United States
                  Environmental Protection
                  Agency
Environmental Monitoring
Systems Laboratory
Las Vegas, NV 89193-3478
                  Research and Development
EPA/600/S4-90/001 July 1990
4>EPA         Project  Summary
                  Direct/Delayed  Response
                  Project:  Quality Assurance
                  Report for Physical  and
                  Chemical  Analyses of
                  Soils  from the  Mid-Appalachian
                  Region  of the United  States
                  G. E. Byers, R. D. Van Remortel, M. J. Miah, J. E. Teberg, M. L. Papp,
                  B. A. Schumacher, B. L. Conkling, D. L. Cassell, and P. W. Shaffer
                   The  Direct/Delayed Response
                 Project was designed to address the
                 concern over potential acidification
                 of surface waters by atmospheric
                 sulfur deposition in the United States.
                 The purpose of these synoptic soil
                 physical and chemical surveys was
                 to characterize watersheds in regions
                 of the United States  believed to be
                 susceptible to  the effects of  acidic
                 deposition. This document describes
                 the implementation of a quality
                 assurance  program  and  the
                 verification of the analytical  data
                 base for the Mid-Appalachian Soil
                 Survey.   It is  directed primarily
                 towards the users of the data base
                 who will be analyzing the data and
                 making various assessments and
                 conclusions relating to the effects of
                 acidic deposition on  watersheds of
                 the region.
                   The  results show  that the
                 measurement quality objectives for
                 detectability, precision, accuracy,
                 representativeness; and complete-
                 ness  were generally satisfied.
                 Measurement  uncertainty  was
                 generally low in relation to  overall
                 data  uncertainty. A series of con-
                 clusions and recommendations are
                 provided at the end of the report. The
                 recommendations will be useful in
the planning of future projects of this
nature.
  This Project  Summary was
developed by EPA's Environmental
Monitoring Systems Laboratory, Las
Vegas, NV, to  announce key findings
of the research project that is fully
documented in a  separate report of
the  same  title (see Project  Report
ordering information at back).
Introduction
  The U.S. Environmental Protection
Agency (EPA),  as a participant  in the
National Acid  Precipitation Assessment
Program, has designed and implemented
a research program to predict the long-
term response of watersheds and surface
waters in the United States to  acidic
deposition.  Based on this research, a
sample of watershed systems  will be
classified according to the time scale in
which each system will reach an acidic
steady state, assuming current levels  of
acidic deposition.   The  Direct/Delayed
Response Project (DDRP) was designed
as the terrestrial component of the EPA
Aquatic Effects Research Program.
  The mapping for the DDRP Mid-
Appalachian Soil Survey  was conducted
in portions of Pennsylvania, Virginia, and
West Virginia  during the spring and
summer of 1988 and the sampling took
place during the fall of 1988.  These

-------
 activities initiated  the  third full-scale
 regional  survey  of  the DDRP. The
 physical and  chemical  properties that
 were measured  in the soil samples are
 listed on the next page.
 Quality Assurance Program
 Quality Assurance Optimization

