September 7, 2017

POINT OF CONTACT

Principle investigator:
     Richard Eckman
     NOAA Air Resources Laboratory Field Research Division
     1750 Foote Dr.
     Idaho Falls, ID 83402
     Richard.Eckman@noaa.gov
     (208) 526-2740

README.TXT - Description of SF6 Tracer Bag Sampling and Data Files

Description of Equipment

Stationary time-integrating sampling of SF6 for PSB2 was performed using programmable 
bag samplers.  These samplers acquired time-sequenced air samples in bags that were 
subsequently analyzed for the concentration of the SF6 tracer.  The samplers collected 
12 samples by sequentially pumping air into each of 12 individual Tedlar bags.  The 
integrated sampling time for each bag in the study was 10 minutes resulting in 12 
individual experiments within each of the eight 2-h Intensive Observational Period (IOP) 
periods. The bag sampler housing is constructed from durable double-wall polypropylene 
manufactured by Mills Industries Inc. and measures 61 cm x 41 cm x 33 cm.  The other 
component of the bag sampler assembly is a cardboard sampler cartridge.  All sampler 
boxes contain 12 microprocessor-controlled air pumps designed to start sequentially 
filling the bags at a time and duration specified for each bag.  The sampling period 
for each bag and the delay before each bag can be independently specified to create a 
sampling program customized for each situation.  The cartridge box contains 12 Tedlar 
bags. Prior to deployment, a sample cartridge was placed into each sampler box and 
connected by R-3603 tubing to the sampler pumps. With its cover in place, each sampler 
box and sampler cartridge assembly had a total mass of approximately 4 kg and was 
powered by a single D-cell battery. The microprocessor and air pump components of the 
sampler design have been used successfully in field experiments for many years and are 
known to be free of artifacts (e.g. Clawson et al. 2004, 2005).  The material used for 
the bag sampler housing represents a recently improved design that was extensively 
tested for reliability and potential sampling artifacts in 2007 and also found to be 
free of artifacts.

Description of Bag Sampling Grid

The bag sampling measurements were the most essential feature of the experiment. Nominally, 
150 (or 151) samplers were deployed for each IOP with one held in reserve for possible 
contingencies. They were mounted atop plastic boxes at 1 m agl and stabilized from 
toppling in the wind by hooking the carrying handle over the metal post marking the 
sampling location. For the daytime IOPs, 36 samplers were placed along each of the 100, 
200, and 400 m arcs. They were placed at 6 deg intervals from 276 deg azimuth to 126 
deg azimuth (i.e., 276, 282,.,120, 126 deg). An additional 16 samplers were deployed 
on the 800 m arc at 6 deg intervals from 0 deg azimuth to 90 deg azimuth (i.e., 0, 6,
.,84, 90 deg). For the nighttime IOPs, 36 samplers were placed along each of the 100, 
200, and 400 m arcs at 6o intervals from 312 deg azimuth to 162 deg azimuth (i.e., 312, 
306,.,156, 162 deg).

For the daytime IOPs, a mobile tower (MOB) was available for vertical sampling. Bag 
samplers were deployed on this tower at 1, 5, 10, 15, 20, and 25 m heights at locations 
on the 100 m arc. Due to the anticipated rapidity of plume rise in the unstable daytime 
conditions, it was assumed it would be unlikely to be able to measure the full plume in 
the vertical at lesser heights or greater downwind distances. For that reason, the only 
daytime vertical sampling was conducted by MOB on the 100 m arc.

For the nighttime IOPs, MOB was moved to site 3 on the 400 m arc with sampling at the 
same heights. Additionally, bag samplers were deployed on four fixed 10 m towers (FIX) 
on the 100 and 200 m arcs at 1, 3, 6, and 9 m heights.

The bag sampler locations were designated with a 4-digit code with the 2-character 
prefix LC. The first digit represents the arc or tower location (1 = 100 m, 2 = 200 m, 
4 = 400 m, 8 = 800 m, 5 = a tower sampler). For the arc samplers digits 2-4 specified 
the azimuth. For example, Location LC2006 would be an arc sampler located on the 200 m arc 
at 6 deg azimuth. Quality control (QC) was integral to the experimental plan and called 
for the use of blank, control, and duplicate samplers. A 5 as the second digit was used 
to designate a duplicate sampler. For example, the duplicate collocated with LC2006 was 
LC2506. There were 4 duplicate samplers on each of the 100, 200, and 400 m arcs and 2 
duplicate samplers on the 800 m arc for a total of 12 per IOP (14 daytime).

Field blank and field control samples were designated with 9 in the first digit with 
digits 2-4 designating azimuth. The field blank and control sample designations did not 
indicate arc but these were documented ahead of time and included in the electronic 
record. For example, LC9042 was the only blank or control sampler located on any of the 
arcs at 42 deg azimuth and was specified to be a field control. For field blank and control 
samples the second digit specified the arc location. There was 1 field blank and field 
control, each, on the 100, 200, and 400 m arcs for all IOPs.

For tower samplers, the second digit identified the tower and digits 3 and 4 specified 
the sampling height (agl). The second digit designations are:
	0  mobile (MOB) tower
	1  fixed (FIX) 10 m tower at 17.5 deg on the 100 m arc
	5  fixed 10 m tower at 53.5 deg on the 100 m arc
	8  fixed 10 m tower at 89.5 deg on the 100 m arc
	9  fixed 10 m tower 53.5 deg on the 200 m arc
For the mobile tower, digits 3 and 4 01, 02, 03, 04, 05, and 06 represent 
the 1, 5, 10, 15, 20, and 25 m heights, respectively. For the fixed towers, digits 3 and 
4 01, 02, 03, and 04 represent the 1, 3, 6, and 9 m heights. For example, LC5502 
would designate the 3 m agl sampler on the fixed 10 m tower at 53.5 deg on the 100 m arc.

Quality control samplers were also deployed. This included 14 field collocated duplicates, 
3 field controls, and 3 field blanks for the daytime IOPs 1-4. The same was done for the 
nighttime IOPs 5-8 but only 12 field duplicates were used since the 800 m arc was not used. 
The arc angle positions of these QC samplers are listed in the table below. 

Arc and arc angle location of field duplicate, field control, and field blank samplers.
	Arc Angle Position (degrees)
Arc (m)	Duplicate		Control		Blank
100	30, 36, 54, 78		42		84
200	6, 12, 24, 66		78		30
400	24, 36, 48, 66		60		12
800	18, 78		

Sampler Cartridge Analysis

Sample cartridges were analyzed at the Tracer Analysis Facility (TAF) in Idaho Falls, ID.  
The TAF hosts four gas chromatographs (GC), each housed within its own autosampler module 
and connected to a computer with the master data acquisition system.  The complete 
configuration with GC, autosampler, and data acquisition system is called an Automated 
Tracer Gas Analysis System (ATGAS). A dedicated small black handheld computer was used 
to set the operational parameters on each ATGAS.

Each GC housed two Supelco 60/80 Molecular Sieve-5A columns (5' x 1/4" and 2' x 1/4"), a 
10-port sample valve, and a sample loop.  These columns were maintained at 65 C inside 
their respective ovens.  Two columns (pre-column and main column) were used to reduce 
analysis time and to vent interfering species, mainly oxygen, which can damage the columns 
and detector.  After the SF6 sample was injected onto and eluted by the first 2-foot 
(610 mm) pre-column, the gas flow was switched to back-flush the pre-column while the 
sample loop was filled with the next sample.  The SF6 continued on to the main 5-foot 
(1520 mm) column where further separation occurred before being passed to the detector. 
Detection of SF6 was accomplished using a Valco Instrument Co., Inc., Model 140BN electron 
capture detector (ECD) containing 5 millicuries of Ni-63.  The ECD operating temperature 
was kept at 170 C.  The ECDs and columns were protected by a Supelco High Capacity Gas 
Purifier tube heated inside an oven to remove oxygen, water, carbon monoxide and carbon 
dioxide in the carrier gas as well as a Supelcarb HC hydrocarbon trap to remove organic 
impurities.  Ultra high purity (UHP) nitrogen served as the carrier gas and filtered 
compressed air was used as the valve actuator gas. Concentration ranges from 2 pptv to 
about 1 ppmv have been analyzed using this methodology.

