboratory bias and accu-
racy.

COMPARISONS OF PRECISION
ESTIMATES

The four types of precision estimates are expected to
relate in numerically consistent ways. By comparing
the observed relationships between precision esti-
mates to the expected relationships, the quality of
the data can be evaluated. Quantitation limits (based
on the variability of field blank and laboratory blank
measurements) were used both for the evaluation of
individual precision estimates (see Section 4) and for
comparing precision estimates.

The quantitation limits provided a means of classify-
ing the data which was necessary to objectively eval-
uate precision. It is important to understand that the
quantitation limit provided only two precision catego-
ries for each precision estimate (pairs with mean val-
ues greater than zero and pairs with mean values
greater than the quantitation limit), whereas the rela-
tive precision is expected to vary along the entire
range of mean sample values (Mericas et  al. 1986).
Therefore, if the sample concentrations in two popu-
lations are sufficiently  different, the relative preci-
sions of the sample populations will differ even when
both populations have mean values above a quantita-
tion  limit. Such differences may be  unimportant
when, for example, the precision estimates for  both
populations  meet  predetermined  precision goals.
ELS-I data quality was evaluated by comparing the
relative  variability  of samples  which  had  means
abovethequantitation limitsagainstthe DQOsestab-
lished  prior  to the  survey. When  precision  is
expressed in absolute terms (as it was for  pH in the
ELS-I), the observed precision can be directly com-
pared against the DQOs.

The inherent differences between the methods used
to calculate precision estimates should also be con-
sidered  when comparing chemical measurements
from the ELS-I. Precision estimates for many samples
that have the same concentration (e.g., among-batch
precision using  audit samples) can be expressed in
terms of the percent  relative standard  deviation
(%RSD).  Among-batch precision  estimates were
therefore directly comparable with the  interlabora-
tory  precision goals which were part of the DQOs.
Precision estimates for measurements that have a
wide range of values (e.g., field duplicate pairs and
laboratory duplicate pairs) can be calculated as the
root mean square (RMS) of %RSD.The RMS of %RSD
estimates the mean of the %RSD over the  observed
range of mean sample pair values. Because the RMS
of %RSD is an estimate of the true %RSD, the within-
batch precision  values from survey data are directly
comparable to the interlaboratory precision goals as
set forth in the DQOs for the ELS-I.

EXPECTED RELATIONSHIPS BETWEEN
PRECISION ESTIMATES

Overall  within-batch  precision  estimates  were
expected to be numerically  larger than analytical
within-batch precision estimates by an amount equal
                                                59

-------
to the variability from sample collection, processing,
and storage (see above). Similarly,  overall  among-
batch  precision estimates  were expected to  be
numerically largerthan analytical among-batch preci-
sion estimates. Analytical and overall among-batch
precision estimates were expected to be numerically
larger than the corresponding analytical and overall
within-batch precision  estimates by amounts that
were equal to the temporal variability.

The exceptions to the four types of expected relation-
ships between precision estimates in the ELS-I are
listed below. Unless specifically noted, the compari-
sons which include overall or analytical within-batch
precision estimates are based on duplicate pairs with
means greater than the quantitation limits. Where the
mean of the audit sample values for total extractable
Al or total Al were less than 0.010 mg L~1,the among-
batch precision estimates for those measurements
were  compared  to the corresponding within-batch
precision  estimates for duplicate pairs with means
less than 0.010 mg L""1. Similarly,  where the mean of
audit  sample DOC values was less than 5.0 mg L"1,
the among-batch precision for those measurements
was compared to the corresponding within-batch pre-
cision estimate  for the duplicate pairs which had
means less than 5.0 mg L~! This convention was also
followed for the comparisons between precision esti-
mates for Al and DOC means which were greater than
0.010 and 5.0 mg L~1, respectively.

Overall within-batch precision estimates were numer-
ically larger than analytical within-batch precision
estimates for all parameters except total extractable
Al, conductance, Mn, total P, and SO4~2 (see Tables 20
and 22). Overall among-batch precision  estimates
from FN2 audit samples were numerically largerthan
overall  within-batch  precision  estimates  for  all
parameters  except total  dissolved F~,  NO3~, initial
ANC  pH, and initial BNC pH (see Tables 20  and 23).
This relationship was observed for both FN2 samples
with means greater than the quantitation limits and
for FN2 samples with means less than the quantita-
tion limits. Overall  among-batch precision estimates
from FN3 audit samples were numerically larger than
overall within-batch precision estimates for 15 of 23
parameters (see Tables 20 and 23). Exceptions to this
relationship include values for ANC,  Cl~, conduc-
tance, air-equilibrated DIG, total dissolved F~, NO3~,
air-equilibrated pH, and SO4~2. Overall among-batch
precision estimates from FH audit samples were
numerically larger than overall within-batch precision
estimates for all parameters except NH4', NO3~, and
SO4~2(seeTables20and 24). Overall among-batch pre-
cision estimates from FL audit samples were numeri-
cally larger than overall within-batch precision esti-
mates for all parameters except total Al, ANC, Mn,
and NH4+,  (see Tables 20 and 25). Lastly, analytical
among-batch precision estimates from both LH and
LLaudit samples were numerically largerthan analyt-
ical within-batch precision for all parameters except
conductance, Mn, and SO4
25).
                        -2
(see Tables 22, 24, and
Exceptions to the expected relationships were gener-
ally associated with the presence of one or more
extreme outliers in the verified data set, with values
close to the detection limit, or with a methodological
problem. Many exceptions involved small differences
in estimated precision. In several cases it was neces-
sary to retain confirmed but questionable values in
the verified data set which were later  deleted during
data validation. The confirmed, questionable values
influenced the statistical evaluation of the ELS-I data.
For subsequent surveys, a data qualifier (XO) was
added to ensure that such values were retained in the
verified  data set but were not included in statistical
calculations.

This brief  discussion of the discrepancies in  the
expected ELS-I precision relationships is a first step
in understanding the mechanisms that affect data
variability. A more comprehensive analysis of these
discrepancies is needed to adequately describe the
potential contributing factors and their implications.
The complexity of such an analysis is a subject which
calls for detailed evaluation in a separate report.
                                                 60

-------
                                           Section 6
                                           Summary
In general, the ELS-I was conducted smoothly and
efficiently with surprisingly few problems given the
magnitude of the survey. This success may be attrib-
uted to the use of a peer-reviewed research plan, a QA
plan, and operations, training, and methods manuals,
as well as to the efforts  of all  of  the  individuals
involved. When problems did occur, they were identi-
fied and generally resolved quickly; this indicates
that the checks and balances fundamental to the QA
program operated effectively.

Implementation of pilot studies  prior to the ELS-I
proved to be worthwhile in that many field sampling
and analytical issues which could have caused prob-
lems during the  full survey were corrected without
loss of survey data. In particular, a nitrate contamina-
tion problem was  identified and partially resolved
prior to initiation of the survey.

The operational QA program appeared to be adequate
to ensure that all samples were collected and ana-
lyzed consistently and that the resulting data were of
known and traceable quality. The field QC and calibra-
tion procedures proved sufficient to detect specific
instrument or operator problems; this was evidenced
by detection of both a faulty pH meter and contamina-
tion of field blanks which was caused  by the  use of
deionized water  from a supply not meeting ELS-I
requirements.

Analytical QC procedures were also sufficient as evi-
denced by the detection and resolution of both a sil-
ica calibration error and of aluminum contamination
which resulted from the use of borosilicate glassware
and from floor sanding. After the ELS-I, the base coor-
dinators recommended improvements in the training
program. This recommendation was addressed by
using the  same personnel on later surveys for the
NSWS.  All  returning  NSWS  personnel received
"refresher" training in addition to training in proto-
cols which were specific to each subsequent survey.

The data  base entry and  verification procedures
enabled virtually all transcription, transposition, and
typographical  errors on  the various  forms  to be
detected and corrected. A  problem was identified
with the error  correction procedures  in that the
method was slow;  the alternative method that was
employed  was labor-intensive. The correction proce-
dures were revised  for later surveys.

Evaluation of the QA and QC sample data indicated
that the data quality objectives for detectibility and
precision were achievable. The  importance of using
field blank measurements during data interpretation
was demonstrated. Little bias was detected among
field stations and analytical laboratories. Evidence
was presented to show that the bias that did exist
could be quantified, and that correction factors could
be applied to the values to aid in data interpretation
(see Appendix A).

These results point to the usefulness of the QA pro-
gram  in assuring consistency and reliability of the
data collected in the NSWS. To that end, the ELS-I QA
program was successful in producing data of  known
and verifiable quality in accordance with the objec-
tives of the NSWS.
                                                61

-------
                                           Section 7
                                          References


American Society for Testing and Materials, 1984. Annual Book of ASTM Standards, v. 11.01, Standard Specifi-
     cation for Reagent Water, D1193-77 (reapproved  1983). ASTM, Philadelphia, Pennsylvania.

Barnes, R. B., 1975. The Determination of Specific Forms of Aluminum in Natural Water. Chem. Geol., v. 15, pp.
     177-191.

Driscoll, C. T, J. P. Baker, J. J. Bisogni, and C. L. Schofield, 1983. Aluminum Speciation and Equilibria in Dilute
     Acidic Surf ace Waters of the Adirondack  Region of New York State, pp. 55-75. In: O. P. Bricker(ed.), Geolog-
     ical  Aspects of Acid Deposition. Butterworth Publishers,  Boston, Massachusetts.

Drouse, S. K., and M.  D. Best, 1986. Comparability of Chemical Data Obtained  Using ICPES and AAS, 1C, or
     Colorimetry. (In  preparation.)

Drouse, S. K., D. C. Hillman, L. W. Creelman, and S. J. Simon, 1986. Quality Assurance Plan for the National
     Surface Water Survey-Eastern Lake Survey (Phase I--Synoptic Chemistry). U.S. EPA, Las Vegas,  Nevada.

Eilers, J. M., D. J. Blick, Jr., and M. Dehaan. 1986. National Surface Water Survey, Eastern Lake Survey- Phase I,
     Validation of the Eastern Lake  Survey - Phase I Data Base. U.S. EPA, Corvallis, Oregon.

Fountain, J., and D. T. Hoff, 1985 (Draft). AQUARIUS User's Guide. Internal Report for Lockheed-EMSCO, Las
     Vegas, Nevada.

Hillman, D. C., J. F. Potter, and S. J. Simon, 1986. Analytical Methods Manual  for the National Surface Water
     Survey - Eastern Lake Survey (Phase I Synoptic Chemistry). U.S. EPA, Las Vegas, Nevada.

Kramer, J. R., 1984. Modified Gran  Analysis for Acid and Base Titrations.  Environmental Geochemistry Report
     No. 1984-2. McMaster University, Hamilton, Ontario, Canada.

Linthurst, R. A., D. H. Landers, J. Eilers, D. F. Brakke, W. S. Overton, E. R Meier, and R. E. Crowe., 1986. Character-
     istics of Lakes in the Eastern United States. Volume I: Population Descriptions and Physico Chemical
     Relationships. EPA-600/4-86-007A, U.S. EPA, Washington, D.C. 275 pp.

May, H. M., P. A.  Helmke, and M. L. Jackson, 1979. Determination of mononuclear dissolved aluminum in near-
     neutral waters. Chem. Geol. 24:259-269.

McQuaker, N. R., P. D. Kluckner, and D. Sandberg, 1979. Chemical analysis of acid precipitation: pH and acidity
     determinations. Environ. Sci. and Technol. v. 17, n. 7, pp. 431-435.

Mericas, C. E., M. D. Best, and R. D. Schonbrod, 1986. Measurement uncertainty  in the National Surface Water
     Survey. (In preparation.)

Morris, F. A., D.  V. Peck, M. B. Bonoff, and K. J. Cabbie, 1986. National  Surface Water Survey  - Eastern Lake
     Survey (Phase I  -- Synoptic Chemistry).  Field Operations Report. U.S. EPA, Las Vegas, Nevada.

O'Dell, J. W., J. D. Pfaff, M. E. Gales, and G. D. McKee, 1984. Technical Addition to Methods for the Chemical
     Analysis of Water and Wastes, Method  300.0, The Determination of Inorganic Ions in Water by Ion Chro-
     matography.  EPA-600/4-85-017. U.S. EPA, Cincinnati, Ohio.

Oliver, B. G., E. M. Thurman, and R. L. Malcolm, 1983. The Contribution of Humic Substances to the Acidity of
     Colored Natural Waters. Geochim. Cosmochim. Acta, v. 47, pp. 2031-2035.

Overton, W. S., P. Kanciruk, L. A.  Hook, J. M. Eilers, D. H. Landers, D.  F. Brakke, D. J. Blick, Jr., R. A. Linthurst, and
     M. D. DeHann, 1986.  Characteristics of Lakes in the Eastern United States. Volume II: Lakes Sampled and
     Descriptive Statistics for Physical and Chemical Variables.  EPA-600/4-86-007C, U.S. EPA, Washington,
     D.C., 374 pp.

                                                62

-------
Permutt, T. J., and A. K. Pollack, 1986. Analysis of Quality Assurance Data for the Eastern Lake Survey. Final
      Report for Lockheed-EMSCO, Las Vegas, Nevada. 47 pp.

Quave, S. A., 1980. Nitrite-nitrate nitrogen in drinking water. Holding time and preservation study. Memo to L. B.
      Lobring, U.S. EPA, Cincinnati, Ohio, September 25, 1980.

Rosen, A. E., and P. Kanciruk. 1985. A generic data entry quality assurance tool. SAS (Statistical Analysis Sys-
      tem, Inc.) SAS User's Group, Interna tional, Proceeding SUGI 10, Reno, Nevada.

SAS Institute, Inc., 1985. SAS User's Guide: Basics, Version 5 Edition. SAS Institute, Inc. Gary, North Carolina
      1290pp.

Skougstad, M. W., M. J. Fishman, L. C. Friedman, D. E. Erdman, and S. S. Duncan (eds.), 1979.  Methods for
      Determination of Inorganic Substances in Water and Fluvial Sediments: Techniques of Water-Resources
      Invest! gations of the United States geological Survey, Book 5. U.S. Government Printing Office, Washing-
      ton, D.C.

Stapanian, M. A., T. E. Lewis, M. D. Best, and C. E. Mericas, 1986. Water chemistry methods in acid deposition
      research: A comparative study of analyses from Canada, Norway, and the United States. (In preparation.)

U.S. Environmental Protection Agency,  1983 (revised). Methods for Chemical  Analysis of Water  and Wastes
      EPA-600/4-79-020.  U.S. EPA, Cincinnati, Ohio.

U.S. Environmental Protection Agency, 1984a.  National Surface Water Survey,  Phase  I. U.S. EPA, Office of
      Research and Development, Washington, D.C.

U.S. Environmental Protection Agency, 1984b. National Surface Water Survey,  Phase I. Research Plan, A Sum-
      mary of Contents. U.S. EPA, Corvallis, Oregon.

Williams, T. J., 1979. An Evaluation of the Need for Preserving Potable Water Samples for Nitrate Testing. J. Am.
      Waterworks Association. March:157-160.

Yfantis, E. A., M. A. Stapanian, G. T. Flatman.'E. R Meier, and D. L. Montague, 1986. Comparison of Univariate and
      Chemometrical  Analysis of Multilabora tory data. Anal. Chem. Acta (in press).
                                               63

-------
                                 APPENDIX A
                 ANALYSIS OF QUALITY ASSURANCE DATA
                      FOR THE EASTERN LAKE SURVEY
                                   Prepared by

                            T. J. Permutt and A, K. Pollack

                              Systems Applications, Inc.
                               101 Lucas Valley Road
                            San Rafael, California 94903
The following document is a final report on the statistical evaluation of quality assurance data for the
ELS-I which was prerpared for Lockheed-EMSCO.

-------
                 Final Report

                    ANALYSIS  OF  QUALITY  ASSURANCE  DATA
                       FOR THE EASTERN LAKE SURVEY

                              SYSAPP-86/040

                             26  February 1986
                               Prepared  for

                              Lynn Creelman
                              Locktieed-EMSCO
                    1050  East  Flamingo Road, Suite  246
                         Las Vegas,  Nevada  89119

                          Purchase Order  IEH392
                               Prepared by

                              T. J. Permutt
                              A. K. Pollack

                        Systems  Applications, Inc.
                          101 Lucas Valley Road
                      San Rafael, California  94903
87770:'51B5r
                                   66

-------
                                  CONTENTS


1    INTRODUCTION	      1
2    DETECTION LIMITS	      4
        Decision Limits	      5
        Detection Limits	      7
        Quant1tat1on Limits	    11
3    DUPLICATE SAMPLES	    15
        Introduction	    15
        Method	    15
        Results	    18
4    NATURAL AUDIT SAMPLES	    23
        Int roduct 1 on	    23
        Method	    23
        Results	    27
5    SYNTHETIC AUDIT SAMPLES	    35
        Int roduct 1 on	    35
        Bias and Precision	    35
        Measured Versus Theoretical Concentrations	    37
        Field Versus Laboratory Audits	    41
6    CONCLUSIONS AND RECOMMENDATIONS	    43
        What the QA Data Show	    43
        Synthetic Audits	    46
        Iron	    47
        Detection Limits	    47
References	    49

BSiesr i
                                     67

-------
                                  FIGURES



2-1  Summary of sulfate measurements of 245 field blanks	     8

2-2  Calculation of the detection limits (0)	    12

4-1  Interlaboratory bias and trend 1n pH  of natural  audits
     from B1g Moose Lake (EMSI and Versar  only)	    23

4-2  Interlaboratory bias and trend 1n pH  of natural  audits
     from Big Moose Lake (four laboratories)	    25

5-1  Concentration of ammonium measured in laboratory high
     synthetic audit samples	    35

5-2  Concentration of dissolved inorganic  carbon (initia1)
     measured in laboratory high audit samples	,	    39

5-3  Concentration of silica measured 1n laboratory high
     synthetic audit samples	    41
                                   68

-------
                                  TABLES'
 2-1  System decision and detection limits	     9

 3-1  Various summary statistics for the standard deviation of
     duplicate pairs 1n a Monte Carlo experiment	    16

 3-2  H1th1n-batch precision for pH measurements by contract
     1 aboratoM es	    18

 3-3  W1th1n-batch precision of laboratory duplicates	    19

 3-4  H1th1n-batch precision of field duplicates	    20

 3-5  Very high RSOs 1n laboratory duplicate pairs with a mean
     measurement above the quantltatlon limit	    21

 4-1  Overall precision estimated from natural audits	    27

 4-2  Overall precision of pH measurements 1n natural audits	    28

 4-3  Relative Interlaboratory bias estimated from natural audits
     and controlling for measurement trend	    29

4-4  Relative Interlaboratory bias estimated from natural audits
     and Ignoring measurement trend	    30

4-5  Interlaboratory bias estimated from natural  audits	    32

4-6  Trends (per month) 1n measurements of natural  audit
     samples (B1g Moose Lake)	    33

5-1  Field and laboratory high synthetic audits	    37

5-2  Field and laboratory low synthetic audits	    38

6-1  Summary of estimates of precision	    43
                                   111
                                    69

-------
                             1   INTRODUCTION
The National Surface Water Survey, designed and begun by the U.S. Environ-
mental Protection Agency 1n cooperation with the National Add Precipita-
tion Program, 1s a three-phase project Intended to document the chemical
and biological condition of lakes and streams considered susceptible to
the effects of acidic deposition.  Phase IA of the program 1s a survey of
lake chemistry.  It consists of the Eastern Lake Survey, with which this
report 1s concerned, and the Western Lake Survey, which 1s currently
underway.  The primary objectives of the phase IA survey are to determine
how many lakes have low pH or alkalinity 1n regions of the United States
potentially sensitive to acidic deposition and to understand the chemical
composition of these lakes.  From the lakes sampled 1n phase IA some will
be selected for Intensive study 1n phases II and III of the project (EPA,
1984).

