903R90010
CBP/TRS 51/90
November 1990
Chesapeake Bay Coordinated
Split Sample Program
Annual Report, 1989
P^^^
Chesapeake
Bay
Program
Printed on Recycled Paper
-------
Chesapeake Bay Coordinated
Split Sample Program
Annual Report, 1989
by Peter Bergstrom
November 14,1990
Computer Sciences Corporation
Chesapeake Bay Program
410 Severn Avenue, Suite 113
Annapolis, Maryland 21403
(301) 267-0061
Printed by the U.S. Environmental Protection Agency for the Chesapeake Bay Program
-------
EXECUTIVE SUMMARY
Hie Chesapeake Bay Program is a Federal-State partnership with a goal
of restoring the Chesapeake Bay. Its ambient water quality monitoring
program uses 10 different analytical laboratories. The Chesapeake Bay
Coordinated Split Sample Program (CSSP), initiated in 1988, assesses the
comparability of the water quality results from these laboratories. This
report sunoarizes CSSP results for 1989, its first full year of operation.
The CSSP has two main objectives: estimating measurement system
variability, and identifying parameters with low inter-organization
agreement. The variability estimates are most useful to data analysts and
modelers who need confidence estimates for monitoring data. The
identification of parameters with low agreement is used as part of the
overall Quality Assurance program. Laboratory and program personnel use
this information to investigate possible causes of the differences, and
take action to increase agreement if possible. Later CSSP results
document any improvements in inter-organization agreement.
Estimates of measurement system variability based on split sample data
show that some parameters have more variable results than others. In some
cases these patterns were consistent when different laboratories and
sampling stations were compared.
Inter-organization agreement was high for 18 of the 23 comparisons
made in two components. Agreement was low enough to recommend
investigation for five parameters: Total Phosphorus (TP), Total Dissolved
Phosphorus (TOP), Particulate Carbon (PC), Particulate Nitrogen (PN), and
Dissolved Organic Carbon (DOC). Recommendations for further investigation
were made when there were three or more inter-organization differences that
were larger than within-organization precision, and there were
statistically significant inter-organization differences at the P < 0.01
level.
In all five cases of low inter-organization agreement, only one of the
four organizations compared had divergent results. In each case, this
organization had a different analytical method or instrument type, and in
two cases (PC and PN) there was also a difference in filter type, in three
cases (TP, TOP, and DOC) method changes have been made to increase inter-
organization agreement. The other two cases (PC and PN) are being
investigated by the organizations involved to find ways to increase
agreement.
The results from the first year of operation show that the CSSP is
successful at achieving its goals. The communication and cooperation among
participants that occurred was essential to getting the split sample
results translated into actions that have increased inter-organization
agreement.
ii
-------
ACKNCWLEDGMENTS
The members of the Analytical Methods and Quality Assurance Workgroup
of the Monitoring Subcommittee made many helpful suggestions used in the
analysis. Rich Batiuk, Tina Fletcher, Joe Macknis, Nina Fisher, Bruce
Neilson, Carl Zimmermann, Bruce Michael, and Steve Sokolowski made helpful
comments on earlier drafts of the manuscript. All of the laboratory and
program personnel involved took the time to familiarize themselves with the
program, run the samples, submit the data, and study the results of the
analyses. Their continuing support is what makes the program work.
iii
-------
-------
TABLE OF CONTENTS
EXECUTIVE SUMMARY ........................... ii
ACKNOWLEDGMENTS ............................ ill
TABLE OF CONTENTS ........................... iv
LIST OF FIGURES ............................ vi
LIST OF TABLES ............................ viii
I. INTRODUCTION ............................ 1
II. MfftHOoS .............................. 4
A. COMPONENTS AND PARTICIPATING LABORATORIES ........... 4
1. Mainstem and Tidal Tributaries component ......... 4
2. Tidal Potomac component ................. 4
3. Fall Line component ................... 4
B. SAMPLE COLLECTION AND SPLITTING METHODS ............ 4
1. Mainstem and Tidal Tributaries component ......... 4
2. Tidal Potomac component ................. 5
3. Fall Line component ................... 5
C. DATA ENTRY AND REDUCTION ................... 6
1. Mainstem and Tidal Tributaries component ......... 6
2. Tidal Potomac component ................. 6
3. Fall Line component ................... 6
D. DATA ANALYSIS AND GRAPHING .................. 6
1. Preliminary test of splitting randomness ......... 6
2. Precision estimates ................... 7
3. Assessing inter-organization agreement .......... 7
III. RESULTS ............................. 9
A. WITHIN-ORGANIZATION PRECISION AND ACCURACY .......... 9
1. Mainstem and Tidal Tributaries component ......... 9
2. Tidal Potomac component ................. 9
3. Fall Line component ................... 15
B. INTER-ORGANIZATION PRECISION ................. 15
1. Mainstem and Tidal Tributaries component ......... 15
2. Tidal Potomac component ................. 15
3. Fall Line component ................... 15
C. INTER-ORGANIZATION AGREEMENT ................. 22
1. Mainstem and Tidal Tributaries component ......... 22
2. Tidal Potomac component ................. 22
3. Fall Line component ................... 22
IV. DISCUSSION ............................ 37
A. WITHIN-ORGANIZATION PRECISION AND ACCURACY .......... 37
B. INTER-ORGANIZATION PRECISION ................. 37
C. INTER-ORGANIZATION AGREEMENT ................. 37
1. Mainstem component .................... 37
iv
-------
2. Potomac component 38
3. Fall line component 38
V. SUMMARY AND CONCLUSIONS 38
VI. REFERENCES 39
-------
LIST OF FIGURES
Figure 1. Chesapeake Bay Coordinated Split Sample Program
Components 2
Figure 2. Schematic of operational flow of analyses, Coordinated
Split Sample Program 3
Figure 3. Split sample data for Total Dissolved Phosphorus (TOP),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 23
Figure 4. Split sample data for Total Phosphorus (TP),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 23
Figure 5. Split sample data for Particulate Nitrogen (PN),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 24
Figure 6. Split sample data for Particulate Carbon (PC),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 24
Figure 7. Split sample data for Silica (SI),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 25
Figure 8. Split sample data for Nitrite (N02),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 25
Figure 9. Split sample data for Nitrite + Nitrate (N023),
from samples collected at Station CBS.3 (Mainstem),
showing cruise means with precision bars 26
Figure 10. Split sample data for Ammonium (NH4),
from samples collected at Station CBS.3 (Mainstem),
showing cruise means with precision bars 26
Figure 11. Split sample data for Particulate Phosphorus (PHOSP),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 27
Figure 12. Split sample data for Orthophosphate (P04F),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 27
Figure 13. Split sample data for Dissolved Organic Carbon (DOC),
from samples collected at Station CBS.3 (Mainstem),
showing cruise means with precision bars 28
vi
-------
Figure 14. Split sample data for Total Dissolved Nitrogen (TON),
from samples collected at Station CB5.3 (Mainstern),
showing cruise means with precision bars 28
Figure 15. Split sample data for Total Nitrogen (TN),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 29
Figure 16. Split sample data for Total Suspended Solids (TSS),
from samples collected at Station CB5.3 (Mainstem),
showing cruise means with precision bars 29
Figure 17. Split sample data for Ammonium (NH4),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars 31
Figure 18. Split sample data for Nitrite (N02),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars 31
Figure 19. Split sample data for Nitrite + Nitrate (N023),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars .32
Figure 20. Split sample data for Total Kjeldahl Nitrogen Whole (TKNW),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars 32
Figure 21. Split sample data for Orthophosphate (P04),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars 33
Figure 22. Split sample data for Total Phosphorus (TP),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars 33
Figure 23. Split sample data for Total Organic Carbon (TOC),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars 34
Figure 24. Split sample data for Total Suspended Solids (TSS),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars 34
Figure 25. Split sample data for Silica (SI),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars 35
Figure 26. Split sample data for Biological Oxygen Demand 5 day (BOD5),
from samples collected at Station PMS-10 (Potomac),
showing cruise means with precision bars 35
VII
-------
LIST OF TABLES
TABLE 1: Method Detection Limits at Mainstem Component Laboratories,
1989-1990 10
TABLE 2: Method Detection Limits at Potomac Component Laboratories,
1989-1990 11
TABLE 3: Method Detection Limits at Fall Line Component Laboratories,
1989-1990 12
TABLE 4: Wi thin-organization and inter-organization precision
estimates, Mainstem Component 13
TABLE 5: Wi thin-organization and inter-organization precision
estimates, Potomac Component 14
TABLE 6: Wi thin-organization precision estimates, Fall Line Component. 16
TABLE 7: Percent Recovery Data, Mainstem Component, 1989 17
TABLE 8: Percent Recovery Data, Potomac Component, 1989-90 18
TABLE 9: Percent Recovery Data, Fall Line Component, 1989-90 19
TABLE 10: Standard Reference Material Results, Mainstem Component,
1989 20
TABLE 11: Standard Reference Material Results, Potomac Component,
1989-90 21
TABLE 12: Mainstem Component (Station CB5.3) Split Sample Results
using Cruise Means (1987 - 1989) 30
TABLE 13: Potomac Component (Station PMS-10) Split Sample Results
using Cruise Means (1989 - 1990) 36
viii
-------
I. INTRODUCTION
The Monitoring Subcommittee of the Chesapeake Bay Program initiated
the Chesapeake Bay Coordinated Split Sample Program (CSSP) in 1988. Its
goal is to assess the comparability of water quality results from the 10
analytical laboratories that participate in the Chesapeake Bay Monitoring
Program (Chesapeake Bay Program 1989). This goal is being achieved by
estimating measurement system variability and identifying any parameters
that have low inter-organization agreement.
