United States - Office of General Enforcement EPA-340/1-80-019
Environmental Protection Washington DC 20460 March 1981
Agency
Stationary Source Enforcement Series
&EPA Correlation of Remote and
Wet Chemical Sampling
Techniques for Hydrogen
Fluoride from Gypsum
Ponds
-------
EPA-340/1-80-019
Correlation of Remote and Wet
Chemical Sampling Techniques for
Hydrogen Fluoride from Gypsum Ponds
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of General Enforcement
Division of Stationary Source Enforcement
Washington, DC 20460
March 1981
-------
GCA-TR-80-76-G
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Division of Stationary Source Enforcement
Washington, D.C.
Contract No. 68-01-4143
Technical Service Area 2
Task Order No. 59
EPA Project Officer EPA Task Manager
John Busik Mark R. Antell
CORRELATION OF REMOTE AND WET
CHEMICAL SAMPLING TECHNIQUES FOR
HYDROGEN FLUORIDE FROM GYPSUM PONDS
Headquarters DSSE
Final Report
By
Howard Schiff
Daniel Bause
John Fitzgerald
Mark McCabe
Dan Montanaro
Verne Shortell
GCA CORPORATION
GCA/TECHNOLOGY DIVISION
Bedford, Massachusetts
-------
DISCLAIMER
This Final Report was furnished to the Environmental Protection Agency
by GCA Corporation, GCA/Technology Division, Burlington Road, Bedford,
Massachusetts 01730, in fulfillment of Contract No. 68-01-4143, Technical
Service Area 2, Task Order No. 59. The opinions, findings, and conclusions
expressed are those of the authors and not necessarily those of the Environ-
mental Protection Agency. Mention of company or product names is not to be
considered as an endorsement by the Environmental Protection Agency.
-------
ABSTRACT
For several years, the Environmental Protection Agency (EPA) has used
the Remote Optical Sensing of Emissions (ROSE) system to characterize the
gaseous pollutants emitted by a variety of point and extended area sources.
The purpose of this program was to extend the data base of this versatile
and promising pollutant sensor by comparing the data generated by the ROSE
system with data generated by standard techniques for the sampling and anal-
ysis of hydrogen fluoride. The program was divided into five phases in-
cluding a literature review, pretest survey, sampling and analytical trials
in the laboratory, preliminary field phase, and the final, collaborative
field phase. The field sampling efforts were conducted along gypsum ponds
at two phosphate fertilizer facilities. For the formal sampling phase, both
the double filter cassette and sodium bicarbonate-coated tube were used for
the point sampling. The point sampling effort was conducted simultaneously
with the operation of the ROSE system. A sampling period of 15 minutes was
compatible with the sensitivity requirements of the analytical methods. The
fluoride collected by the wet chemical methods was analyzed colorimetrically
using a semiautomated method with lanthanum-alizarin complexone for the
colorimetric reagent. Two data reduction methods, a peak area and peak
height procedure, were used to compute the HF concentrations from the spectra
obtained by the ROSE system. In 32 independent tests of comparable ambient
HF concentrations, the overall average HF concentration was 37.6 ppb (ROSE
system, peak area method), 36.1 ppb (ROSE system, peak height method) and
36.4 ppb (wet chemical techniques). The standard deviation between the ROSE
system data and the manual sampling results was 11.9 ppb and 9.7 ppb for the
peak area and peak height computation procedures, respectively.
iii
-------
CONTENTS
Abstract iii
Figures v
Tables viii
Acknowledgments ix
1. Introduction 1
Statement of the Problem 1
Project Phases 1
2. Summary and Conclusions 3
3. Literature Review for Manual Sampling and Analysis of HF 4
Description of HF Sampling Methods 4
Description of Laboratory Procedures for Hydrogen Fluoride
Analysis 6
Conclusions. 7
4. Pretest Site Surveys 8
Introduction ... 8
Site Surveys 8
Conclusions 10
5. Laboratory Phase 14
HF Generation System 14
Presampling Preparation 16
Sampling Procedures 16
Analytical Procedures. 18
Results 19
Conclusion 21
6. Preliminary Field Sampling Phase. .... 25
Presampling Preparation 25
Sampling Locations 25
Results and Conclusions 31
7. Formal Field Phase 34
Introduction 34
Sampling Locations 34
Sampling and Analytical Procedures 34
Calculations 45
Results 66
References . . . 97
Bibliography 98
Appendices
A. Project Participants 99
B. Laboratory Results and Calculations 100
C. ROSE Results 144
D. Sampling Equipment Calibrations 146
iv
-------
FIGURES
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
CF Industries gypsum ponds
Gypsum pond at CF Industries
Gypsum pond at Agrico
Agrico gypsum ponds
HF generation system
HF collection methods and sampling train .