Four regions of the United States were selected for study In phase IA:
the Northeast, the Southeast, the upper Midwest, and the mountainous
West.  These regions were chosen because they are known to contain an
abundance of low-alkalinity lakes, which are considered sensitive to
acidic deposition.  The Eastern Lake Survey covered the Northeast, South-
east, and upper Midwest regions and was completed 1n late 1984.

The sampling plan 1s a stratified design.  Each region 1s divided Into
subreglons, and within a subreglon lakes are divided Into three alkalinity
classes.  In addition, there are minor stratification variables:  lake
size, elevation, and watershed size.  An extensive quality assurance/
quality control (QA/QC) program was designed and carried out to assure
high-quality data and to Identify any problems 1n the collection, pro-
cessing, and analyzing of the lake water samples (Drouse et a!., 1985).
While the primary purpose of the QA/QC data was to Identify and correct
potential data quality problems during the survey, they also serve to
assess the data quality that was achieved, which 1s the subject of this
report.

Two-liter lake samples were collected by helicopter crews and processed by
one of eight field laboratories.  At the field laboratories the samples
were split Into seven allquots, labeled  (by coded Identification number,
not by lake name), and shipped to one of four contract laboratories for

85185r 2
                                   70

-------
 chemical  analysis.   Each field  laboratory  was  supposed  to  send  all  of  Its
 samples  (audit  and  routine)  to  one  contract  laboratory,  but  this  protocol
 was  not  always  maintained.   Four  contract  laboratories  analyzed the lake
 samples:   EMSI  analyzed  samples collected  by four of  Its field  labora-
 tories from 8 October  to 7  November 1984;  Versar analyzed  samples collec-
 ted  by two of Its field  laboratories and two of EMSI's  field laboratories
 from 7 October  to 16 November 1984; Global analyzed samples  collected by
 Its  field laboratory and two of Versar's field laboratories  from  20 Octo-
 ber  to 29 November  1984; and one  U.S.  Geological Survey  field laboratory
 collected samples after  all  other field laboratories  completed  their
 sampling, from  2 December to 14 December 1984.

 The  sample load for a  field  laboratory was restricted to 24  routine lake
 samples  per day.  A batch Is defined as the  set of samples collected by a
 single field laboratory  on  a single day (I.e., 24 routine  samples per
 batch).   Each batch was  Identified  by  a three-digit code.  Over the dura-
 tion of the Eastern Lakes Survey, a total  of 127 batches were analyzed.
 One  duplicate sample was obtained at one lake  (usually the first) sampled
 1n each batch.

 Each lake was visited  only  once during the survey and each sample was
 processed by one field station  and  one contract laboratory.  Consequently,
 some Important  aspects of the quality  of the data cannot be  judged  from
 the  routine data themselves.  Systematic differences  between laboratories
 could result from minor  differences 1n Instrumentation or  procedure or
 from real  differences  between the sets of  lakes measured by  the different
 laboratories.   Similarly, variations among measurements by one  laboratory
 could reflect real  differences  between lakes or simply the Imprecision of
 the  measurements.   The routine  data do not allow these possibilities to be
 distinguished.  The quality  assurance  data,  however,  make  H possible to
 estimate  the likely magnitudes  of various  kinds of error.  Variations
 larger than  these 1n the  routine measurements may confidently be supposed
 to represent  actual   differences between lakes.

 Several kinds of blind QA samples are  used to estimate the components of
 the  error 1n  the wnole system from  collection to analytical  determina-
 tion.  For  one, there are duplicates of the  routine samples.  These dupli-
 cates are  of two types,  called  field and laboratory duplicates.  Field
 duplicates  are  second samples collected from one lake 1n each batch and
 processed  1n  parallel with the  routine samples.  Laboratory  duplicates are
 splits made  at  the  analytical laboratories so that the analysis, but not
 the  routine  sample  processing,  1s duplicated.  Both types of duplicate
measurements  Indicate, by the difference between the measurements of a
 pair, the  repeatability  of measurements by a given laboratory on a  given
 day.  The  laboratory duplicates Indicate the the repeatability  of the
 analytical  determination  alone, the field duplicates  Indicate the repeata-


 esissr 2
                                    71

-------
b1!1ty of the entire system of measurement from collection to analysis.
Neither type tells us anything about variation across laboratories  or
across days, such as would be produced by errors 1n calibration.

For this we have the natural  and synthetic audit samples.   Natural  audit
samples were prepared dally from stored lots of water originally  collected
from B1g Moose Lake 1n the Adlrondacks and Lake Superior.   Synthetic audit
samples were prepared dally by diluting concentrate lots of known composi-
tion.  There are both field and laboratory synthetic audits:  field syn-
thetic audits were shipped to the field stations 1n two-liter containers
and were divided Into allquots at the field stations, whereas laboratory
synthetic audits arrived at the field stations ready for reshlpment to the
contract laboratories and thus skipped the processing at the field  sta-
tions.  The natural audits are field audits.  These samples thus  furnish
repeated measurements of the same thing, by the same and by different
laboratories, assuming stability of the natural audits 1n storage and
consistency 1n preparation of the synthetic audits.  Unlike the dupli-
cates, therefore, they give Information about relative bias between
laboratories and about reproduclblHty of measurements from day to  day.
In addition, there 1s some Independent Information about what 1s  1n the
synthetic audit samples, namely the recipes for them.  Unfortunately, this
Information turns out to be very unreliable.

The quality assurance plan also provided for blank samples to reveal any
problems of contamination and also study the detectablllty of analytes at
low concentrations.  There were laboratory blanks, which are not  discussed
1n this report, and field blanks, which were samples of delonlzed water
processed at the field stations like routine samples.

This report discusses methods we developed for estimating measurement
errors from the quality assurance data as well as the results.  He
consider first the evaluation of detectablHty from  field blank data, then
the analyses of duplicate data and natural and synthetic audits.   Our work
1s concerned with 24 parameters measured by the contract laboratories.
Mineral acidity and carbonate alkalinity were not analyzed because most
of the measurements were missing or considered unreliable.   Nitrate  1s
Included, but 1t 1s being studied 1n much greater detail by  Liggett
(1985).
 85185^  2
                                    72

-------
                            2    DETECTION  LIMITS
 For the purpose of evaluating the detectab111ty of various parameters at
 low levels, the quality assurance plan for the Eastern Lake Survey pro-
 vided for some 245 field blank samples.  A Van Dorn sampler was filled
 with delonlzed water, and allquots were prepared at the field stations 1n
 the same way as routine samples.  The blank samples were then shipped to
 the contract laboratories along with routine samples.  The labels were
 coded so that the laboratories were blind as to which samples were blank.

 Because the field blank samples were processed as much like the routine
 samples as possible, measurements of the blanks are subject to errors from
 all the sources that affect the routine measurements.  Internal blanks
 within the contract laboratories were also measured and used to calculate
 Instrument detection limits.  These Instrument detection limits are
 unlikely to be achieved 1n routine samples because the Internal blanks are
 Immune to three kinds of error that affect routine (and field blank)
 measurements.  First, the Internal duplicates are not analyzed blind, and
 analyses of samples known to be blanks may be expected to have rather less
 variable results than blind analyses.  Second, the field blank samples,
 like the routine samples, are exposed to varying degrees of contamination
 1n processing before they reach the laboratory, whereas the Internal
 blanks are not.  Third, errors 1n calibration of an Instrument do not
 affect the calculated Instrument detection limit.  The Instrument detec-
 tion limit may be calculated from the standard deviation of 10 measure-
ments of Internal blanks on the same day.  If the calibration 1s wrong
 that day, the mean, but not the standard deviation, will be affected.  In
 contrast, the field blanks are analyzed on many different days, and so
 variation 1n field blanks reflects errors In calibration.

 Our definitions and method of calculating system detection and decision
 limits from measurements of field blanks follow In spirit those of Hubaux
and Vos (1970).  The details are different for a number of reasons.
 Hubaux and Vos were concerned with allowing for calibration error when the
measurements are all from one calibration; we are not because the measure-
ments are from many calibrations so that their variation Includes calibra-
tion error.  They were also concerned with allowing for uncertainty about
the long-run distribution of measurements of blanks when only a few blanks
are measured; we are not because there are 245 field blanks.  On the other

•siesr 3
                                  73

-------
hand, by using nonparameteMc techniques, we are able to dispense with
their assumption that measurements have normal  distributions.  This 1s
fortunate because the field blank data clearly do not follow normal dis-
tributions.

Following Hubaux and Vos, we distinguish two kinds of limits:  decision
limits and detection limits.  We also calculate a third kind of limit,
sometimes called a quantHatlon limit.  The decision limit 1s a measure-
ment that reliably Indicates a level  above background.  The detection
limit 1s a concentration that would probably be detected 1f that concen-
tration were present.  The quantHatlon limit Is a concentration that can
be measured with reasonable precision.  These concepts are explained 1n
turn below.
DECISION LIMITS

We define the decision limit, following Hubaux and Vos, as "the lowest
signal that can be distinguished from the background."  It 1s thus a mea-
surement so high that 1t rarely occurs 1n blank samples.  If this high a
measurement does occur, 1t 1s therefore a more or less reliable Indication
that the sample 1s not a blank; the more rarely such a measurement occurs
1n blanks, the more reliable an Indication 1t 1s.  If the decision limit
1s set at the upper a quantlle of the distribution of blanks, a measure-
ment above the decision limit can be said to be significantly different
from background at level a.  This means that the probability of erron-
eously calling a measurement significant that 1s really background 1s only
It remains only to choose an error rate o that we are willing to accept
and then to estimate the upper a quantlle of the background distribu-
tion.  Provided a 1s not too small (at least several times 1/245), we can
estimate the quantlle simply, by the observed quantlle of the 245 measure-
ments.  For example, for o - .05, the decision limit Is taken to be the
95th percentlle of the measurements of blanks.  Measurements higher than
tnls occur only 5 percent of the time for blank samples, so such a
measurement 1s a significant Indication (at the .05 level) that the sample
1n question 1s not a blank.

There 1s no consensus on what the probability of error a should be.  It
might seem desirable to make o very small, perhaps a fraction of 1 per-
cent.  This approach has two drawbacks.  First, 1t 1s Impossible to reli-
ably estimate very high quantlles without very many observations.  For
example, suppose that a certain kind of error, say an unusual contamina-
tion, affected one sample 1n 100 and that decision limits were based, as
they often are, on sets of 10 samples.  Suppose that the decision limit


85185T 3
                                   74

-------
was  to  be the  99tn  percentlle.   In nine out of every 10 sets of 10
samples, none  of the samples would be affected by the unusual error, so
that the 99th  percentlle would be underestimated.   In one out of 10 sets,
a measurement  would be affected, and then such an error would be assumed
to happen In about  one out of 10 measurements; the  99th percentlle would
thus be overestimated.  Quite simply, we get a very poor Idea of what
happens 1n one out  of 100 measurements (the 99th percentHe) by looking at
10 measurements.

Second, even when there are enough measurements to  estimate a very high
quantlle, It may not be desirable to do so.  To continue the example, with
245  observations we are likely to see two or three  of those unusual
errors.  The 99th percentHe of the 245 measurements will therefore
probably be one of  those unusual measurements.  The question 1s, do we
want the reported decision limit to reflect this rather extreme behavior
of the measuring process at Its worst, or somewhat more typical behavior?
Our  answer 1s, the more typical behavior.  Accordingly, we use the 95th
percentHe of  the measurements of blanks as the decision limit.  Any
measurement above our calculated decision limit is thus significantly
different from background at a significance level (or error rate) of
a -  .05.

It 1s usual  to calculate a decision limit from the mean (7) and standard
deviation of the measurements of blanks.  For example, the decision limit
might be taken to be 7 + 3 s and the detection limit, which will be dis-
cussed later,  6 s.  More commonly, 3 s 1s used as the detection limit, and
this corresponds to a decision limit of 7 + 1.5 s.  So defined, the
decision limit can be calculated from very few measurements, even as few
as two.  However, sucn a definition has two serious disadvantages.  First,
1t 1s Impossible to know the error rate.  The measurements are sometimes
assumed to follow a normal distribution; 1f they did, 7 + 1.5 s would be
about the 93rd percentlle and the error rate would be about 7 percent.
Unfortunately, the validity of this assumption cannot be checked unless
there are so many measurements that the assumption 1s unnecessary.  On the
other hand,  even 1f the distribution 1s not normal, Long and Wlnefordner
(1983) point out that a mathematical  result called Chebyshev's Inequality
limits the error rate.  Unfortunately, the limit Is crude:  anywhere from
none up to 4/13 of a distribution can be at least 1.5 standard deviations
above the mean.

The  second disadvantage of basing the decision limit on the mean and stan-
dard deviation 1s that these statistics are very sensitive to outliers.
If there are a few very high measurements, the standard deviation 1n
particular may be more heavily Influenced by these few than by all  the
rest.  Here  again, the decision limit would reflect the extreme rather
than the typical  behavior of the system.   Nevertheless, 1n addition to the


BSiesr 3
                                   75

-------
95th percentlle, for purposes of comparison we calculate a decision limit
•s 1.5 standard deviations above the mean of the measurements of blanks.

The case of sulfate Illustrates all these points.  Figure 2-1 summarizes
the 245 measurements of sulfate 1n field blanks.  The distribution 1s very
skewed to the right; 1t does not look like a normal distribution at all.
The highest value (2.61 mg/1), 1s more than eight times the second highest
(0.31 mg/1), which 1s 1n turn twice the third highest (0.16 mg/1).  The
mean 1s 0.043 mg/1 and the standard deviation 1s 0.17 mg/1.  Both are
heavily Influenced by the outliers:  eliminating just the one largest
measurement would reduce the mean to 0.032 mg/1 and the standard deviation
to 0.034 mg/1.   The 95th percentHe 1s 0.093 mg/1.  There are several
measurements fairly close to that above and below; the nearest are 0.085
and 0.094 mg/1.  On the other hand, 1.5 standard deviations above the mean
is 0.30 mg/1; this would be the 93rd percentile of a normal distribution,
but 1t 1s above the 99th percentile for this distribution.  There 1s only
one measurement anywhere near 0.30 mg/1, and only one other that 1s
higher.  Thus the 7 + 1.5 s decision limit mainly reflects the two largest
measurements out of 245.

Table 2-1 shows the system decision limits for all parameters (except pH)
analyzed by the contract laboratories, as well as the detection limits,
which we discuss next.  Both the parametric version based on mean and
standard deviation and the nonparametric version we recommend are shown.
DETECTION LIMITS

Hubaux and Vos call a detection limit "the limit at which a given analyti-
cal procedure may be relied upon to lead to detection...."  "Detection"
here can be understood to mean a measured concentration above the decision
limit, for 1f the measurement 1s above the decision limit it can be
reliably asserted that the analyte 1s present 1n more than background
amounts, i.e., that 1t has been detected.  The question thus becomes, how
high does the true concentration have to be before the measured concentra-
tion can be relied on to be above the decision limit?  If D 1s the
detection limit, then when the true concentration 1s D, the measured
concentration should be above the decision limit except with some small
probability 6.  Thus, two kinds of error are relevant to the definition of
a detection Hm1t.  First, false detection occurs when the measurement 1s
above the decision limit but the sample 1s In fact a blank.  The decision
limit was chosen to limit the probability of this kind of error to a.
Second, failure to detect occurs when the sample 1s not a blank but the
measurement 1s below the decision limit.  The higher the true concentra-
tion, the less probable 1s this kind of error.  The true concentration at
which the probability 1s just e is the detection limit.


85185r 3
                                   76

-------
        MOMENTS
                                                         OUANTILES(DEF'3>
                                                                                                         EXTREMES
1 245
WAN •.0428653
SID DEV •.168*75
'KEVNESS 14.7423
USS 7.33883
CV 393.939
T:MEAN-» 3.9733Z
jCN RANK MM*
HIM ~- • 223
2.65»«
2.35*
2.05*
1.75*
CO
1.45*
1.15*
• .»5»
0.55*
•
^ •
».25«
*•
•

SUM WCTS 245 1 MX MAX 2. SI 99X 0.15 LOWEST HICNCST
*""..„ . I».4S3 75X 03 0.046 95X 0.093 -0.023 0.137
VARIANCE •.•282494 S0XMED ».03 90X 0.066 -0.0Z3 0.142
KURTOSIS 225.66 2bXOI 0.01 I»X 0 -0.023 0.IS
«S -r 6.89285 0X MIN -0.023 5X -0.004 -0.02 0.31
"0 MEAN •.0107379 |X -0.023 -0.02 2.61
PROBXTI 0.0001 RANGE 2.633
PROBXSI «.0001 03-01 0.036
MODE 0.03
HISTOGRAM * BOXPLOT NORMAL PROBABILITY PLOT
1 • 2.65* •
2.35
2.05
1.75
1.45
1.15
•.as
• .55

I •
• .25
• t







****
#*#*++ •
******


• MAY REPRESENT UP TO 5 COUNTS -Z -1 • «I *2
FIGURE 2-1.  Summary of sulfate measurements of  245  field blanks.