   As  a  result  of conclusions and
 recommendations  from  the  DDRP
 Northeastern region and Southern Blue
 Ridge Province soil surveys, a number of
 improvements were made in the quality
 assurance  program  for  the  Mid-
 Appalachian  soil survey.  A  single
 preparation laboratory  was established in
 Las Vegas, Nevada, in close proximity to
 the  EPA  Environmental Monitoring
 Systems Laboratory -  Las Vegas (EMSL-
 LV)  quality  assurance (QA) staff. This
 allowed  the development  of,  and
 adherence to, strictly defined sample
 preparation  protocols  and the  ability to
 track and  control  progress  at the
 laboratory on a real-time basis.
   Mineral  and  organic  samples were
 placed in separate batches because  of
 differences in analyte concentrations and
 in the soiksolution ratios required for
 analysis.   As a result,  the  analytical
 laboratories were able to  perform
 instrument calibration  and sample
 analyses within narrower  linear dynamic
 ranges,  allowing  the QA staff  to  make
 more reliable  assessments  of the
 resulting data quality.
   Several changes in analytical methods
 and procedures were initiated in the Mid-
 Appalachian survey. For example,  the
 exchangeable aluminum in 1M potassium
 chloride  parameter was  replaced  by
 exchangeable  aluminum  in  1M
 ammonium chloride  as  part  of the
 measurement of exchangeable cations.
 This  change in  extraction  solution
 reduced  the  analysis  time for
 exchangeable aluminum while  retaining
 similar  experimental  conditions.  The
 quantity  of  mineral soil  used in the
 exchangeable  cation analyses  was
 increased relative to the volume  of
 extracting solution  in  order to increase
 the cation concentrations in  solution
 which facilitated  instrumental analysis.
 The amount of cellulose filter pulp used
 in extractions was decreased and  a pulp
 prewash step was added to  reduce
contamination by exchangeable  cations
in the pulp material. The determination of
extractable  cations  in 0.002M calcium
chloride using a mechanical extractor
was changed to a 3 overnight extraction
using a mechanical shaker. Also, pH was
measured on this extract rather than on a
 separate  extract as  used in previous
 surveys. Other changes included various
 solution  modifications  in the  measure-
 ment of total exchangeable acidity and
 the  use of  vanadium pentoxide  as  a
 combustion accelerator in the determin-
 ation of total carbon and  nitrogen.  Total
 carbon,  nitrogen,  and sulfur  were
 analyzed by a single analytical laboratory
 in response to the  Mid-Appalachian soil
 survey  quality  assurance requirements
 for tighter  measurement  quality and the
 specialized  nature of  the  analytical
 instrumentation required  for these
 analyses.
   The  statistical approach in the  Mid-
 Appalachian survey  made  use  of  a
 balanced hierarchical design that allowed
 various components  of  measurement
 uncertainty to be estimated with respect
 to the larger  population  uncertainty.  A
 step function technique was  developed
 by the QA staff to evaluate data quality  in
 terms  of  predefined objectives  and  in
 relation to the routine sample data.
   An  effort  was  made  to  use
 measurement quality samples in the  field,
 at the  preparation laboratory,  and at the
 analytical laboratories in such a way  as to
 provide  optimal  real-time  control   and
 evaluation of data quality.  Of particular
 importance was the addition of field  audit
 samples  at the sampling  phase to  allow
 the estimation of sampling  and system-
 wide measurement uncertainties.
   A Laboratory Entry and Verification In-
 formation System (LEVIS) was  developed
 for use  by the  QA staff  and by the
 analytical  laboratories.  The  LEVIS
 program  facilitated the entry,  edit,  and
 review of raw data and the calculation of
 final data values.  The  program  also
 performed  verification checks for the
 measurement  quality   samples  and
 produced QC summary reports.
   Many significant changes in the batch
 acceptance criteria were initiated. Among
 them  were the tightening  of  contract-
 required  detection limits  and   precision
 requirements. Several new measurement
 quality samples were introduced to check
 both precision and accuracy. Acceptable
 accuracy windows for  laboratory audit
 samples  were established and used as
 contractual  requirements for the labora-
tories. A template was also developed by
the QA staff to assist in setting  major and
 minor flags, and  helped to eliminate the
subjectivity present in the previous DDRP
surveys in regard to reanalysis  requests.

Data Quality Assessment
   The quality assurance  (QA) program
for  soil  sampling, sample  preparation,
and sample analysis were  designed to
 satisfy  measurement  quality  objectives
 (MQOs) for the resulting  data  and to
 assess  the  variability  of  sampling,
 preparation, and analytical  performance.
 The MQOs for this survey were primarily
 directed toward  the  attributes  of
 detectability, precision,  accuracy,  and
 completeness. Representativeness  and
 comparability of the  data  were also
 assessed, although quantitative   MQOs
 were not imposed.

 Detectability
   The  instrument  detection limit is the
 lowest value that an analytical instrument
 can reliably  detect above instrument
 background  concentrations.  During
 sample analysis,  the  overall  instrument
 detection  limits for a  given  parameter
 were defined as three  times  the  pooled
 standard  deviation   of  at   least  15
 nonconsecutive calibration blanks run on
 three separate  days.  Acceptable initial
 instrument  detection  limits  were
 established prior to  any sample analysis
 and subsequent values were determined
 and reported  on  a  batch-by-batch basis.
 Contracts with the analytical laboratories
 specified maximum  allowable  instrument
 detection  limits and, if a  reported batch
 detection limit was invalid, the batch was
 reanalyzed for that parameter.
   Secondary checks on the reliability of
 the  instrument detection  limits were
 made using independently determined
 values.  These independent instrument
 detection limits were calculated for each
 parameter as three times the standard
 deviation of a series of low level  quality
 control check samples.
   System  detection  limits   were
 estimated using  low concentration field
 audit samples  which reflected  the
 uncertainty  introduced during soil
 sampling, preparation, extraction, and
 analysis. These limits allowed  data users
 to identify soil samples which  had a
 measured concentration  that  was
 statistically different  from  the  reagent or
 calibration blanks. The  system detection
 limit for each parameter  was  calculated
 as three times the standard deviation of
 the low-range field audit samples.
 Precision

   Precision  is the  level  of  agreement
 among   multiple  measurements  of  the
same soil characteristic.  Measurement
 imprecision is  distinct  from the  overall
variability  in  the  population  itself.
 Determination  of   measurement im-
precision and its  sources in  the Mid-
Appalachian soil survey relied strongly on
analysis  of the  measurement quality
samples and was  a  function of the

-------
      Physical  and Chemical Properties Measured in the Direct/Delayed Response Project
                                          Mid-Appalachian Soil Survey