All fittings associated with the GCs were leak checked prior to the start of the project. 
Any leaks identified were corrected. GC2 exhibited drift and erratic response problems 
when analyzing samples for IOPs 1 and 2. The detector for GC2 was replaced with a rebuilt 
detector on August 2nd for the analysis of subsequent IOPs and the performance was then 
significantly more stable. The Valco 140BN detector on GC4 was replaced by a SRI 110 
detector between the analyses for IOPs 4 and 5. The design of the SRI 110 detector is 
the same as the Valco 140BN but the controller box is included as a module with the 
detector rather than being a separate unit. The SRI 110 detector generally performed 
well although it was somewhat less sensitive to low concentrations as will be described 
later.

The ATGAS computer software (Carter, 2003) was developed in-house and was used to analyze 
the tracer gas chromatograms, calculate concentrations, and perform quality control 
functions.  The software incorporates a history file system that records all operations 
performed on each ATGAS. 

Sampler Handling and Chain of Custody

A history file in the master ATGAS computer maintained a complete, comprehensive record 
for each sampler cartridge.  The scheme for maintaining the comprehensive history file was 
based upon unique bar coded serial numbers attached to both samplers and sample cartridges 
and the use of bar code scanners.  In addition, prior to the start of the project, each 
field sampling location was identified and tagged with a location number that consisted of 
a weatherproof bar code label.  These were affixed to the metal fence posts installed at 
each sampling location.  A file with a list of the locations was uploaded to the ATGAS 
computer in the TAF.  The bar code labels for the samplers, cartridges, and locations were 
used to automatically generate a chain of custody record for each sample. 

In preparation for each test, a sample cartridge was placed inside each sampler and then 
transported to the field.  Samplers were deployed at each location, the tubing was 
connected, clips were opened, and a sampling program downloaded into the memory of each 
samplers microprocessor.  The latter was accomplished with the use of a small hand-held 
computer (Videx Timewand II).  The Timewands were programmed with sample start and stop 
times for each bag prior to each test using a dedicated laptop computer in the TAF.  They 
were then used in the field to download the sampling program and acquire and record the 
location number, sampler number, and cartridge number.  The complete field download records 
were later retrieved from the Timewands and transferred into the history file on the ATGAS 
computer in the TAF prior to the start of cartridge analysis.

Personnel responsible for deploying the samplers in the field received classroom and 
hands-on training in Idaho Falls prior to the experiment.  It was also required that 
handwritten Sampler Servicing Record sheets be completed in the field for each removed or 
installed cartridge.  These records were created to provide the TAF analyst with details 
of potential problems pertaining to each cartridge and sample bag.  In combination with 
the history files, these records were invaluable as a reference for sample check-in and 
later for QC flagging of data.  The Sampler Servicing Records were given to the laboratory 
analyst after sampler collection and delivery were performed.  All record sheets were 
organized and placed in a binder for future reference.

The sample cartridges were transported back to the TAF at the completion of each IOP and 
analyzed within a few days of sampling.  They were all checked in prior to analysis using 
a bar code scan.  During this process each bag was inspected and the following flags were 
entered into the computer for each bag:
 	B   =  Too big (overfilled)
 	G   =  Good
	L   =  Low
	F   =  Flat
	D   =  Damaged clip or bag
	I   =  Improper hookup (tubes crossed, clip open, etc.)
These flags were used later for querying, sorting and generating final QC flags as well as 
for monitoring sampler performance and checking for mistakes by field personnel. Each 
cartridge was again scanned when it was attached to the ATGAS prior to analysis. This 
linked the GC identity and the acquired chromatogram and calculated concentration data to 
the computerized data previously collected in the field that specified the project 
identification, test number, grid location number, grid location coordinates, sampling 
start time, the sample time per bag, and sampling type (primary or quality control sample).
The record also included the cartridge check-in record and cleaning records.  Thus a 
complete computer-generated chain of custody is available for each bag sample as well as 
automatically linking all field, chromatogram, concentration, and quality control data 
into one comprehensive data record that could be readily reviewed.  This minimized the 
possibility of errors caused by mistakes in manually recording, copying, or entering of 
location information and provided an invaluable source of information in the event of a 
discrepancy or a question about the data.

Potential Sampling Artifacts

Latex tubing was used in the past but it was found to be very susceptible to degradation, 
cracking, and having tubes pinched off even with the clips open. R-3603 tubing has been 
found to be largely immune to that type of problem. However, with continued usage, it was 
found that the R-3603 tubing has the potential to be associated with a different set of 
sampling artifacts. Foremost among these is how the tubes sometimes fail to completely 
seal when the clip is closed. This was relatively infrequent but it did call for attention 
to making sure clips were fully closed to ensure the tube was pinched closed. If the clips 
were open slightly, sample bags could be corrupted during the line purge cycle prior to 
the start of the analysis on the gas chromatograph. It is believed that this was almost 
always identified and the data flagged accordingly but the data user should be aware that 
there could be instances where this is not the case. There could be some instances where 
a bag sample was slightly diluted during the purge cycle resulting in an indeterminate 
lowering of the concentration. The available lines of evidence involving the replication 
of sampling and analyses indicate that, while some samples might have been affected, the 
overall picture provided by the concentrations and their areal distribution should be an 
accurate representation. 

Another artifact was identified by the often large differences measured between collocated 
field duplicates during the nighttime IOPs. Previous experience with duplicate sampling 
demonstrated consistently good agreement between the concentrations measured in the 
collocated bag samplers. However, that experience was mainly during daytime tests. 
Temperatures were generally near or below freezing during the PSB2 nighttime tests. 
It is conjectured that the artifact was possibly due to cold temperature related leaks 
in the air pumps and/or tubing fittings. Part 16 in the Quality Control Procedures section 
below describes results of testing regarding this artifact.

Quality Control Procedures and Measurement Quality Objectives

The following are detailed descriptions of the quality control and quality assurance 
methods followed for the sampling, analysis, and reporting of the PSB1 time-integrated 
bag sampler tracer data.  Protocols established in the Environmental Protection Agencys 
(EPA) Guidance for Data Quality Assessment (U.S. EPA 2000a), the general requirements for 
the competence of calibration and testing laboratories of International Standards 
Organization/IEC Guide 25 (ISO 1990), the quality systems established by the National 
Environmental Laboratory Accreditation Conference (U.S. EPA 2000b), and the Department of 
Defense Quality Systems Manual for Environmental Laboratories (U.S. DOD 2002) provided a 
basis for quality assurance and quality control procedures followed during analysis.  
Instrument and method limits of detection (ILOD/MLOD) were calculated based upon 40 CFR 
Part 136, Appendix B and the American Chemical Society (ACS) Committee on Environmental 
Improvements paper titled, Principles of Environmental Analysis (Keith et al. 1983).  
ACS principles relative to detection limit calculations in 40 CFR Part 136, Appendix B are 
documented in Revised Assessment of Detection and Quantitation Approaches (U.S. EPA 2004).
Although our research-based automated analysis of tracer gases has no specified method 
performance or regulatory criteria, compliance with the established quality control 
procedures stated above were followed, where applicable, to provide high quality data that 
is both accurate and reliable.

The laboratory procedures followed were designed to ensure meeting the stated Measurement 
Quality Objectives (MQO) for the project are listed below.

Measurement quality objectives (MQO) for the bag sampling Data Quality Indicators

Data Quality Indicator		Objectives (MQO)	How Determined

Instrument sensitivity		Instrument Limit of	Lab blanks and low concentration
				Detection (ILOD)<4 pptv	calibration checks

Between instrument precision	RSD(1) < 10%		Lab background checks

Low end instrument bias		< 1 pptv		Lab blanks

Instrument precision		|RPD(2)| < 5%		Lab duplicates above MLOQ
				RSD < 10%		Lab controls above MLOQ

Instrument accuracy		|RPD(3)| < 20%		Required by calibration check
							and recalibration protocol

Low end method bias(4)		< MLOQ(5)		Field blanks

Method sensitivity		Method limit of 	Calculated from field blanks,
				detection (MLOD)	low concentration field controls,
				< 12 pptv		field duplicates, or background
							samples

Method precision		|RPD(2)| < 15%		Field duplicates above MLOQ
				RSD < 15%		Field controls

Transport&storage		< MLOQ			Transport blanks
contamination

Completeness %			90%			Percentage of samples producing
							good measurements

(1) RSD is relative standard deviation: standard_deviation/average
(2) RPD is relative percent difference: for duplicates is 
    (measure_1  measure_2)/average_of_1&2
(3) RPD is relative percent difference: for known concentrations is 
    (measure  actual)/actual
(4) Method is entire sampling method including sampling and analysis.
(5) Method Limit of Quantitation