Estimates of measurement system variability are useful to data users
such as statisticians and modelers who need confidence bounds for
monitoring data. Although split sample results do not include sampling
variability, they are the best estimate available of total system
variability for Chesapeake Bay water quality monitoring data.
Identifying parameters with low agreement enables the organizations
involved to investigate any significant differences and take actions to
raise inter-organization agreement. This might involve changing field
methods, laboratory methods, or both. Because results of field split
samples are affected by both field and laboratory variability, the terms
"inter-organization" and "within-organization" are used rather than "inter-
laboratory" and "within-laboratory." The organization includes all the
elements of the measurement system: field sampling, sample handling,
laboratory analysis, data handling, and the state or municipal agency that
supervises the water quality monitoring program.
The CSSP has four components, each including three to five
laboratories that analyze samples from similar salinity regimes and
concentration ranges (Fig. 1). Laboratories in each component analyze
triplicate field split samples collected quarterly, following the specified
analysis flow (Fig. 2). Laboratory personnel send the analytical results
to the EPA Chesapeake Bay Liaison Office (CBLO) in Annapolis for data entry
and analysis by Computer Sciences Corporation (CSC/CBLO) staff.
This report summarizes the results of three of the four CSSP
components for 1989. These components are 1) the Mainstem and Tidal
Tributaries component, which analyzes estuarine samples from the mainstem
of the Chesapeake Bay; 2) the Tidal Potomac component, which samples the
tidal fresh portion of the river; and, 3) the Fall Line component, which
samples the Susquehanna River fall line station. The Virginia Mainstem
and Tidal Tributaries component, which samples the tidal fresh portion of
the James River, did not operate between January 1989 and February 1990, so
it is not included in this report.
This report includes split sample data from September 1987 through
March 1990. Early 1990 data were included when there were not enough data
through 1989 for statistical analysis.
-------
1
4*
O
U
2
i
Q.
C/3
I
13
'3
4)
3
-------
LARGE
VESSEL
(Aliquots split in
sequence: all Aliquot
1 bottles, then 2, etc.)
Figure 2. Schematic of the Operational Flow of Analyses,
Coordinated Split Sample Program
Normal Laboratory
Quality Control
Procedures
Replicate Analyses
**
Triplicate Aliquots
(sent to each laboratory)*
-
Analyze for
Routine Parameters
Analyze for
Routine Parameters
*(in-matrix estimate
of field precision)
Spike Sample
***
Analyze for
Routine Parameters
Analyze for
Routine Parameters
** (in-matrix estimate
of lab precision)
Analyze for
Percent Recovery
***(in-matrix estimate
of accuracy)
Deionized/distilled
water dilution
EPA Standard
Reference Material
Matrix water
dilution (if saline)
Analyze for
SRM Parameter
Analyze for
SRM Parameter
-------
II. METHODS
A. COMPONENTS AND PARTICIPATING LABORATORIES
1. Mainstem and Tidal Tributaries component
This component includes the three Chesapeake Bay mainstem analysis
laboratories: Chesapeake Biological Laboratory (CBL) in Solomons, MD,
Virginia Institute of Marine Science (VIMS) in Gloucester Point, VA, and
Old Dominion University (ODU) in Norfolk. It also includes one tributary
laboratory, Maryland Department of Health and Mental Hygiene (MDHMH) in
Baltimore. Starting in June, 1990, the Virginia tributary laboratory,
Division of Consolidated Laboratory Services (DCLS) in Richmond, was added
to this component. Split sample results from this component were included
in two previous reports (Bergstrom 1989, 1990s).
2. Tidal Potomac component
This component includes three analytical laboratories: MDHMH, DCLS,
and the EPA Central Regional Laboratory (CRL) in Annapolis. Analyses at
CRL are conducted by District of Columbia Department of Consumer and
Regulatory Affairs (DCRA) personnel (referred to as CRL/DCRA). Split
sample results from this component were included in two previous reports
(Bergstrom 1989, 1990b).
3. Fall Line component
The Fall Line component includes five analytical laboratories: MDHMH,
DCLS, the United States Geological Survey (USGS) in Denver, the
Pennsylvania Department of Environmental Resources (PADER) in Harrisburg,
and the Occoquan Watershed Monitoring Laboratory (OWML) in Manassas, VA.
DCLS does not currently participate in this component of the CSSP, but they
may do so in the future. Split sample results from this component were
included in one previous CSSP report (Bergstrom 1990c).
All of these laboratories also participate in two-way split sample
programs with USGS. Data from these programs were not included here
because the goals and methods of the USGS program are different from those
of the CSSP, and many of the two-way split results have been analyzed by
others (e.g., Kenney 1990) or will be analyzed in the future.
B. SAMPLE COLLECTION AND SPLITTING METHODS
1. Mainstem and Tidal Tributaries component
A field crew from the Maryland Department of the Environment (MDE)
collected quarterly water samples from the surface layer at Station CB5.3,
near Smith Point on the Maryland-Virginia line. The field crew followed
the splitting procedures in the CSSP Implementation Guidelines (CBP 1989).
The MDE field crew processed and distributed samples to the two Maryland
laboratories (CBL and MDHMH), while a VIMS field crew processed and
distributed samples to the two Virginia laboratories (VIMS and ODU).
-------
Starting in June 1989, each laboratory analyzed a minimum of four
samples per cruise: three aliquots split in the field and a laboratory
replicate for one of the aliquots. One exception was CBL, which received
four bottles from the field crew. Their "laboratory" replicate was split
from Aliquot 1 in the field until March 1990. Some laboratories did more
than one laboratory replicate: ODU analyzed all three aliquots in
triplicatenine samples per cruise. Because CSC/CBLO staff received these
additional replicate data after data entry was almost complete, they were
only included for Cruise 111.
2. Tidal Potomac component
A field crew from DCRA collected quarterly water samples from the
surface layer at Station PMS-10, at Key Bridge on the Potomac River. The
field crew followed the splitting procedures in the CSSP Implementation
Guidelines (CBP 1989) starting in May 1989. The field crew left whole
water samples in ice-filled coolers at the designated dock. Personnel
from each laboratory retrieved the coolers. The field crew did not filter
any samples. MDHMH did laboratory filtration for Total Suspended Solids
(TSS) only. In the laboratory, DCLS and CRL/DCRA personnel also filtered
samples for Ammonium (NH4), Nitrite (N02), Nitrite + Nitrate (N023),
Orthophosphate (PO4F), Total Dissolved Phosphorus (TOP), Dissolved Organic
Carbon (DOC), and Silica (SI). CRL/DCRA and DCLS both used pre-rinsed
Gelman cellulose membrane filters with 0.45 micron pore size. Samples were
received by the laboratories either the same day they were collected or the
following day. The March 1990 samples were not picked up for DCLS, so
there are no DCLS results for that split sample.
Starting in May 1989, each laboratory analyzed a minimum of four
samples per cruise: three aliquots (field replicates) split in the field
and a laboratory replicate for one of the aliquots.
3. Fall Line component
A field crew from USGS-Towson sampled the Susquehanna River fall line
station and distributed samples to each laboratory (except DCLS, which did
not participate). The field crew used USGS sampling procedures, including
flow-weighted cross-sectionally integrated samples. Splitting was done
with a churn splitter. Field filtration was done with a 0.45 micron
membrane filter and the nutrient samples shipped to USGS and OWML were
preserved with mercuric chloride according to USGS standard protocol. All
samples were immediately placed and kept on ice. Samples for MDHMH and
PADER were delivered to the laboratory the day of collection. Samples for
USGS and OWML were sent via priority mail in ice-filled coolers. USGS
samples usually arrived in two days. The samples sent to OWML arrived five
days later with no ice around them, so their analytical results were not
reported.