       Air-dry moisture content

       Total sand

       Very coarse sand

       Coarse sand

       Medium sand

       Fine sand

       Very fine sand

       Total silt

       Coarse silt

       Fine silt

       Total clay

       pH in deionized water

       pH in 0.002M calcium chloride

       pH in O.OM calcium chloride

       Ca in 1.01M ammonium chloride

       Mg in 1 .OM ammonium chloride

       K in  1.0M  ammonium chloride
Na in 1.0M ammonium chloride

Al in 1.0M ammonium chloride

Ca in I.OM ammonium acetate

Mg in 1 .OM ammonium acetate

K in  1 .OM ammonium acetate

Na in 1 .OM ammonium acetate

CEC in 1.0M ammonium chloride

CEC in 1 .OM ammonium acetate

Ex. Acidity by barium chloride-TEA

Ca in 0.002M calcium chloride

Mg in O.002M calcium chloride

K in 0.002M calcium chloride

Na in 0.002M calcium chloride

Fe in 0.002M calcium  chloride

Al in 0.002M calcium chloride

Ext.  Fe in pyrophosphate

Ext. Al in pyrophosphate
Ext. Fe in acid oxalate

Ext. Al in acid oxalate

Ext. Si in acid oxalate

Ext. Fe in citrate dithionite

Ext. Al in citrate dithionite

Ext. sulfate in deionized water

Ext. sulfate in sodium phosphate

Sulfate isotherm 0 mg sulfur

Sulfate isotherm 2 mg sulfur

Sulfate isotherm 4 mg sulfur

Sulfate isotherm 8 mg sulfur

Sulfate isotherm 16 mg sulfur

Sulfate isotherm 32 mg sulfur

Total carbon

Total nitrogen

Total sulfur
intralaboratory  within-batch  precision
MQOs  defined in the QA Plan.  Overall
variability stemming from  measurement
and population  sources  was  estimated
frorji the routine data.          |
    The precision  MQOs  were a  two-
tiered  system.   Below   a  specific
concentration, called  the  "knot",
precision was  defined  as  an  absolute
standard deviation in parameter reporting
units;  above the knot,  precision  was
defined as  a percent relative  standard
deviation.  To  address  the  issue  of
concentration-dependent variance,  the
range of soil analyte  concentrations was
partitioned  into appropriate  intervals
within  which the  error variance  was
relatively constant. A step function was
fitted across the intervals to represent the
error variance for the entire concentration
range.  Different step functions were  used
to  assess  variability  in  selected
measurement  quality  samples,  and
variability in the routine population of soil
samples collected was also estimated.

Accuracy
    Accuracy is the  level  of agreement
between an  observed value  and  the
"true"  value of a soil characteristic.  Data
from the laboratory audit samples  were
used to estimate analytical accuracy and
data from the field  audit samples were
used to assess accuracy from a system-
wide  measurement perspective. Each
audit sample type was assigned a range
of acceptable values for each parameter
in the form of an accuracy window, which
was  derived from  an  MQO-based
confidence interval  placed  around  a
weighted  estimate of the mean calculated
using analytical data from the  previous
DDRP surveys.
    The  three aspects of  accuracy
investigated  were bias,  laboratory
differences,  and  laboratory  trends.
Analytical bias was considered to be the
quantitative measure of accuracy used in
the  estimation   of   measurement
uncertainty. Laboratory differences were
assessed in relation to known reference
values and, in conjunction with laboratory
trends,  served  as  quantitative  and
qualitative evaluations  of   analytical
laboratory performance.

Representativeness
    The representativeness objectives of
the  survey  were qualitative  and
quantitative  in  nature. The  general
objectives  were  that:  (1) the  pedons
sampled  by the field sampling crews  be
representative of the soil sampling class
characteristics, (2) the, samples that were
     collected would  be homogenized  and
     subsampled properly by the  preparation
     laboratory  personnel,  and  (3) the  field
     duplicate samples adequately represent
     the  range  and frequency distribution  of
     analyte  concentrations found  in the
     routine samples.

     Comp/eteness
         The completeness objectives of the
     survey we're to ensure that  (1) all soil
     pedons  designated for  sampling  were
     actually sampled,  (2) all samples
     received by  the  preparation  laboratory
     were  processed, and  (3) all  samples
     received by  the  analytical laboratories
     were  analyzed and that 90  percent  or
     more of  the required measurements were
     made on all of the samples. Enough data
     were  provided  to  allow statistically
     significant conclusions to be drawn.  Data
     qualifiers, or flags, for completeness were
     inserted  in the data base to indicate any
     missing values.

     Comparability
         Comparability of data from the three
     DDRP surveys was  approached  as  a
     complex issue having  several levels  of
     detail  which should be considered. Level
     1 comparability was established on the
     basis  of statistical  evaluation methods,

-------
 measurement  quality  samples,  and the
 sample  collection,  preparation,  and
 analysis methods used. Level 2 compara-
 bility was established by the acceptability
 and useability of the verified data bases
 as defined by the data users.  Level 3
 comparability  allowed  the  direct
 quantitative comparison of data  for each
 parameter of interest.