Quality control issues pertaining to procedures for sample handling in the field and 
chain of custody were described in the previous section.  Pre-project and laboratory QC 
procedures are described below and consisted of the following 21 steps:
1.	Pre-project maintenance of bag samplers.
2.	Testing of all sample bags.
3.	Pre-project cleaning and analysis checks of all sample bags.
4.	Development of analysis protocols for the expected sample concentration ranges.
5.	Use of a written standard operating procedure (SOP).
6.	Pre-project calculation of instrument limit of detection (ILOD) and instrument 
        limit of quantitation (ILOQ).
7.	Holding time studies.
8.	Daily calibration of the ATGAS.
9.	Initial ATGAS Calibration Verification (ICV).
10.	Continuing ATGAS Calibration Verification (CCV) and analysis of laboratory controls.
11.	Atmospheric background checks of SF6 at the tracer analysis facility (TAF).
12.	Analysis of laboratory (instrument) blanks.
13.	Analysis of laboratory duplicates.
14.	Analysis of field blanks.
15.	Analysis of field controls.	
16.	Analysis of field duplicates.
17.	Software quality control checks.
18.	Data verification.
19.	Post-project determination of MLOD and MLOQ.
20.	Final data review.
21.	Data handling.
22. 	Summary of Data Completeness

1.   Pre-project maintenance of the bag samplers.

Prior to deployment to the field, each of the original 135 bag samplers was extensively 
tested to ensure proper operation in the field and to ensure the collection of an adequate 
sample volume.  This mainly involved checking the function of the microprocessor and pumps. Fifteen new bag samplers were built for PSB1 using Texas Instruments MPS-430 series microprocessors and similarly tested.

2.   Testing of all sample bags.

Experience has shown that almost all leaks in sample bags occurred around the fitting used 
for attachment to the sample tubing. To rectify this problem prior to PSB1, the seam 
between the fitting and the bag was permanently sealed in all sample bags using Pliobond 
30. All bags were also inspected and if there were any holes or suspected holes besides 
the fitting seam they were discarded prior to gluing. Previously bags had been checked for 
leaks using the procedure detailed in Clawson et al. (2008) but the bag sealing resulted 
in a lower failure rate than had been achieved by the leak checking procedure of the past.

3.   Pre-project cleaning and analysis checks of all sample bags.

After the bags were leak checked but prior to deployment to the field, all bags in the 
sampler cartridges were cleaned.  The bags were cleaned by repeatedly filling them with 
UHP nitrogen and then evacuating them on the cartridge cleaning apparatus. The apparatus 
consisted of a nitrogen tank and vacuum connected to a system that fills and evacuates the 
sample bags by changing valves.  Seventy-two bags in 6 cartridges were cleaned at one time. 
The computer mounted underneath the cleaning apparatus was used to create cartridge 
cleaning records.  This information was then uploaded into the ATGAS history file.  
An 8-step cleaning protocol was used to clean the bags:
	1.	Connect all tubes to the cleaning machine.
	2.	Open all clips.
	3.	Make sure the cleaning machine valves are set so that nitrogen can flow 
                into all connected cartridges.
	4.	Evacuate bags.
	5.	Fill all bags with nitrogen and then evacuate.  Repeat until all bags have 
                been evacuated 5 times.
	6.	Fill all bags with nitrogen for analysis.
	7.	Scan all cartridge bar codes with the bar code scanner and upload the data
		to the ATGAS PC.
	8.	After analysis, place the cartridges back on the cleaning machine, 
                evacuate the nitrogen, disconnect the tubes and wait 30 seconds before 
                closing clips.

This protocol was developed after significant testing to ensure that bags containing 
concentrations in the expected high range of up to 150,000 pptv or more could be cleaned 
to less than background levels.  After cleaning, the bags were filled with UHP nitrogen 
and analyzed to ensure there was no contamination from previous tests or from long-term 
storage.  Any bags with a concentration greater than 5 pptv were re-cleaned and re-analyzed.
All but 23 out of 6744 bags (562 cartridges) were successfully cleaned below 5 pptv in the 
initial cleaning and none were greater than 10 pptv.  The vast majority were below the 
instrument limit of detection and within 0.1-0.2 pptv of zero.  The 23 exceptions were 
successfully re-cleaned and analyzed.  All bags were stored evacuated until their use. 

4.   Development of analysis protocols for the expected sample concentration ranges.

Analysis protocols were developed to optimize instrument performance, accuracy and 
efficiency during the project.  In particular, each GC was configured to optimize the 
detection of the lowest possible concentrations in line with the expectation that the 
planned tracer release rates would result in mostly low to moderate concentrations and 
relatively fewer very high concentrations.  Larger volume sample loops were selected in 
anticipation of measuring mostly lower concentrations.  However, smaller volume sample 
loops were also evaluated to characterize the dynamic range available for measuring high 
concentrations on each GC in the event these were encountered.  Analysis parameters were 
adjusted to account for the magnitude of concentration ranges that were expected.  One set 
of parameters dealt with the worst case scenario carryover issue resulting from measuring 
extremely low concentration samples immediately following extremely high concentration 
samples.  Nitrogen purge and vacuum times and the number of purge-vacuum cycles of the GC 
were set to ensure no carryover of high concentrations.  Other parameters controlling the 
timing of the injection, switch to back-flush, and total length of the analysis cycle were 
set to ensure that oxygen and other contaminants were back-flushed before reaching the ECD 
to avoid any interferences.  Electron capture detector attenuation adjustments were also 
tested at different concentration levels to provide quick adjustments to the instruments 
in the case of unexpected concentration ranges. 

5.  Use of a written standard operating procedure (SOP).

A written SOP entitled, Standard Operating Procedure for Sampling and Analysis of Sulfur 
Hexafluoride Using Programmable Integrating Gas Samplers (PIGS) and Automated Tracer Gas 
Analysis Systems (ATGAS) was used by all personnel performing SF6 analysis so that all 
analyses were performed consistently. The SOP contained the following sections:
	1.    Scope and Application.
	2.    Summary of Method.
	3.    Health and Safety Warnings.
	4.    Interferences.
	5.    Personnel Qualifications.
	6.    Equipment and Supplies.
	7.    ATGAS Setup.
	8.    Sample Collection.
	9.    Cartridge Check-In.
	10.  Analysis Preparation.
	11.  Analysis.
	12.  Sample Handling and Holding Times.
	13.  Data Analysis and Calculations.
	14.  Quality Control and Quality Assurance.
	15.  Data and Records Management.
	16.  Trouble-shooting.
	17.  References.

6.   Pre-project calculation of instrument limit of detection (ILOD) and instrument 
limit of quantitation (ILOQ).

Prior to the start of the project, the ILOD and ILOQ were established for each ATGAS to 
provide information on instrument performance.  The ILOD is the instruments limit of 
detection and is defined as the lowest concentration that can be determined to be 
statistically different from zero.  It is a measure of instrument sensitivity and based 
upon the specific instruments ability to differentiate a low level concentration standard 
from instrument noise.  One bag filled with a low level standard was analyzed on each of 
the 12 autosampler ports on each ATGAS.  The analysis at each port was preceded by the 
analysis of a higher concentration standard of at least 10,000 pptv to evaluate any 
possible carryover effects.  The ILOD was calculated as three times the standard deviation 
of a low level standard that was analyzed twelve times.  The ILOQ is the instruments 
limit of quantitation and is defined as the lowest concentration that can be determined 
within 30% of the actual concentration.  The ILOQ was calculated as ten times the standard 
deviation of the same low level standard analyzed 12 times.  Since using different 
concentrations will yield different ILOD and ILOQs, the analyst selected the lowest 
concentration standard to meet as many of the following criteria as possible:
	Has a relative standard deviation (RSD), i.e., the standard deviation divided by 
the mean multiplied by 100 of less than 15%.
	Has a signal to noise (S/N; the mean divided by the standard deviation) between 
3 and 10 (a higher value does not invalidate the result; rather it indicates that a lower 
concentration standard can be used).
	Has a percent recovery (analyzed value divided by the certified value multiplied 
by 100) between 90% and 110%.
 
Results for the pre- and post-project estimation of ILOD and ILOQ for each ATGAS are 
shown in the table below.  While some of the recoveries were slightly outside of 
specification, all initial pre-project ILOD were less than 1 pptv and much less than 
the stated measurement quality objective (MQO) of less than 4 pptv.  The maximum initial 
ILOQ was 3.02 pptv for GC4.  No carryover effects were observed.