-------
The Fall Line component generally followed CSSP guidelines (CBP
1989), except each laboratory received less than three aliquots (field
replicates) split in the field. The field crew could not split a
sufficiently large volume of water accurately to provide three aliquots.
Each laboratory received two aliquots in October 1989 and one aliquot in
March 1990. Only MDHMH reported a laboratory replicate for one of the
aliquots.
C. DATA ENTRY AND REDUCTION
1. Mainstem and Tidal Tributaries component
Laboratory or program personnel submitted raw data on handwritten CSSP
Data Submission forms, except VIMS personnel submitted their data on
diskette starting in June, 1989. CSC/CBLO staff transcribed and entered
the handwritten data, and data originators verified the printouts. Field
and laboratory precision and means over the three aliquots were calculated
with SAS programs (SAS Institute 1985).
2. Tidal Potomac component
Laboratory or program personnel submitted raw data on handwritten CSSP
Data Submission forms. CSC/CBLO staff transcribed and entered the
handwritten data. Data submitters verified the printouts through March
1990. Field and laboratory precision and means for the three aliquots were
calculated with the SAS procedure MEANS (SAS Institute 1985).
3. Fall Line component
USGS-Towson personnel submitted the Upper Bay data on diskette for
October 1989 data and in hard copy for March 1990 data. CSC/CBLO staff
transcribed and entered the handwritten data, transferred the diskette
data. Data submitters verified printouts of the data. Field precision was
calculated with the SAS procedure MEANS (SAS Institute 1985).
D. DATA ANALYSIS AND GRAPHING
1. Preliminary test of splitting randomness
Data were checked for the randomness of the splitting procedures. If
splitting was done correctly, the results for one of the aliquots should
not be consistently higher or lower than the results for the other
aliquots. Since the aliquots are split sequentially, non-random splitting
would probably result in higher results for solids and particulates in
Aliquot 3, which is drawn from the lower part of the splitting vessel.
Splitting randomness was checked with the Friedman two-way non-parametric
analysis of variance, comparing the results for the three aliquots for each
parameter and sampling date. None of the parameters had statistically
significant results (P > 0.03), showing that splitting was done randomly.
-------
2. Precision estimates
The standard deviation (SD) and coefficient of variation (CV,
standard deviation/mean x 100) of field triplicate results estimated
within-organization precision. Since the field replicates are usually
split by a different organization from the one doing the laboratory
analysis, they do not measure only "within-organization" field and
laboratory precision, but they are considered to approximate it for the
purposes of this report. The CSSP results also include laboratory
replicates, split in the laboratory just before analysis. These were not
used to estimate within-organization precision because they do not include
field variability, and are almost always less variable than the field
replicates.
The means of the field triplicate results were then used to calculate
the SD and CV of the results from different organizations for each sampling
date, which estimated inter-organization precision. For the Mainstem
component, precision estimates were calculated separately for groups of
four and three laboratory means, including all the laboratories or only the
mainstem laboratories (see above).
The SD was positively correlated with the mean for several parameters,
and usually the CV was not affected by the mean. However, the CV was
sometimes affected by concentration as well. It was positively correlated
with the mean in a few cases, and negatively correlated with the mean in a
few other cases. The negative correlation usually occurred when the mean
concentrations were low. Thus, neither precision estimate should be used
in other analyses without checking for concentration effects. Because the
primary purpose of the CSSP is to assess inter-organization agreement, a
detailed analysis of precision estimates is beyond the scope of this
report. The author is currently analyzing fall line split sample data to
produce precision estimates to use in calibrating the Chesapeake Bay River
Input computer model. Preliminary results of this analysis can be sent
upon request.
The Method Detection Limit (MDL) was also used to estimate within-
organization precision, especially in the graphs of the data (see below).
At many of the laboratories, the MDL is calculated from three times the
standard deviation of seven replicates of a low-level sample, so it
estimates within-organization precision.
3. Assessing inter-organization agreement
Graphs of the split sample results show which differences were larger
than the within-organization precision. Based on a discussion with the
Analytical Methods and Quality Assurance Workgroup (AMQAW) on 4/24/90,
within-organization precision for CSSP analyses is estimated by the larger
of: 1) the Method Detection Limit (MDL, Tables 1-3); or, 2) the standard
deviation of the three aliquots for each sample which estimates field
precision. Graphs of the cruise means for each laboratory show this
estimate as "precision bars." Any laboratory means with non-overlapping
-------
precision bars have differences that are larger than within-organization
precision.
Statistical significance was assumed when the significance level (P) <
0.01, rather than when P < Q.,05 as in the previous report (Bergstrom 1989).
Standard quality control procedures use the P - 0.01 level as the "control"
or action level for precision and accuracy charts (e.g., Montgomery 1985).
A decision rule was developed to identify which parameters had inter-
organization differences that were large and consistent enough to warrant
investigation by the organizations involved. The decision is based on
graphs with precision bars and the results of statistical tests.
Investigation was recommended if:
1) three or more sampling dates had inter-organization differences
that were larger than within-organization precision; and,
2) an appropriate statistical test had a probability (P) < 0.01 that
the differences were due to chance alone, equivalent to 99% confidence
that the observed difference was real.
The graphs show the magnitude of differences, while the statistical test is
more sensitive to consistency of the differences over time. Based on
results from the Mainstem component of the CSSP (see below), parameters
identified by the combination of these two criteria usually have different
field and/or laboratory methods at one of the laboratories involved.
A non-parametric statistical test was used to analyze the split sample
data. This test assumes matched (positively correlated) samples, since
this is inherent in the split sample design. Comparisons were done with
the Friedman two-way, non-parametric, repeated measures analysis of
variance (ANOVA), which requires a minimum of 4 complete samples to achieve
P - 0.01 (Siegel 1956). The P < 0.01 standard is currently unattainable
for most of the parameters in the Potomac component due to small sample
sizes. The test was done with the Macintosh software package StatView 512+
(Brainpower, Inc. 1986) using exact P values from Siegel (1956) when sample
sizes were small. Below detection limit data were included if they only
affected results from one laboratory, but no comparison was made if two or
more laboratories had below detection limit data, because this made the
rankings of the data ambiguous.
The analyses for the Mainstem component were done on two groups of
laboratories, because the four laboratories included in this component
differed in analytical methods. The three mainstem laboratories use
different analytical methods than the tributary laboratories. Mainstem
laboratories measure the dissolved and particulate fractions of nitrogen,
carbon, and phosphorus and calculate the total fractions, while tributary
laboratories measure the total and dissolved fractions and calculate the
particulate fractions (D'Elia et al. 1987). Data analyses and graphs were
done for all four laboratories and also for the three mainstem
laboratories, because of this difference in analytical methods.
-------
Participate Carbon (PC) comparisons excluded MDHMH results because their
method only measures organic carbon.
Split sample results can also be analyzed by comparing the magnitude
of inter-organization CV to the mean within-organization CV of the
different organizations. This approach was used with Chesapeake Bay fall
line split sample data by Kenney (1990), adapted from a procedure for
performance testing given by Taylor (1987). It was not used in this
report for three reasons: Taylor's approach does not provide decision rules
for identifying parameters which should receive further investigation; CV
comparisons could be biased by the dependence of CV on mean concentration,
or by a few high CV values; and, this comparison method does not take the
consistency of inter-organization differences into account. The
evaluation of the consistency of inter-organization differences in this
report is designed to facilitate actions by the organizations involved to
increase inter-organization agreement.
III. RESULTS
A. WITHIN-ORGANIZATION PRECISION AND ACCURACY
Two estimates of within-organization precision were used in this
analysis: the method detection limits (MDLs), listed in Tables 1-3; and the
precision of field replicates, the three aliquots split in the field and
analyzed by the same laboratory. The Methods section describes how the
precision estimates were calculated. Percent recovery data and results
from Standard Reference Materials (SRMs) estimated within-organization
accuracy.
1. Mainstem and Tidal Tributaries component
Table 4 gives the mean SD and CV of field replicates under "Mean
within-organization precision." These precision estimates varied among
parameters in data from the same organization, as well as among
organizations for the same parameter, with CV values ranging from 1% to
45%. In a few cases, the means were inflated by a single high result, and
sample sizes varied slightly among organizations.