Ion chromatography flow scheme
Sampling points at CF Industries gypsum ponds
Sampling points at Agrico Chemical Co.'s gypsum ponds ....
Line of sight at CF Industries
Sampling line at CF Industries (with light source for ROSE
system in background)
Line of sight at Agrico
Sampling line at Agrico (with ROSE van in foreground)
Sampling train and filter cassette
Sodium bicarbonate-coated tube
HF sampling trains utilized in the final field phase (a) pre-
filter and alkali treated filter (b) sodium bicarbonate
coated glass tube (c) vacuum system for sampling trains. . .
The ROSE optical system
The ROSE system light source and telescope
ROSE van and receiver telescope
Page
9
11
11
12
15
17
20
26
28
35
36
37
38
39
39
41
42
43
44
-------
FIGURES (continued)
Number
20
21
22
23
2 3D
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
Quartz— iodine light source and telescope
Concentration calibration curve for the R(5) line of HF. . . .
(A) Background spectrum, 900-meter path; (B) gypsum pond
spectrum, Agrico, 630-meter path; and (C) subtracted
Original spectrum run — ROSE 43, GCA 32
Background spectrum adjusted for constant water vapor; used
to correct run — ROSE 43 GCA 32
Subtracted spectrum; run number ROSE 43, GCA 32
Original spectrum, run number ROSE 44, GCA 33
Adjusted background spectrum; run — ROSE 44, GCA 33
Subtracted spectrum; run — ROSE 44, GCA 33
Data reduction method - subtracting peak height of background
Spectra of C02 at Agrico (absorption at 2056.7 cm ) ....
Spectra of C02 at Agrico (absorption at 2056.7 cm"*) ....
Spectra of N20 at Agrico
Spectra of C02 - absorption at 4837.25 cm"1 (Agrico runs
41-44)
Spectra of C02 - absorption at 4837.25 cm"1 (Agrico runs
45-48)
Changes in concentration with time at CFI for each manual
sampling site and for the ROSE data based on peak area . .
Changes in concentration with time at CFI for each manual
sampling site and for the ROSE data, based on peak height.
Changes in concentration with time at CFI for the average of
the manual sampling sites and for the ROSE data (peak
area method)
44
46
49
50
51
52
53
54
55
56
57
59
61
62
63
64
65
73
74
75
vi
-------
FIGURES (continued)
Number
39 Changes in concentration with time at CFI for the averages of
the manual sampling sites and for the ROSE data (peak
height method) 76
40 Changes in concentration with time at Agrico for each manual
sampling site and for the ROSE data (peak area method) ... 77
41 Changes in concentration with time at Agrico for each manual
sampling site and for the ROSE data (peak height method) . . 78
42 Changes in concentration with time at Agrico for the averages
of the manual sampling sites and for the ROSE data (peak
area method) 79
43 Changes in concentration with time at Agrico for the averages
of the manual sampling sites and for the ROSE data (peak
height method) 80
44 Comparison of HF concentrations measured by two techniques at
CFI 7/24/79. (Peak area method for ROSE data) 82
45 Comparison of HF concentrations measured by two techniques at
CFI (7/24/79). (Peak height method for ROSE data) 83
46 Comparison of HF concentrations measured by two techniques at
CFI site (7/25/79). (Peak area method for ROSE data).... 84
47 Comparison of HF concentrations measured by two techniques at
CFI site (7/25/79). (Peak height method for ROSE data). . . 85
48 Comparison of HF concentrations measured by two techniques at
Agrico site (7/26/79). (Peak area method for ROSE data) . . 86
49 Comparison of HF concentrations measured by two techniques at
Agrico site (7/26/79). (Peak height method for ROSE data) . 87
50 Composite comparison of HF concentrations measured by two
techniques. (Peak area method for ROSE data) 88
51 Composite comparison of HF concentrations measured by two
techniques. (Peak height method for ROSE data) 89
vii
-------
TABLES
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Testing Schedule for Laboratory Phase
Results of the Laboratory Sampling Phase.