 Uncertainty Estimates
     The  term  "uncertainty" was used to
 describe  the  sum  of all  quantifiable
 sources of error associated with a given
 portion  of the measurement  system.
 Uncertainty estimates, or  delta values,
 were calculated for each parameter using
 the square root of the sum of the within-
 and  between-batch  variances  and
 squared  bias term.   Four delta values
 were calculated for each parameter. The
 delta1,  values  represent  analytical
 laboratory  uncertainty  and  were
 estimated using the  laboratory  audit
 samples. The delta2 values represent the
 confounded  uncertainty  of  sample
 preparation  and  analysis,  and  were
 estimated  using  the  preparation
 duplicates. The delta3 values represent
 the confounded  overall  measurement
 uncertainty of field  sampling,  sample
 preparation, and  sample  analysis, and
 were estimated using the field duplicates.
 The della4 values represent  uncertainty
 due to the spatial heterogeneity of  the
 routine sample population confounded
 with the overall measurement uncertainty
 of  field sampling, sample preparation,
 and sample analysis, and were estimated
 using the sampling class/horizon groups
 of  routine samples.   The sampling
 class/horizon   groups  refer  to
 configurations of the samples that were
 believed  to have similar  physical  and
 chemical  properties  in  relation  to soil
 responses to acid deposition.

 Measurement Quality Samples
    Quality evaluation (QE)  samples
 were  used to assess  overall measure-
 ment uncertainty  and  to provide  an
 independent check on the  quality control
 (QC) procedures.  The QE samples were
 known to the  QA  staff  but were either
 blind  or  double-blind to  the sampling
 crews  and preparation  or  analytical
 laboratory personnel.  Six  types  of QE
 samples  were  used  in  the Mid-
 Appalachian  soil  survey:  (1) field
duplicates (soil samples were  collected
 by each sampling  crew from one  horizon
of  every  third pedon  sampled and  were
placed randomly in the sample batch with
the other samples from the same pedon);
(2) field  audits  (duplicate,  median-
 concentration,  mineral  soil samples or
 triplicate,  median-concentration, organic
 soil samples were sent by the  QA staff to
 the sampling crews for processing as if
 they were routine samples); (3) low-range
 field audits  (low concentration, mineral
 soil  samples were  sent  with the field
 audits to the sampling crews  by the QA
 staff); (4) preparation duplicates (a pair of
 preparation duplicates, one split from the
 field duplicate sample and  one split from
 its  associated routine  sample,  was
 created  at the preparation laboratory and
 placed randomly in the sample batch); (5)
 laboratory audits  (duplicate,  median
 range, mineral soil samples or triplicate,
 median-range,  organic soil  samples,
 identical to the field audits were sent by
 the  QA staff to the preparation laboratory
 for inclusion in each batch); and (6) low-
 range  laboratory  audits (low  con-
 centration, mineral soil samples identical
 to the low-range field audits were sent to
 the preparation laboratory by the QA staff
 for  inclusion with  each  mineral soil
 batch).
     The composition of the QC samples
 was  known  to  the analyst,  and  the
 analytical  results from each  laboratory
 were  required  to  satisfy the. batch
 acceptance criteria as the samples were
 analyzed.   Immediate  feedback on  the
 functioning of  the  analytical system
 allowed sample processing and analytical
 deficiencies to be  resolved  quickly,
 resulting  in  minimal error from these
 sources. Nine types of QC samples were
 used in the Mid-Appalachian soil survey:
 (i) QC  audit samples were  median-
 concentration mineral soil audit samples
 provided  directly  to  the analytical
 laboratories  together  with respective
 accuracy  windows, and  were  used to
 control  bias and  reduce  between-
 laboratory  and  between-batch  measure-
 ment uncertainty; (2) analytical duplicates
 were  splits of a single sample  and  were
 used  to  control  analytical  within-batch
 precision;  (3) calibration   blanks  were
 used  as  a check  for sample  contami-
 nation, analytical carryover effects,  and
 baseline  drift in the analytical instrument
 immediately after calibration; (4) reagent
 blanks underwent the same  treatment as
 the routine  samples and  served as  a
 check for reagent contamination; (5) QC
 check samples contained the analyte of
 interest in the mid-calibration range  and
 served as a check on the accuracy  and
 consistency of the  instrument calibration
 throughout the sample batch analysis; (6)
 detection limit GC  check samples  were
 low  concentration  samples   that
eliminated the necessity of determining
the detection limit every day and allowed
 accuracy to be estimated at the low end
 of the calibration range; (7) matrix spikes
 were  sample  aliquots to which known
 quantities  of  analyte were  added for
 determining the sample  matrix effect on
 analytical  measurements; (8)  adsorption
 spike  solutions were either the extraction
 solution  used  in determining  calcium
 chloride-tractable  calcium  or  the
 extraction  solution standards  used for
 generation, of the sulfur adsorption  iso-
 therms, and served  as a  check for
 reagent  contamination; and  (9)  ion
 chromatography resolution samples con-
 tained  known  concentrations of sulfate,
 phosphate, and nitrate, and were used to
 provide evidence  of  acceptable  peak
 separation for the sulfate analyses.
     Technical  system  audits or on-site
 evaluations were also conducted.  Each
 analytical laboratory underwent an audit
 after successfully  analyzing  a set of
 performance evaluation samples prior to
 receiving a contract. Second-round audits
 were performed after each laboratory had
 completed most of the analyses on two
 sample batches.  During third-round
 auditing, two laboratories underwent two
 additional  audits.  The third  laboratory
 underwent only one additional audit.