Percent recovery data (Table 7), although limited, show that most
values were near 100%. Results from SRMs (Table 10) from VIMS and ODU also
had percent recovery values (SRM results/expected x 100) near 100%. All
but three were between 90 and 110%.
2. Tidal Potomac component
Table 5 gives the mean SD and CV of field replicates under "Mean
within-organization precision." All CV values were at or below 20%, except
for TSS data from MDHMH. Percent recovery data (Table 8) and results from
Standard Reference Materials (SRMs, Table 11) estimated within-organization
accuracy. Percent recovery data are limited since DCLS did not submit
them, but most values were close to 100%. Results from SRMs from CRL/DCRA
-------
TABLE 1: Method Detection Limits at Mainstem Component Laboratories, 1989-
1990.
Method detection limits (mg/1)
Parameter1
NH4
N02
N023
TDN
PN (PON)
P04F
TOP
PHOSP
TP
DOC
PC
TSS
SI
CBL'
0.003
0.0002
0.0002
0.002
0.0105
0.0006
0.001
0.0012
-
0.24
0.063
1.5
0.01
DCLS"
0.04
0.01
0.04
-
-
0.01
0.01
-
0.01
1.0
-
1
0.1
MDfflW
0.008
0.002
0.02
-
-
0.004
0.01
-
0.01
0.5
(0.8, 5/89-
3/90
-
1
0.1
ODU*
0.0056
0.001
0.0025
0.05
0.05
0.005
0.005
0.007
-
0.5
0.24
2
0.023
VIMS"
0.01
0.0015
0.0021
0.040
0.029
0.0005
0.005
0.003
-
1.0
0.104
5
0.007
NR4 ammonium, NO 2 - Nitrite, NO23 - Nitrite + Nitrate, TDN - Total
Dissolved Nitrogen, PN Particulate Nitrogen, PO4F - 0rthophosphate
filtered, TOP Total Dissolved Phosphorus, DOC - Dissolved Organic
Carbon, PC - Particulate Carbon, TSS - Total Suspended Solids, SI -
Silica (as SI).
Calculated from: 3 x standard deviation of 7 replicates of the lowest
concentration sample encountered. Limits verified by CBL and ODU.
Calculated as 2* of full scale, except carbon limits from Method MDL
in 40 CFR Ft 113 App. B (7-1-87 Ed.).
10
-------
TABLE 2: Method Detection Limits at Potomac Component Laboratories, 1989-
1990.
Method detection limits (mg/1)
Parameter
NH4
N02
N023
TKNW
P04
TOP
TP
DOC
TOC
TSS
SI
BODS
CRL/DCRA
0.04
0.01
0.04
0.2
0.007
0.01
0.01
1.0
1.0
4
0.2
1
DCLS
0.04
0.01
0.04
0.1
0.01
0.1
(0.01)1
0.1
(0.01)1
1.0
1.0
1
0.1
1
MDHMH
0.008
0.002
0.02
0.1
0.004
0.01
0.01
0.5 (0.8 from 5/89-3/90)
0.5 (0.8 from 5/89-3/90)
1
0.1
0.5
Th«ir low-l«v«l phosphorus system achi«v«s 0.01 mg/1,
but this syst«» -was not r»
-------
xftuus j: necncx
1990.
Parameter1
NH4
M02
N023
TKNW/TKNF
P04F
TOP
TP
DOC
i uececcion ionics ac ra-LJ. lone uanponenc ijaooracories, j.y«»-
Method detection limits (mg/1)
FADER
0.02
0.004
0.04
0.2
0.005
0.02
0.02
-
DCLS
0.04
0.01
0.04
0.1
0.01
0.1
0.1
1.0
MDHMH"
0.008
0.002
0.02
0.1
0.004
0.01
0.01
0.5
(0.8,
uses
0.01
0.001
0.10
0.20
0.001
0.001
0.01
5/89-
3/90
OWML
0.01
0.01
0.01
0.10
0.01
-
0.01
TOC
1.0
1.0 0.5 0.1
(0.8, 5/89-
3/90
TSS
SI
BODS
1 1
0.1
1
1
0.1
0.5
1
0.1
1
0.04
1
NH4 - a»»oniu», NO 2 - Nitrite, NO23 - Nitrite + Nitrate, TKNP Total
Kj«ld»hl Nitrog.n Filt«r«d, TKRW - Total Kj«ld«hl Nitrogen Whole,
PO4F Orthophosph*t filtered, TDP - Total Dissolved Phosphorus, TP
Total Phosphorus, TOC - Total Organic Carbon, TSS » Total Suspended
Solids, SI - Silica (as SI).
Calculated as 2% of full scale, except carbon Units froa Method MOL
in 40 CFR Pt 113 App. B (7-1-87 Ed.).
12
-------
4J
d
d
o
3-
o
u
d
3
*>
2
a
d
o
4
a
u
b
Oi
d
o
H
JJ
M
.04
d
u
0
u
w
a
o
q
q
0
H
M
1
D<
Li
f
a
5
*
.
3
2
H
d
e
r*
H
U
U
a
d
o
Jj
4
M
I
U
b
d
2
IB
|
~*
b
B
o
m
.
,3
b
O
h,
B
o
OT
Z
a
o
H
H
U
b
a
d
o
H
4J
M
H
d
O1
u
I
*»4
5
'j
d
2
X
a
£,
B
o
in
§
o
£
B
a
OT
9
g
m
B
a
OT
K
S5
B
a
OT
b
U
b
Bi
0
*
fl
qp
Ol
o
o
0
f*
*
0
in
^
o
o
^
o
0
0
o
o
0
to
o
**
o
o
e
fi
0
^4
f^
O
e
o
4
0
»
o
o
o
i
to
.
fl
m
to
ft
o
o
e
I
i
"
i-
o
o
o
o
1
1
1
to
01
^
o
0
0
o
T
-
o
o
o
o
01
g
m
T
0
r-
0
o
e
ft
fi
ft
i
r4
O
o
"»
m
in
0
o
e
«
fl
^
ft
O
O
o
01
0»
Ol
0
o
0
0
to
fl
0
o
o
fl
Ol
g
to
",
o
ml
o
0
^
o
M
o
o
Ol
in
0
in
o
e
**
e
in
**
m
w
m
0
o
m
Ol
^i
^
^i
o
0
o
H
O
m
o
o
s
H
to r. i
Ol Ot
in 0
Ol
O| ^
O O 1
o o
fl ft 1
oi r>
01 ft
m 0
fl Ol
O O 1
e o
*
to r- I
o m
in
fl O
o o I
0 0
1
to to
Ol ft 1
0| fl
fl fl
Ol
Ol 01
O O 1
o o
1
01 01 |
fi in
o
fi m
fl to
O fl i
0 0
o » "
fl fl fl
o m o
o o o
e o o
h.
Z Z O
Bi H Bi
49
*
^4
in
Ol
m
o
o
o
O|
^
0
fl
ft
Ol
o
o
m
fi
Ol
o
0
0
"
fl
fl
fl
H
O
o
o
**
o
Ol
o
fl
o
o
*
~
in
f*
o
o
o
Bi
in
.
f»
ot
0
0
o
o
fl
^
to
fl
fl
o
o
-t
Ol
o
o
a
o
in
w
01
r«
O
o
o
^
f
m
^
fi
01
o
o
fl
^
o
0
e
Bi
M
§
Bi
to
.
01
fl
^
0
o
0
o
ot
to
to
to
01
o
o
«
o
01
o
o
o
to
fl
fl
0
m
o
o
o
"*
r-
^J
01
^4
0
Ol
o
o
o
H
0
o
0
o
Bi
H
V to fl to
1
0 0 to 1-
m ft » fi
r-
0 « 0 to
01 fl 0 O
fl O to O
o in
1 1
01 0
in i*i
Ol
oi r»
r- o
1 1
to O
.
oi f m
o » »
o m
to oi m
O ft O
0 01 0
» to fl to
oi fi H r-
oi r» » o
Ot 04
0m o
fl O 01 O
O O -9 O
^ 1 ^ 0
O Ol 0|
1
01 Ol to
01
m fi
^ fi *v
O fl O
1
0 fl O
to 0 o oi
fl fl
fi oi oi r-
01 in fi o
oi o r- o
o o o o
U I/I
a Bi H in
*> «
fi **4
* d
U 0
M «
0
0
U b
d
s!
u **
.
d M
0 -H
**4 O
4 O*
. 14
'.?
o d
^2
« -*4
o *
c
t)
"
4
Q ^
"* "3
O
* «
23
^s
s:
fT fi
2 <
g-o
89 J
^
iS
ss
H
O g
b
a o
-1 d
4 H
*J
3 *
4J ~
H O
0
fl «
d
0
" 9
9 O
O fl
T3 ~*
0 0
i^ -H
4J
a «
H
>
4J
O «4
SB o
jj
*
d
°
^
'
M
*
II
d
o
H
n
u
b
Oi
d
o
M
M
H
q
CP
u
^
i,
4J
H
|:
«
o>
w d
«
H Oi
"o, «
2-
u
*<
n
O A
2
3"
" n
fi
9
01 -U
u u
"*
5 *
^j
*"
s
d S
u n
B
b 0
0
So
g
9 o
0 .H
TJ -H
U
M
h
a Oi
d
2
S3
M
*
41*
O
N
9
B
fl
2
o
u
a
J3
4J
4J
3
O
J3
4J
S
..