Recovery of HF by Three Sampling Devices
Precision - Laboratory Phase
Preliminary Field Phase: HF Concentrations
Windspeed and Wind Direction
Analysis of Citrate Filters for Fluoride
Preliminary Field Phase: Intersampling Device Precision at
Same Site
HF Concentration Data for Manual Sampling and Analysis by
Spectrophotometric Method
HF Concentration Data for Manual Sampling and Analysis by 1C. .
Results of ROSE Measurements — CFI
Results of ROSE Measurements — Agrico
Windspeed and Wind Direction Data and Ambient Temperature . . .
HF Concentration Data Grouped in Sequence Obtained
Comparison of Manual Sampling Data at CFI (7/24/79)
Comparison of Manual Sampling Data at CFI (7/25/79)
Comparison of Manual Sampling Data at Agrico (7/26/79)
Statistical Analyses Based on Difference Values for Manual
Sampling Data and ROSE Data (Peak Area Method)
Statistical Analyses Based on Difference Values for Manual
Sampling Data and ROSE Data (Peak Height Method)
Pag(
21
22
23
24
29
30
31
33
67
68
69
70
71
72
91
92
93
94
95
viii
-------
ACKNOWLEDGMENTS
The ROSE system measurements and data analyses were carried out by
Dr. William F. Herget, U.S. Environmental Protection Agency, Environmental
Sciences Research Laboratory, Research Triangle Park, North Carolina.
Dr. Herget*s contributions to this report are appreciated.
The cooperation of Mr. William Schimming, Director of Environmental
Affairs, CF Industries, Bartow, Florida; Mr. Harold Long, Manager of Environ-
mental Control, and Mr. Maurice Johnson, Environmental Control, Agrico Chemical
Co., South Pierce, Florida is gratefully acknowledged.
Dr. Jay S. Jacobson and Larry Heller of The Boyce Thompson Institute for
Plant Research at Cornell University supplied advice and the HF generation
apparatus used in the laboratory phase of the program.
The analyses of the formal field phase samples were conducted in the
Agrico Environmental Laboratory by Mr. Ed Germain and Mr. Charles Kinsey.
Their help.is gratefully appreciated.
We appreciate the support and guidance of the EPA Task Manager, Mr. Mark
Antell.
ix
-------
SECTION 1
INTRODUCTION
STATEMENT OF THE PROBLEM
For several years the Environmental Protection Agency (EPA) has used the
Remote Optical Sensing of Emissions (ROSE) system to characterize the gaseous
pollutants emitted by a variety of point and extended area sources. The
ROSE system consists of a Fourier transform infrared (FTIR) interferometer
with telescopic optics and has been installed in a van. The system is used
either with a remotely located infrared light source to make long path (up
to 1.5 km) atmospheric absorption measurements or in a single-ended mode to
measure the infrared emission signal from gases exiting industrial stacks
at elevated temperatures.
For the purpose of developing a technical basis for enforcement action to
abate human health hazards, it may be necessary to determine concentrations
of toxic gaseous pollutants in the vicinity of sources. The ROSE system used
in the "active long path mode" is conceptually capable of evaluating the
breathing zone pollutant concentrations. The purpose of this task was to ex-
tend the data base of this versatile and promising pollutant sensor by comparison
of data generated by the ROSE system with data generated by standard techniques
for the measurement of hydrogen fluoride (HF). This work will enhance the
ability of the EPA to rely, in enforcement actions, upon data generated by the
ROSE system.
PROJECT PHASES
A phased approach was adopted for conducting the project. The five
phases are indicated below and will be presented in more detail in the sub-
sequent sections.
• Literature Review
• Pretest Survey
• Sampling and Analytical Trials in Laboratory
• Preliminary Field Phase
t Collaborative Field Sampling and Analytical Phase
-------
The purpose of the literature review was to determine which techniques
for the sampling and analysis of HF would facilitate the collaborative sampling
program. Three possible sampling procedures (i.e., the double filter cassette,
the sodium bicarbonate-coated glass tube, and the prefLiter and impinger system)
were selected for further investigation.
The Pretest Survey was conducted to accomplish the following:
• Locate two phosphate chemical plants with a geography compatible
with the ROSE system and chemical sampling methods.
• Determine sites at each plant which are adjacent to gypsum ponds
and have an unobstructed path length of about 400 meters. This
would provide a high signal to noise ratio for the ROSE system.