-------
TABLE 2-1.  System decision and detection limits.
System Detection Limit
Parametric
Parameter* (3 s)
Calcium
Magnesium
Potassium
Sodium
Manganese
Iron
Aluminum (extractable)
Chloride
Sulfate
Nitrate
Silica
Fluoride (total)
DOC
Ammonium
Acidity (peq/1)
Alkalinity (peq/1)
Conductivity (pS/cm)
DIC (equilibrated)
DIC (Initial)
Phosphorus (total)
Aluminum (total )
0.043
0.012
0.11
0.060
0.028
0.037
0.012
0.11
0.52
1.0
0.70
0.010
0.76
0.17
45
17
2.8
0.17
0.33
0.011
0.084
Nonparametrlc
(2
-------
 The detection  limit  tnus  depends  on  the  decision  limit  and, through  1t, on
 the distribution  of  measurements  of  blanks.   The  detection Hm1t also
 depends  on  the distribution  of  measured  concentrations  for various small,
 positive true  concentrations:   1n fact,  the detection limit 1s that  true
 concentration  for which the  e quantHe of measured concentrations 1s the
 decision limit.   Unlike the  distribution of blanks, the distribution of
 measurements at the  detection limit  has  not been  observed.  To observe 1t
 would  require  audit  samples  with  true concentration equal to the detection
 limit, and  to  get these entails knowing  the detection limit 1n advance.
 Instead, the distribution  at the  detection limit  must be Inferred from the
 distribution of blanks.

 Let us therefore  look  again  at  the distribution of sulfate measurements 1n
 blanks,  which  was presented  1n  Figure 2-1.  Some  new points are now  of
 Interest.   The observations  are mostly positive:  91 percent are more than
 zero,  the median  1s  0.030  mg/1, and  the  mean  1s 0.043 mg/1.  There are two
 reasons  for this  result.   First,  while the field  blank  samples start out
 as  delonlzed water,  they go  through  the  same  processing as the lake
 samples  before they  get to the  laboratory and thus may  be contaminated 1n
 the process.   It  1s  likely that many blanks actually do contain measurable
 amounts  of  sulfate by  the  time  they  are  analyzed.  Second, the measurement
 error  for a blank  sample 1s  very  likely  to be positive.  It 1s not
 Impossible  to  measure  a concentration as negative because of the calibra-
 tion of  the Instrument:  If  an  unknown produces a signal less than that of
 the calibration blanks, the  measured concentration 1s negative.  Still,
 errors are much more likely  to  be  positive than negative:  there are many
 ways to  overestimate a concentration that 1s  nearly zero, and not many
 ways to  underestimate  1t.  Furthermore,  even when the calibrated Instru-
 ment produces  a negative reading,  this reading might not appear In the
 reported data:  1t might be  recorded as  zero, or the Instrument might be
 recalibrated and  this  reading might  be discarded.

 These effects  on the distribution  of blanks are Important precisely
 because  the distribution of  measurements at the detection limit 1s not
 likely to be affected in quite  the same way.  Suppose that the true  con-
 centration of  sulfate in a lake sample 1s some small positive amount x and
 that this sample 1s measured many  times.  What does the distribution of
 these measurements look like?   It  Is reasonable to assume that contamina-
 tion affects this sample the same way 1t affects blanks since contamina-
tion simply adds sulfate to  the sample, no matter how much was 1n it to
 start with.  (This would not be reasonable for parameters that are not
additive, e.g., conductivity.)  However, 1t 1s not reasonable to assume
that measurement error affects this  sample the same way It affects
blanks.  Along with positive errors, negative errors of up to x are  now
possible, and 1f most errors are smaller than x, negative and positive


 esiesr 3

                                   10
                                   79

-------
errors will be about equally likely.   Furthermore,  the  reporting  problems
with negative measurements will  have  little effect  since  there  are  few
negative measurements:  only an  unusually large (more than  x) negative
error produces a negative measurement.  On the other hand,  the  upper  part
of the distribution of measurements of this sample  might  resemble that for
blanks, shifted by x, since the  sources of positive errors  are  the  same.

As x moves away from zero, then, the  shape of the lower half of the dis-
tribution changes.  The lower half 1s likely to look more and more  like a
mirror-Image of the upper half.   The  upper half keeps about the same  shape
as for x - 0, a blank.

The problem 1s to find an x for which the lower B quantlle  of the distri-
bution 1s the decision limit.  With a true concentration  of x,  the  mea-
surements will be above the decision  limit except with  probability  e, and
so the analyte will be "detected" with probability 1 -  B.  This is  the
detection limit.

Figure 2-2 shows a solution.  It Is assumed that 1n the distribution of
measurements at true concentration D, the half above the median is  the
same as the upper half of the distribution of measurements  of blanks, but
shifted by D.  The lower half 1s a mirror-Image.  The detection limit D is
the sum of A, the distance from the median of measurements  of blanks to
the decision limit, and B, the distance from the lower  B  quantlle to the
median at true concentration D.  By symmetry, B equals  C, the distance
from the median to the upper B quantlle, and by assumption, this  distance
Is the same at D as for blanks.  Under these assumptions, the decision
limit can be calculated from the distribution of measurements of  blanks as

                        D - A + B

                          « A + C
where o^ and q. are the upper a and B quant lies and q 5 1s the median.  We
use B • o • .05, giving a detection limit of 2(P95 - ^g), I.e., twice the
difference between the 95th percent He and the median of the measurements
of blanks.

As with decision limits, 1t 1s more usual to calculate detection limits
from the standard deviation of the measurements of blanks.  The detection
limit is often reported as 3 s, where s  1s the standard deviation.  This
may be taken to correspond to our procedure with a decision limit of
T + 1.5 s, where T 1s the mean, and with the mean replacing the median.
The error rates, assuming a normal distribution, would be o » B * .07.  Of
BSiesr 3

                                   11
                                   80

-------
           Distribution of measurements
           of blanks (true concentration zero)
Distribution of measurements
at true concentration D
                                                                        Measured
                                                                        concentration
                                 Decision limit
          FIGURE 2-2.  Calculation of the detection limit (D).
          for discussion.
                See the text
65185
                                        12
                                         81

-------
course the same detection limit gives different combinations of o and 6
with different decision limits.  For example, with 7 + 3 s as the decision
limit and 3 s as the detection limit, and stm assuming normality, a -
0.003 but B « 0.5, which can hardly be called reliable detection.

Basing detection limits on the standard deviation has the same disadvan-
tages as apply for decision limits.  The standard deviation 1s very sensi-
tive to outliers.  Since the distributions are not normal, the error rates
cannot be known.  In addition, to assume normality 1s to assume the same
distributional shape for blanks and nonblanks; but, as we have argued, the
one 1s skewed and the other may be almost symmetric.  Still, we compute a
3 s detection limit for purposes of comparison.

For parameters that are not additive, like conductivity, the Interpreta-
tion of the detection limit 1s a little complicated; Indeed, the concept
of detection limits may not be very useful for conductivity.  Conductivity
different from that of a blank will be detected reliably 1f the conduc-
tivity of the sample exceeds that of a blank by more than the detection
limit.  For concentrations, on the other hand, the difference between the
true concentration when measured and the background added 1n processing 1s
simply the true concentration 1n the sample to begin with.  So the Inter-
pretation for concentrations 1s more straightforward:  a concentration
that 1s above the detection limit before processing will be detected
reliably.  The detection limits for all chemical parameters (except pH)
analyzed by the contract laboratories are given 1n Table 2-1.  Again we
show both the traditional  detection limit based on standard deviation and
the nonparametHc version  we recommend.
QUANTITATION LIMITS

If the true concentration 1s at the detection limit, the measured concen-
tration will probably be above the decision limit:  this 1s how those two
limits were chosen.  However, 1f the median background concentration 1s
small, the decision limit 1s only about half the detection limit.  So
negative relative errors of as much as 50 percent will occur with proba-
bility B 1n measuring such low true concentrations.  Smaller relative
errors are expected at higher true concentrations.  If the absolute stan-
dard deviation were the same as for blanks, the relative standard devia-
tion would be more than 10 percent for any true concentration up to 10 s
and less than 10 percent for any above 10 s.  This 1s only saying that s
1s more than 10 percent of any number up to 10 s, which has sometimes been
called the limit of quantltatlon (e.g., by Long and Wlnefordner, 1983).
We use 1t 1n the next section 1n computing estimates of precision:
samples believed to have true concentrations below the quantltatlon limit
are excluded from estimates of precision since high precision 1s not
expected at such low concentrations.  Quantltatlon limits are shown 1n
Table 2-1.
esiesr 3

                                   13
                                     82

-------
                           3   DUPLICATE SAMPLES
INTRODUCTION

At one lake each day, usually the first, each helicopter crew collected a
second sample.  This duplicate sample was processed by the field station
1n the usual way and shipped to the contract laboratory with a label coded
so that 1t was Indistinguishable from the routine samples 1n the batch.
Differences between measurements of these field duplicate samples reflect
analytical error and also errors at the lake and at the field station.
For example, the sampling procedure might vary slightly from one sample to
the next, or one of the samples might be contaminated 1n preparing the
allquots at the field station.

In addition, duplicate samples 1n each batch were created at the contract
laboratory by splitting one sample.  Differences 1n measurements of these
laboratory duplicates reflect only analytical errors at the contract
laboratory, and only variation within a batch and not from day to day.
For example, unless the Instrument 1s recalibrated within the batch, any
error 1n calibration will not be seen as a difference within a duplicate
pair.

Field and laboratory duplicate pairs are thus quite different 1n origin
and provide Information on different aspects of data quality.  Neverthe-
less, the form of the data 1s the same—pairs of measurements of the same
thing.  Our method of analysis 1s also the same for both, and so we
discuss them together here.
METHOD

Since the analysis 1s simplest for the three measurements of pH by the
contract laboratories, we begin there.  Errors 1n pH measurements are
believed to be about the same size for any pH over the relevant range; 1n
any case, since pH 1s already a logarithm, there 1s no point 1n taking
relative standard deviations.  The standard deviation of each duplicate
pair 1s
•siesr i,

                                   14
                                   83

-------
which can be written more simply as

                              s -  |R - D|//2

where R and D are the routine and duplicate measurements.  Now the ques-
tion arises of how to combine the standard deviations for duplicate pairs
to get an estimate of precision.  According to statistical theory, the
right summary statistic 1s the pooled or root-mean-square standard devia-
tion

                 RMS standard deviation - (••••

where n 1s the number of pairs.

To show the different effects of using this and other summary statistics
for the standard deviation, we did a Monte Carlo experiment.  Consider
an Imaginary lake with pH 5.08 and measurements of pH subject to errors
normally distributed with mean zero and standard deviation 0.03.
(Actually, this approximately describes B1g Moose Lake 1n the natural
audit measurements.)  We drew 10 sets of 100 pairs of samples from such an
Imaginary lake.  We calculated the standard deviation of each pair, and
then we computed four summary statistics for the 100 standard deviations
1n each set:  the root mean square (RMS), the mean, the median and the
geometric mean.  The results are shown 1n Table 3-1.  As we would expect
from statistical theory, the 10 root mean squares are clustered around the
actual measurement error of 0.03, and all three other summary statistics
are systematically too small.

For parameters other than pH there are two complications.  First, errors
are expected to be roughly proportional to concentration, at least at hlgti
concentrations.   It 1s therefore usual to report precision 1n the form  of
relative standard deviation  (RSD).  We calculate RSD for  each duplicate
pair by dividing  Us standard deviation by Its mean, and  again we use root
mean square as a  summary statistic for RSDs.  Second, the proportional
relationship breaks down at  very low concentrations:  as  concentration
approaches zero, measurement error does not vanish.  Therefore, RSDs may
be  very high at low concentrations.  What  1s  of  Interest  1s  the RSD  at
concentrations that are  not  extremely low.  We therefore  compute  the  RMS
RSD for pairs with mean  greater than the  quantltatlon limit,  defined  as  10
times  the  standard deviation of measurements  of  field blank  samples.

We  note that the  theoretical argument for using  the  RMS  of  RSDs,  unlike
standard deviations, 1s  only approximate.   However,  the  approximation 1s


esissr n

                                    15
                                     84

-------
                   TABLE 3-1.  Various summary statistics
                   for the standard deviation of duplicate
                   pairs 1n a Monte Carlo experiment.
RMS
0.030
0.027
0.029
0.033
0.030
0.031
0.030
0.031
0.030
0.026
Mean
0.024
0.022
0.024
0.026
0.025
0.024
0.024
0.025
0.024
0.022
Median
0.020
0.019
0.021
0.025
0.020
0.020
0.020
0.021
0.019
0.024
Geometric
Mean
0.018
0.014
0.017
0.017
0.017
0.017
0.017
0.018
0.016
0.016
                   Ten sets of 100 duplicate pairs are
                   represented.  All 2000 measurements
                   have a normal distribution with mean
                   5.08 and standard deviation 0.03.
                   The standard deviation of each pair
                   was calculated, and then four summary
                   statistics were calculated for the
                   100 standard deviations 1n each set.
esiesr s


                                   16
                                   85

-------
standard, and 1t Is reliable whenever RSDs are much less than 1 (100
percent), which 1s the case here.  Use of the RMS RSD has been recommended
by the Environmental Protection Agency for Its QA work (EPA, 1983).
RESULTS

Table 3-2 shows the estimates of precision for the three pH parameters.
As expected, the figures for field duplicates are slightly higher than
those for laboratory duplicates, reflecting some variability 1n the field
procedure or perhaps some real difference 1n the pH of water collected 1n
successive samples.  Most of the variability 1s already present 1n labora-
tory duplicates, however, and so must be attributed to analytical Impre-
cision.

Tables 3-3 and 3-4 show the estimates of relative precision of measure-
ments other than pH from field and laboratory duplicate pairs.  The pre-
cision (RMS RSD) 1s calculated for all pairs with a positive mean measured
concentration and for just those pairs with a mean above the quantltatlon
limit.  Again, the estimates from field duplicate pairs are generally a
little higher than those from laboratory duplicate pairs, as expected.
Extractable aluminum, sulfate, conductivity, and phosphorus are excep-
tions.

The estimates of precision from laboratory duplicate pairs for half a
dozen parameters, Including all these exceptions, are each heavily Influ-
enced by a single very high RSD.  Table 3-5 Identifies these outliers and
shows the RSD recalculated when they are eliminated.
BSlSSp K


                                   17
                                   86

-------
               TABLE 3-2.  W1thin-batch precision for pH
               measurements by contract laboratories
               (RMS standard deviation).

                                   Laboratory     Field
                                   Duplicates   Duplicates
                   Parameter	(N « 127)     (N •» 125)

               pH (equilibrated)     0.077        0.085

               pH (alkalinity)       0.042        0.052

               pH (acidity)          0.065        0.075
85185 5

                                 18
                                  87

-------
TABLE 3-3. Hlthln-batch precision of laboratory duplicates.



Parameter8
Calcium
Magnesium
Potassium
Sodium
Manganese
Iron
Aluminum
(extractable)
Chloride
Sulfate
Nitrate
Silica
Fluoride (total)
DOCC
Mean > 5 mg/1
Mean < 5 mg/1
Ammonl um
Acidity (peq/1)
Alkalinity (yeq/1 )
Conductivity (yS/cm)
DIG (equilibrated)
DIG (Initial)
Pnospnorus (total )
Aluminum (total)
Pairs

No. of
Pairs
123
121
125
123
102
120
123

127
127
122
127
127
126


121
112
116
125
126
127
125
126
with Mean > 0


RMS RSD (%)
0.89
0.63
18
13
21
14
13

23
11
13
2.5
2.5
6.8


7.1
28
30.
11
7.1
3.7
13
19

Quant1tat1on
Unit
(10 sB)b
0.15
0.039
0.37
0.20
0.093
0.12
0.039

0.39
1.7
3.4
2.3
0.034
2.5


0.56
150
57
9.0
0.56
1.1
0.037
0.28
Pairs With

No. of
Pairs
123
121
98
119
36
71
45

85
104
0
64
120
96
41
55
19
5
86
122
99
93
20
9
Mean > 10 SR


RMS RSD (X)
0.89
0.63
1.2
0.97
1.6
3.0
18

8.2
12
--
2.2
2.5
2.3
2.4
2.2
1.4
10.
2.1
11
2.4
2.4
19
5.4
a All  parameters are measured 1n mg/1  unless otherwise stated.
b SB 1s standard deviation of measurements of blank samples.
c Precision goals for DOC are different above and below 5 mg/1.
     85185P 5
                                        19
                                        88

-------
TABLE 3-4. H1tn1n-batcn precision of field



Parameter8
Calcium
Magnesium
Potassium
Sodium
Manganese
Iron
Al um1 num
(extractable)
Chloride
Sulfate
Nitrate
Silica
Fluoride (total)
DOCC
Mean > 5 mg/1
Mean < 5 mg/1
Ammonium
Acidity (ueq/1)
Alkalinity (yeq/1 )
Conductivity (viS/cm)
DIC (equilibrated)
QIC (Initial)
Phosphorus (total )
Aluminum (total)
Pairs

No. of
Pairs
125
125
125
125
82
118
112

123
125
116
120
125
125


113
118
121
125
124
125
122
125
wltn Mean > 0


RMS RSD (%)
2.3
2.3
4.7
4.2
57
97
39

15.
6.6
140
44
8.2
9.8


34
700
20.
1.9
12
6.9
37
35
duplicates.

Quant1tat1on
Limit
(10 sB)b
0.15
0.039
0.37
0.20
0.093
0.12
0.039

0.39
1.7
3.4
2.3
0.034
2.5


0.56
150
57
9.0
0.56
1.1
0.037
0.28

Pairs

No. of
Pairs
125
125
82
121
6
32
9

85
115
1
50
62
105
59
46
0
3
90
125
94
85
4
1

With Mean > 10 SR


RMS RSD (%}
2.3
2.3
3.7
4.3
11
10.
11

17
6.5
65
2.7
9.0
10.
12.
5.8
__
74
10.
1.9
5.0
3.7
9.7
24
8 All  parameters are measured 1n mg/1  unless otherwise stated.
b Sg 1s standard deviation of measurements of blank samples.
c Precision goals for DOC are different above and below 5 mg/1.
                                        20
                                        89

-------
TABLE 3-5.  Very high RSDs 1n laboratory  duplicate  pairs with  a mean
•easuement above the quantHatlon  limit.
                                            __

                                           Highest      Root Mean Square
                     Highest RSD             RSD        RSD Eliminating
   Parameter          (percent)   Batch   (percent)    this Pair  (percent)

Aluminum (extractable)   118       518       9.7               3.3

Chloride                  73       105       7.1               2.1

Sulfate                  117       317       5.9               1.3

Acidity                   21       605       1.7               1.3

Conductivity             116       510       3.7               0.72

Phosphorus (total)        84       500       6.8               2.4
   85185  5

                                     21

-------
                         4   NATURAL AUDIT SAMPLES
 INTRODUCTION

 Large  quantities  of water were collected from B1g Moose Lake and Lake
 Superior  some months before the  Eastern Lake Survey.  Two-liter samples of
 this water were shipped dally to the field stations, where they were
 handled like the  routine samples; the field stations prepared allquots and
 shipped them to the contract laboratories with coded labels along with the
 day's  routine samples.