JJ
0
Ol
§
8
b
(X
0
Oi
a
u
q
H
^
J
1
ttl
3
"3
>
A
Oi
J3
S
JQ
o
^J
e
H
O
U fl
01
a
U! 0
C 9
IB
*<
13
-------
o
M
H
U
U
aj
S §
i
M
U f4
u I
U
O
o.
- a
« -H
m
o>
O
Ul
o
o
a a
4 O
Ol
O
s
I §
I O
H
H U
e a
u o
? 3
u
N
4J -H
5 g
o Z
S I
a -H
O JB
H 41
« -s
H
« S
Q
Ul
O
o
m
o
o
o
A
IN
O
O
r>
g g
04
H
U)
V)
H
! " o o
o -a
I ^
<=>(«
2 *» 5 2
s^^i
-* w «
flsa-3
o " o-
315325
o -S *» a 3
6 W H -H Q,
00 «
w u .
** -o * JS
- S1 "
u g u
O-o^O
** § "3 «**
"2 *J ° " S
^ * M *^ ^
3 -H 5 a Q.
5 > S *
u ' u
»-* u
5 S -8 ~ -
««:*-
tm
"*" 3 5 5 »
»':"§
O " -H H
^ BI « a
q
-------
show good agreement with the expected results, and all but one result (the
last P04F result) were within the 95% confidence intervals for the SRMs.
3. Fall Line component
Field precision was estimated from the standard deviation of field
duplicate aliquots analyzed for the October 11, 1989 Upper Bay sample
(Table 6). Although based on results from a single sampling date, the
results are generally similar to those from the Potomac component (Table
5). Percent recovery data (Table 9) were only reported by FADER and MDHMH;
most values were close to 100%. No SEN results were reported for this
component.
B. INTER-ORGANIZATION PRECISION
The tables of inter-organization precision give both the standard
deviation (SO) and coefficient of variation (CV) of the mean results from
split samples for each sampling date. The data user will have to decide
which precision estimate is best for their application (see Methods for a
definition and discussion of each estimate).
1. Mainstem and Tidal Tributaries component
Table 4 lists SD and CV estimates under "Mean inter-organization
precision." Although parameters with low inter-organization precision
tended to have low inter-organization agreement, this correlation was not
found in all cases. Three of the five parameters with statistically
significant inter-organization differences (TP, TOP, and DOC; see below)
had high CV values (> 50%), while the other two parameters in that group
(PC and PN) had lower CV values. Three parameters with high agreement
(NH4, PHOSP, and TSS) also had mean inter-organization CV values over 50%
when four organizations were included.
2. Tidal Potomac component
The mean inter-organization precision estimates for Potomac data
(Table 5) varied from CV values of 5% for NH4 to 41% for TSS, and TKNW had
the next highest CV value. Due to small sample sizes, more data are needed
to determine if there are consistent differences among parameters in CV.
The parameter that had the lowest inter-organization agreement based on
graphing and ANOVA, N023 (see below), had one of the lowest CV values.
3. Fall Line component
No estimates were possible due to limited and missing data for 1989.
15
-------
a
*
a
i
o
u
B
3
i
h.
^
»
2
2
H
*
0
H
e
H
U
b
ft
§
H
*>
N
H
a
8.
w
I
a
A
H
*
^
g
B
0
n
H
U
b
a
o
*f4
4J
M
H
S
tr
?**
B
4J
H
S
b
^
!
2
Ui
(9
S
U
3
a
Q
jjr
w
B
a
n
a
U)
B
a
U)
ot4ooiNO>r4or4«oo
ioo««oaoor-iflo
(S H m »< «
o P* »4 -
'U o -
1! " a S
H On
H « -1
B
* M. O **
5>g «
b u o
o 3 -H
b B * "ft
o a
U -H a T>
B 4J 3
M'? "
J -S -o rl
"3
^^15
> * «
"3 *"
2l S-
« 3 > -
"* Ml O *"*
^4 * 3
- *
* 2
s u
a"° a
S at *>
7*4
1 .2 °
o « o. o
3 " 0
*u _j >-* C
0 g *B .2
.5 « M
a " « -H
X U
4J JJ X
o o a b
z e v) Qi
16
-------
TABLE 7: Percent Recovery Data, Mainstem component, 1989.
Parameter Laboratory Percent Recovery
TOP
TP
SI
N02
N023
NH4
PHOSP
DOC
P04F
TON
CBL
HDHMH
VIMS
MDHMH
CBL
MDHMH
VIMS
CBL
MDHMH
VIMS
CBL
MDHMH
VIMS
CBL
MDHMH
VIMS
CBL
VIMS
CBL
MDHMH
CBL
MDHMH
VIMS
CBL
VIMS
June 89
97
100
-
_
93
87
99
92
96
96
114
108
107
144
96
100
100
107
79
90
120
98
98
92
Sept 89
102
100
87
112
_
95
-
99
100
100
105
116
91
105
102
101
_
89
_
89
_
100
90
100
79
Dec 89
_
102
87
100
_
93
88
103
^^
95
^
94
101
_
_
122
^
72
100
w
85
Not*: Percent recovery data are only possible for directly measured
paraaeters. There is no way to spike parameters don* in th* CHN analyzer
(Particulate Carbon, PC and Particulat* Nitrogen, PN). ODU did not report
any p*rc*nt r*cov*ry data, and CBL did not report p*rc*nt recovery data
after Sept. 89. TOP » Total Dissolved Phosphorus, TP » Total Phosphorus,
SI Silica, NO2 - nitrite, N023 » nitrite + nitrate, NH4 - ammonium, PHOSP
Particulate Phosphorus, DOC » dissolved Organic Carbon, P04F »
0rthophosphat filtered, TDN » Total Dissolved Nitrogen.
17
-------
TABLE 8: Percent Recovery Data, Potomac Component, 1989-90.
Parameter Laboratory
Percent Recovery
5/1/89 6/12/89
9/11/89 f/8/90
3/5/90
NH4
N02
N023
TKNW
P04
TCP
TP
DOC
TOC
SI
CRL/DCRA
MDHMH
CRL/DCRA
MDHMH
CRL/DCRA
MDHMH
CRL/DCRA
MDHMH
CRL/DCRA
MDHMH
CRL/DCRA
CRL/DCRA
MDHMH
CRL/DCRA
CRL/DCRA
MDHMH
CRL/DCRA
MDHMH
140
110
98
105
102
100
144
116
82
103
136
107
108
108
100
85
95
117
98
96
106
115
104
91
99
111
93
102
116
100
106
108
67
100
96
112
95
83
76
86
92
92
75
102
110
102
100
85
102
109
98
136
100
100
100
108
115
93
95
86
76
96
95
98
104
98
94
102
100
105
104
103
138
Not*: Percent recovery data are not possible for TSS analysis. DCLS did
not report any percent recovery data. TOP ~ Total Dissolved Phosphorus, TP
Total Phosphorus, SI - Silica, NO2 - Nitrite, NO23 - Nitrite + Nitrate,
TKNW - Total Kjeldahl Nitrogen Whole, NH4 Ammonium, DOC - Dissolved
Organic Carbon, TOC - Total Organic Carbon, PO4 « Orthophosphate.
18
-------
TABLE 9: Percent Recovery Data, Fall Line Component, 1989-90.