 Data  Verification
     The analytical laboratories  and the
 QA staff used LEGS for data verification.
 Phase  one of the  LEVIS  program"
 included the data entry component and
 two verification  components.  The
 analytical   laboratories   entered  and
 evaluated data as it was produced using
 a  QC  summary report  of  batch  data
 characteristics and sixteen soil chemistry
 relationships..  The data were  then
 transferred to the  central computer at
 EMSL-LV.  The  QA staff  evaluated
 preliminary  and formally  submitted
 batches from the analytical  laboratories
 using precision  and  accuracy  windows
 which  were checked  by  LEVIS.   The
 LEVIS  program  flagged data  which did
 not meet the MQOs. Following  the initial
 data evaluation  for  a batch, the  QA staff
 prepared a summary document of  all
 flagged parameters, indicating whether a
 flag  originated  from  an   unacceptable
 value for the QC, chemistry, precision, or
 accuracy  criteria.  The  number  and
 severity of  the flags for each parameter
were checked using the  QA  reanalysis
template to  determine if  reanalysis was
required.
    After completion and  receipt of  all
final batch  data from  the analytical
laboratories,  two internal consistency
checks  were performed  to  check for
possible outliers in  the routine  data. A

-------
correlation approach was used to assess
internal  consistency  in  which  the
coefficients  of  determination  were
obtained  by performing  weighted linear
regressions.   From the  regressions,
studentized residuals and difference of fit
statistics were calculated to  identify
extreme data  values that could be
considered  outliers.   Approximately  3
percent  of  the  data   underwent
confirmation  using this  consistency
check. The second internal consistency
check  used the data  structured into  a
pedon/horizon data set using the original
soil  profile  sequence.  Data for each
pedon were  visually  scanned  by  soil
scientists  for  consistency  of  the
parameter values from  one horizon to the
next. Approximately 2 percent of the data
underwent confirmation  using  this
consistency check.

Data Management
    The  field  sampling  and  sample
preparation data were  entered into  SAS-
AF TM  raw  data files  on personal
computers  at EMSL-LV.  The analytical
data were entered into  LEVIS on  personal
computers  located  at the  laboratories.
Data verification was accomplished by a
systematic  evaluation  of  completeness,
precision,  Internal consistency,  and
coding accuracy.  Apparent  discrep-
ancies were appended with flags unless
they could be corrected. After verification
was complete,  the data  bases  were
frozen  and  sent to  the  Oajk. Ridge
National Laboratory ' in Tennessee (to
undergo  validation  in  cooperation  with
personnel  at  the EPA Environmental
Research  Laboratory  at Corvallis,
Oregon, and at EMSL-LV. The validation
procedures included a specific assess-
ment of outlying data points for  inclusion
or omission in validated  data sets based
on assigned confidence levels.

Results and Discussion
Detecfa/b;7/ty

    The  calculated instrument detection
limits were  less  than  the contract-
required limits  for every  parameter.
Therefore,  the MQOs  for detectability  in
the  Mid-Appalachian Soil  Survey  were
completely satisfied.  System detection
limits, calculated from  the low-range field
audit  samples,  were  used to  assess
system-wide  detectability.    Using   a
criterion  of 80 percent  or  more  of the
routine sample concentrations exceeding
the system detection limits as a basis for
assessment,  most of the parameters
were suitable for all data uses throughout
the concentration  range.  The six
exceptions  were exchangeable  calcium
and  sodium in  ammonium  acetate and
ammonium chloride, iron extracted  in
calcium chloride or acid  oxalate, silicon
extracted in acid oxalate, and total sulfur.
Data  users should use  caution  when
assessing  these parameters,  as
significant portion of the  routine sample
concentrations  were  less  than the
corresponding  system  detection  limits
and  may  be difficult  to  distinguish  in
regard to overall system detection.

Precision
    The analytical  within-batch precision
objectives were satisfied for most of the
parameters. Occasionally, an  objective
was  not satisfied for an  upper or lower
tier of a parameter, such  as  magnesium
in calcium chloride or  silicon in  acid
oxalate. When  the  two tiers were pooled
over  the total  concentration  range for
each of the parameter groups, however, a
precision index showed that all parameter
groups satisfied the "overall"  precision
objectives.           ,
    The  precision  of  the  preparation
duplicates was  about the same, on the
average, as the laboratory audit samples
and satisfied the MQOs except for iron  in
acid  oxalate.  This indicates  that  the
preparation  laboratory  performed  very
well  in  subsampling  the  bulk soil
samples. For the field duplicates, only a
few cases, such as calcium in ammonium
acetate,   exceeded  the  precision
objectives).  In  addition, the relatively low
precision for the field duplicates for some
parameter groups suggests  that  the
component of  error from soil  sampling,
which includes spatial  heterogeneity
within the  sampled horizons, is a  large
portion of the data collection error.