 Measurements of these natural audit samples give Information about two
 aspects of the measuring process and consequently about the quality of the
 routine data.  First, they are measurements by different laboratories of
 the same  thing, provided the actual composition of the samples 1s con-
 stant.  If the measurements by these laboratories differ systematically,
 we cannot tell which of two laboratories 1s right, 1f either, because we
 have no way of knowing exactly wnat 1s 1n the natural audit samples.  We
 can. however, estimate the relative bias, I.e., the adjustment required to
 make measurements from the different laboratories comparable.  Second,
 random variations 1n repeated measurements of the same thing are an Indi-
 cation of the precision of the measurements, again assuming the actual
 composition 1s constant.
METHOD

Measurement bias between laboratories, measurement precision, and changes
1n the samples over time are all related.  All three are visible 1n Figure
4-1, whlcn shows the measurements of Initial pH from the acidity tltratlon
of samples from B1g Moose Lake  (type FN2) performed by EMSI and Versar.
On the whole, 1n Figure 4-1 the E's are above the V's:  there 1s an upward
bias of the measurements by EMSI relative to those by Versar, or, equlva-
lently, a downward bias by Versar relative to EMSI.  There 1s an upward
trend 1n the measurements over time:  later ones are rather higher than
earlier ones on average.  There 1s also some scatter, or Imprecision, 1n
measurements by the same laboratory at about the same time.
esissr 7

                                   22
                                  91

-------
o
10
                                                                  EMSI
                                                                         Versar
6

9

12

15

18
October
21

24
Batch
27
Date
30

2
5 8 11
	 November — -
14
     FIGURE 4-1.   Inter!aboratory bias  and trend 1n pH of natural  audits from
     Big Moose Lake (EMSI  and Versar only).
     85185
                                        23
                                       92

-------
 The  upward  trend 1n  measurements  over time  1s  somewhat  hidden  by the
 Interlaboratory  bias.  The  latest measurements are by Versar and so
 probably a  little lower  than  1f they had  been  by  EMSI.  The trend would
 have looked stronger 1f  the latest measurements had been higher, and they
 probably would have  been higher but for the bias.  Indeed the  trend 1n
 each of  the laboratories separately 1s stronger than when they are com-
 bined.

 Conversely, the  Increase 1n measured values over  time hides some of the
 bias.  Since the measurements by  Versar are on average  slightly later than
 those by EMSI, they  are  probably  slightly higher  than 1f they  had been
 performed earlier.   Thus, the bias shows up a  little less 1f the averages
 of all measurements  by EMSI and by Versar are  compared  than 1f measure-
 ments by the two laboratories around the same time are  compared.

 It seems reasonable  to estimate bias and measurement trend simultaneously
 by the technique known as analysis of covariance.  Separate but parallel
 least-squares lines  are  fit for the two laboratories.   Bias 1s estimated
 by the vertical  distance between  the lines:  this 1s the average differ-
 ence between contemporaneous measurements.  The trend 1s estimated by the
 common slope, which  1s the  slope  of the measurements corrected for bias.

 Precision can be estimated  two ways.  If the estimates  of bias and trend
 are  assumed to be correct, measurements can be corrected for both.  The
 precision of such corrected measurements 1s estimated by the standard
 deviation of the audit measurements around their  respective least-squares
 lines.   On  the other hand,  the estimated bias and trend are small compared
 to the residual  scatter.  It might, therefore, be just  as well not to
 correct  the measurements.   In this case the precision 1s estimated simply
 by the standard  deviation of the  audit measurements.

 The  picture 1s much  less  neat when the other two  laboratories  are added
 (Figure  4-2).  Versar's  and EMSI's measurements are 1n  October and early
 November, but USGS's are  1n December.  The analysis of  covariance extrapo-
 lates the linear  trend from October and November Into December.  If that
were true,  the USGS measurements  should be very high, unless they are
 biased.   Accordingly, a  very large negative bias 1s estimated  for USGS's
to account  for their being about  the same as Versar's on average.

There 1s  no evidence 1n the data  of a linear upward trend Into December.
On the other hand, there 1s no way to distinguish a change 1n  the trend
from a bias at USGS because there are no other measurements contemporane-
ous with  USGS's.   The best we can do 1s to estimate bias Ignoring trend,
by analysis of variance,  as well  as controlling for trend by analysis of
covariance.  The  disparity of the estimates of bias for USGS by the two
•5185 7


                                24



                                  93

-------
u
10
                                                          EMS I
                                                                            USGS
      1   6   11   16   21   26  31    5   10  15   20   25  30    5   10  15
      	October	    	November	    -December-

                                    Batch Date



      FIGURE  4-2.   Inter!aboratory  bias and trend 1n pH of natural audits from

      B1g  Moose  Lake  (four  laboratories).
      B5185
                                       25
                                       94

-------
methods 1s a warning of the uncertainty of these estimates.  Incidentally,
the analysis of variance provides a third estimate of precision—correct-
ing for bias but Ignoring trend.
RESULTS

Precision

Estimates of precision for parameters other than pH are shown 1n Table 4-1.
Since precision for these parameters 1s believed to be roughly propor-
tional to concentration, 1t 1s reported as a relative standard deviation
(RSD), which 1s the appropriate standard deviation divided by the mean.
(The standard deviation can be the ordinary standard deviation. Ignoring
bias and trend; within-laboratory standard deviation from the analysis of
variance, correcting for bias; or residual standard deviation from the
analysis of covaMance, correcting for bias and trend.)  For Lake Superior
there are only seven samples, too few to estimate bias or trend; thus only
the estimated precision Ignoring bias and trend 1s reported.  Table 4-1
also shows the means and the quantltatlon limit.  If the concentration 1s
below or near this limit, a high RSD 1s expected since large relative
errors may occur at low concentrations.  Iron, for example, 1s simply not
found 1n determlnable quantities 1n the natural audit samples.  The esti-
mates of precision for Iron should therefore not be taken as representa-
tive of the precision with which Iron 1s measured at higher concentra-
tions.

On the whole, correcting for bias alone or for both bias and trend makes
little difference In the estimates of precision.   The biases and trends
are small compared to the random variation even though they are sometimes
large enough and systematic enough to be statistically significant.

For pH, since precision 1s more or less constant over the relevant range,
there 1s no need for RSDs.  (Anyway, absolute differences 1n pH already
represent relative differences 1n hydrogen 1on concentration.)  Therefore
the estimates of precision for pH are 1n pH units rather than percents.
Also, there 1s no problem of quantltatlon limits for the pH measure-
ments.  The estimates of precision for pH are 1n Table 4-2.
Bias

The estimates of bias, both correcting for trend and Ignoring trend, are
shown 1n Tables 4-3 and 4-4.  Like precision, bias might be supposed to be
roughly proportional to concentration; accordingly, we show bias as a
percentage of the mean.  Again, pH measurements are an exception and are
85185 7

                                 26
                                    95

-------
TABLE 4-1.  Overall precision estimated from natural  audits.
Parameter*
Calcium
Magnesium
Potassium
Sodium
Manganese
Iron
Aluminum
(extractable)
Chloride
Sulfate
Nitrate
Silica
Fluoride (total)
DOC
Ammonium
Acidity (peq/1)
Alkalinity (ueq/1)
Conductivity (pS/cm)
QIC (equilibrated)
DIC (Initial)
Phosphorus (total )
Aluminum (total)


Ignoring
Bias and
Trend
9.2
2.7
4.3
5.1
26
66
34

51
7.3
25
8.1
4.1
9.3
54
40.
260
3.6
60.
27
140
28
B1g Moose Lake
Rsn(t)
Correcting
for
Bias
6.5
1.9
4.4
4.7
23
65
32

50.
7.4
26
7.3
3.9
8.7
50.
36
246
2.8
53
17
140
22
(N - 41)

Correcting
for Bias
and Trend
6.4
1.9
4.4
4.4
23
65
32

50.
7.2
26
7.4
3.9
8.8
50.
34
249
2.8
53
15
128
21

Mean
1.9
0.35
0.49
0.67
0.070
0.02
0.18

0.61
6.9
1.6
4.3
0.077
3.3
0.059
51
2.4
27
0.19
0.42
0.002
0.31
Lake Superior
RSO(t)
Ignoring
Bias and
Trend
5.5
3.6
12
6.4
360
140
86

2.0
3.3
5.8
6.0
5.1
9.7
130
63
2.3
1.7
3.8
4.2
140
120
(N - 7)
Mean
13
2.8
0.49
1.3
0.003
0.003
0.002

1.4
3.3
1.4
2.7
0.035
1.4
0.007
30.
850
96
9.7
9.9
0.001
0.021
Quantltatlon
Limit
0.15
0.039
0.37
0.20
0.093
0.12
0.039

0.39
1.7
3.4
2.3
0.034
2.5
0.56
150
57
9.0
0.56
1.1
0.037
0.28
* All parameters are measured In mg/1 unless other-wide  stated.
     85185 5

-------
TABLE 4-2.  Overall precision of pH measurements 1n natural audits.

                                Big Moose Lake (N * 41)Lake Superior
                           	Precision	         (N * 7)	
                                                        Correcting            Precision
                           Ignoring Bias   Correcting    for Bias           Ignoring Bias
    Parameter       Mean     and Trend      for Bias    and Trend    Mean     and Trend

pH (equilibrated)   5.18        0.27          0.27         0.28      8.23        0.08

pH (alkalinity)     5.07        0.04          0.04         0.03      7.76        0.12

pH (acidity)        5.08        0.05          0.04         0.03      7.79        0.11
      85185 5
                                      28
                                          97

-------
TABLE 4-3.  Relative Interlaboratory bias (expressed as a percent of the mean) estimated  from  natural  audits and controll-
ing for measurement trend (estimate t standard error of estimate).

Calcium
Magnesium
Potassium
Sodium
Manganese
Iron
Aluminum
(extractable)
Chloride
Sulfate
Nitrate
Silica
Fluoride (total)
DOC
Ammonium
Acidity
Alkalinity
Conductivity
DIC (equilibrated)
OIC (Initial)
Phosphorus (total)
Aluminum (total)
EMS I - Global
0.6 i 4.4
0.1 t 1.3
-0.2 t 3.0
0.7 t 3.0
-40.5 t 15.5
47.6 t 44.5
21.1 t 21.5

14.8 t 33.7
-0.5 t 4.9
-0.4 t 17. -4
-12.9 J 5.1
-4.3 t 2.7
-4.5 ± 6.0
-51.9 t 34.1
77.9 ± 22.9
-27.8 ± 169.7
3.4 J 1.9
-102.5 t 36.0
-79.8 i 10.4
107.6 ± 87.2
47.7 ± 14.6
EKSI - USGS
-32.0 ± 6.3
-5.7 t 1.8
-0.1 t 4.3
4.1 t 4.3
22.0 t 22.3
0.2 t 64.0
-31.3 * 31.0

-3.6 t 48.6
-13.7 t 7.1
-18.5 i 25.1
-0.7 t 7.3
-0.9 t 3.9
-15.6 t 8.6
-30.1 t 49.1
119.1 ± 33.0
385.8 t 244.4
-9.2 t 2.8
-15.1 i 51.9
6.0 t 15.0
161.0 t 125.7
90.0 t 21.1
EHSI - Versar
-1.1 i 2.2
-3.8 t 0.6
2.1 t 1.5
3.2 t 1.5
6.3 t 7.7
12.5 t 22.1
-16.9 t 10.7

-30.2 i 16.8
-1.5 t 2.4
5.1 i 8.7
3.2 ± 2.5
-2.9 t 1.3
2.5 ± 3.0
40.5 t 16.9
4.7 ± 11.4
146.4 t 84.3
0.5 i 1.0
2.0 ± 17.9
-14.0 + 5.2
-1.0 i 43.3
23.6 1 7.3
Global - USGS
-32.6 t 6.1
-5.8 t 1.8
0.1 t 4.2
3.4 t 4.2
62.6 t 21.6
-47.3 t 61.9
-52.4 ± 30.0

-18.4 + 47.0
-13.2 i 6.9
-18.1 i 24.3
12.2 J 7.0
3.4 + 3.7
-11.1 t 8.4
21.8 t 47.5
41.2 t 31.9
413.6 t 236.4
-12.6 t 2.7
87.4 i 50.2
85.7 + 14.5
53.4 t 121.5
42.3 1 20.4
Global - Versar
-1.8 l 4.4
-3.9 i 1.3
Z.3 ± 3.0
2.4 ± 3.0
46.8 i 15.4
-35.1 t 44.3
-38.0 t 21.4

-45.0 t 33.6
-1.0 t 4.9
5.4 i 17.4
16.1 i 5.0
1.4 t 2.7
7.0 t 6.0
92.4 ± 34.0
-73.2 t 22.9
174.2 ± 169.1
-2.9 t 1.9
104.5 t 35.9
65.8 + 10.4
-108.6 t 87.0
-24.1 t 14.6
USGS - Versar
30.8 ± 6.3
1.9 t 1.8
2.2 t 4.3
-0.9 t 4.3
-15.8 t 22.1
12.2 t 63.6
14.4 t 30.8

-26.6 t 48.2
12.2 J 7.0
23.6 t 24.9
3.9 + 7.2
-2.0 ± 3.8
18.1 i 8.6
70.7 t 48.7
-114.4 t 32.8
-239.4 t 242.6
9.7 ± 2.8
17.1 t 51.5
-20.0 l 14.9
-162.0 t 124.7
-66.4 t 20.9

-------
TABLE 4-4.  Relative 1nterlaboratory bias (expressed as a percent of the mean) estimated fro* natural audits and  Ignoring
measurement trend (estimate t standard error estimate).

Calcium
Magnesium
Potassium
Sodium
Manganese
Iron
Aluminum
(extractable)
Chloride
Sulfate
Nitrate
Silica
Fluoride (total)
DOC
Ammonium
Acidity
Alkalinity
Conductivity
DIC (equilibrated)
DIC (Initial)
Phosphorus (total)
Aluminum (total)
EMSI - Global
3.0 4 4.1
0.8 i 1.2
0.4 ± 2.1
-2.4 i 2.9
-50.0 4 14.4
66.3 t 40.8
34.4 4 20.0

-3.0 t 31.2
2.6 4 4.6
10.4 i 16.2
-11.9 * 4.6
-4.5 ± 2.4
-3.7 4 5.4
-42.1 ± 31.0
55.8 1 22.3
-50.7 t 153.5
4.2 4 1.8
-113.5 t 32.8
-68.2 t 10.4
6.9 1 87.3
37.8 t 13.7
EMSI - USGS
-25.5 4 4.1
-3.8 t 1.2
1.4 t 2.7
-4.7 t 2.9
-4.0 t 14.4
51.9 i 40.8
5.4 ± 20.0

-52.9 1 31.2
-5.4 + 4.6
11.3 t 16.2
2.0 4 4.6
-1.4 t 2.4
-13.4 ±5.4
-3.2 t 31.0
57.9 t 22.3
322.7 t 153.7
-7.1 ± 1.8
-45.4 t 32.8
38.0 t 10.4
-117.3 4 87.3
62.7 ± 13.7
EMSI - Versar
-1.0 + 2.2
-3.8 4 0.6
2.2 t 1.5
3.0 t 1.6
5.9 * 7.8
13.3 t 22.1
-16.3 t 10.9

-30.9 t 16.9
-1.4 f 2.5
5.5 t 8.8
3.3 4 7.5
-3.0 4 1.3
2.5 4 2.9
40.9 4 16.8
3.8 t 12.1
145.4 4 83.2
0.5 4 1.0
1.5 i 17.8
-13.5 4 5.6
-5.3 4 47.4
23.1 ± 7.4
Global - USGS
-28.5 J 5.3
-4.5 4 1.6
1.1 t 3.6
-2.2 4 3.9
46.0 4 18.9
-14.4 4 53.4
-29.0 4 2fi.2

-49.9 4 40.8
-7.9 + 6.0
0.9 4 21.3
13.9 4 6.0
3.0 4 3.2
-9.7 4 7.1
39.0 4 40.6
2.1 i 29.2
373.3 4 200.9
-11.2 4 2.3
68.1 i 42.9
106.2 ± 13.6
-124.2 4 114.3
24.9 ± 17.9
Global - Versar
-4.0 4 4.1
-4.5 4 1.2
1.8 4 2.7
5.5 4 3.0
55.8 4 14.5
-53.0 4 40.9
-50.7 4 20.1

-27.9 4 31.3
-3.9 4 4.6
-4.9 4 16.3
15.1 4 4.6
1.5 4 2.4
6.2 4 5.4
83.1 4 31.1
-52.0 t 22.4
196.1 4 154.1
-3.7 4 1.8
115.0 4 32.9
54.7 4 10.4
-12.2 4 87.7
-14.6 ± 13.8
USGS - Versar
24.5 4 4.1
-0.0 t 1.2
0.7 t 2.7
7.7 4 3.0
9.9 4 14.5
-38.6 + 40.9
-21.7 4 20.1

21.9 4 31.3
4.0 4 4.6
-5.8 4 16.3
1.2 4 4.6
-1.5 4 2.4
15.9 4 5.4
44.1 4 31.1
-54.1 t 22.4
-177.2 4 154.1
7.6 4 1.8
46.9 t 32.9
-51.5 4 10.4
112.1 4 87.7
-39.5 t 13.8

-------
Shown separately 1n Table 4-5.  Each estimate of relative bias between two
laboratories 1s accompanied by Its standard error, a measure of the sta-
tistical uncertainty 1n the estimate.  Again, we caution that tne uncer-
tainty for estimates Involving USGS may be much greater because bias and
trend cannot be distinguished 1n the data from this laboratory.  The
magnitude of this sort of uncertainty 1s suggested by the difference
between corresponding estimates Ignoring and controlling for trend.

There are three different standards against which the biases should be
measured; tne biases are on the whole rather small on all three scales.
First, the biases are small compared to the precision of the Individual
measurements.  This 1s why correcting for bias did not Improve precision
much.  Correcting for bias Improves the precision of Initial dissolved
Inorganic carbon from 27 to 17 percent, the largest change among the
parameters 1n this study but still only about a third.  For most
parameters the effect 1s much smaller.  Thus, bias 1s small 1n the sense
that 1t does not contribute much to the Inaccuracy of Individual measure-
ments.

Second, bias 1s a small percentage of the average measured concentration
1n most cases.  Most of the exceptions fall Into three categories:  (a)
estimates Involving USGS, which are not very reliable; (b) estimates for
parameters for which precision 1s also relatively poor, e.g., chloride;
and  (c) estimates for parameters that are not present 1n the samples 1n
amounts large enough to measure accurately, e.g., Iron.