• Determine the feasibility of using the proposed wet chemical sam-
pling methods at the sites.
• Determine the availability of onsite laboratory space and instru-
mentation for the fluoride analysis.
The facilities at both CF Industries and Agrico Chemical Company, located
in the Bartow, Florida phosphate complex, were found to meet the physical
criteria above.
A protocol for the Laboratory Phase utilizing the three proposed manual
sampling trains was developed and implemented. The Laboratory Phase was de-
signed to determine the reproducibility and sensitivity of each sampling method
under controlled conditions of hydrogen fluoride concentrations. The results
of the Laboratory Phase were evaluated and sampling and analytical methods were
selected.
A Test Plan was developed for the Preliminary Field Phase based upon the
results of the Pretest Survey and preliminary laboratory work. The objectives
of this phase were to evaluate the compatibility of the selected sampling pro-
cedures with the sites; to determine the range of ambient HF concentrations at
each site; and to determine whether a minimum sampling period of 16 minutes for
each method was compatible with the sensitivity requirements of the analytical
methods.
A Test Plan for the Formal Collaborative Sampling Phase was then developed
to measure the ambient HF concentrations at the two sites, using both the double
filter cassette and sodium bicarbonate-coated tube and sampling simultaneously
with the EPA ROSE system. The fluoride collected by the wet chemical methods
was analyzed spectrophotometrically with a Technicon Autoanalyzer. Some samples
were also analyzed by ion chromatography (1C).
-------
SECTION 2
SUMMARY AND CONCLUSIONS
The three methods (the ROSE system and the two manual sampling methods)
utilized for the sampling and analysis of HF along the edge of a gypsum pond
gave good agreement. Two data reduction methods were used to compute the HF
concentrations from the spectra obtained by the ROSE system. One method for
determining the HF concentration is based upon elimination of the H20 inter-
ference and determination of the area under the HF absorption line. In the
peak height method, the sample and background spectra were plotted, and the
net peak absorbance of the line center due to HF was measured by subtracting
the H20 absorbance found at the peak maximum for HF. Most of the HF concen-
trations changed by ±3 ppb or less when the peak area and peak height data
were compared. However, the HF concentrations that were apparently higher
at Agrico using the peak area method were found to be within the range of the
point sampling values when the peak height method was used for data reduction.
In 32 independent tests of comparable ambient HF concentrations, the overall
average HF concentration was 37.6 ppb (ROSE system, peak area method), 36.1
ppb (ROSE system, peak height method), and 36.4 ppb (manual techniques). The
standard deviation between the ROSE system (peak area method) and the manual
sampling methods (for any single measurement) was 11.9 ppb, while the standard
deviation based on the ROSE data computed from the peak height was 9.7 ppb.
Both manual sampling methods used in this study, the filter cassette
with the citrate-treated and sodium-hydroxide-treated filters, and the bicar-
bonate-coated glass tubes, were effective for the collection of gaseous HF.
Laboratory results indicated that the two methods collected 100 percent of
the HF generated for each run. The precision, as measured by the relative
standard deviation for replicate experiments, was less than 8 percent for
each manual sampling method. Replicate measurements in the field (prelim-
inary experiments) showed more variation with a between-method relative
standard deviation of 37.4 percent. This increased variability is ascribed
to increased random error.
An analysis of the sources of error for the ROSE method revealed that
the maximum error on any single HF measurement (average of 100 interferograms)
is ±25 percent. This analysis is based, in part, upon the variation in the
peak absorption for the spectra of C02 and N20, since these gases should
have essentially constant concentrations. The error in the ROSE measurement
is also consistent with the observation that on multiple reduction of the
same HF data, the maximum variation on HF concentration or identical runs
was never greater than 25 percent.
-------
SECTION 3
LITERATURE REVIEW FOR MANUAL SAMPLING
AND ANALYSIS OF HF
Prior to the commencement of the collaborative sampling program, it was
necessary to determine the chemical techniques to be used for sampling and
analysis, which would facilitate the comparison of the ROSE system with standard,
wet chemical methods. Ambient sampling for HF is complicated by the low concen-
trations and the reactivity of the compound, as well as the presence of sub-
stances which interfere with the analysis. The selection of the methods to be
utilized was based on the compatibility with the sampling program, including
sensitivity (minimum sampling time), reproducibility, ease of handling, and free-
dom from interferences.