Pre-Project (3.11 pptv)
		GC1	GC2	GC3	GC4		
Number		12	12	12	12		
Mean		3.44	3.22	2.80	3.41		
Recovery	110.6	103.5	89.9	109.8		
S.D.		0.087	0.091	0.27	0.30		
RSD%		2.54	2.83	9.52	8.84		
S/N		39.4	35.3	10.5	11.3		
ILOD		0.26	0.27	0.80	0.91		
ILOQ		0.87	0.91	2.66	3.02		
Lab Blank
	GC1	GC2	GC3	GC4 (1-4)	GC4 (5-8)	
Number	198	206	212	108		92	
Mean	0.003	0.003	-0.657	0		-0.121	
S.D.	0.043	0.029	0.708	0.100		0.379	
ILOD	0.128	0.086	2.125	0.301		1.138	
ILOQ	0.426	0.286	7.083	1.005		3.793	
Lab Control (pptv)
	GC1 (3.11)	GC2 (3.11)	GC3 (3.11)	GC3 (10.1)	GC4 (10.1)	All (3.11)
Number	147		124		151		141		137		422
Mean	3.14		3.22		3.03		10.17		9.90		3.13
Recovery%	101.1	100.3		97.4		100.7		98.0		100.8
S.D.	0.185		0.41		0.60		1.01		1.22		0.44
RSD%	5.9		7.8		19.8		10.0		12.3		14.1
S/N	17.0		12.9		5.1		10.0		8.1		7.1
ILOD	0.56		1.24		1.80		3.04		3.7		1.33
ILOQ	1.85		4.12		6.00		10.13		12.2		4.42
Post-Project
	GC1 (3.11)	GC2 (3.11)	GC3 (3.11)	GC4 (10.1)		
Number	12		12		12		12		
Mean	3.25		3.14		3.2		3.12		
Recovery%	106.6	112.1		97.3		105.4		
S.D.	0.095		0.15		0.16		0.26		
RSD%	2.92		4.78		5.00		8.33		
S/N	34.21		20.93		20.00		12.00		
ILOD	0.285		0.45		0.48		0.78		
ILOQ	0.95		1.5		1.6		2.6		

		
7.  Holding time studies.

Holding time studies are determinations of the length of time a sample can be held in its 
container before the sample concentration changes appreciably.  Holding time studies are 
conducted whenever the method or sampling container is changed in any way prior to 
commencement of a project. These studies are used to determine what effect degradation of 
the materials will have on sample results. Knowledge of the length of time the samples 
can be held will help in planning the analysis schedule for the samples in the field.  
Holding time studies on the Tedlar sample bags performed in 2004 showed no appreciable 
change in sample concentration for up to six months if stored indoors and away from 
temperature extremes.  Artifact studies on the Pliobond-sealed bag sample fittings and 
R-3603 tubing were performed in 2011 and early 2013 and showed no evidence of sample 
contamination or bag leakage.  All samples were initially analyzed within a week of 
sampling for this project.

ATGAS	Loop Volume	Calibrated Range	Number of Standards
1	1 ml		ILOD-75,100 pptv	19
2	1 ml		ILOD-142,300 pptv	22
3	500 ul		ILOD-210,700 pptv	final 23
4	1 ml		ILOD-52,600		18

8.  Daily calibration of the ATGAS.

In order to quantify the concentration of the samples, each of the four ATGASs was 
calibrated at the beginning of each analysis day using 10 to 18 NIST-traceable SF6 
standards.  The number of standards used was dependent upon the concentration range 
available to each ATGAS as they were configured for this experiment and the expected 
range of concentrations.   Each ATGAS was configured to optimize the ability to detect 
very low concentrations, principally by choice of a sufficiently large sample loop. This 
low end optimization had the effect of restricting the ability to quantify higher 
concentrations without changing sample loops.  The analytical ranges for each ATGAS as 
configured for the experiment are shown in the table above.  Differences relate to sample 
loop size and the specific performance characteristics of each ATGAS.

The commonly used calibration standards ranged from 3.11 pptv to 75,100 pptv and 
covered most of the range of field sample concentrations encountered. In general, the 
nighttime IOPs had higher concentrations and therefore required a broader, higher 
analytical ranges. Over the course of the project, three standards were depleted and 
replaced by standards with similar concentrations. The 504 pptv standard was replaced 
with a 505 pptv standard, the 9730 pptv standard was replaced with a 10010 pptv standard, 
and the 10.1 pptv standard was replaced with a 9.95 pptv standard. The latter happened 
late in the project and was used mainly in follow up testing. Some standards were getting 
low by the end of the project but it was not necessary to replace them. The UHP air zero 
point used in the calibration contained a slight trace of SF6 as demonstrated by the very 
small peaks often seen in the chromatograms.  Concentrations of samples were calculated 
using a point-to-point fit calibration of the standards.  The calibration curve was 
examined for "wild fits" and error messages were displayed in the event of the occurrence 
of some anomaly so that the analyst could more closely examine the calibration curve and 
decide if it was appropriate to use.

9.  Initial ATGAS calibration verification (ICV).

After each calibration was completed and reviewed, the curve was validated by analyzing 
the same calibration standards as if they were field samples.  This validation 
demonstrated that sample concentrations within the calibration range could be quantified 
correctly.  The recoveries were required to be within 10% of the certified value or the 
standards were re-analyzed.  If the recoveries still did not meet the acceptance limits, 
the bags were refilled and analyzed again.  If the recoveries were still not acceptable, 
the instrument was re-calibrated and ICV was attempted again. 

10. Continuing ATGAS calibration verification (CCV) and laboratory controls.

The validity of the ATGAS instrument calibration curves were regularly checked by 
re-analyzing calibration standards as if they were field samples.  This procedure, 
called continuing calibration verification (CCV), was performed to provide evidence that 
instrument drift had not caused the calibration to be unable to correctly quantify sample 
results within the MQO acceptance levels.  Standards were chosen to cover the concentration 
range of samples that had been analyzed since the last calibration verification. The 
standards were required to have a recovery of 10% of the certified value (20% for 
< 50 pptv) for the CCV to be considered valid (Table 4).  If any of the standards were not 
within the acceptance window, the instrument was re-calibrated and the curves were 
re-validated.  All data within the unacceptable concentration range, from the point of 
the last acceptable CCV, were flagged and re-analyzed.

There was a tendency for the responses of the GCs to become more stable with continued 
operation but all of them exhibited some susceptibility to drift of the calibration, 
especially in the first few hours of operation. The frequency of CCVs ranged from less 
than 1 to about 3 h depending on the GC and how long it had been in operation with a 
relatively stable calibration for any given day. In general, calibration checks were 
done more frequently in the first few hours of operation and less frequently after that 
if the GC was exhibiting stable behavior. Recalibrations were usually done if the 
response had drifted significantly (> about 6-8%) as there was a tendency that once 
drift had commenced it often continued and raised the prospect of performing analyses 
that would have to be redone due to violating the MQO requirements of 10% (> 50 ppt) 
or 20% (< 50 ppt). Furthermore, the intent was to keep all results within 10%.  Following 
any recalibration, responses were often stable within 5% for the remainder of the day.  
In some cases it was not necessary to recalibrate after the initial calibration although 
it was common for GCs 2 and 3 to be recalibrated once a few hours into the day and then 
remain stable for an extended period of time. GCs 1 and 4 were the most susceptible to 
problematic calibration drift but they sometimes had stable calibrations. Considerable 
time was spent in calibration and recalibration of the GCs to ensure achieving MQO. 
There was also some analysis time lost due to the necessity to rerun some sets of sample 
cartridges due to failure to achieve the requisite CCV recoveries.

The CCV serve as laboratory control samples and measures of instrument precision and 
instrument accuracy.  Results for the combined laboratory control samples (CCV) are 
summarized in the table below.  All CCV data are included except for when a failed CCV 
was repurposed to be a new calibration. In such cases those records were eliminated. 
Excepting the 3.11 pptv standard, all of the RSD were below the 10% limit specified in 
the MQOs and indicated good instrument precision.  The higher RSD result for the 3.11 
pptv standards arises from the relative lack of low end sensitivity for GCs 3 and 4. 
In fact, GC4 often had trouble detecting the 3.11 pptv peak. That explains the decision 
to eventually omit the 3.11 standard as part of the calibration curve for GC4. The lack 
of low end sensitivity in GC3 was less acute but it still contributed to an increase in 
RSD at 3.11 pptv. The lack of sensitivity to such low concentrations was not critical in 
that ambient background concentrations of SF6 were much higher, generally around 8-9 pptv. 
The average recoveries are indicative of excellent accuracy across the full range of 
concentrations used and are easily within the 10010% (or 20%) requirement.