Finally, 1n many but not all cases the estimated bias 1s small compared to
the standard error of the estimate.  For example, the relative bias for
calcium measurements by EMSI and Versar 1s about 1 percent, but this esti-
mate 1s subject to an uncertainty of about 2 percent.  In such cases the
evidence that there Is any bias at all 1s not statistically significant.
Trend

Table 4-6 shows the estimates of linear trends 1n the measurement of the
parameters over time.  Like the biases, the trends are small compared to
(a) the precision of the measurements, (b) the measurements themselves,
and (c) usually, but not always, the uncertainty 1n estimating them.
Statistically significant trends (t-test, p < .05), all upward, are seen
for sodium, acidity, Initial dissolved Inorganic carbon, total phosphorus,
and Initial pH from both acidity and alkalinity tUratlons.
85185T 7


                                   31
                                    100

-------
TABLE 4-5a.  Interlaboratory bias (expressed as pH units) estimated from natural audits:  controlling for measurement
trend (estimate t standard error of estimate).

pH
pH
pH

(equilibrated)
(alkalinity)
(acidity)
TABLE 4-5b. Ignoring

pH
pH
pH

(equilibrated)
(alkalinity)
(acidity)
EMSI
0.1
-0.02
0.04
- Global EMSI
t o.
40.
± 0.
2 0.02
02 0.13
02 0.16
- USGS
i 0.
i 0.
t 0.
3
03
03
EMSI - Versar
0.1 i 0.1
0.02 ± 0.01
0.06 ± 0.01
Global
0.1
0.15
0.12
- USGS
t 0.3
t 0.03
± 0.03
Global - Versar
0.0 ± 0.2
0.04 t 0.02
0.03 t 0.02
USGS - Versar
-0.1 + 0.3
-0.11 t 0.03
-0.10 ± 0.03
measurement trend.
EMSI
0.1
-0.05
0.00
- Global EMSI •
1 0.2 0.2
t 0.02 0.03
* 0.02 0.05
• USGS
t o.
2
t 0.02
t 0.02
EMSI - Versar
0.1 t 0.1
0.02 i 0.01
0.06 + 0.01
Global
0.0
0,08
0,05
- USGS
f 0.2
i 0.03
1 0.03
Global - Versar
0.0 J 0.2
0.08 + 0.02
0.06 i 0.02
USGS - Versar
0.0 + 0.2
-0.01 i 0.02
0.01 + 0.02

-------
         TABLE 4-6.  Trends (per month) 1n measurements of
         natural  audit samples (Big Moose Lake).
85185 5
Parameter*
Calcium
Magnesium
Potassium
Sodium
Manganese
Iron
Aluminum (extractable)
Chloride
Sulfate
Nitrate
Silica
Fluoride (total)
DOC
Ammonium
pH (equilibrated)
pH (alkalinity)
pH (acidity)
Acidity (ueq/1)
Alkalinity (weq/1)
Conductivity (uS/cm)
DIC (equilibrated)
DIC (Initial)
Phosphorus (total )
Aluminum (total)
Estlnated
Trend
-0.08
-0.004
0.00
0.03
0.012
-0.007
-0.04
0.20
-0.4
-0.3
-0.1
0.000
0.0
-0.011
0.0
0.07
0.07
20.
1
-0.4
0.04
-0.09
0.003
0.05
Standard Error
of Estimate
0.06
0.003
0.01
0.01
0.008
0.007
0.03
0.15
0.3
0.2
0.2
0.002
0.1
0.015
0.1
0.01
0.02
9
3
0.4
0.05
0.03
0.001
0.03
         * All  parameters measued 1n mg/1 unless otherwise noted.
                                   33
                                 102

-------
                            SYNTHETIC AUDIT SAMPLES
 INTRODUCTION

 Four  kinds  of  synthetic audit samples were used 1n the Eastern Lakes Sur-
 vey.   All synthetic audits were mixed and diluted from stock solutions
 according to one of two recipes, one giving relatively high concentra-
 tions, the  other low.  "Field synthetic audits," like the natural audits,
 were  sent dally 1n two-liter containers to the field stations, where they
 were  processed Into allquots by the usual procedure.  "Laboratory syn-
 thetic audits" arrived at the field station as allquots; the field sta-
 tions, without opening the allquots, simply relabeled them and shipped
 them  with the  routine samples to the contract laboratories.

 In addition to their use 1n quality control, the synthetic audits have
 three uses  1n  quality assessment:  estimation of precision and Inter-
 laboratory  bias; comparison of measured with theoretical concentrations;
 and evaluation of performance of field stations by comparison of field and
 laboratory  audits.  Each of these three subjects 1s discussed 1n turn.
BIAS AND PRECISION

In principle, the synthetic audits, like the natural audits, can be used
to estimate precision and Inter!aboratory bias.  Differences 1n measure-
ments of the same thing 1n the same laboratory at different times Indicate
how precise those measurements are.  Systematic differences 1n measure-
ments of the same thing between different laboratories Indicate a relative
bias between the laboratories.  Unfortunately, there 1s ample evidence
that the synthetic audit data are not repeated measurements of the same
thing.

Figure 5-1 shows the concentrations of ammonium measured 1n laboratory
high synthetic audit samples by three contract laboratories.  With one
exception (a measurement of about 1.9 mg/1 by Versar on 18 October) the
measurements on Individual days form very tight groups, but the groups are
spread far apart.  That 1s, measurements on a given day are 1n close
agreement both within and among laboratories, but on different days they
85185r e


                                  34
                                   103

-------
                                                      E - EMSI
                                                      V - Versar
                                                      G - Global
                                                      U - USGS
    11  15   19   23   27   31   4    8   12  16   20   24  28    2    6   10
              October                      November                December
                                 Batch Date
FIGURE 5-1.  Concentration of ammonium measured in laboratory high synthetic
audit samples.
 65185                          35
                                  104

-------
are very different.  Probably the concentrations were very different on
different days, and Imprecision 1n measurement and bias among laboratories
were very small by comparison.

We would like to be able to quantify this comparison.  An appropriate
statistical technique 1s analysis of variance with fixed and random
effects.  Basically, differences between days are attributed to errors in
preparation of the audit samples, and differences within days are attribu-
ted to errors 1n measurement.  But we need to consider many variables, or
effects, as they are called in statistics.  There may be bias between
laboratories.  Measurements within a laboratory on the same day may be
less variable than those on different days.  There may be variation 1n
preparation of the several samples on the same day, though less than
across days.  There may be systematic as well as random errors in prepara-
tion across time.  With enough data all these effects could be estimated,
but in this study there are not enough data.  We did fit some random-
effects models, but we consider the results wholly unreliable.  The esti-
mates are extremely sensitive both to single data points (like the
exceptional 18 October ammonium concentration mentioned earlier) and to
the statistical method of estimation chosen from several equally reason-
able alternatives.  With such large errors 1n preparation there 1s simply
not enough information 1n the synthetic audits to estimate errors in mea-
surement.
MEASURED VERSUS THEORETICAL CONCENTRATIONS

Comparing the concentration measured 1n the synthetic audit samples with
the theoretical concentrations furnishes very useful confirmation that the
system 1s measuring what 1t was meant to measure.  Tables 5-1 and 5-2 show
the theoretical values and the means of the measurements of field and
laboratory, high and low synthetic audit samples (calcium and chloride are
divided by lot because the theoretical concentrations changed between
lots).  For most parameters there 1s reasonable 1f not very close agree-
ment between the measured and theoretical values.

The most notable exception 1s Iron, which 1s essentially absent from the
field audit samples but present 1n the laboratory audits.  (Although there
1s no theoretical concentration, a similar difference 1s seen between
field and laboratory audits for extractable aluminum; a smaller difference
1s seen for manganese).  Apparently Iron 1s removed by the processing 1n
the field.  Also, Initial dissolved Inorganic carbon is consistently above
Us theoretical value (Figure 5-2).  However, some confusion 1s added by
an apparent change in preparation around 23 October.  This change 1s also
85185 8
                                 36
                                 105

-------
TABLE 5-1. Field and laboratory high synthetic audits.
Parameter*
Calcium
Lot 4 (n - 20. 14)
Lots 5, 6 (n - 16. 7)
Magnesium
Potassium
Sodium
Manganese
Iron
Aluminum (extractable)
Chloride
Lot 4 (n - 20, 14)
Lots 5, 6 (n - 16, 7)
Sulfate
Nitrate
Silica
Fluoride (total)
DOC
Ammonium
pH (equilibrated)
pH (alkalinity)
pH (acidity)
Acidity (ueq/1)
Alkalinity (ueq/1)
Conductivity (uS/cm)
DIC (equilibrated)
DIC (Initial)
Phosphorus
Aluminum (total)
Field
Mean

1.66
1.95
2.37
3.03
11.84
1.19
0.001
0.037

3.66
4.00
14.39
1.816
9.23
0.445
10.01
1.256
7.72
7.08
7.12
73
477
104.5
4.72
5.94
0.057
0.199
(n - 36)
Standard
Deviation

0.11
0.16
0.09
0.19
1.03
0.16
0.099
0.027

1.45
0.73
0.80
0.666
1.26
0.042
1.96
0.162
0.54
0.25
0.27
56
82
4.2
1.28
1.98
0.013
0.059
Laboratory (n « 21)
Mean

1.55
1.96
2.36
3.02
11.87
1.39
0.175
0.161

3.39
5.81
14.47
1.431
9.42
0.435
10.25
1.342
7.87
7.05
7.11
85
486
104.3
4.79
5.94
0.061
0.195
Standard
Deviation

0.18
0.10
0.08
0.15
1.04
0.04
0.019
0.030

0.24
5.34
0.69
0.672
1.39
0.012
1.68
0.176
0.15
0.26
0.25
68
95
4.1
1.31
1.90
0.010
0.067
Theoretical

1.54
2.39
2.43
2.97
12.41
1.50
0.15
__

2.72
4.22
14.09
1.707
10.70
0.452
10.0
1.25
—
--
--
--
—
—
—
3.10
0.075
0.19
  Measured 1n mg/1 unless other noted.
65185 5
                                 37
                                    106

-------
TABLE 5-2. Field and laboratory low synthetic audits.

Calcium
Lot 4 (n - 20, 20)
Lots 5, 6 (n « 23, 23)
Magnesium
Potassium
Sodium
Manganese
Iron
Aluminum (extractable)
Chloride
Lot 4 (n « 20, 20)
Lots 5, 6 (n - 23, 23)
Sulfate
Nitrate
S111C8
Fluoride (total)
DOC
Ammonium
pH (equilibrated)
pH (alkalinity)
pH (acidity)
Acidity (peq/1)
Alkalinity (yeq/1)
Conductivity (uS/cm)
DIC (equilibrated)
DIC (Initial)
Phosphorus (total )
Aluminum (total)
Field (n - 36)
Standard
Mean Deviation

0.144 0.027
0.169 0.012
0.424 0.031
0.231 0.043
2.71 0.14
0.091 0.009
0.001 0.012
0.005 0.004

0.329 0.234
0.365 0.081
2.27 0.23
0.547 0.362
1.00 0.31
0.042 0.009
0.982 0.476
0.158 0.043
7.34 0.14
6.87 0.11
6.96 0.19
16.3 23.3
112 8.3
19.0 1.3
1.35 0.12
1.65 0.21
0.021 0.006
0.023 0.005
Laboratory (n « 21)
Mean

0.137
0.161
0.426
0.225
2.69
0.092
0.070
0.018

0.300
0.347
2.28
0.471
1.02
0.040
1.190
0.186
7.29
6.86
6.93
16.1
110
18.7
1.35
1.65
0.022
0.031
Standard
Deviation

0.015
0.014
0.019
0.023
0.25
0.014
0.016
0.006

0.042
0.109
0.10
0.172
0.20
0.002
0.724
0.041
0.13
0.12
0.15
27.3
10.3
1.2
0.15
0.24
0.009
0.028
Theoretical

0.13
0.19
0.45
0.20
2.79
0.10
0.06
--

0.22
0.34
2.28
0.466
1.07
0.042
1.0
0.17
—
._
—
—
—
--
--
0.96
0.027
0.02
* Measured 1n mg/1 unless otherwise noted.
85185 5
                                  38
                                    107

-------
•


T
^
Et
0
o»
H4
c
•"*
a
i


K


I


.

t V V

11 15 19 2
	 October
I . E - EKSI
' V « Versar
G - Global
U - USGS

fl
V
V
V
V
V

theoretical value

3 27 31 4 8 12 16 20 24 28 2 6 10
                              Batch Date
FIGURE 5-2.  Concentration of dissolved Inorganic carbon (Initial)
measured 1n laboratory high audit samples.  Measurements before and
after 23 October (vertical line) seem to be different.
851B5
                          39
                                   108

-------
noticed 1n several other parameters, especially sodium, silica (Figure
5-3), acidity, alkalinity, and pH, and may have been caused by a change 1n
the concentration of the stock solution of sodium silicate.

We do not think any detailed statistical analysis of the differences
between measured and theoretical concentrations 1s called for.  The like-
liest reason for any difference 1s not a problem 1n processing or analysis
but rather that the theoretical concentrations are not what was actually
1n the samples.
FIELD VERSUS LABORATORY AUDITS

It was hoped that by comparing the variability of the field audits, which
were processed by field stations, with that of the laboratory audits,
which were not, one could distinguish the amount of variability Introduced
at the field stations from the Imprecision of the analytical labora-
tories.  As 1t happens, both of these sources of variability are swamped
by variability 1n the preparation of synthetic audit samples.  The
standard deviations of measurements of the four types of synthetic audit
samples (field and laboratory, high and low) are In Tables 5-1 and 5-2.
On the whole the field audits are not any more variable than the labora-
tory audits.  That 1s, preparation of the allquots 1n the field Introduces
no more variability than preparation 1n the laboratory.  Again, finer
analysis 1s made Impossible by the magnitude of the errors 1n preparation
of synthetic audit samples.
8S185 8


                                40
                                   109

-------
r— *• *
                                                             E  -  EMSI
                                                             V  «=  Versar
                                                             G  -  Global
                                                             U  -  USGS
     11   15  19   23  27
     	October	
31
     8   12  16   20  24
        November	
Batch Date
28
2   6   10
—December
    FIGURE 5-3.  Concentration of silica measured 1n laboratory high
    synthetic audit samples.  Measurements before and after 23 October
    (vertical line) seem to be different.
    85185
                                   41
                                    no

-------
                    6   CONCLUSIONS  AND  RECOMMENDATIONS
WHAT THE QA DATA SHOW

One of the purposes of the QA data was to allow us to estimate  the
Imprecision of routine lake sample measurements.  In  this  report we  have
provided many estimates of various kinds of precision.  In this section we
relate all these estimates.  Table 6-1 shows our estimates of w1th1n-batch
laboratory precision, w1th1n-batch precision (Including errors  made  both
at the field stations and at the analytical laboratories), and  overall
precision.  The estimates of precision are for measurements above the
quantltatlon limit for each parameter, as we know that there are large
relative errors below the quantltatlon limits.

The findings of our analysis may be Illustrated by considering  together
the Information given by all the quality assurance data about one para-
meter.  We use calcium as the example.  The estimate  of measurement
precision from laboratory duplicate pairs for calcium 1s 0.89 percent.
Thus, the reprodudbUKy of determinations of calcium within any single
laboratory and single batch 1s excellent.  When precision  1s estimated
from field duplicate pairs, the estimate rises to 2.3 percent.  So,  when
the entire system of collection, field processing, and laboratory analysis
1s considered (and when the contract laboratory 1s blind to the dupli-
cates) reprodudblHty 1s much less but still very good.  This  reprodud-
bH1ty 1s still for measurements within the same laboratory and within  the
same batch.  These are likely to be done by a single  analyst, possibly  off
a single calibration, certainly on a single day, and  so under broadly
similar laboratory conditions.

Information about variation from day to day and from  laboratory to  labora-
tory comes from the natural audits.  Within the same  laboratory the  esti-
mated precision for the determination of calcium 1n B1g Moose  Lake  samples
1s 6.5 percent.  The measurements by the USGS laboratory are some 25 per-
cent higher than the others, either because of bias or because  of a  change
1n the samples over time; consequently 1f all B1g Moose Lake measurements
are lumped together without regard to laboratory or date,  the  estimate  of
precision rises to 9.2 percent.
Bsiesr 9

                                   42

-------
TABLE 6-1. Summary of estimates of precision (percent RSD, except for pH,
which 1s expressed as SD 1n pH units).
Natural Audits
B1g Moose Lake


Parameter*
Calcium
Magnesium
Potassium
Sodium
Manganese
Iron
Aluminum
(extractable)
Cnlorlde
Sulfate
Nitrate
S111ca
Florlde (total)
DOC
Ammo n1 urn
Acidity
Alkalinity
Conductivity
DIC (equilibrated)
DIC (Initial)
Phosphorus (total)
Aluminum (total)
pH (equilibrated)
pH (alkalinity)
pH (acidity)
Correcting
for Bias
and Trend
6.4
1.9
4.4
4.4
23
65
32

50.
7.2
26
7.4
3.9
8.8
50.
34
249
2.8
53
15
128
21
0.28
0.03
0.03
Ignoring
Bias and
Trend
9.2
2.7
4.3
5.1
26
66
34

51
7.3
25 S.S
8.1
4.1
9.3
54
40.
260
3.6
60.
27
140
28
0.27
0.04
0.05
Lake
Superior
Ignoring
Bias and
Trend
5.5
3.6
12
6.4
360
140
86

2.0
3.3
5.8
6.0
5.1
9.7
130
63
2.3
1.7
3.8
4.2
140
120
0.08
0.12
0.11


Field
Duplicates
2.3
2.3
3.7
4.3
11
10.
11

23
6.5
65
2.7
9.0
10.
._
74
10.
1.9
5.0
3.7
9.7
24
0.085
0.052
0.075


Laboratory
Duplicates
0.89
0.63
1.2
0.97
1.6
3.0
3.3

2.1
1.3
__
2.2
2.5
2.3
1.4
1.3
2.1
0.72
2.4
2.4
2.4
5.4
0.077
0.042
0.065
85185 5
                                   43
                                  112

-------
The rise from 0.89 to 2.3 to 6.5 to 9.2 percent as  we move from the
laboratory duplicates through the field duplicates  to the natural  audits
1s not surprising, and 1t demonstrates the Importance of the quality
assurance program.  As 1t happened, the Imprecision of a perfectly cali-
brated Instrument under constant conditions made only a small  contribution
to overall uncertainty, compared to errors 1n calibration and  variation
over time 1n conditions or procedures 1n the laboratory and 1n the
field.  These other effects can only be judged by audits treated as  much
as possible like routine samples under field conditions.  The  blind  field
natural audits serve this purpose well.  The estimates of precision  from
field natural audits are therefore probably the best Indicators of uncer-
tainty in the routine samples.