Parameter
NH4
N02
N023
TKNW
TKNF
PO4F
TOP
TP
DOC
TOC
SI
Laboratory
PADER
MDHHH
PADER
NDHMH
PADER
NDHMH
PADER
NDHMH
PADER
MDHNH
PADER
NDHMH
PADER
NDHMH
PADER
NDHMH
MDHMH
NDHMH
PADER
MDHNH
MDHNH
PADER
MDHMH
Percent Recovery
10/11/89 3/28/90
100
100
100
96
100
101
106
100
128
100
102
97.5
100
100
100
104 94
95
100
92 112
103
- -
Hot*: Percent recovery data are not possible for TSS analysis. DCLS, USOS
and OHML did not report any percent recovery data. TDP » Total Dissolved
Phosphorus, TP - Total Phosphorus, SI « Silica, NO2 - Nitrite, NO23 -
Nitrite + Nitrate, TKNW - Total Kjeldahl Nitrogen Whole, TKNF - Total
Kjeldahl Nitrogen Filtered, NH4 « Ammonium, DOC » Dissolved Organic Carbon,
TOC Total Organic Carbon, P04 - 0rthophosphate .
19
-------
TABLE 10: Standard Reference Material Results, Mainstem Component, 1989.
Para- Date Laboratory Distilled matrix Estuarine matrix
meter Results % Recovery Results % Recovery
TOP
N023
NH4
PHOSP
DOC
PO4F
TON
6/89
9/89
12/89
12/89
6/89
9/89
12/89
12/89
6/89
9/89
12/89
12/89
12/89
12/89
6/89
9/89
12/89
6/89
9/89
12/89
12/89
VIMS
VIMS
VIMS
ODU
VIMS
VIMS
VIMS
ODU
VIMS
VIMS
VIMS
ODU
ODU
ODU
VIMS
VIMS
VIMS
VIMS
VIMS
VIMS
ODU
0.072
0.074
0.074
0.213
0.206
0.203
0.0397
0.164
0.183
0.181
0.041
0.37
4.07
0.050
0.047
0.051
0.253
0.217
0.323
0.343
96.0
98.7
98.7
106.5
103.0
101
99
.5
.3
82.0
91,
90,
102.5
94.9
93.3
100.0
94.0
102.0
101.2
86.6
129.2
107.2
0.080
0.074
0.073
0.208
0.205
0.189
0.189
0.196
0.201
0.218
0,
0,
050
048
0.053
0.266
0.287
0.231
106.7
98.7
97.3
95.9
102.5
94.5
94.5
98.0
100.5
109.0
100.0
96.0
106.0
106.2
114.8
92.2
Not*: CBL and HDHMH have not reported any SRM results. Results r given
for diluted samples, diluted either with disti11ed/deionized or estuarine
water. Confidence intervals are not given due to the dilution. ODU
results for PHOSP are for a PO4P standard diluted with IN HC1. TOP - Total
Dissolved Phosphorus, NO23 » Nitrite + Nitrate, NH4 - a»»oniu», PHOSP
Particulate Phosphorus, DOC » Dissolved Organic Carbon, PO4P »
Orthophosphate filtered, TDK Total Dissolved Nitrogen.
20
-------
TABLE 11: Standard Reference Material Results, Potomac
nt, 1989-90.
Para-
meter
Date
Laboratory
Results (mg/1)
Expected
Distilled
95% CI1
NH4
N023
TKNW
PO4F
TOP
TP
TOC
6/89
9/89
1/90
3/90
5/89
6/89
9/89
1/90
3/90
9/89
1/90
3/90
5/89
6/89
9/89
3/90
1/90
3/90
6/89
9/89
5/89
6/89
9/89
1/90
3/90
CKL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
CRL/DCRA
1.90
1.90
1.90
1.90
1.43
1.43
1.43
1.43
1.43
5.
4.
00
78
4.78
.35
.35
.35
.35
1.00
1.03
,50
.50
6.1
6.1
6.1
6.1
6.1
1.88
1.80
1.84
1.93
1.32
1.32
1.38
1.35
1.29
5.
5,
4,
70
05
90
0.327
0.331
0.326
0.401
1.00
0.924
.53
.60
9.2
7.3
7.1
7.5
6.9
1.68
1.68
1.68
1.68
1.28
1.28
1.28
1.28
1.28
- 2.12
- 2.12
- 2.12
- 2.12
- 1.56
- 1.56
- 1.56
- 1.56
- 1.56
3.88 - 6.02
3.70 - 5.77
3.70 - 5.77
0.
0,
0.
0.
33
33
33
33
0.89
0.89
0.37
0.37
0.37
0.37
- 1.21
- 1.21
,30
,30
,50
,50
,50
,50
76
76
9.32
9.32
9.32
9.32
3.50 - 9.32
Provided by EPA Environmental Monitoring and Support Laboratory (EMSL)
Cincinnat i.
Not*: DCLS and MDHMH have not reported any SRM results. NH4 » ammonium,
N023 " nitrite + Nitrate, TKNW » Total Kjeldahl Nitrogen Whole, PO4P -
0rthophosphat filtered, TOP » Total Dissolved Phosphorus, TP - Total
Phosphorus, TOC » Total Organic Carbon.
21
-------
C. INTER-ORGANIZATION AGREEMENT
1. Mainstem and Tidal Tributaries component
Cruise means with precision bars for each parameter are shown in Figs.
3-16. Five parameters out of 14 analyzed had three or more cruises with
non-overlapping error bars: Total Dissolved Phosphorus (TDP), Total
Phosphorus (TP), Particulate Nitrogen (PN), Particulate Carbon (PC), and
Dissolved Organic Carbon (DOC).
Friedman ANOVA performed on cruise means assessed inter-organization
agreement and determined which parameters had statistically significant
inter-organization differences. The Friedman ANOVA results (Table 12)
showed that the same five parameters (TDP/ TP, PN, PC, and DOC) had
statistically significant inter-organization differences (P < 0.01). The
first two parameters had significantly higher results from MDHMH, while
the two participate parameters had significantly higher results from CBL.
DOC results were significantly lower at MDHMH, due to a series of below
detection limit values starting in June 1989 (Fig. 13) after they started
using a new DOC instrument.
2. Tidal Potomac component
Cruise means with precision bars for each parameter plotted against
time show inter-organization agreement (Figures 17-26). Two parameters
out of 10 graphed, N023 and TP, had three or more sampling dates with non-
overlapping precision bars. N023 samples (Fig. 19) were from unfiltered
samples at MDHMH and filtered samples at the other two laboratories. TP
results had non-overlapping precision bars between MDHMH and CRL/DCRA
results on three dates (Fig. 22). The other parameters had inter-
organization differences that tended to be smaller than within-
organization precision. Two parameters are not shown due to missing data:
TDP and DOC.
The Friedman ANOVA results (Table 13) showed that none of the nine
parameters analyzed had statistically significant inter-organization
differences at the P < 0.01 level, although this significance level was
only possible when N - 4. N023 results were the closest to statistical
significance (P - 0.042). Below detection limit values at DCLS prevented
the analysis of TP results, and missing data ruled out analysis of BOD5
results.
3. Fall Line component
No analyses were possible due to limited and missing data for 1989.
22
-------
FIGURE 3. Split sample data for Total Dissolved Phosphorus (TDP), from samples collected
at Station CBS.3 (Mainstem), showing cruise means with precision bars.
0.10
^i 0.08
E
%
M
J 0.06
Q TDP_CBL
TDP~MDH
O TDP_ODU
A TDP~VIM
"3 0.04 '
66 70 74 78 82 86 90 94 98 102 106 110
CRUISE and DATE
FIGURE 4. Split sample data for Total Phosphorus (TP), from samples collected at Station
CB5.3 (Mainstem), showing cruise means with precision bars.
TP_CBL
TP'MDH
TP_ODU
Ar TP VIM
98 102 106 110
s
CRUISE and DATE
23
-------
FIGURE 5. Split sample data for Paniculate Nitrogen (PN), from samples collected at Station
CB5.3 (Mainstem), showing cruise means with precision bars.
OJO
0.25
£ 0.20
e
| 0.15 '
Z
i o.io
0.05
0.00
PN_CBL
PN'MDH
PN_ODU
PN'VIM
(No precision shown for MDHMH)
66 70 74 78 82 86 90 94 98 102 106 110
i £ p s 1 * i §
v) S OQS 2 to a
CRUISE and DATE
FIGURE 6. Split sample data for Particulate Carbon (PQ, from samples collected at Station
CB5.3 (Mainstem), showing cruise means with precision bars.
.
4^
«
eu
1.4
1.1
1.0
0.9 '
0.8 '
0.7
0.6
0.5 1
0.4
0 J
0.2
0.1
0.0
66
PC_CBL
PC~ODU
PC VIM
70 74 78 82 86 90 94 98 102 106 110
CRUISE and DATE
24
-------
FIGURE 7. Split sample data for Silica (SI), from samples collected at Station CB5.3 (Main-
stem), showing cruise means with precision bars.