Concentration			%Recovery
Actual	Measured	S.D.		RSD%	S/N	Number
	(Avg.)	S.D.				
0	-0.005		2.23				558
3.11	3.12		0.44	100.8	14.1	7.1	422
10.1	10.11		0.91	100.1	9.0	11.1	540
19.19	19.67		1.22	102.4	6.2	16.1	516
35.1	34.9		2.11	99.4	6.0	16.6	501
88.7	89.7		3.54	101.2	3.9	25.3	500
301	302.6		19.6	100.5	6.5	15.5	487
504	509.0		27.9	101.0	5.5	18.2	269
505	511.7		23.7	101.3	4.6	21.6	199
799	806.5		50.3	100.9	6.2	16.0	459
1550	1582.4		106.4	102.1	6.7	14.9	442
3140	3182.6		170.8	101.4	5.4	18.6	423
4980	5073.4		269.4	101.9	5.3	18.8	388
8270	8364.9		275.5	101.1	3.3	30.4	301
9730	9912.4		343.4	103.0	3.5	28.9	103
10010	10213.4		315.9	102.0	3.1	32.3	155
16370	16571.9		618.7	101.2	3.7	26.8	239
21720	21790.6		1015.1	100.3	4.7	21.5	150
36900	36997.8		783.7	100.3	2.1	47.2	102
52600	52434.6		1646.0	99.7	3.1	31.9	78
75100	75394.8		1005.5	100.4	1.3	75.0	28
90100	90540.8		982.6	100.5	1.1	92.1	17
103600	103019.0	758.0	99.4	0.7	135.9	7
142300	143822.0	1019.4	101.1	0.7	141.1	2


11.   Atmospheric background checks of SF6 at the tracer analysis facility (TAF).

A background atmospheric check of SF6 in the TAF consisted of analyzing three samples 
of the room air in the TAF, after calibration but prior to running regular samples, 
on each GC every analysis day,.  This information was used to determine if there was 
any leakage in the analysis system when compared to the instrument blanks that were 
subsequently analyzed.  The data provided for an inter-comparison between GCs that 
were being used on the same day to check the between instrument precision.  The results 
were also used to reveal discrepancies between GCs to indicate a problem that otherwise 
might go undetected.  The results shown in the table below indicate that, with the 
possible exception of GC1, there was good precision between the four GCs.  The average 
concentration for all background checks was 9.8 pptv with a standard deviation of 1.7 
pptv.  The RSD specification was not satisfied with all RSD greater than the 10% MQO. 
The primary reason for this is that there were a few days where the ambient room air 
had slightly elevated SF6 concentrations, usually in the 10-20 pptv range. This was 
above the nominal ambient background that was usually about 8-10 pptv, thus the larger 
standard deviations and RSD. The slightly elevated concentrations were found to be a 
transient phenomenon, when present, and the results could vary by GC depending on the 
timing of the room air analyses. In the absence of any transient elevated background, 
the agreement between GCs was even better than indicated in the table.

Room Air	#	Mean	S.D.	RSD
GC1		57	10.72	2.38	22.2
GC2		60	9.66	1.32	13.7
GC3		55	9.81	1.29	13.2
GC4		53	9.14	1.15	12.5
All		225	9.84	1.71	17.3

12.   Laboratory (instrument) blanks.

A laboratory or instrument blank was analyzed on each ATGAS each analysis day to verify 
that there was no contamination or leaks within the analysis system as compared to the 
background checks analyzed that day, that there was no carry-over from previously 
analyzed high concentration standards, and to ensure carrier gas purity.  The blank sample 
consisted of a cartridge of 12 bags that were each filled with ultra high purity (UHP) 
nitrogen.  The concentration results of all bags were required to be less than the lowest 
calibration standard and close to a concentration of 0 pptv.  If the concentration of 
one or more of the bags was higher than the acceptable range, the bag was re-filled and 
re-analyzed.  If the concentration still was not within acceptable limits, the instrument 
was re-calibrated and re-verified or the samples were flagged and re-analyzed.  If there 
were still indications of contamination, the problem was identified and fixed before 
analysis continued.

The average results indicate no contamination or leakage problems within any of the 
ATGASs as well as no carryover issues and meet the MQO of <4 pptv.  The higher mean 
and standard deviation for ATGAS 3 reflect its sensitivity to the effect of very small 
changes in baseline on the peak integration at very low level concentrations.  This 
features also shows up in some calculations of the ILOD and ILOQ for ATGAS 3.

13.   Laboratory duplicates.	

Analyses of laboratory duplicates were performed each day to provide evidence of 
instrument precision.  Each day at least one primary field bag sampler cartridge was 
analyzed in duplicate on each ATGAS.  The sample cartridge and its duplicate were 
analyzed at least 3 hours apart in order to ensure an appropriate estimation of 
instrument precision over time.  The duplicate cartridges were selected to encompass 
as much variation and range of concentration as possible within the concentration 
range bracketed by the calibration curve for each ATGAS.  The mean of the absolute 
value of the relative percent differences (RPD),

RPD = (100*(measure#1 - measure#2)/average(#1 and #2)) 

were required to be within 5%.  Any result not within the acceptable limits was flagged 
and re-analyzed.  If the result was still not within acceptable limits, the analysis was
terminated until the ATGAS precision could be re-established.

The |RPD| laboratory duplicate results are shown in the table below. GCs 1, 2, and 3 
showed good precision over time and satisfied the MQO objective for lab duplicates. 
GC4 did not satisfy the RPD MQO objective suggesting a greater tendency for temporal 
drift in response that was not fully accounted for by routine CCV and re-calibrations.  

Summary of RPD results for laboratory duplicates.
Laboratory Duplicates			
			Mean %	Mean %
GC		#	RPD	|RPD|
1		160	1.08	2.25
2		212	-0.03	1.84
3		209	0.06	2.10
4 (IOPs 1-4)	59	2.10	5.58
4 (IOPs 5-8)	65	-2.01	6.20

14.   Field blanks.

Field (method) blanks were sampled and analyzed to indicate if there was any 
contamination or leakage introduced by any part of a bag samples history from sampling, 
handling, and transport through to the final analysis.  For example, isolated instances 
of high concentrations of SF6 in the field blanks can indicate holes in the sampling bag, 
clips not properly closed, wrong location number, or other operational problems.  
Consistently high concentrations would indicate a sampling method that could not measure 
null concentrations accurately.

Three field blank samplers were deployed during each IOP. field blank consisted of a 
sampler containing a cartridge filled with ultra high purity (UHP) nitrogen. Each sampler 
was deployed at its designated location and collocated with a regular sampler with the 
tubes connected and clips left open.  Software requirements of the sampling program made 
it necessary for the pump on the first bag to turn on for one short pulse.  However, after 
that, all pumps were left off and there was no additional filling of any of the bags.  
At the end of each test, the clips on the blank cartridges were closed and the cartridges 
were collected, transported, and stored along with all the regular sample cartridges.  
With the exception of the special sampling program, the field blanks were treated 
identically to the regular samples.

A summary of the results is presented in the table below.  The means and standard 
deviations for IOPs 1, 2, and 4 are all near zero and very low indicating no contamination 
or sample handling problems. The means for IOPs 3, 6, and 8 are fairly low but hint at the 
possibility of slight contamination or other artifact. The much larger means for IOPs 5 and 
7 are believed to represent sample contamination due to infiltration of high concentration 
plume into some of the bag samples. The highest measured blank was 13.7 pptv during IOP5. 
During these IOPs, high concentrations were measured at many locations on the bag sampling 
array for extended periods of time. It is possible that small amounts of tracer diffused 
into the sample bags past the pump seals and through the open, unclipped tubing. This 
contamination issue will be discussed below in detail in part 16 of this section (field 
duplicates). Despite this, these results suggest that the influence of high concentration 
plumes on low concentration bags was still minimal as even IOPs 5 and 7 were well below 
ambient background concentrations.

The consequences of these observations are considered more fully in the determination of 
final MLOQ for the project results (step 19 below).  Briefly, the field blank results 
adversely affected some of the project MQOs in that they indicated an MLOQ sometimes 
greater than the nominal MLOQ.