Along with calcium, the Big Moose Lake natural audits probably give
reasonable estimates of overall precision 1n routine samples for mag-
nesium, potassium, sodium, aluminum (extractable and total), chloride,
sulfate, silica, fluoride, dissolved organic carbon, and all three mea-
surements of pH by contract laboratories.  The levels of other parameters
1n 81g Noose Lake samples are below the quantltatlon limit, and the  rela-
tive standard deviations at such low concentrations should not be con-
sidered representative.  These parameters fall Into three classes.

     Dissolved Inorganic carbon (equilibrated and initial).  There Is
     quantifiable DIG 1n natural audit samples from Lake Superior.  The
     RSD for these samples can be used as an estimate of overall precision
     in place of that for Big Noose Lake.

     Alkalinity.  The alkalinity of samples from Lake Superior 1s well
     above the quantltatlon limit; Indeed, at 850 ueq/1 it 1s probably
     much higher than that of most lakes in the survey.  The precision of
     Lake Superior measurements is therefore not representative of routine
     measurements.  We therefore have no good estimate of overall pre-
     cision for alkalinity.  The closest we can come 1s the within-batcn
     estimate of 10 percent from field duplicates,  recognizing that  w1tn-
     1n-batch calculations usually underestimate overall imprecision.  On
     the other hand, the low synthetic audits, 1n spite of preparation
     error, also have RSDs around 10 percent, so that overall  Imprecision
     cannot be much higher.  The figure of 10 percent 1s therefore a rea-
     sonable estimate of precision of measurements  of alkalinity 1n  rou-
     tine samples.

     Nitrate, ammonium, acidity, phosphorus, Iron,  and manganese.  The
     levels of these parameters in the natural audits are below the
     quantltatlon limits.  So are the levels 1n almost all the lakes for
     which there are field or laboratory duplicates, and so presumably in
     almost all the routine samples.  Thus these things were not measured


esissr 9

                                   44
                                   113

-------
     very precisely simply because there was not enough of them to
     measure.  One could say that the lakes contained no measurable levels
     of these parameters, or that the system was Insufficiently sensitive
     to measure what was 1n the lakes; the matter will  be judged according
     to the quantltatlon limits.
SYNTHETIC AUDITS

The synthetic audit samples turned out not to be useful  for estimating
errors 1n measurement because these were much smaller than errors 1n
preparation.  In retrospect this 1s not surprising:   1t  1s hard to prepare
something more accurately than one can measure 1t.

The synthetic audits, however, are the only samples  that furnish one
Important kind of data, but not enough of them--measurements of split
samples by different laboratories.  Such measurements are very useful 1n
estimating Inter!aboratory bias.  For most parameters the stability of the
natural audits 1s sufficient for this purpose, but 1t would be preferable
to be able to compare measurements on the same day.   Also, the natural
audits are field audits and so Include effects of processing by field
stations, while laboratory synthetic audits could give Information about
bias between laboratories without Involving field station effects.

We recommend that 1n the future synthetic audits always  be treated as
split samples.  Dally lots should be prepared and divided Into allquots,
and the allquots should be assigned to different contract laboratories at
random.  The QA data base should Identify the dally lot  for each sample so
that comparisons between laboratories can be made using  data from the same
lot.  Variation among lots 1s then of little Importance.  Indeed 1t might
even be desirable to vary the composition of samples between lots, both to
prevent recognition of blind audit samples and to see precision and
accuracy at various concentrations.
DETECTION LIMITS

To calculate detection limits from measurements of blank samples, we had
to make certain unverlflable assumptions about the response of the measur-
ing system at low concentrations.  In particular, we assumed that the
calibration of the system was linear  (with slope 1) although the Intercept
need not be zero.  Thus, some background might b« added to samples 1n
processing, but the amount of background should be Independent of the
concentration.
 85185T 9


                                   45



                                      114

-------
He  believe that 1f detection limits are Important, they should be deter-
mined empirically.  That 1s, audit samples with a concentration near the
putative detection limit for an analyte ought to be measured, and 1t
should be seen whether the analyte 1s reliably detected or not.  Decision
limits, on the other hand, can be straightforwardly defined and accurately
estimated from blank data alone.
I5185r 9


                                   46


                                    115

-------
                                REFERENCES
Drouse, S. K., P. C. Hlllman, L. W. Creelman, J. F. Potter, and S. J.
     Simon.  1985.  [Draft] "Quality Assurance Plan for the National
     Surface Water Survey Project."  Lockheed-EMSCO. Las Vegas, NV.

EPA.  1983.  "Calculation of Data Quality Indicators."  U.S. Environmental
     Protection Agency, Office of Research and Development.

EPA.  1984.  National Surface Water Survey;  National Lake Survey-
     Phase I.  U.S. Environmental Protection Agency, Office of Research
     and Development, Washington, D.C. (NAPAP Project Reference Number
     EI-23).

Hubaux, A., and G. Vos.  1970.  Decision and detection limits for linear
     calibration curves.  Analytical Chemistry.  42:849.

Liggett, W.  1985.  [Draft] "The Instability of Nitrate Concentrations 1n
     Lake Samples."  National Bureau of Standards, Galthersburg, MD.

Long, 6. L., and J. D. Wlnefordner.  1983.  Limit of detection:  a closer
     look at the IUPAC definition.  Analytical Chemistry.  55:712A.
esiesr
                                   47
                                   116

-------
                               APPENDIX B
    INSTRUMENTAL DETECTION LIMITS, SYSTEM DETECTION LIMITS.
            AND SYSTEM DECISION LIMITS BY LABORATORY,
                   EASTERN LAKE SURVEY — PHASE I
Values were reported by the contract analytical laboratories or were estimated from field blank data
and calibration blank data.
                                   117

-------


Parameter
Al , extractable, mg L"1
Al , total, mg L"1
Ca, mg L-1
Cl~, mg L"1
Conductance, uS cm~l
DIG, air-equilibrated, mg L~*
DOC, mg L-1
F", total dissolved, mg L"1
Fe, mg L"1
K, mg L-l
Mg, mg L~l
Mn, mg L-1
Na, mg L~l
NH4+, mg L'1
N03", mg L"1
P, total , mg L~l
Si02, mg L"1
S042", mg L"1
TABLE B-
Reported
EMSI
0.002
0.002
0.01
0.00
-0.4C
0.02
0.1
0.003
0.03
0.00
0.00
0.01
0.00
0.01
0.006
0.001
0.03
0.01
1.
Instrumental
Global
0.002
0.002
0.00
0.00
-0.2C
0.04
0.1
0.001
0.02
0.00
0.00
0.01
0.00
0.00
0.003
0.001
0.02
0.01

Detection
USGS
0.003
0.002
0.00
0.00
0.5C
0.03
0.1
0.002
0.01
0.01
0.00
0.00
0.00
0.01
0.006
0.001
0.02
0.03

Limits
Versar
0.003
0.004
0.01
0.01
0.5C
0.003
0.1
0.002
0.01
0.01
0.00
0.00
0.00
0.01
0.006
0.001
0.02
0.03
aThree times the standard deviation of ten^nonconsecutive laboratory
 calibration blank measurements (see Drouse et al., 1986).
DA11 batches (aliquots 3 and 5).
cFor conductance, the mean of six nonconsecutive blank measurements
 was required to be less than 0.9 uS cm~l.
                                 118

-------
                                      611
                                          *(9  PUB e sq.onb.LLB)  sau.Dq.Bq LLVq
                                                             •sq.uauiajnsBaui
>)UBLq uo.Lq.Bjqi.LBD A~joq.BJoqBL i° 'aj^puaojad u.q.09 = OS
-------


Parameter
Al, extractable, mg L-1
Al, total, mg L'l
ANC, ueq L'1
Ca, mg L"1
C1-, mg L'1
Conductance, uS cm~l
DIC, air-equilibrated, mg L~l
DIC, initial, mg L"1
DOC, mg L'1
F~, total dissolved, mg L~l
Fe, mg L~l
K, mg L'1
Mg, mg L~l
Mn, mg L~l
Na, mg L-1
NH4+, mg L'1
N03~, mg L~lb
N03~, mg L~lc
NO-f, mg L~ld
P, total, mg L'1
Si02, mg L'1
S042-, mg L'1
TABLE B-3.

EMS I
0.011
0.59
12
0.08
0.24
1.8
0.20
0.34
0.6
0.072
0.02
0.04
0.00
0.02
0.20
0.46
0.096
0.049
0.118
0.018
0.14
0.14

System Detection
Global
0.00
0.088
6.0
0.10
0.04
1.0
0.34
0.84
1.0
0.060
0.00
0.06
0.02
0.14
0.02
0.10
0.082
0.091
0.066
0.022
0.34
0.06

Limit*
USGS
0.004
0.708
3.4
0.02
0.06
1.0
0.12
0.08
0.6
0.002
0.00
0.02
0.02
0.00
0.04
0.05
0.048
0.048
—
0.002
0.02
0.10


Versar
0.006
0.036
6.2
0.06
1.4
0.8
0.16
0.28
0.4
0.000
0.00
0.02
0.00
0.00
0.20
0.08
0.270
0.045
0.341
0.010
0.40
0.10
aSystem detection limit = 2(Pgg - PCQ).  where Pgg =  95th  percentile,
 and PSQ = 50th percentile,  of field blank measurements.
bAll batches (aliquots 3 and 5).
cAliquot 3 after filtration  protocol change.
^Aliquot 5.
                                     120

-------


Parameter
AT, extractable, mg L"1
AT, total, mg L"1
ANC, ueq L'1
Ca, mg L"1
C1-, mg L'1
Conductance, uS cm~l
DIG, air equilibrated, mg L"1
DIG, initial, mg L"1
DOC, mg L~l
F~, total dissolved, mg L~l
Fe, mg L~l
K, mg L'1
Mg, mg L~l
Mn, mg L"1
Na, mg L'1
NH4+, mg L"1
N03", mg L~lb
N03", mg L"lc
N03~, mg L"ld
P, total, mg L"1
Si02, mg L"1
S042-, mg L"1
aSystem decision limit = Pg5,
measurements.
TABLE

EMSI
0.008
0.030
10.7
0.04
0.13
0.6
0.28
0.46
0.5
0.038
0.01
0.02
0.00
0.01
0.10
0.24
0.056
0.019
0.068
0.010
0.08
0.10
where Pgg
B-4.
System Dec
Global
0.004
0.045
3.9
0.05
0.02
0.6
0.33
0.69
0.6
0.020
0.00
0.03
0.01
0.07
0.01
0.05
0.041
0.046
0.033
0.014
0.17
0.04
= 95th percenti

ision Limit3
uses
-0.002
0.365
-8.9
0.01
0.04
1.4
0.24
0.27
0.5
0.001
0.00
0.01
0.01
0.00
0.00
0.02
0.024
0.024
—
0.001
0.01
0.10
le of field


Versar
0.004
0.023
3.1
0.03
0.10
1.4
0.26
0.30
0.4
0.000
0.00
0.01
0.00
0.00
0.10
0.04
0.131
0.029
0.167
0.005
0.15
0.08
blank
bAll batches (aliquots 3 and 5).
cAliquot 3 after filtration protocol change.
^Aliquot 5.
                                     727

-------
                             APPENDIX C
                  OVERALL WITHIN-BATCH PRECISION
                BY LABORA TORY FOR 23 PARAMETERS,
                  EASTERN LAKE SURVEY — PHASE I
Values were estimated from field duplicate and field blank data
                                  123

-------
TABLE C-l.

Overall Wi thin-Batch
Pairs with

Parameter
Al , extractable, mg L"1
all values
x <= 0.010
x > 0.010
Al , total , mg L-1
all values
7 <= 0.010
x > 0.010
ANC, ueq L"l
Ca, tng L"l
CT, mg L'1
Conductance, pS cm"1
DIC, mg L-1
air-equilibrated
initial
DOC, mg L'1
all values
x <= 5
7 > 5
Fe, mg L"1
F~, total dissolved, mg L~l
K, mg L-1
Mg, mg L"j
Mn, mg L"1
Na, mg L'1
NH^, mg L"|
N03~, mg L'1
all batches
aliquot 3
aliquot 5
PH
air-equilibrated
initial ANC
initial BNC
P, total, mg L'l
all values
7 <= 0.010
7 > 0.010
Si02, mg L"1
SO/', mg L'1

n

47
26
21

51
0
51
50
51
51
51

51
51

51
29
22
49
51
51
51
46
51
51

50
12
38

51
51
50

48
33
15
51
51
Mean > 0
RMS of SRSDa

47
61
16

39
—
39
19
2.6
21
1.8

3.9
7.6

14
8.0
20
141
9.4
6.5
3.3
69
1.9
18

308
6.1
354

0.07
0.04
0.06

40
45
27
20
2.9

EMSI
Precision
Pairs with
Mean > lOse
n

1
0
1

3
0
3
47
51
37
51

44
44

40
18
22
4
34
51
51
4
49
0

7
4
5

—
—
—

0
0
0
42
51
RMS of ZRSDa

24
—
24

36
—
36
10
2.6
24
1.8

2.2
5.3

15
6.5
20
14
10
6.5
3.3
13
1.9
—

0.85
1.7
0.83

--
—
—

—
—
—
20
2.9






Quantisation
Limit
n

99



99


99
99
99
99

99
99

99


99
99
99
99
99
99
97

98
21
77

--
--
—

99


99
99
10sR

0.046



0.187


29.1
0.13
0.45
5.9

0.52
0.91

2.7


0.18
0.029
0.10
0.03
0.06
0.28
0.80

0.182
0.009
0.184

— —
—
—

0.005


0.04
0.31
aRoot mean square of percent relative standard
pH).
deviation (RMS of absolute SD for
                                        124

-------
                                   TABLE C-2.
                                                  GLOBAL
Overall Within-Batch
Pairs with
Mean > 0
Parameter
Al , extractable, mg L"1
all values
7 <= 0.010
x > 0.010
Al, total, mg L'1
all values
7 <= 0.010
x > 0.010
ANC, ueq L'l
Ca, mg L"1
CT, mg L'1
Conductance, uS cm"1
DIC, mg L"1
air-equilibrated
initial
DOC, mg L"1
all values
x <= 5
7 > 5
Fe, mg L"1
F~, total dissolved, mg L'1
K, mg L"1
Mg, mg L"1
Mn, mg L"1
Na, mg L"1
NH4+, mg L'1
N03~, mg L"1
all batches
aliquot 3
aliquot 5
PH
air-equilibrated
initial ANC
initial BNC
P, total , mg L"1
all values
x <= 0.010
7 > 0.010
Si02, mg L"1
S042', mg L'1
n

17
15
2

17
3
14
16
17
17
17

16
17

17
9
8
16
17
17
17
11
17
17

15
10
5

17
17
17

17
10
7
17
17
RMS of %RSDa

36
38
9.8

21
30
19
3.1
4.1
4.4
4.1

11
6.7

5.9
7.8
2.6
21
3.4
2.0
2.0
46
3.0
21

38
27
53

0.03
0.06
0.07

44
38
51
3.7
1.3
Precision
Pairs with
Mean > IOSR
n

2
0
2

5
0
5
14
17
17
17

14
12

12
4
8
4
17
17
17
0
17
0

7
3
4

—
—
__

0
0
0
16
17
RMS of 2RSDa

9.8
—
9.8

23
—
23
1.5
4.1
4.4
4.1

11
3.7

5.4
8.6
2.6
16
3.4
2.0
2.0
—
3.0


2.6
2.0
3.0

—
—
__

—
—
__
3.8
1.3
Quantisation
Limit
n

31



30


31
31
31
31

31
31

31


31
31
31
31
31
31
31

31
23
8

—
—
__

31


31
31
10SR

0.016



0.128


47.9
0.20
0.08
2.6

0.97
1.64

1.9


0.13
0.007
0.09
0.05
0.23
0.06
0.15

0.114
0.114
0.119

—
—
	

0.044


0.54
0.12
aRoot mean square of percent relative standard deviation (RMS of absolute SD
 for pH).
                                       725

-------
TABLE C-3.