The literature has been reviewed with respect to the above sampling and
analytical requirements of the program. An extensive review of sampling and
analytical procedures for fluorida was published by Jacobson and Weinstein in
1977. This review, coupled with the NERAC computer searches of the post-1976
literature, formed the basis of the literatrue review.
Sampling procedures published by ASTM-* or ISC^ were selected, since they
have been subjected to extensive laboratory and field evaluation. In addition,
consultations with Dr. Jay Jacobson and Mr. Richard Mandl of the Boyce Thompson
Institute for Plant Research were very helpful in the determination of the
sampling procedures to be used. Brief descriptions of the selected procedures
are given below.
DESCRIPTION OF HF SAMPLING METHODS
In addition to the manual sampling methods, automated methods which com-
bined sampling and analysis were also reviewed. These automated methods were
eliminated since the cost of multiple units was relatively high and the units
would be used only for this study.
Prefilter and Impinger Method—ASTM No. D3267
Air is drawn through a short Teflon probe and a citric acid-treated pre-
filter to remove particulate. Two impingers, a standard and a modified
Greenburg-Smith, both containing a sodium hydroxide solution or water, follow
to remove gaseous fluorides. Sampling trains without a prefilter do not separ-
ate gaseous and particulate components. These may contribute additional fluor-
ide ions or complex with collected gaseous fluoride ions and make analysis dif-
ficult. This technique is not readily conductive to short sampling periods for
-------
low hydrogen fluoride concentrations. To overcome this problem, the impinger
solutions must be evaporated to a smaller volume, risking the loss of collected
fluoride.
Double Tape Sampler—ASTM No. D3266
This method automatically separates and collects acidic, gaseous, and par-
ticulate fluoride forms by means of a double paper tape system. Air is drawn
across a citric acid-treated prefilter tape for particulate removal and then an
alkali (sodium hydroxide)-treated filter to remove gaseous acidic fluoride.
The instrument may be programmed for sampling times varying from a few minutes
to several hours. After sampling, the tapes are stored in a compartment which
is protected from fluoride contaminated ambient air. The advantages are auto-
mated collection and ease of sample recovery. The main disadvantage is cost.
Prefilter and Alkali-Treated Filter (Double Filter Cassette)
This method is a modification of the preceding ASTM procedure. The mod-
ification was necessitated by the lack of availability of the automatic double
paper tape sampler. A citric acid-treated prefilter is followed by an alkali-
treated (sodium-hydroxide) filter. The first filter will remove particulate
and the second will remove acidic fluoride gases. Advantages of this method
include ease of sample handling and recovery. Elution of the fluoride content
from the dry filter requires small amounts of water.
Bicarbonate-Coated Glass Tube and Particulate Filter—ASTM No. D3268
A 4-ft borosilicate glass tube, the inside of which is coated with sodium-
bicarbonate, is held vertically above a 47 mm citric acid-treated Whatman 42
filter. Gaseous fluorides are removed by chemical absorption on the wall of
the tube while particulates are drawn through and collected on the filter. An
advantage of this technique is the relative ease of recovery. Collection of
fluoride requires a small volume of eluent which results in a concentrated so-
lution for analysis. In this way, a minimum sampling time, necessary for com-
parison with the ROSE system, can be achieved. The method is also low in cost.
Since the filter follows the gaseous collection device, this technique is recom-
mended for use in the presence of particulate which may react with and remove
HF on a prefilter. Difficulty in handling seems to be the major drawback.
Quartz Tube with Carbonate-Coated Silver Beads
This system works on the same principle as the bicarbonate-coated glass
tube. A quartz tube containing sodium bicarbonate-coated silver beads follows
a separator (a Herpetz cap or heated membrane filter) to exclude large particles.
Advantages are ease of handling and concentration of gaseous fluoride. The
method, however, does not provide complete separation of particulate and gaseous
fluorides. It is also relatively expensive.
The previously described sampling methods have been utilized in the field
for the collection of gaseous fluoride,5-8
-------
DESCRIPTION OF LABORATORY PROCEDURES FOR HYDROGEN FLUORIDE ANALYSIS
Ion Exchange—ASTM No. D3269
o 4
The ion exchange column is ASTMJ and Intersociety Committee approved as
a method for the isolation and concentration of fluoride in a sample. The fluor-
ide ion is preferentially sorbed on an anion exchange resin while interfering
substances and the solvent pass through. A small volume of eluent is then re-
quired to desorb the fluoride. An automated ion chromatograph incorporates the
isolating capabilities of the ion exchange column with a conductimetrie detection
system. Low concentrations of fluoride ions can be measured with a minimum amount
of sample preparation. This reduces the risk of sample contamination or loss of
fluoride during evaporation. The ion chromatograph, however, is not portable.