Accuracy

    Accuracy in the Mid-Appalachian soil
survey was  evaluated  by  estimating
analytical bias with respect to a reference
value, defined as  the  mean of  an
accuracy window for a given parameter.
Laboratory differences  and trends  were
assessed  by comparing mean values for
the  laboratories, combined across  audit
samples, to the pooled reference values.
     Analytical  bias was negligible when
compared with the system detection limit
for  all  parameters for  which system
detection  limits werei established. The
only case of significant bias occurred for
the   cation  exchange  capacity   in
ammonium  acetate  for   the  two
laboratories which used the  titration
procedure.  Bias  for the  two newly-
established parameters  for  the  Mid-
Appalachian survey,  i.e.,  aluminum  in
ammonium  chloride and silicon in  acid
oxalate,  were both  much less than  their
respective  detection  limits.   The
percentage  of observations that  were
outside the  respective accuracy windows
and the magnitude of their contribution to
the overall  bias estimates  were  also
calculated. The results show a very  wide
range in the ratios of  bias  for values
outside the  window  compared to the total
analytical  bias. Only  11  of the 50
parameters  had  values  outside the
window that contributed  more than  one-
third to  the bias: coarse and  fine silt,
magnesium  and aluminum  in calcium
chloride,  magnesium  and  cation
exchange capacity in ammonium acetate,
aluminum in sodium pyrophosphate and
citrate dithionite,  iron  and silicon in acid
oxalate, and the  32 mg  S/L sulfate
isotherm parameters.
    Laboratory   differences   were
expressed  as a percentage  from  the
reference value for each laboratory for
each parameter. When compared across
laboratories,  all  laboratory  differences
were ten percent or less except for clay,
potassium  and sodium in  ammonium
acetate, and iron and aluminum  in
calcium  chloride. In all  five cases, the
concentrations  were very low.  When
combined  into parameter groups, all
laboratory differences were six percent or
less except for  cations  in  ammonium
acetate  for Laboratory 1 and cations in
ammonium chloride for Laboratory 4.
    Approximately 54 percent of  the
parameters  (25  of  46  parameters)
showed  significant  laboratory differences
by Scheffe's pairwise comparison  test.
Four of the 25 parameters (total silt, pH in
0.002M calcium  chloride, cation
exchange capacity in ammonium acetate,
and silicon  in acid oxalate) showed  all
three laboratories to be significantly
different from each other. There were no
general  trends for any specific  laboratory
for most of the parameter groups. For the
cations  in  calcium chloride,  however,
Laboratory  2 was  significantly different
from the other laboratories in  all  cases
shown and Laboratory 4  was significantly
lower for the sulfate isotherm parameters
than the other laboratories  in the four
cases that showed differences.
    Moving averages of the laboratory
audit samples were  plotted to identify
situations  when  a  particular laboratory
showed an upward or downward  trend
over time  for a given parameter.
Generally,  the  trends  did  not   show
extreme divergence with respect  to the
accuracy  window  acceptance criteria.
However, certain data users may find the

-------
 trends to  be noteworthy  for a specific
 data analysis.
     No single laboratory was consistently
 superior to the others for all  parameters
 or  parameter groups regarding  low
 differences.  Each  laboratory  appears  to
 have  individual strengths for specific
 analytical  methods. This  is probably a
 reflection  of  the  combination   of
 experience,  instrumentation,  and  labor-
 atory  management practices within each
 laboratory, resulting  in  a patchwork  of
 differences on a parameter group basis.
 Representativeness
     All pedons sampled were within the
 range of morphological  characteristics
 outlined in  their  respective  sampling
 classes.  The  homogenization and
 subsampling  procedures  at  the
 preparation laboratory  produced
 representative analytical soil samples  of
 known and  accepted  quality. Overall
 precision for the preparation duplicates
 was approximately equivalent to that  of
 the  laboratory audit samples,  hence, the
 procedures were shown to  be suitable for
 creating representative subsamples.
     The analyte concentrations in  the
 field  duplicates  generally  were
 representative  of the  range  and
 frequency  distribution of routine sample
 concentrations.  The exceptions  were
 seen  mainly in  the  cations  and
 oxtractable sulfate  parameters and were
 mostly   representative   of   the
 concentration range,  but not of  the
 distribution within the > tange.  Analyte
 concentrations in  the  preparation
 duplicates  were generally  representative
 of the corresponding field duplicates. The
 audit samples, as expected, were usually
 representative only  of the overall range of
 data from the routine samples.