Overall Wi thin-Batch
Pairs with

Parameter
Al , extractable, mg L"l
all values
x <= 0.010
x > 0.010
Al, total, mg L"1
all values
x <= 0.010
x > 0.010
ANC, ueq L'l
Ca, mg L"1
CT, mg L~!
Conductance, uS cm"1
DIC, mg L"1
air-equilibrated
initial
DOC, mg L"1
all values
"x <= 5
x > 5
Fe, mg L"l
F~, total dissolved, mg L"1
K, mg L"l
Mg, mg L~J
Mn, mg L"1
Na, mg L"1
NH4+, mg L'1
N03", mg L L
all batches
aliquot 3
aliquot 5
pH
air-equilibrated
initial ANC
initial BNC
P, total , mg L"1
all values
7 <= 0.010
x > 0.010
Si02, mg L"1
S042", mg L'1

n

6
1
5

10
0
10
8
10
10
10

10
10

10
5
5
9
10
10
10
7
10
10

9
9
0

10
10
9

10
6
4
10
10
Mean > 0
RMS of ZRSDa

6.6
0.00
7.2

34
—
34
28
0.47
2.3
0.65

10
11

3.7
3.4
4.0
67
12
1.4
0.49
58
1.0
23

68
68
—

0.06
0.10
0.18

18
22
11
3.7
5.3
USGS
Precision
Pairs with
Mean > lOsg
n

0
0
0

0
0
0
6
10
10
10

8
9

9
4
5
6
10
10
10
2
10
2

3
3
0

—
—
__

9
5
4
8
10
RMS of ZRSDa

—
—
—

—
—
—
1.9
0.47
2.3
0.65

6.9
9.6

3.9
3.8
4.0
7.1
12
1.4
0.49
18
1.0
3.2

6.6
6.6
—

--
—
--

19
24
11
2.3
5.3




Quantitation
Limit
n

18



18


18
18
18
18

18
18

18


18
18
18
18
18
18
18

18
18
0

--
—
--

18


18
18
= = = = =
10sp

0.024



0.850


17.3
0.03
0.10
1.9

0.38
0.34

1.1


0.00
0.005
0.03
0.03
0.01
0.02
0.05

0.060
0.060
—

--
—
—

0.004


0.14
0.38
aRoot mean square of percent relative standard deviation (RMS of absolute SD
 for pH).
                                     126

-------
ill
                                 •(Hd
OS svuosqe ^o SWH) uo.ueuap pjepue^s 3A.nei.aJ auao
fr9 *Z
Z9'£


220'0

—
—
—

S9£'T
T£2'0
90fr *T

T£'0
2T'0
OO'O
KTO
89'0
TWO
20'0


L'Z

99'0
£fr'0

fr'8
LZ'Q
9T'0
6'Efr


860'0



020'0

JSOT
q.LUi|
L6
L6


L6

—
—
—

09
L£
L6

L6
L6
L6
L6
L6
L6
L6


L6

L6
L6

L6
1.6
L6
L6


L6



L6

u
n
uo.UBaiq.UBno






TT 82 6'6 Lt
6'T
Ofr
—
Ofr

—
—
—

—
fr6
--

9'L
K9
TT
SZ'O
6'£
2'£
£T
E'fr
8'fr
9'fr

L'V
9'9

£t»'0
0£
T£'0
£T
L2
—
Li

L£
—
L£

Easbs jo swa
HSOT < UB8W
Ljq.LM SJLBd
UOj.Sj.39Jd I
HVSH3A

2T
9
0
S

—
—
—

0
2
0

T
Sfr
8T
Z.fr
6T
frT
2£
fr2
T2
9fr

9t»
fr£

Zfr
82
Zfr
2£
OT
—
OT

OT
0
OT

u


me
TZ 2fr
92 6£
69 8
t?£ ^f

SO'O Zfr
SO'O Zfr
TT'O £t>

08Z S2
ZZ£ frT
L99 6£

£9 S£
f 9 IM
TT 8T
9Z'0 Zfr
£'£ Lt
9'9 Zfr i-l
9fr frfr
£*fr fr2
9'fr £2
9't- LV

8'fr Zfr
L\ Uv

£fr'0 it
t2 LV
U'O Zt>
T2 Zfr
62 Tfr
t?9 9
S£ Lv

62 9T
29 Z2
29 £fr

Basaz jo swa u
0 < UB3W
U,q.LM S-UEd
g-u.Lnq.LM LLBU9AQ
ijad j.o guBnbs uB9iu q.
ooyE
T.l 6ui './OS
T-l ^^ '
OTO'O <
OTO'O =>
sen LEA L
j_l 6ui • |.Bq.o
ON8 LB.U.L
ONV LBH.I
p9q.Bjqi |.inb9 — i

S q.onb.L
E qonbi.
saL)oq.eq [
T.l Bui '
T 1 6ui •
T.-1 6"J
•[-1 Bui
T-l Bui
2OIS
x
x
L"B
q 'd
UL
U.L
IB
' Hd
LB
LB

_£ON
.'HN
'BN
'uw
*BW
I-l 6111 «x
Bui 'pgALOssip [Bq,oq.
T-l Bui
9 <
S =>
S9n[BA L
T-l Bui

p9q.BjqiLLnb9-j
T-l Bui
T-Uio sri 'aouBion
T-l Bui
T-l Bui
T-l barl
OTO'O <
OTO'O =>
S9H[BA L
T_l 6ui ' [.eqoq.
OTO'O <
OTO'O =>
S9n[EA L
Y_l 6ui '9[qBq.DBJqx9
ueiauiwBd



•_j
'9J
X
X
IB

Ul
LB
'310
puoo
'-13
'BO
'ONV
x
x
LB
'LV
x
X
LB
•LV





•fr-0 318V1

-------
                              APPENDIX D
                 ANALYTICAL W/TH/N-BATCH PRECISION
                 BY LABORATORY FOR 23 PARAMETERS,
                   EASTERN LAKE SURVEY — PHASE I
Values were estimated from contract analytical laboratory duplicate and calibration blank data
                                    129

-------


TABLE D-l.




EMS I
Analytical Within-Batch Precision
Pairs with

Parameter
Al, extractable, mg L~l
all values
x <= 0.010
x > 0.010
Al, total, mg L'l
all values
x <= 0.010
7 > 0.010
ANC, ueq L'l
Ca, mg L'l
CT, mg L-l
Conductance, uS cm"l
DIC, mg L"l
air-equilibrated
initial
DOC, mg L'l
all values
x <= 5
x" > 5
Fe, mg L"l
F", total dissolved, mg
K, mg L-l
Mg, mg L"l
Mn, mg L~l
Na, mg L~l
NH4+, mg L'J
N03~, mg L"1
all batches
aliquot 3
aliquot 5
PH
air-equilibrated
initial ANC
initial BNC
P, total , mg L"1
all values
7 <= 0.010
x > 0.010
Si02, mg L"1
S042-, mg L'1

n

49
6
43

49
1
48
47
49
49
49

49
49
49

25
24
47
L-l 49
49
49
44
49
49

47
10
37

49
49
49

49
24
25
49
49
Mean > 0
RMS of %RSDa

3.2
3.3
3.2

17
7.6
17
14
1.1
10
0.81

3.5
3.7
1.6

1.7
1.4
8.5
1.7
1.7
0.74
11
0.90
4.1

3.2
3.6
3.1

0.02
0.03
0.07

16
22
5.5
3.1
0.96

n

43
0
43

48
0
48

49
49
49

27
43
49

25
24
36
49
49
49
20
49
26

29
4
24

—
—
_ _

23
0
23
49
49
Pairs with
Mean > lOsg
RMS of ZRSDa

3.2
—
3.2

17
—
17

1.1
10
0.81

1.7
3.3
1.6

1.7
1.4
4.1
1.7
1.7
0.74
2.7
0.90
2.0

2.7
0.62
2.6

—
—
	

5.4
—
5.4
3.1
0.96
Quantisation
Limit
n

49



49


—
49
49
49

48
48
49



49
49
49
49
48
49
49

49
11
38

—
—
__

49


49
49
10sB

0.010



0.011


_.
0.03
0.03
2.4

0.73
0.58
0.4



0.07
0.013
0.02
0.01
0.04
0.02
0.06

0.076
0.072
0.078

—
—
	

0.011


0.12
0.19
aRoot mean square of percent relative standard deviation (RMS of absolute SD
 for pH).
                                       130

-------
TABLE D-;




Parameter
Al, extractable, mg L"1
all values
x <= 0.010
x > 0.010
Al, total, mg L'1
all values
x <= 0.010
7 > 0.010
ANC, ueq L~l
Ca, mg L'l
CT, mg L~l
Conductance, uS cm~l
DIC, mg L~l
air-equilibrated
initial
DOC, mg L"1
all values
7 <= 5
7 > 5
Fe, mg L'l
F", total dissolved, mg L~l
X, mg L'l
Mg, mg L'~l
Mn, mg L"l
Na, mg L~l
NH4+, mg L'J
N03", mg L"1
all batches
aliquot 3
aliquot 5
pH
air-equilibrated
initial ANC
initial BNC
P, total , mg L~l
all values
x <= 0.010
x > 0.010
Si02, mg L'1
S042', mg L'1


Analytical Wii
Pairs with

n

19
11
8

18
7
11
18
15
19
17

18
19
18

11
7
14
19
17
13
13
15
15

16
12
4

19
19
19

18
5
13
19
19
Mean > 0
RMS of SR!

31
19
42

39

GLOBAL
hin-Batch Precision
Pairs with
Mean > lOsg
Da n

8
0
8

11
61 0
10 11
67
1.0 15
56 16
11 16

17 16
6.7 19
16 18

20 11
4.5 7
38 11
3.3 16
49 14
0.78 13
39 13
37 13
13 13

36 15
41 12
4.5 4

0.03 0
0.06 0
0.08 0

4.7 14
6.8 1
3.6 13
3.6 19
11 16
RMS of 2RSDa

42
—
42

10
—
10

1.0
4.4
1.2

4.0
6.7
16

20
4.5
5.3
3.6
1.2
0.78
39
0.41
14

3.7
41
4.5

—
—
_.

4.1
8.3
3.6
3.6
2.2
Quantisation
Limit
n

19



19


—
19
19
19

19
19
19



19
19
19
19
19
19
19

19
15
4

—
—
__

19


19
19
10SR

0.013



0.011


—
0.06
0.06
0.8

0.37
0.37
0.0



0.06
0.002
0.02
0.00
0.00
0.04
0.02

0.080
0.000
0.035

—
—
__

0.006


0.02
0.05
aRoot mean square of percent relative standard deviation (RMS of absolute SO
 for pH).
                                     131

-------




TABLE D-3.




USGS




Analytical Wi thin-Batch Precision


Parameter
Al , extractable, mg L"l
all values
x <= 0.010
x > 0.010
Al , total , mg L"l
all values
x <= 0.010
x > 0.010
ANC, ueq L'l
Ca, mg L~l
C1-, mg L-l
Conductance, uS cm~l
DIG, mg L"l
air-equil ibrated
initial
DOC, mg L'1
all values
x <= 5
x > 5
Fe, mg L"l
F~, total dissolved, mg L"1
K, mg L-1
Mg, mg L"l
Mn, mg L~l
Na, mg L~l
NH4+, mg L-l
N03", mg L"1
all batches
aliquot 3
aliquot 5
pH
air-equilibrated
initial ANC
initial BNC
P, total , mg L"l
all values
x <= 0.010
x > 0.010
Si02, mg L"1
SO/', mg L~!


n

7
1
6

10
0
10
3
10
10
10

10
10

10
6
4
10
10
10
10
10
10
10

10
10
0

10
10
10

9
0
9
10
10
Pairs with
Mean > 0
RMS of ?RSDa

7.5
16
4.9

4.5
—
4.5
5.9
0.66
2.6
0.51

3.6
2.1

10
12
0.89
2.1
2.9
0.91
0.33
0.73
1.6
3.7

1.4
1.4
—

0.26
0.10
0.13

3.0
—
3.0
0.87
1.3


n

4
0
4

10
0
10

10
10
10

7
10

8
4
4
10
10
10
10
10
10
10

9
9
0

—
—
	

9
0
9
10
10
Pairs with
Mean > 10sR
RMS of JRSDa

3.9
—
3.9

4.5
—
4.5

0.66
2.6
0.51

3.9
2.1

1.7
1.9
0.89
2.1
2.9
0.91
0.33
0.73
1.6
3.7

1.5
1.5
—

—
—
	

3.0
—
3.0
0.87
1.3
Quantisation
Limit
n

10



10


__
10
10
10

10
10

10


10
10
10
10
10
10
10

10
10
0

--
—
	

10


10
10
lOSp

0.019



0.004


_ _
0.00
0.11
1.1

0.41
0.26

1.4


0.02
0.012
0.00
0.00
0.02
0.00
0.03

0.013
0.013
—

—
--
	

0.000


0.04
0.27
aRoot mean square of percent
 for pH).
relative standard deviation (RMS of absolute SD
                                       132

-------
                                   TABLE D-4.
                                                   VERSAR
Analytical
Within-Batch Precision
Pairs with Pairs with
Mean > 0 Mean > lOsp
Parameter
Al, extractable, mg L'l
all values
x <= 0.010
x > 0.010
Al, total, mg L-1
all values
7 <= 0.010
7 > 0.010
ANC, ueq L~l
Ca, mg L"1
CT, mg L-l
Conductance, uS cm~l
DIC, mg L"1
air-equilibrated
initial
DOC, mg L"l
all values
x <= 5
x > 5
Fe, mg L'1
F", total dissolved, mg L~l
'K, mg L-l
Mg, mg L"l
Mn, mg L~l
Na, mg L~l
NH/, mg I'}
N03~, mg L'1
all batches
aliquot 3
aliquot 5
PH
air-equilibrated
initial ANC
initial BNC
P, total , mg L"1
all values
x <= 0.010
x > 0.010
Si02, mg L"1
SO/', mg L'1
n

48
4
44

49
0

48
49
49
49

49
49

49
43
6
49
49
49
49
35
49
47

49
19
30

49
49
49

49
1
48
49
49
RMS Of

4.2
3.7
4.2

6.6
—
6.6
16
0.55
3.1
17

1.6
1.7

2.5
2.5
2.5
2.9
2.8
1.4
0.52
24
0.94
7.3

4.8
5.0
2.4

0.02
0.01
0.02

13
0.00
13
1.4
17
ZRSDa n

7
0
7

49
0
49

49
48
48

49
49

44
38
6
49
49
49
49
35
49
18

39
17
23

—
—
__

49
1
48
21
49
RMS of ZRSDa

2.9
—
2.9

6.6
—
6.6

0.55
3.1
17

1.6
1.7

2.6
2.6
2.5
2.9
2.8
1.4
0.52
24
0.94
1.9

4.2
4.8
4.0

—
—
	

13
0.00
13
1.5
17
Quantitation
Limit
n

49



49


—
49
49
49

48
48

49


49
49
49
49
49
49
49

47
17
30

—
—
__

49


49
49
10sp

0.053



0.013


__
0.02
0.16
1.3

0.15
0.15

1.0


0.01
0.012
0.06
0.01
0.01
0.15
0.21

0.233
0.177
0.261

—
—
__

0.005


3.56
0.18
aRoot mean square of percent relative standard deviation (RMS of absolute SD
 for pH).
                                      133

-------
                                APPENDIX E
         OVERALL AND ANALYTICAL AMONG-BATCH PRECISION
                 BY LABORA TORY FOR 23 PARAMETERS,
                     EASTERN LAKE SURVEY - PHASE I
Values were estimated from field natural. Lot 2 (FN2) and field natural. Lot 3 (FN3) audit sampled data,
field and laboratory high synthetic (FH, LH) audit sample data, and field and laboratory low synthetic
(FL and LL) audit sample data.
                                      135

-------
                                   TABLE E-l.
                                                     EMSI
                                         Overall  Among-Batch  Precision
                                FN2 Samples (n = 18)
         Parameter
Mean
2RSD*
                        FN3 Samples (n = 3)
Mean
SRSD*
Al, mg L-l
total extractable
total
ANC, ueq L'1
Ca, mg L"l
CT. mg L-l
Conductance, uS cm"l
DIC, mg L-l
air-equilibrated
initial
DOC, mg L'1
Fe, mg L~l
F", total dissolved,
K, mg L~l
Mg, mg L"1
Mn, mg L~l
Na, mg L~l
NH4+, mg L'1
N03~, mg L"l
PH
air-equil ibrated
initial ANC
initial BNC
P, total , mg L'l
Si02, mg L"1
S042-, mg L'1
====================

0.175
0.357
4.3b
1.87
0.51
26.7

0.17b
0.39b
3.2
0.02b
mg L'l 0.076
0.49
0.34
0.07
0.68
0.07b
1.449

5.25
5.07
5.11
0.001
4.36
6.89
====================

25
16
74
7.9
2.0
3.1

52
16
4.3
75
2.6
4.6
1.9
6.4
2.6
12
2.6

0.40
0.03
0.05
158
4.0
2.1
===========

0.002b
0.032b
841.0
57.84C
1.40
97.4

9.61
9.86
1.5b
0.00b
0.035
0.52
2.74
O.OO15
1.34
0.01b
1.46

8.18
7.86
7.88
0.001
2.66
3.32
===============

143
125
1.0
135C
1.4
1.1

4.4
4.2
10

2.0
3.5
1.5
__
3.2
201
3.3

0.10
0.08
0.06
187
3.7
0.83
aPercent relative standard deviation (absolute SD for pH).  Not applicable for
 pairs with Y = 0.
bMean less than quantisation limit (see Table C-l).
cWhen one confirmed but questionable value of 147.7 mg L~l is omitted, mean =
 12.91 mg L'1 and RSD = 5.7%.
                                       136

-------
                                   TABLE E-2.
                                                     GLOBAL
                                         Overall Among-Batch Precision3
                                FN2 Samples (n = 3)
         Parameter
Mean
 ZRSD&
Al, mg L'1
total extractable
total
ANC, ueq L"1
Ca, mg L"1
CT, mg L'1
Conductance, uS cm~l
DIC, mg L'1
air-equilibrated
initial
DOC, mg L"1
Fe, mg L~l
F~, total dissolved, mg L-1
K, mg L~!
Mg, mg L"1
Mn, mg L"1
Na, mg L~l
NH4+, mg L'1
N03~, mg L'1
PH

0.112
0.242
5.5C
1.81
0.52
25.6

0.39C
0.68C
3.4
o.oic
0.079
0.49
0.34
0.10C
0.69
0.09C
1.471


59
7.9
85
3.1
2.2
3.5

80
19
11
173
5.5
2.0
1.7
64
0.83
43
2.5

  air-equilibrated
  initial ANC
  initial BNC
P, total, mg L~l
Si02, mg L'1

S042', mg L"1
3———»————3——S™3SS
aNo FN3 samples were analyzed by Global.
DRelative standard deviation (absolute SD for pH).
cMean less than quantisation limit (see Table C-2),
5.14
5.13
5.11
0.001C
4.88

6.72
  0.03
  0.02
  0.02
284
  5.3

  1.8
                                      737

-------
                                   TABLE E-3.
                                                      USGS
                                         Overall  Among-Batch Precision3
                                FN2 Samples (n = 3)
         Parameter
                              Mean
            2RSDb
Al, mg L'1
  total extractable
  total
ANC, ueq L~l
Ca, mg L"1
Cl~, mg L~l
Conductance, uS cm~l
DIC, mg L-1
  air-equilibrated
  initial
DOC, mg L-1
Fe, mg L~l
F~, total dissolved, mg L"1
K, mg L-l
Mg, mg L"1
Mn, mg L~l
Ha, mg L'l
NH,
N0
pH
   mg L"1

   mg L'1
  air-equilibrated
  initial ANC
  initial BNC
P, total , ma L~l
SiO?, mg L-1
SO
'2-
 2-
    mg  L
           _i
 0.165
 0.166C
-3.4C
 2.36
 0.83
28.6

 0.26C
 0.23C
 3.7
 0.01
 0.077
 0.49
 0.35
 0.07
 0.71

 0.07

 1.457
 5.10
 5.04
 5.06
 0.003C
 4.27

 7.27
 11
 70
395
 13
 51
  6.8

 38
 37
 25
  7.7
  6.6
  1.1
  2.1
  0.81
  4.7

 50

  3.1
  0.05
  0.05
  0.01
 96
  0.66

  9.7
3No FN3 samples were analyzed by USGS.
^Percent relative standard deviation (absolute SD for pH)
cMean less than quantitation limit (see Table C-3).
                                       138

-------
                                    TABLE  E-4.
                                                      VERSAR
                                          Overall  Among-Batch  Precision
                                 FN2  Samples  (n  =  17)
                                                          FN3 Samples  (n = 4)
          Parameter
                                 Mean
ZRSDa
air-equilibrated
initial  ANC
initial  BNC
 total,  mg L"1
    mg L~*
5.13
5.05
5.05
0.002b
4.22
                                            0.04
                                            0.05
                                            0.03
                                          127
                                           10
S042", mg L
           -1
                                                         Mean
                                                           8.26
                                                           7.68
                                                           7.72
                                                           0.002b
                                                           2.78b
Al. mg L-l
total extractable
total
ANC, ueq L'l
Ca, mg L~l
CT, mg L-l
Conductance, uS cm"l
DIG, mg L"l
air-equilibrated
initial
DOC, mg L~l
Fe, mg L"l
F", total dissolved, mg L"l
K, mg L-l
Mg, mg L'l
Mn, mg L"1
Na, mg L"l
NH4+, mg L'1
N03", mg L'1
PH