Field blank results for each test.
IOP	#	Mean	S.D.	MLOQ
1	36	0.02	0.06	0.6
2	36	-0.17	0.36	3.6
3	36	0.19	1.41	14.1
4	36	0.00	0.00	0.0
5	35	2.79	3.74	37.4
6	36	0.15	0.54	5.4
7	35	0.79	1.38	13.8
8	36	0.12	0.70	7.0

15.   Field controls

Three field control samplers were deployed during each IOP.  The cartridge for each 
control sampler was filled with NIST-certified tracer concentrations ranging from 
14.79 pptv to 5170 pptv.  Bags 1-3 contained 14.79 pptv, bags 4-6 contained 283.9 pptv, 
bags 7-9 contained 1571 pptv, and bags 10-12 contained 5170 pptv.  Each sampler was 
deployed at its designated location and collocated with a regular sampler with the 
tubes connected and clips left open.  Software requirements of the sampling program 
made it necessary for the pump on the first bag to turn on for one short pulse.  
However, after that, all pumps were left off and there was no additional filling of 
any of the bags.  At the end of each test, the clips on the control cartridges were 
closed and the cartridges were collected, transported, and stored along with all the 
regular sample cartridges.  With the exception of the special sampling program, the 
field controls were treated identically to the regular samples.

The field control samplers served two primary purposes.  First, they checked for any 
biases or inaccuracies introduced during the sampling, handling, and storage of the 
samples.  Second, recall that the standards used to calibrate the GCs (up to 210,700 
pptv) were all NIST traceable.  The tracer concentrations used to fill the control 
bags also came from NIST-certified standards but they were different from those used 
in the calibration of the ATGASs.  As a consequence, the field control samples serve 
as a semi-independent measure of quality control of the overall process, essentially 
a method audit.

The results for the field control samples expressed in terms of the individual IOPs 
are shown in the table below. While many of the results were quite good, there were 
some significant problems identified with the field control samples. First, the lower 
recoveries for the 1571 and 5170 pptv standards in IOP1 can be attributed to an 
operational oversight. An improper size tubing fitting was in place at the time these 
standards were filled and there was insufficient time to locate the correct fittings 
prior to the IOP1 deployment. The size discrepancy resulted in an incomplete seal and 
slight dilution of the some of the noted sample bags during filling.

Second, the number of samples for the 283.9 and 1571 pptv standards was 8 instead of 9 
in IOP4 due to an operator error during bag filling.

Third, one of the control sample cartridges was mistakenly set to run as a regular 
sample during IOP7. This operator error in the field resulted in further filling bags 
that were already filled with their respective control standard concentrations. This 
corrupted the control samples at that site and explains the drop from 9 to 6 for the 
number run during IOP7. The regular sampler that was collocated with the control 
sampler was programmed to run as the control sample. As a consequence, the pumps failed 
to run and the bags were flat thus voiding the results from that regular sampler as well.

Fourth, and perhaps most significantly, the recoveries for the 1571 and 5170 pptv 
standards for the control samples in IOPs 7 and 8 are poor and anomalously low. The 
corresponding RSD and RPD are similarly very poor. The cause of this is not certain 
but it might be related to cold temperature sampling artifacts. The MQO objectives 
for the 14.79 pptv standard in IOP5 were not met, possibly for the same reason. A 
review of Table 11 suggests that the failures to meet MQO objectives were biased toward 
the nighttime IOPs. The potential cold temperature sampling artifacts are discussed in 
greater detail in the following section in the context of field duplicate discrepancies.

Combined ATGAS field control results expressed in terms of standard concentration 
and IOP number.
Standard	IOP1	IOP2	IOP3	IOP4	IOP5	IOP6	IOP7	IOP8	All
14.79									
#		9	8	9	9	9	9	6	9	68
Mean		15.26	14.96	14.83	14.98	17.0	15.31	15.32	14.96	15.33
S.D.		0.26	0.27	0.54	0.45	4.14	0.28	0.75	0.48	1.63
Avg. Recovery	1.03	1.01	1.00	1.01	1.15	1.04	1.04	1.01	1.04
Mean RPD%	3.2	1.2	0.3	1.3	14.9	3.5	3.6	1.1	3.7
Mean |RPD|%	3.3	1.5	2.8	2.7	15.4	3.5	4.7	2.6	4.6
RSD%		1.7	1.8	3.7	3.0	24.4	1.8	4.9	3.2	10.6
S/N		58.6	56.1	27.3	33.1	4.1	55.5	20.4	31.0	9.4
283.9									
#		9	9	9	8	9	9	6	9	68
Mean		282.3	300.6	300.4	300.6	297.3	295.2	298.8	295.9	296.2
S.D.		20.9	4.1	7.0	5.5	6.6	26.7	6.2	2.3	13.8
Avg. Recovery	0.99	1.06	1.06	1.06	1.05	1.04	1.05	1.04	1.04
Mean RPD%	-0.6	5.9	5.8	5.9	4.7	4.0	5.2	4.2	4.3
Mean |RPD|%	6.6	5.9	5.8	5.9	4.7	8.5	5.2	4.2	5.9
RSD%		7.4	1.4	2.3	1.8	2.2	9.1	2.1	0.8	4.7
S/N		13.5	72.5	42.9	54.7	45.1	11.0	48.1	128.3	21.4
1571									
#		9	9	9	8	9	9	6	9	68
Mean		1468.1	1546.9	1524.2	1513.7	1524.3	1531.9	1215.8	1105.5	1437.0
S.D.		148.3	20.9	26.0	73.1	18.8	78.7	394.8	355.5	236.0
Avg. Recovery	0.93	0.98	0.97	0.96	0.97	0.98	0.77	0.70	0.91
Mean RPD%	-6.6	-1.5	-3.0	-3.6	-3.0	-2.5	-22.6	-29.6	-8.5
Mean |RPD|%	6.6	1.8	3.0	3.7	3.0	4.2	22.6	29.6	8.8
RSD%		10.1	1.35	1.7	4.8	1.2	5.1	32.5	32.2	16.4
S/N		9.9	74.1	58.7	20.7	81.1	19.5	3.1	3.1	6.1
5170									
#		9	9	9	9	9	9	6	9	69
Mean		4259.4	5075.9	4861.5	5022.6	4893.5	5091.0	3362.1	4376.6	4672.5
S.D.		1611.4	138.7	124.2	78.2	145.2	117.1	1339.0	885.4	892.3
Avg. Recovery	0.82	0.98	0.94	0.97	0.95	0.98	0.65	0.85	0.90
Mean RPD	-17.6	-1.8	-6.0	-2.85	-5.35	-1.5	-35.0	-15.3	-9.6
Mean |RPD|	17.6	2.7	6.0	2.85	5.35	2.25	35.0	15.3	9.8
RSD		37.8	2.7	2.55	1.6	3.0	2.3	39.8	20.2	19.1
S/N		2.6	36.6	39.1	64.2	33.7	43.5	2.5	4.9	5.2

16.   Field duplicates.

Fourteen field duplicate samplers were deployed for each daytime IOP and 12 for each 
nighttime IOP during PSB2. The duplicate samplers were handled identically to the 
primary samplers with which they were collocated.  Both samplers were mounted at the 
same height and affixed to the same post. The sample inlets for the two samplers 
faced opposite directions from the mounting post with inlet separation between samplers 
being a little more than 1 m.  A summary of the results is provided in the table below.

A salient feature of the table is the much different |RPD| results between the daytime 
and nighttime IOPs. For the daytime, only IOP1 failed to meet the MQO objective of |RPD| 
< 15%. In contrast, only IOP6 met the |RPD| objective for the nighttime IOPs. In fact, 
the remaining IOPs were significantly higher than the objective. The high nighttime 
|RPD| values resulted in the overall (All) MQO objectives not being met. That was true 
for all distances.  However, a distinct pattern emerges when the data are broken down by 
distance and day or night. First, the discrepancy between duplicate samplers was much 
higher at night than during the day. The daytime |RPD| mostly satisfied the MQO objective 
while the nighttime |RPD| were generally significantly greater than the MQO objective. 
Second, there appears to be a distinct trend in both the daytime and nighttime |RPD| 
with decreasing values at increasing downwind distance.

These |RPD| results were uncharacteristically large compared to previous tracer studies
conducted by FRD. The reader is referred to the Technical Memorandum for this project 
for a more comprehensive description of the field duplicate results, their potential 
significance, and some follow up studies investigating potential explanations for the
differences.