 Comp/eteness
    Sampling of the specified  pedons
 had a completeness level of 100 percent,
 and  processing was accomplished  for
 100  percent of  the  field  samples
 satisfying the sample receipt  criteria  at
 Iho  preparation  laboratory  Analytical
 completeness in the verified data base
 was  100 percent  for all  parameters.
 Sufficient validated  data were  generated
 to make conclusions for each  parameter
 in the Mid-Appalachian  survey  data
 bases, with a completeness level  of 98
 percent or higher for all parameters.
 Compara6///*y
   The  statistical methods   were
comparable among the  three DDRP
surveys.  Additional  audit samples  were
used to optimize QE and QC activities in
 the Mid-Appalachian  survey although
 they  did  not  directly  affect  the
 comparability of  the data  from  the
 surveys.
     In most  cases,  the sampling,
 preparation and analytical  methods and
 protocols  for  the  three  surveys were
 comparable, although there were several
 parameter protocol modifications in the
 Mid-Appalachian  survey.   The only
 modification  which  may  affect
 comparability was  the  change  of
 aluminum extractant from  1M potassium
 chloride to 1M ammonium chloride. As a
 result  of more stringent MQOs in the Mid-
 Appalachian survey, overall measurement
 quality generally  was better in the Mid-
 Appalachian  data bases. Therefore, it  is
 possible that  some of the Northeastern
 region and Southern Blue Ridge Province
 data may not be suitable for the same
 data analysis procedures  performed  on
 the  Mid- Appalachian data without some
 form of caveat. Two examples  are the
 organic soil sulfate isotherm data  and the
 specific  surface  data  from the
 Northeastern and Southern  Blue Ridge
 Province surveys  which are  generally
 considered to be suspect by data  users.
     Significant differences  among  the
 surveys were identified using the median-
 concentration B mineral soil audit sample
 common to the three surveys. Only nine
 parameters  showed differences  among
 surveys at the 0.01 significance level:
 aluminum  extracted  with  potassium
 chloride versus  ammonium chloride,
 water-jextractable sulfate, the 0,2, and  4
 mg  S/l sulfate  isotherms,  calcium  in
 calcium chloride, iron in citrate dithionite,
 total carbon, and coarse silt. Data users
 should exercise caution when using data
 from the  nine parameters exhibiting
 significant differences.
     As  part  of  the  DDRP,  an
 interlaboratory  methods  comparison
 study  was  conducted which compared
 soil  analysis  data from  two  laboratories
 using  the  DDRP  methods  to  16
 statistically  randomly-chosen  external
 laboratories.  These  laboratories used
 their own  methods  which  were  similar,
 but not identical to,  the DDRP methods.
 The  results  of the  study show the
 generally  good comparability  of the
 DDRP  data  with  data  from  other
 independent laboratories.

 Uncertainty Estimates

    Within-batch imprecision estimates
 increased,  as  expected, from analytical
 (deltas values) to sample  preparation
 (delta2  values) to  field  sampling  (delta
values) to sampling class/horizon  groupl
 (delta4  values).  The  between-batch
 precision estimates were generally low.
     The overall measurement uncertainty
 in the routine samples was based on the
 deltas values which were estimated  from
 the  field  duplicate samples.  The delta3
 values,  in  relation  to  the associated
 delta4, values, were  used  to provide the
 data users with a basis for assessing the
 contribution of the measurement system
 to  the  data uncertainty.  Using  this
 procedure, measurement uncertainty was
 negligible for 90 percent  (45  of the 50
 parameters  measured) of  the  data.  The
 five exceptions  were the  fine silt  and
 coarse silt fractions  which had no strict
 precision MQOs, sodium in both calcium
 chloride  and ammonium  acetate,  and
 silicon in adid oxalate. The sodium  and
 silicon parameters showed high relative
 measurement uncertainty due to  the
 inordinate effect  of measurement error in
 the  low  concentration  range  near  the
 detection lirrjit.

 Conclusions and
 Recommendations

     As a result  of previous conclusions
 and  recommendations from the  DDRP
 Northeastern and  Southern Blue  Ridge
 Province   surveys,  a  number  of
 improvements were made in the quality
 assurance   program  for the Mid-
 Appalachian  survey.   The  quality
 assurance  data are presented in  a
 manner  considered  to be  the most
 appropriate for use by the primary data
 users. The  data quality  evaluation
 procedures and the report format resulted
 from  regular interactions with  the data
 users, and the assessment by several
 external  reviewers.  Each  user has  a
 subjective concept  of data quality as  well
 as a knowledge  of the specific level of
 data quality required for his/her own use.
 The  users are therefore encouraged to
 become  familiar  in detail  with the text,
 figures,  and  tables  to facilitate  the
 identification  of  data  satisfying their
 specific requirements.
    The   consolidation  of  sample
 preparation facilities at a single laboratory
 facilitated  quality  assurance of  the
 samples from the field sampling through
 the  sample  analysis  phase.   The
 separation  into  different  batches  of
 mineral and  organic samples should be
continued.  In addition, measurement
quality samples should  continue to  be
distributed among batches  and  analytical
laboratories in such  a way as to  provide a
balanced  design  for  assessment
purposes.