0.205
0.286
0.8b
1.89
0.69
26.5

0.17b
0.45b
3.16
0.02b
0.078
0.48 b
0.35
0.06
0.66
0.04b
1.386b


35
25
797
1.2
6.3
0.74

24
14
6.4
33
4.2
4.6
1.9
7.7
6.6
93
7.4


0.002b
0.014b
853.7
13.10
1.38
95.7

9.74
9.99
1.3b
O.OQb
0.035b
0.47b
2.84
0.01
1.28
0.01b
1.363b


43
86
3.0
3.2
2.3
1.8

3.8
4.7
6.3

6.9
16
4.1
200
7.9
68
5.3

                              0.05
                              0.09
                              0.10
                            122
                              7.2
                                6.99       10              3.22          4.0

apercent relative standard deviation (absolute SD for pH}.  Not applicable for
 pairs with Y = 0.
bMean equal to or less than quantisation limit (see Table C-4).
                                      139

-------
                                   TABLE  E-5.
                                                    EMS I
                                  Overall  Among-
                                  Batch  Precision

                                FH Samples (n  =  14)
                                                   Analytical  Among-
                                                    Batch  Precision

                                                   LH  Samples  (n  = 9)
         Parameter
                         Mean
          2RSDa
pH
  air-equilibrated
  initial ANC
  initial BNC
P, total, ma L~l
SiO,, mg L"1
SO,
   2-
mg L
    -1
 7.62
 6.96
 7.00
 0.054
 9.01

14.6
 0.58
 0.25
 0.26
16
14

 2.1
               Mean
                                                       81
                                                       94
                                                       06
                                                       062
                                                       08
14.73
           %RSDa
Al , mg L"l
total extractableb
total
ANC, ueq L'1
Ca, mg L"1
Lot 4 (n = 11, 9)
Lots 5 and 6 (n = 3, 0)
Cl", mg L'l
Lot 4 (n = 6, 9)
Lots 5 and 6 (n = 3, 0)
Conductance, uS cm"!
DIC, mg L-l
air-equilibrated
initial
DOC, mg L~l
Fe, mg L~lb
F~, total dissolved, mg L"1
K, mg L-l
Mg, mg L~l
Mn, mg L"l
Na, mg L"l
NH.+, mg L"1

N03~, mg L"1

—
0.233
465.8

1.65
1.73

3.33
4.03
103.3

4.79
7.43
9.8
—
0.438
3.07
2.32
1.13
11.83
1.25

1.661

—
23
23

9.3
3.3

3.0
1.1
1.4

32
28
24
—
5.5
2.5
1.8
16
8.6
16

29

0.153
0.219
483.9

1.50
—

3.36
—
102.3

4.99
7.26
10.4
0.18
0.439
3.08
2.33
1.38
11.67
1.35
1.353


20
35
19

14
—

1.5
—
2.1

33
27
20
11
2.9
2.4
2.1
3.2
9.9
9.1

45
 0.14
 0.30
 0.31
 4.3
15

 2.4
^Percent relative standard deviation (absolute SD for pH).  Not applicable for
 pairs with X = 0.
t>Laboratory synthetic audit samples only (see Section 4).
                                      140

-------
                                   TABLE E-6.

                                                   GLOBAL
                                  Overall Among-
                                  Batch Precision

                                FH Samples (n = 5)
                          Analytical Among-
                           Batch Precision

                          LH Samples (n = 2)
         Parameter
Mean
%RSDa
Mean
2RSDa
Al, mg L"1
total extractable0
total
ANC, peq L"1
Ca, mg L"1
Lot 4 (n = 3, 1)
Lots 5 and 6 (n = 2,
CT, mg L"1
Lot 4 (n = 3, 1)
Lots 5 and 6 (n = 2,
Conductance, uS cm~l
DIC, mg L"1
air-equilibrated
initial
DOC, mg L'l
Fe, mg L'lb
F", total dissolved, mg
K, mg L"1
Mg, mg L"l
.Mn, mg L'1
Na, mg L"l
NH4+, mg L"1
N03~, mg L'1
PH
air-equilibrated
initial ANC
initial BNC
P, total, mg L"1
Si02, mg L"1
S042-, mg L'1

—
0.175
482.7

1.69
1) 1.90

3.15
1) 4.10
100.5

4.94
5.26
10.8
—
L-l 0.440
3.08
2.27
1.17
12.04
1.23
1.313

7.89
7.15
7.10
0.048
10.13
13.96

—
6.3
15

2.4
4.1

1.8
5.7
4.5

26
18
20
—
3.2
4.5
2.9
18
14
10
60

0.08
0.22
0.25
19
13
2.8

0.129
0.177
392.1

1.65
1.82

3.27
3.95
103.1

5.97
6.46
9.9
0.18
0.430
2.97
2.24
1.44
13.07
1.27
1.464

7.94
7.00
7.04
0.054
11.55
14.65

21
0.80
49

—
—

—
—
3.9

9.4
3.7
4.3
7.9
2.3
3.6
3.8
2.0
2.8
0.17
15

0.05
0.01
0.05
9.3
4.0
0.77
aPercent relative standard deviation (absolute SD for pH).
 pairs with X = 0 or where n = 1.
bLaboratory synthetic audit samples only (see Section 4).
                            Not applicable for
                                       141

-------
                                   TABLE E-7.
                                  Overall Among-
                                  Batch Precision

                                FH Samples (n = 3)
                                                   USGS
                                               Analytical Among-
                                                Batch Precision

                                               LH Samples (n = 1)
         Parameter
                      Mean
ZRSDa
so.
   2'
       mg L
"1
                     14.53
 3.2
 Mean
14.10
                                                                       2RSDa
Al, mg L'1
total extractable'3
total
ANC, peq L"1
Ca, mg L'1
Lot 4 (n = 0, 0)
Lots 5 and 6 (n = 3, 1)
CT, mg L'1
Lot 4 (n = 0, 0)
Lots 5 and 6 (n = 3, 1)
Conductance, uS cm'1
DIC, mg L'1
air-equil ibrated
initial
DOC, mg L"1
Fe, mg L~lb
F~, total dissolved, mg L"1
K, mg L""1
Mg, mg L"1
Mn, mg L'1
Na, mg L"1
NH4+, mg L-1
N03", mg L"1
pH
air-equilibrated
initial ANC
initial BNC
P, total, mo L'1
SiOo, mg L

—
0.175C
542.7

—
2.00

—
3.50
110.6

6.33
6.55
10.0
—
0.455
3.11
2.33
1.32
12.47
1.32
1.700

7.71
7.29
7.53
0.061
10.07

—
3.2
2.8

—
13

—
17
1.4

13
26
18
—
0.46
4.9
2.8
6.4
0.90
10
29

0.06
0.35
0.24
23
1.6

0.159
0.175
554.7

__
1.81

__
2.96
111.1

6.01
6.03
10.1
0.12
0.456
3.10
2.28
1.47
12.34
1.42
2.02

7.65
7.38
7.38
0.075
10.24

__
__
__

__
__

__
__
__

__
__
__
__
—
__
__
__
—
—
—

__
—
__
—
—
aPercent relative standard deviation (absolute SD for pH).  Not applicable for
 pairs with Y = 0 or where n = 1.
^Laboratory synthetic audit samples only (see Section 4).
cMean less than quantisation limit (see Table C-3).
                                      142

-------
                                   TABLE  E-8.
                                                   VERSAR
                                  Overall Among-
                                  Batch Precision
PH
  air-equilibrated
  initial ANC
  initial BNC
P, total, mg L~l
Si02,

V
      mg

     , mg
          mg
         L'1
 7.75
 7.14
 7.16
 0.062
 8.94

14.32
 0.64
 0.20
 0.21
26
14

 8.4
                          Analytical Among-
                           Batch Precision
FH Samples (n = 14}
Parameter
Mean
2RSD3
LH Samples (n = 9)
Mean 2RSDa
AT, mg L~l
total extractableb
total
ANC, ueq L~l
Ca, mg L~l
Lot 4 (n = 6, 4)
Lots 5 and 6 (n = 8, 5)
Cl", mg L"l
Lot 4 (n = 6, 4)
Lots 5 and 6 (n = 8, 5)
Conductance, uS cm"l
DIC, mg L"l
air-equilibrated
initial
DOC, mg L"l
Fe, mg L~lb
F", total dissolved, mg L"l
K, mg L-l
Mg, mg L-l
Mn, mg L"1
Na, mg L"l
NH4+, mg L'1
NO.', mg L'1

—
0.179
471.4

1.68
2.03

4.52
4.15
105.9

4.22
4.57
9.9
—
0.451
2.96
2.45
1.23
11.64
1.26
1.800

—
37
13

2.2
2.1

57
23
4.2

18
21
16
.-
14
9.2
2.2
6.0
7.5
12
21

0.177
0.176
502.5

1.65
2.02

3.49
6.75
105.7

4.19
4.50
10.2
0.18
0.430
2.96
2.43
1.38
11.76
1.35
1.57

12
37
16

2.0
1.0

13
92
4.5

17
20
16
6.3
2.4
6.4
1.9
1.9
8.1
18
44
   95
   14
   14
   060
 9.20
 0.13
 0.20
 0.21
22
14
^Percent relative standard deviation (absolute SD for pH),
 pairs with T = 0.
bLaboratory synthetic audit samples only (see Section 4).
14.20         6.5

Not applicable for
                                      143

-------
                                   TABLE E-9.
                                                     EMS I
                                  Overall Among-
                                  Batch Precision

                                FL Samples (n = 16)
                         Analytical Among-
                          Batch Precision

                         LL Samples (n = 16)
         Parameter
Mean
2RSDa
S042-, mg L'1
2.38
  9.8
                                                            Mean
            %RSDa
Al , mg L"l
total extractableb
total
ANC, peq L'1
Ca, mg L~l
Lot 4 (n = 11, 10)
Lots 5 and 6 (n = 5, 6)
Cl", mg L"l
Lot 4 (n = 11, 10)
Lots 5 and 6 (n = 5, 6)
Conductance, uS cm"!
DIC, mg L"l
air-equilibrated
initial
DOC, mg L'l
Fe, mg L'lb
F-, total dissolved, mg L"l
K, mg L-l
Mg, mg L"l
Mn, mg L"l
Na, mg L'l
NH4+, mg L'1
N03", mg L'1
pH
air-equilibrated
initial ANC
initial BNC
P, total, ma L~l
Si 0?, mg L'1

—
0.025b
112.8

0.14
0.17

0.39C
0.32C
18.9

1.34
1.83
0.9C
—
0.041
0.22
0.41
0.09
2.75
0.18C
0.547

7.24
6.83
6.93
0.022
1.00

__
16
6.1

8.4
3.6

87
2.6
7.9

8.2
7.2
45
—
24
11
8.8
7.0
3.9
23
8.3

0.05
0.07
0.13
35
43

0.017
0.033
115

0.13
0.17

0.29
0.31
18.5

1.32
1.75
1.28
0.08
0.040
0.22
0.42
0.10
2.77
0.20
0.609

7.23
6.82
6.90
0.021
0.97

34
54
6.8

9.9
9.5

7.3
1.7
5.4

8.5
15
78
21
6.8
10
2.9
6.3
3.4
17
60

0.05
0.09
0.12
24
16
aPercent relative standard deviation (absolute SD for pH)
 pairs with Y = 0.
DLaboratory synthetic audit samples only (see Section 4).
°Mean less than quantisation limit (see Table C-l).
2.30        1.8

Not applicable for
                                      144

-------
                                  TABLE E-10.
                                                   GLOBAL
                                  Overall  Among-
                                  Batch  Precision

                                FL Samples (n  = 7)
                                                   Analytical  Among-
                                                    Batch Precision

                                                   LL Samples  (n = 8)
         Parameter
                          Mean
          %RSDa
SO,
   2-
mg L
    -1
2.21
3.3
               Mean
2.21
Al, mg L'1
total extractableb
total
ANC, ueq L-1
Ca, mg L"1
Lot 4 (n = 1, 2)
Lots 5 and 6 (n = 6, 6)
Cl", mg L"1
Lot 4 (n = 1, 2)
Lots 5 and 6 (n = 6, 6)
Conductance, uS cm~l
DIC, mg L"1
air-equil ibrated
initial
DOC, mg L"1
Fe, mg L~lb
F", total dissolved, mg L"1
K, mg L'1
Mg, mg L~J
Mn, mg L~l
Na, mg L"1
NH4+, mg L"1
N03", mg L"1
pH
air-equilibrated
initial ANC
initial BNC
P, total, mg L~l
SiOo, mg L"1

—
0.020C
115.13

0.16C
0.17C

0.26
0.33
18.0

1.37
1.66
1.3C
—
0.041
0.23
0.42
0.09
2.66
0.15
0.245

7.40
6.96
7.00
0.019C
1.29

—
38
6.2

—
4.4

—
8.4
7.1

13
12
57
—
5.9
12
10
19
1.7
20
105

0.02
0.10
0.13
35
9.1

0.015
0.019
115.15

0.17
0.16

0.30
0.33
17.8

1.35
1.75
1.5
0.06d
0.040
0.24
0.42
0.10
2.51
0.16
0.29

7.40
6.99
7.00
0.024
1.26

33
38
6.4

4.3
9.9

2.4
7.9
8.0

19
15
50
40
5.4
10
8.2
6.3
19
11
92

0.02
0.11
0.12
82
11
2.5
aPercent relative standard deviation (absolute SD for pH).  Not applicable for
 pairs with Y = 0 or where n = 1.
^Laboratory synthetic audit samples only (see Section 4).
cMean less than quantisation limit (see Table C-2).
dMean equal to or less than quantisation limit (see Table D-2).
                                      145

-------
                                  TABLE E-ll.
                                                     USGS
                                  Overall Among-
                                  Batch Precision

                                FL Samples (n = 4)
                                                    Analytical Among-
                                                     Batch Precision

                                                    LL Samples (n = 3)
         Parameter
                           Mean
                                           2RSDa
Ca, mg L"1
  Lot 4 (n = 0, 0)
  Lots 5 and 6 (n = 4, 3)
Cl~, mg L"1
  Lot 4 (n = 0, 0)
  Lots 5 and 6 (n = 4, 3)
Conductance, uS cm"1
DIC, mg L"1
  air-equilibrated
  initial
DOC, mg L'1
Fe, mg L"lb
F~, total  dissolved, mg L"
K, mg L"1
Mg, mg L'1
Mn, mg L"1
Na, mg L"1
NH,
NO-
PH
mg L

mg L
          '1
          "1
  air-equilibrated
  initial ANC
  initial BNC
P, total, mg L'1
Si02, mg
          mg
         L"1
SO,
   2-
       mg L
     '1
                                 0.17
 0.40
21.1

 1.41
 1.44
 1.2

 0.040
 0.32
 0.43
 0.09
 2.76

 0.16

 0.267
 7.25
 6.93
 7.22
 0.020
 1.03

 2.23
                                      10
 24
  1.6

  7.7
  1.9
 49

  1.3
 27
  3.3
  1.6
  2.1

  8.9

108
  0.15
  0.12
  0.50
 19
  1.1

  4.8
                            Mean
 0.17


 0.33
20.8

 1.44
 1.43
 1.3C
 0.05
 0.040
 0.23
 0.43
 0.10
 2.75

 0.17

 0.355
                                                              08
                                                              80
                                                              11
                                                      0.022
                                                      1.04
                                                      2.23
                             2RSDa
Al, mg L"1
total extractableb
total
ANC, ueq L"1

—
0.018C
112.6

—
12
14

0.014d
0.080
106.5

28
120
2.9
                                         8.6
12
 1.2

 8.3
14
 7.1
12
 1.4
13
 4.0
 2.8
 2.7

 8.5

81
              0.07
              0.14
              0.33
             24
              2.4

              4.1
aPercent relative standard deviation (absolute SD for pH).  Not applicable for
 pairs with J = 0.
bLaboratory synthetic audit samples only (see Section 4).
cMean equal to or less than quantitation limit (see Table C-3).
^Mean less than quantitation limit (see Table D-3).
                                      146

-------
                                  TABLE E-12.
                                                  VERSAR
                                  Overall Among-
                                  Batch Precision
Analytical  Among-
 Batch Precision
FL Samples (n = 16)
Parameter
Al, mg L~l
total extractableb
total
ANC, ueq L'1
Ca, mg L"1
Lot 4 (n = 11, 11)
Lots 5 and 6 (n = 5, 5)
CT, mg L'1
Lot 4 (n = 11, 11)
Lots 5 and 6 (n = 5, 5)
Conductance, uS cm"!
DIG, mg L'1
air-equilibrated
initial
DOC, mg L~l
Fe, mg L~lb
F~, total dissolved, mg L~l
K, mg L"1
Mg, mg L~l
Mn, mg L~l
Na, mg L"1
NH4+, mg L'1
N03', mg L'1
pH
air-equilibrated
initial ANC
initial BNC
P, total , mg L~l
Si02, mg L"1
S042-, mg L'1
Mean

—
0.025C
108.9

0.15C
0.16C

0.28C
0.43
19.0

1.35
1.52
0.9C
—
0.044
0.22
0.44
0.09
2.68
0.14°
0.458C

7.43
6.85
6.91
0.02QC
0.87C
2.21C
%RSDa

—
20
7.0

24
5.7

11
26
3.8

9.3
8.9
41
__
24
9.3
2.3
5.8
7.1
36
32

0.16
0.12
0.09
25
21
12
LL Samples
Mean

0.022d
0.027
103.2

0.14
0.15

0.31
0.42
18.9

1.36
1.54
0.9
0.07
0.04
0.22
0.44
0.08
2.68
0.19d
0.481

7.33
6.85
6.89
0.022
0.94
2.32
(n = 16)
»SDa

17
32
11

5.3
2.9

18
51
3.8

9.0
8.7
30
10
5.0
9.1
1.8
24
6.2
28
18

0.15
0.12
0.11
22
22
5.9
aPercent relative standard deviation (absolute SD for pH).  Not applicable for
 pairs with Y = 0.
^Laboratory synthetic audit samples only (see Section 4).
cMean equal to or less than quantisation limit (see Table C-4).
dMean less than quantitation limit (see Table D-4).
                                      147

-------