Summary of field duplicate sampler results.
IOP	#	Avg. % RPD	Avg. % |RPD|
1	130		-2.3	18.0
2	141		-3.4	13.8
3	129		1.9	12.7
4	143		-1.7	6.4
5	141		-7.6	33.0
6	140		-1.7	9.3
7	140		-17.3	44.0
8	140		6.5	32.1
All	1104		-3.3	21.3
All 100 m	378	-2.9	24.3
All 200 m	365	-4.5	22.2
All 400 m	361	-2.3	17.1
All 100 m Day	190	-3.1	15.6
All 200 m Day	177	0.6	12.2
All 400 m Day	175	-1.2	9.2
All 100 m Night	188	-2.8	33.1

17.   Software quality control checks.

Several important quality checks were built into the software to efficiently aid the TAF 
analyst in ensuring that the ATGAS instruments were functioning correctly during analysis. 
	Since the concentration is dependent upon the temperature of the ATGAS ovens, it 
        is critical that oven temperatures do not fluctuate widely during analysis.  
        Temperature acceptance limits were set (2 C) and the software produced a pop-up 
        window to alert the analyst in case of unacceptable oven temperature readings. 
        All samples obtained using the incorrect oven temperatures were re-analyzed.
	To check for instrument drift, the software alerted the analyst to validate the 
        calibration curve when more than three hours had elapsed from the last CCV.  The 
        analyst had the option of overriding the alert or checking the calibration and 
        re-starting the 3-hour clock.  This option was always exercised except on a few 
        occasions near the end of the analysis day when only 1-2 more cartridges required 
        analysis.  Even then this was only done on ATGASs that had previously been 
        exhibiting consistently stable response for extended periods of time during that 
        day.
	In order to verify the calibration curve in the area of interest and to save time, 
        the software produced on the computer screen a record of the highest and lowest 
        concentrations measured since the last CCV.  The analyst had only to re-analyze 
        calibration samples within that range.  However, the complete calibration range 
        was routinely done to most fully evaluate the current status of instrument 
        response and performance. 
	Several data flags were shown immediately on the computer screen to aid the 
        analyst in deciding whether the data for each bag was good or re-analysis was 
        necessary.  For example, the low pressure flag alerted the operator to a problem 
        with the analysis that was almost invariably due to pinched tubing restricting 
        sample flow.
	The software kept track of which ATGAS field duplicate was analyzed on and 
        directed the analyst to use the same GC for the duplicate cartridge.  This helped 
        to quantitate the variability of the field analysis without adding the extra 
        variability of analyzing on a separate ATGAS.  However, due to limitations 
        imposed by the restricted calibration ranges of ATGASs 2 and 4, it was not 
        uncommon for the field duplicates to be done on different ATGASs.
	The software alerted the analyst if any calibration points did not meet 
        pre-determined acceptance criteria.  The analyst could then review the 
        calibration curve to determine the acceptable course of action.

18.   Data verification.

Data verification was performed to ensure that the samples met all QC acceptance limits 
and that all samples had been analyzed for that particular test.  Transcription and 
calculation errors were reduced by automated data reduction techniques such as automated 
flagging of results outside acceptable limits, raw data summary sheets, auto-generated 
quality control sheets, auto generation of chromatogram plots including calibration curves,
and electronic transfer of data from the ATGASs to Excel spreadsheets.  The analyst and 
at least one other person familiar with the data analysis process reviewed all data 
packages.  All data packages were batch processed per run on each ATGAS.  All data 
packages included the raw data sheets, quality control sheets that summarized the results 
of all QC data generated for that batch, plots of all chromatograms and calibration curves,
a copy of the laboratory notebook pages for that analysis, and a data verification sheet
to ensure the verifier checked all QC parameters.  Software produced an Analysis Summary 
that was utilized to ensure that there was at least one acceptable result for each bag 
for each location that was downloaded for each IOP. Any samples noted by the software 
were re-analyzed and the Analysis Summary report was re-run until all samples had been 
analyzed or a justifiable reason had been determined for a missing sample.  Cartridges 
were not cleaned until all available samples had been analyzed.

19.   Post-project determination of ILOD, ILOQ, MLOD, and MLOQ.

ILOD and ILOQ were previously defined in the quality control procedures (step 6 above).  
In that section a procedure was described for obtaining a preliminary pre-project 
estimate of the ILOD and ILOQ using a very low concentration calibration standard.  
These results were reported in Table 5.  There are additional ways to estimate ILOD 
and ILOQ.  These include the use of laboratory blanks and the low level laboratory 
control standards used for calibration and CCV.  These alternative determinations 
together with a post-project repeat of the initial procedure were also shown in the
table above.  In general, all of the various estimates for ILOD were low and well 
below the stated MQO of 4 pptv. The lab blank estimate of ILOD for GC3 was relatively 
high at 2.1 pptv. As noted earlier, this is due to the sensitivity of GC3 to the effect 
of very small changes in baseline on the peak integration at very low concentrations. 
Similarly, the ILOD estimate for GC3 based on lab control (CCV) results was also 
elevated at 3.04 pptv. The other exception was for the ILOD on GC4 (3.7 pptv) based 
on the lab control results. This was based on the 10.1 pptv standard, instead of the 
3.11 pptv standard, since the detection of the lower concentration standard by GC4 
was erratic.

The method limit of detection (MLOD) and method limit of quantitation (MLOQ) are 
estimates of the lowest field concentration level that can be determined with some 
degree of certainty.  Unlike ILOD and ILOQ, MLOD and MLOQ incorporate all the sources 
of variability and uncertainty introduced during each phase of the sampling, handling, 
and analysis.  The MLOD is defined as the lowest field concentration measurement that 
can be determined to be statistically different from zero.  It is based upon the 
methods ability to differentiate a low-level concentration standard from the combined 
effects of instrument and method noise.  The MLOD and MLOQ are calculated exactly the 
same as ILOD and ILOQ except that method variability is factored into the determination 
by using results from samples that have been put through the rigors of field sampling.  
The MLOD is calculated as 3 times the standard deviation of a low level standard.  
The MLOQ is defined as the lowest concentration that can be determined within 30% of 
the actual concentration.  The MLOQ is calculated as 10 times the standard deviation 
of the same low level standard.

There are several ways to attempt to estimate MLOD and MLOQ.  These include field blanks, 
low concentration field controls, and field duplicates. Ambient background samples of 
all regular field samples can also be used to estimate MLOQ. However, these samples do 
not incorporate all sources of variability observed during experiments.  Specifically, 
background samples, by definition, were not exposed to the higher level concentrations 
measured by many of the samplers that were strongly impacted by the tracer plume.  
Sampler cartridges located on parts of the grids that were heavily impacted by the tracer 
plume were seen to occasionally have their lower concentration bags affected (e.g., some 
of the field blank bags). There is also the problem of setting a cutoff value separating 
truly background samples from those that were slightly influenced by the plume. For these 
reasons, the ambient background method was not calculated. Estimates of MLOQ were made 
using each of the other methods. The table below summarizes the results of the analysis 
for the estimate of MLOQ.

Estimates of MLOQ using the field duplicates technique provided estimates ranging by 
IOP from 7.5 to 17.8 pptv for duplicate pairs less than 12 pptv with an overall IOP 
average of 11.3 pptv. Again, there was a notable difference between the overall day 
and night results (9.6 and 12.8, respectively). For duplicate pairs less than 20 pptv 
estimates of MLOQ ranged from 7.5 to 35.7 pptv by IOP with a much larger discrepancy 
between day and night. Estimates of MLOQ using field blanks ranged from zero to 37.4 
pptv with an overall average of 17.5 pptv. The overall average is skewed by the results 
of IOPs 3, 5, and 7. The result for IOP3 is dominated by one value of 8.4 pptv with the 
other 35 values being zero. IOPs 5 and 7 had several non-zero blank values resulting, 
again, in a much higher estimate of MLOQ for the overall nighttime average. That is 
likely due to the fact that IOPs 5 and 7 had the highest measured concentrations during 
PSB2 and they occurred over much of the sampling array. All of the estimates of MLOQ 
by the field controls were < 7.5 pptv with the exception of IOP5 (41.4 pptv). Again, 
it is possible this higher result is due to a sampling artifact.
    
The significant differences in MLOQ results suggest using different values for day and 
night. For reasons given earlier, it is preferable to use the lowest practicable 
concentrations for the calculation of MLOQ. That would discount the estimates of MLOQ 
using duplicate pairs <20 pptv and recommend the usage of duplicate pairs <12 pptv. 
The daytime average MLOQ based on field duplicates is then 9.6 pptv. Excepting the 
one outlier field blank value in IOP3 all of the daytime blank results are < 5 pptv. 
The daytime MLOQ based on the low control standard is similarly low. Thus, a case 
can be made for a daytime MLOQ not to exceed about 9-10 pptv. A daytime MLOQ was 
selected to be 9 pptv. That is consistent with MLOQ determined in past studies and 
very near what appears to be the current ambient global background for SF6.