CA
CA
SI_CBL
SI~MDH
SI_ODU
SI'VIM
Note: Cruises 81 and 94 had
below detection limit values.
CRUISE and DATE
FIGURE 8. Split sample data for Nitrite (NO2), from samples collected at Station CB5.3
(Mainstem), showing cruise means with precision bars.
0.025
0.020
NO2_CBL
NO2~MDH
At NOZ'VIM
0.000
66 70 74 78 82 86 90 94 98 102 106 110
CRUISE and DATE
25
-------
FIGURE 9. Split sample data for Nitrite + Nitrate (NO23), from samples collected at Station
CB5.3 (Mainstem), showing cruise means with precision bars.
-S
2
-O NO23_CBL
- N023~MDH
-O NO23_ODU
A NO23~VIM
0.20
0.18
0.16
0.14
0.12 '
0.10
0.08
0.06
0.04
0.02
0.00
CRUISE and DATE
FIGURE 10. Split sample data for Ammonium (NH4), from samples collected at Station CB5.3
(Mainstem), showing cruise means with precision bars.
%,
|
<
NH4_CBL
NH4~MDH
NH4_ODU
NH4~VIM
78 82 86 90 94 98 102 106 110
CRUISE and DATE
26
-------
FIGURE 11. Split sample data for Paniculate Phosphorus (PHOSP), from samples collected at
Station CB5.3 (Mainstem), showing cruise means with precision bars.
PHOSPCBL
PHOSPMDH
PHOSPODU
£r PHOSPVIM
CRUISE and DATE
FIGURE 12. Split sample data for Orthophosphate (PO4F), from samples collected at Station
CB5.3 (Mainstem), showing cruise means with precision bars.
0.050
eo
Q.
i
a.
o
P04F_CBL
P04F~MDH
PO4F"viM
Note: All cruises had below
detection limit values.
CRUISE and DATE
27
-------
FIGURE 13. Split sample data for Dissolved Organic Carbon (DOC), from samples collected at
Station CB5.3 (Mainstem), showing cruise means with precision bars.
5.0
e
5
g
i
o
1
i
.2
Q
4.5
4.0
3.5
3.0 '
2£
2.0 '
1.5
1.0 1
0.5
0.0
DOC_CBL
DOC~MDH
DOC~ODU
Only one cruise (94) had data
for a 3 -way comparison.
66 70 74 78 82 86 90 94
& 8 3 S 8
98 102 106 110
O Q Jg 2 «o Q
CRUISE and DATE
FIGURE 14. Split sample data for Total Dissolved Nitrogen (TON), from samples collected at
Station CB5.3 (Mainstem), showing cruise means with precision bars.
0.60
0.55
0.50 -j
I 0-45
0.40
M
i
2
5 OJ5 '
Q
3
OJO
OJ15 '
0.20
TDN_CBL
TDN~MDH
TDN'ODU
TON VIM
No precision shown for MDHMH, too wide
66 70 74 78 82 86 90 94 98 102 106 110
r~ 23 2op2i o\o2!
CRUISE and DATE
28
-------
FIGURE 15. Split sample data for Total Nitrogen (TN), from samples collected at Station
CB5.3 (Mainstem), showing cruise means with precision bars.
i .2 ^
1.1
1.0
0.9 1
0.8
0.7 '
0.6
0.5
0.4 H
0.3
0.2 1
0.1
0.0
TN_CBL
TN'MDH
TN_ODU
TN~VIM
66 70 74 78 82 86 90 94 98 102 106 110
CRUISE and DATE
FIGURE 16. Split sample data for Total Suspended Solids (TSS), from samples collected at
Station CB5.3 (Mainstem), showing cruise means with precision bars.
26
T3
1
i
3
24
22 1
20
18 1
16
14
12
10
8-
6
4'
2-
0
TSS_CBL
TSS_MDH
TSS_ODU
TSS~VIM
66 70 74 78 82 86
90
I
94
98 102 106 110
Os
2
CRUISE and DATE
29
-------
TABLE 12: Mainstem Component (Station CB5.3) Split Sample Results using
Cruise Means (1987 - 1989).
Parameter N
Laboratory Means (mg/1)
P values1
NH4
N02
N023
TEN
PN
TN
P04F
TOP
PHOSP
TP
DOC
PC
TSS
SI
8
62
32
6
6
7
83
8
5
6
4
6
3
6
CBL
0.0181
0.00756
0.110
0.398
0.152
A
0.525
0.0036
0.0071
A
0.0138
0.0227
A
2.81
0.840
A
5.1
0.297
MDHMH
0.0370
0.0089
0.102
0.421
0.114
0.633
0.0091
0.048
B
0.0320
0.0722
B
1.15<
A
_5
10.6
0.442
ODU
0.0176
-
0.102
0.388
0.112
0.479
-
0.015
B
0.0181
0.0322
3.63
B
0.686
17.5
0.382
VIMS
0.0203
0.00828
0.102
0.439
0.105
B
0.523
0.00071
0.0084
A
0.0124
0.0218
0.588
B
15.6
0.404
4 way / 3 way
0.051/0.285
- /0.184
0.608/0.36
0.532/0.43
0.0074/0.0055
0.28 /0.085
- /MDL3
0.0011/0.079
0.47 /0.17
0.0055/0.142
- /0.0046
- /0.0017
0.148/0.194
0.016/0.052
Underlined P values w*r«> statistically significant ( P_ < 0.01, Friedman
2-way ANOVA" on cruise Means, using exact probabilities for 3-way
comparisons. Laboratory Beans with different letters below then had
statistically significant pairwise differences (expe r iment wi s e P_ <
0.01), otherwise they did not differ. When there were data from four
labs, 3-way comparisons omitted MDHMH data.
Only using data from cruises that had values above detection limits.
Too many values were below the MDL to make a comparison.
MDHMH data were mostly below detection due to a faulty instrument.
MDHMH data were excluded because they included only organic carbon.
30
-------
FIGURE 17. Split sample data for Ammonium (NH4), from samples collected at
Station PMS-10 (Potomac), showing cruise means with precision bars.
I
E
I
0.20
0.18 -
0.16 -
0.14 -
0.12 -
0.10 -
0.08 '
0.06 -
0.04 -
0.02 -
0.00
NH4_CRL
NH4_DCL
NH4~MDH
£
i
Gd
I
FIGURE 18. Split sample data for Nitrite (NO2), from samples collected at Station
PMS-10 (Potomac), showing cruise means with precision bars.
0.040
N02_CRL
NO2_DCL
NO2 MDH
0.000
I
u
en
|
y?
31
-------
FIGURE 19. Split sample data for Nitrite + Nitrate (NO23), from samples collected at Station
PMS-10 (Potomac), showing cruise means with precision bars. MDHMH samples
were unfiltered, other samples were filtered (see text).
2.0
1.8
1.6
1.4
L2
1.0
0.8
0.6
i: 0.4 -
Z
0.2 H
2
0.0
NO23_CRL
NO23_DCL
NO23 MDH
i
CA
I
FIGURE 20. Split sample data for Total Kjeldahl Nitrogen Whole (TKNW), from samples
collected at Station PMS-10 (Potomac), showing cruise means with precision
^ ^
b
I
2
^5
s?
1.0
o^
0.8
0.7
0.6
OJ
0.4
0.3
OJ
0.1
0.0
TKNW_CRL
TKNW.DCL
TKNW'MDBI
1
g
I
32
-------
FIGURE 21. Split sample data for Orthophosphate (PCM), from samples collected at Station
PMS-10 (Potomac), showing cruise means with precision bars.
PO4F_CRL
- PO4F_DCL
PO4W MDH
FIGURE 22. Split sample data for Total Phosphorus (TP), from samples collected at Station
PMS-10 (Potomac), showing cruise means with precision bars.
TP_CRL
TP_DCL
TP MDH
33
-------
FIGURE 23. Split sample data for Total Organic Carbon (TOG), from samples collected at
Station PMS-10 (Potomac), showing cruise means with precision bars.
U
i
«
u
w
2
&
73
7.0
6.5
6.0
53
5.0
4.5
4.0
3.5
3.0
2.0 '
1.0
0.5
0.0
TOC_CRL
TOcfDCL
TOC MDH
|
Cd
CA
FIGURE 24. Split sample data for Total Suspended Solids (TSS), from samples collected at
Station PMS-10 (Potomac), showing cruise means with precision bars.
JO
"o
I
CO
s
TSS_CRL
TSS_DCL
TSS'MDH
3A
-------
FIGURE 25. Split sample data for Silica (SI), from samples collected at Station PMS-10
(Potomac), showing cruise means with precision bars.