-------
    The  use of a  computerized  data
entry  and  verification  system which
allowed the  calculation  of final  data
values  and  produced a list of flags and
data entry errors for each  sample batch
greatly facilitated  the  verification  and
reanalysis decision-making process.  A
similar  program  should  be tailored for
each future survey, with the addition of a
method that  will identify  and  confirm
outlying data points for real-time control
purposes.
    Considerable effort  was  expended
throughout the three surveys to  improve
detectability  of  various parameters  and
significant improvement was made for the
exchangeable  cations   and  sulfur
parameters. Additional methods  research
is needed to improve detectability further.
In addition, both  instrument and  system-
wide detectability should be  defined  in
the  MQOs.  As part of the implementation
of such MQOs,  it is recommended that
low  concentration audit samples, entered
into the  system, during  the sampling
phase,  continue  to be  utilized  as
substitutes for soil blank samples.
    The  precision  results  indicate  that
the  analytical  within-batch  precision
objectives were satisfied  in most cases.
Occasionally an  objective was  not
satisfied for  an  upper or lower  tier  of a
parameter.  It is  recommended  that the
lower and upper tier precision objectives
for  exchangeable  aluminum  in   am-
monium chloride and  extractable silicon
in acid oxalate be modified as described
in the report.  The preparation and  field
duplicate samples generally satisfied the
precision objectives. A precision  index  of
parameter groups revealed  that  all
groups were of acceptable precision. ,
    Audit sample  accuracy windows
were  developed  for use  in  the  Mid-
Appalachian  survey from data collected
in the previous two surveys. The  use of a
quality  control audit sample  should be
continued in future surveys. In  addition
liquid  audit  samples  should  be
incorporated  into the quality assurance
program and  be used to differentiate soil
extraction  error  from instrument  error.
Emphasis  should be placed on the use of
audit  sample control  charts by  the QA
staff to identify abnormal scatter outside
the accuracy windows during the  batch
analysis.
    No single laboratory was consistently
superior to the  others for  all parameters
or parameter groups. To control  inter-
laboratory differences in future surveys, it
is  recommended  to  continue the
selection  of  a  specific  laboratory  to
perform analysis on a parameter basis for
those  parameters  or parameter groups
that reveal inherently high differences  of
where specialized  instrumentation  is
used,  e.g., total elemental  analysis. Also,
a  stringent  performance  evaluation
process should  be continued to select the
best available contract analytical labora-
tories.
    Quality is a continuum and  the need
for improved data quality dictates the
data  quality  objectives  chosen.  These
objectives may or  may not be attainable
with the current technology. It is recom-
mended that the analytical  procedures for
specific parameters  be  revised, tested,
modified, and clarified where appropriate.
    In evaluating representativeness  of
the quality assurance  samples,  it  is
evident that the field  duplicates  and pre-
paration duplicates were representative of
the routine sample  concentration  range
for most parameters.   It is recommended
that the field  sampling  and preparation
laboratory protocols  c'ontinue to specify
statistically valid methods  for selecting
the field and  preparation  duplicates. The
method should be   reiterated to the
sampling and laboratory personnel during
the pre-sampling training  session. The
QA field  auditor should  ensure that a
sufficient amount of   soil is collected for
each  bulk  sample  in  the  field to allow a
preparation duplicate to be subsampled.
    Data comparability across the three
surveys  was  generally  good.  It  is
recommended,  however,  that a methods
comparison be performed for the two soil
extraction   methods  used  for
exchangeable  aluminum, i.e.,  1M  am-
monium  chloride  and  1M  potassium
chloride.
    The step function statistical approach
has  been  shown  to be an effective
procedure for evaluating  measurement
quality  issues  in  environmental  data
spanning a wide concentration range. It is
recommended that additional research
and  development be  undertaken  to
identify  an  optimal  step  function
procedure that  is  fully  compatible  with
the measurement quality sample design.

-------
   G. £ Byers, R. D. Van Remortel, M. J. Miah, J. E. Teberg, M. L Papp, and B. A.
         Schumacher are  affiliated with the Lockheed Engineering & Sciences
         Company, Las  Vegas, Nevada. B. L.  Conkling is affiliated with  the
         Environmental Research Center of the  University of Nevada at  Las Vegas.
         D.L.  Cassell and P.W.  Shaffer are affiliated with the NSI Technology
         Services Corporation,  Corvallis, Oregon.
   L J. Blume and D. T. Heggem  are  the EPA  Project Officers (see below).
   The complete report,  entitled "Direct/Delayed Response Project:  Quality
         Assurance Report for Physical and Chemical Analyses of Soils from the
         Mid-Appalachian Region of the United  States," (Order No. PB  204710/AS;
         Cost: $39.00, subfect to change) will be available only from:
             National Technical Information Service
             5285 Port Royal Road
             Springfield, VA 22161
             Telephone: 703-487-4650                  •
   The EPA Project Officers can be contacted at:
             Environmental Monitoring Systems  Laboratory-Las Vegas
             U.S. Environmental Protection Agency
             Las Vegas, NV 89193-3478
*U.S. CcverwKnt frInline Offices 1990-748-012/20066
 United States
 Environmental Protection
 Agency
Center for Environmental Research
Information
Cincinnati OH 45268
 Official Business
 Penalty for Private Use S300
 EPA/600/S4-90,t)01

-------