The nighttime assessment is more problematic. The average nighttime MLOQ by field 
duplicates, field blanks, and low control standard is 12.8, 22.8, and 22.7 pptv, 
respectively. In this nighttime analysis, the field blank result will be given lesser 
weight as it emphasizes the variability around zero due to the infiltration of small 
quantities of tracer. The effect of tracer infiltration suggested by the field blanks 
would be small on any samples near ambient background levels of 8-10 pptv and 
increasingly negligible as concentrations increased. The maximum blank value was 13.7 
pptv in IOP5 and very few values exceeded 5 pptv. All of the nighttime estimates of 
MLOQ by field control standard are low except for IOP5 (41.4 pptv) and the overall 
nighttime average is 16.3 pptv. Based on the stated assumption and the overall results 
for the field duplicates and control an argument can be made for a nighttime MLOQ in 
the range of roughly 12-16 pptv. The decision was somewhat arbitrary but a value of 15 
pptv was selected as the final nighttime MLOQ. 

While an argument could be made for a higher daytime MLOQ and an even better argument 
might be made for a higher nighttime MLOQ, the final MLOQ were determined to be 9 and 
15 pptv for daytime and nighttime samples, respectively. All daytime values (IOPs 1-4) 
less than 9 pptv and all nighttime values (IOPs 5-8) less than 15 pptv have been 
flagged as estimates in the final database.

20.   Final data review.

All field data were verified to make sure there was a result for every location, 
cartridge, and sample bag and that all results were flagged appropriately.  The 
following examples of verification plots and summaries were chosen to illustrate the 
diligence with which each data point is reviewed.  Every quality control sheet for each 
data package was reviewed to ensure proper flagging of final data.  Bubble/dot plots 
were created and reviewed to ensure all data were reasonable and consistent with 
respect to the overall concentration pattern and the nearby neighbors of each bag 
sample.  Any suspicious data point was traced back through the analysis and deployment 
records to determine if it was indeed a valid result.  The sampler servicing records 
maintained by all field sampler deployment personnel for noting any problems, were 
used to check any outliers or anomalies in the data.  Cartridge time history plots 
as well as individual chromatograms were also reviewed to determine any suspicious data 
points.  Any suspicious data point was traced back through the analysis and deployment 
records, sometimes with the aid of the master history file, to determine if it was 
indeed a valid result.  All field QC was scrutinized.  All suspicious data were 
appropriately flagged.

The finalized data set was then analyzed using a program used to determine if all flags 
were added correctly and if the sample results could possibly be QC results. Any results 
appearing on this sheet were verified and changes to the data base were made as 
necessary.

21.   Data handling.

All results were printed on hard copy as a backup in case of loss of the data files 
and to aid in the data verification process.  The data packages were filed for future 
reference and to be readily available during the project for immediate review.  
Backup copies of the raw ATGAS data were made occasionally and at the end of the 
project to prevent total loss of data in the case of a computer failure.  All final 
QC and sample results were printed on hard copy and placed in a binder to be stored 
with any reference materials in the project archive. 

22.  Summary of Data Completeness and Contribution by GC

The table below summarizes bag sampling data completeness for each test as well as 
for the entire project.  The MQO of 90% (Table 4) was exceeded in every case. GC3 ran 
the highest fraction of the total number of samples. This was due to its wide 
analytical range available without resorting to sample loop changes, the shortest 
analytical cycle time, and relatively consistent stable operation that minimized the 
need to recalibrate or rerun samples. GC2 ran the next most samples. It had a slower 
cycle time than GC3 but had the most stable operation once the detector was replaced 
between IOPs 2 and 3. GCs 1 and 4 were overall the least productive, primarily due to 
greater difficulty in sustaining stable calibrations, but also longer cycle times than 
GC3. Regardless of GC, however, data had to bracketed by satisfactory ICV and CCV 
recovery criteria to be acceptable.

		IOP
GC		1	2	3	4	5	6	7	8	Total
1		458	323	215	421	310	409	326	408	2870
2		225	367	550	421	466	374	441	420	3264
3		576	567	600	433	485	455	434	444	3994
4		301	303	195	285	299	334	371	300	2388
Total		1560	1560	1560	1560	1572	1572	1572	1572	12528
Valid		1512	1523	1543	1534	1501	1487	1480	1547	12127
Flag3 (see below)	2	4	0	1	13	12	6	14	52
Flag4		46	33	17	25	58	61	86	11	337
Flag5		0	0	0	0	0	9	0	0	9
Flag6		5	2	5	1	5	3	7	1	29
Not Analyzed	0	0	0	0	12	0	0	0	12
Completeness%	96.7	97.6	98.9	98.3	95.4	94.6	94.1	98.4	96.8

Summary table of data completeness by IOP with contribution to analyses by individual GC.


Final Bag Sampler Data Files and Format

The main final bag sample tracer data files provided with this report contain 12 columns:

1.	test (IOP) number
2.	bag number (1-12)
3.	begin time (yyyymmdd)
4.	start time (hhmmss)
5.	sampling period (seconds/bag)
6.	dist (distance from release point in meters)
7.	angle (angle in degrees along respective arc from north)
8.	agl (meters, above ground level)
9.	latitude
10.	longitude
11.	concentration (SF6 pptv)
12.	quality control flag
The files are in csv format with fixed width fields.  The data files are named 
SAGE16Sampler##.csv where ## is the number of the Individual IOP test. The bag 
sampling Readme file accompanying this report summarizes the contents of this chapter 
on the bag sampling.

A supplemental set of tracer data files containing field duplicate analyses is also 
provided. This was not customarily included in the databases for past field studies. 
They are included here because of the sometimes large differences observed in 
collocated field duplicate sampling during PSB2 and the potential significance of 
these differences on interpretation of the data and any attempts to estimate 
uncertainty in the measurements. These files are named SAGE16##DUP.csv where ## 
is the IOP number. There are 12 columns in these files:

1.	Project		SAGE16
2.	Test		IOP number
3.	Loc. 1		Location identification of primary sampler
4.	Cart. 1		Sample cartridge at Loc. 1
5.	Sampler 1	Sampler at Loc. 1
6.	Loc. 2		Location identification of duplicate sampler collocated with Loc. 1
7.	Cart. 2		Sample cartridge at Loc. 2
8.	Sampler 2	Sampler at Loc. 2
9.	Bag		Bag number
10.	Concen. 1	Concentration at Loc. 1
11.	Concen. 2	Concentration at Loc. 2
12.	RPD		Calculated |relative % difference|

Some of the files contain additional information in the next few columns related to 
sample reruns attempting to confirm the original result for selected field duplicate 
pairs.

Final Data File Quality Control Flags

All of the data were flagged with one of seven possible quality flags: These are:

0	> MLOQ; good data to be used without qualification.
1	< MLOD (4 pptv)
2	< MLOQ (9 pptv day or 15 pptv night) and > MLOD. Treat as an estimate.
3	missing  field problem (check in was F, I, or B), also missing analyses are 
        included here; data values set equal to -999.
4	missing  lab problem; data values set equal to -999.
5	estimate because of laboratory problem, data values set equal to -999.
6	possibly suspect based on spatial/temporal comparisons with nearby results but 
        for which there is no other basis to believe there is a problem with the sample

Flag 1' applies primarily to anomalously low ambient samples. Ambient background samples 
were generally in the range from 7-10 pptv. Values less than 5 pptv were preemptively 
designated with -999 since anything less than 5 pptv for an ambient sample is unlikely.

Flag 2' applies primarily to ambient background samples and those samples that were 
affected by the plume but still had concentrations below the MLOQ.  Flag 3' was applied 
to any data that was suspect due to field-related problems.  This includes improperly 
connected bags, clips in the open position when they were checked in before laboratory 
analysis, flat bags, and overfilled bags.  Flat bags, or low bags with suspect values, 
were the most common problem in this category.  The reasons for flat bags include the 
sampling program failed to download from the Timewand into the sampler, the sampler 
failed to function properly, or the operator forgot to open the clips. In some cases 
operator error was the cause of the failure of the sampling program to download correctly. 
Flags 4' and 5 were applied to any data that was suspect due to problems with the 
laboratory analysis.  An example of this was clips being open during the purge cycle of 
the analysis resulting in bag-filling and sample dilution.

Flag 6 was applied to concentrations that were judged to be somewhat inconsistent or 
anomalous with respect to neighboring concentrations in space and/or time. However, 
there was no other evidence to suggest that there might be a problem with the result. 
They are included as Valid and it is recommended that these values be accepted for 
use with an awareness of this flag. Samples with this flag were generally identified 
using either the bubble or cartridge time series plots.