CM
en
at
a
C/3
4.0
3.5
3.0
SI_CRL
SI_DCL
SI'MDH
FIGURE 26. Split sample data for Biological Oxygen Demand 5 day (BODS), from samples
collected at Station PMS-10 (Potomac), showing cruise means with precision bars.
CRL/DCRA data from 05MAR90 were added to the figure.
O
B
i
I
O
^»
"o
BOD5.DCL
BOD5~MDH
I
I
ac
a.
u
03
35
-------
TABLE 13: Potomac Component (Station PMS-10) Split Sample Results using
Cruise Means (1989 - 1990).
Parameter1
N
Laboratory Means (mg/1)
P value2
NH4
N02
N023
TKNW
P04
TP .
TOC
TSS
SI
3
4
4
3
3
3
4
3
3
CRL/DCRA
0.097
0.015
1.255
0.676
0.027
0.065
4.48
12.3
1.66
DCLS
0.098
0.013
1.395
0.478
0.039
0.10
3.62
8.8
1.66
MDHMH
0.092
0.012
1.525
0.544
0.038
0.082
3.12
11.1
1.57
0.36
0.93
0.042
0.36
0.19
MDL3
0.65
0.53
0.53
Total Dissolved Phosphorus (TOP), Dissolved Organic Carbon (DOC), and
Biological Oxygen Demand 5 day (BODS) could not be analyzed due to
ising data .
No P_ values were statistically significant ( P_ < 0.01, Friedman 2-way
ANOVA on cruise means uning exact probabilities).
Too many values were bellow the HDL to Make a comparison.
36
-------
IV. DISCUSSION
All of these results are preliminary, due to small sample sizes and
start-up problems inherent in any new sampling program. The results are
not necessarily or exclusively a reflection of laboratory performance.
They may suggest that CSSP design features, specific parameters, or field
or laboratory methods should receive extra attention in future quality
assurance work.
Readers should also keep in mind that all the components include state
laboratories that analyze a large number of samples per day comprising a
wide range of sample concentrationsfrom estuarine samples to wastewater
treatment plant samples. The number of samples they analyze reduces the
time available for researching advanced techniques and equipment. Some of
the laboratories have also had difficulty obtaining and paying for the
Standard Reference Material samples that are included in some CSSP results.
A. WITHIN-ORGANIZATION PRECISION AND ACCURACY
The estimates of within-organization precision and accuracy generally
show similar results from the different organizations involved. One
exception is the two parameters calculated by subtraction in the Mainstem
component, PN and PHOSP results from MDHMH (Table 4). These were more
variable than the PN and PHOSP results calculated directly by the other
three organizations. The same pattern was noted by D'Elia et al. (1987)
and used as an argument for using the direct methods. One parameter, TON,
was calculated by addition in the MDHMH data and directly by the other
organizations. It had.similar precision among the four organizations.
B. INTER-ORGANIZATION PRECISION
The estimates of inter-organization precision are designed to assess
the measurement system variability in each component after sampling has
occurred. They should be used with caution for that purpose, since both SD
and CV values can be affected by concentration. They should not be used to
assess inter-organization agreement. Unless there are consistent inter-
organization differences over several sampling dates, larger SD and/or CV
values may not indicate any problem with inter-organization agreement. For
example, some tests may be more sensitive to variations in splitting
procedures than others; this variability may not be under the laboratory's
control. The SD and CV results will be more useful when sample sizes are
larger.
C. INTER-ORGANIZATION AGREEMENT
1. Mainstem component
In the Mainstem component, five parameters (of 14 compared) showed
statistically significant inter-organization differences that were larger
than wi thin-organization precision. All involve a method difference at the
laboratory that had divergent results. MDHMH uses a different method for
TP and TOP from the three mainstem laboratories. The MDHMH method does not
37
-------
appear sufficiently sensitive to measure the low TP and TOP concentrations
at Station CB5.3. Recent digestion method changes should increase
sensitivity to low TP and TOP concentrations. CBL uses a different CHN
analyzer (used for PC and PN) and a different filter (25 mm diameter
Whatman GF/F) than ODU or VIMS. ODU and VIMS both use identical GIN
analyzers and 13 mm diameter Gelman AE filters. Experiments are being
planned to study the possible effects of these different instruments and
filters on PC and PN results. The problem with DOC analyses at MDHMH was
caused by a new DOC instrument that did not work with estuarine samples.
It has now been remedied by using their old DOC instrument, which works
better with estuarine samples.
2. Potomac component
In the Potomac component, none of the nine parameters analyzed met the
criteria for recommending further investigation. Two of the parameters,
N023 and TP, had inter-organization differences that were larger than
within-organization precision on three or more sampling dates. However,
none of them had statistically significant differences. Unfiltered samples
at MDHMH may have accounted for the N023 differences, since MDHMH had
higher results than the other two laboratories. MDHMH will receive
filtered samples starting in late 1990. The DCLS results for TP were all
below their detection limit because their low-level phosphorus system was
not requested for these samples.
3. Fall line component
No analyses were done due to the limited data for 1989. The sample
distribution problems were solved in 1990, and each laboratory should start
receiving triplicate aliquots once a cone splitter is obtained.
V. SUMMARY AND CONCLUSIONS
Estimates of within-organization and inter-organization precision were
presented which may be useful in statistical analyses and computer
modeling of Chesapeake Bay water quality. In three cases, parameters that
showed low inter-organization precision also had low inter-organization
agreement, but the two were not always correlated. Other parameters such
as TKNW and TSS usually had fairly low inter-organization precision but
generally high inter-organization agreement. Estimates of within-
organization accuracy, including percent recovery results and Standard
Reference materials (SRMs), showed high accuracy.
Assessments of inter-organization agreement found high agreement for
18 of the 23 comparisons made in two components. Agreement was low enough
to recommend investigation for five parameters (TP, TOP, PC, PN, and DOC).
In all five cases, the organization with divergent results had a different
analytical method or instrument type, and in two cases (PC and PN) there
was also a difference in filter type. In three cases (TP, TOP, and DOC),
method changes have been made to increase inter-organization agreement, and
38
-------
the other two cases are being investigated by the organizations involved to
find ways to increase agreement.
VI. REFERENCES
Bergstrom, P. 1989. Split sample water quality results from laboratories
participating in the Chesapeake Bay Program: 1985-1989. Chesapeake
Bay Program, Coordinated Split Sample Program, Report Series: No. 1.
Bergstrom, P. 1990a. Coordinated Split Sample Program Interim Report,
Wainstem Component, 1989. Chesapeake Bay Program, Coordinated Split
Sample Program, Report Series: No. 2.
Bergstrom, P. 1990b. Coordinated Split Sample Program Interim Report,
Potomac Component, 1989. Chesapeake Bay Program, Coordinated Split
Sample Program, Report Series: No. 3.
Bergstrom, P. 1990c. Coordinated Split Sample Program Interim Report,
Fall Line Component, 1989. Chesapeake Bay Program, Coordinated Split
Sample Program, Report Series: No. 4.
Brainpower, Inc. 1986. Statview 512+. Brainpower, Inc., Calabasas, CA.
Chesapeake Bay Program. 1989. Coordinated Split Sample Program
Implementation Guidelines, Revision 1. EPA Chesapeake Bay Program,
Annapolis, MD.
D'Elia, C., R. Magnien, C. Zimmermann, P. Vaas, N. Kaumeyer, C. Keefe, D.
Shaw, and K. Wood. 1987. Nitrogen and phosphorus determinations in
estuarine waters: A comparison of methods used in Chesapeake Bay
Monitoring. Report to EPA Chesapeake Bay Program, Annapolis, MD.
Kenney, S. 1990. Maryland Department of the Environment Chesapeake Bay
Water Quality Monitoring Program River Input Monitoring Component,
Inter- and Intra-laboratory Comparisons. MDE Water Management
Administration, Chesapeake Bay and Special Projects Program, Watershed
Restoration Projects Division Administrative Report, Baltimore, MD.
Montgomery, D. C. 1985. Introduction to statistical quality control.
Wiley & Sons, NY.
SAS Institute, Inc. 1985. Statistical Analysis System (SAS) Usersfs
Guide: Basics, Version 5 edition. SAS Institute, Gary, NC.
Siegel, S. 1956. Non-parametric statistics for the behavioral sciences.
McGraw-Hill Book Co., NY.
Taylor, J. K. 1987. Quality assurance of chemical measurements. Lewis
Publishers, Chelsea, MI.
39
-------
ui o O
-o 52. n>
3
------- |