Samples must be brought back from the field to the laboratory and risk fluoride
loss due to prolonged storage. Fluoride complexed with other species will not
be detected unless the fluoride is converted to the ionic form.
Willard-Winter Distillation—ASTM No. D3269
This technique employs steam distillation from a strong acid, sulfuric or
perchloric, in the presence of silica to separate interfering substances. Fluor-
ide is collected as fluosilicic acid. This method is used only for separation,
not for measurement. If complexing of the fluoride ion; i.e., CaF, Ca2FPOit, etc.,
is suspected, a NaOH fusion is required prior to the distillation and subsequent
analysis.
Spectrophotometric Procedures—ASTM No. D3269
A reagent, composed of an element such as aluminum, iron, thorium, zir-
conium, lanthanum, or cerium, which reacts with inorganic fluoride to produce a
compound or complex with a low dissociation constant, and an indicator dye, un-
dergoes a shift in absorption spectrum in the presence of fluoride. Zirconium-
Eriochrome Cyanine R, Zirconium-SPADNS, and Lanthanum-Alizarin Complexone are
the three commonly used reagents. The first two experience fading when complexed
with fluoride and obey Beer's law over the range from 0 to 1.4 yg F/ml with a
detection limit of the order of 0.02 yg F/ml. The Lanthanum-Alizarin Complexone
reagent differs from the above reagents since there is an increase in absorb-
ance of the solution proportional to the amount of fluoride present. This is
more sensitive and covers a lower range, 0 to 0.5 yg F/ml, with a detection limit
of approximately 0.015 yg F/ml.
Semiautomated Method with Microdistillation—ASTM No. D3270
The sample solution is mixed with sulfuric acid and pumped into the poly-
tetrafluoroethylene coil of a microdistillation device maintained at 170°C.
The acidified sample is carried to a fractionation column by a stream of air.
The fluoride and water vapor are condensed and pumped continuously from the dis-
tillate collector, while the solids and spent acids are removed from the system.
The distillate is mixed continuously with a colorimetric reagent and passed
through the flow cell of a spectrophotometer. The equipment required for this
procedure is commercially available, and this system was employed at Agrico
Chemical Co. for the fluoride analyses.
-------
Titrimetric Procedures—ASTM No. D3269
The sample solution containing an indicator dye; e.g., - Alizarin Red S,
Eriochrome Cynanine R, or SPADNS is buffered at pH 3.0. Upon addition of thor-
ium nitrate, insoluble thorium fluoride is formed. When the end point is reached,
the excess thorium reacts with the indicator dye causing a change in color which
can be detected visually or by instrumental techniques. The method is capable of
high sensitivity; but it is slow and tedious and the results are highly dependent
on the analyst.
Potentiometric Method—ASTM No. D3269
The method requires the use of an ion specific electrode for the measurement
of fluoride. Ionic strength and pH must be controlled, and the sample must be
free from agents which complex fluoride. The potential of the sample in milli-
volts is recorded and converted to yg/ml of fluoride using a calibration curve.
The detection range is 0.019 yg F/ml to 19,000 yg F/ml. Slow response time and
nonlinearity of the calibration curve cause measurements below 0.1 yg F/ml to be
less accurate. Care must be exercised because ion specific electrodes have a
limited life span.
CONCLUSIONS
Three sampling and two analytical techniques were chosen, based on the lit-
erature review. The manual sampling methods using a prefilter and impinger, the
double filter cassette, and the bicarbonate-coated glass tube were investigated
in more detail in the laboratory phase. Fluoride analysis for the laboratory
phase was accomplished primarily by ion chromatography, although some solutions
were also analyzed by a spectrophotometric method using Lanthanum-Alizarin Com-
plexone as the chromotropic reagent. For the preliminary field phase all of the
fluoride analyses were done by 1C, with some of the solutions also analyzed at
Agrico Chemical Co. using the semiautomated method described previously. All
of the fluoride samples were analyzed at Agrico Chemical Co. for the collabora-
tive field phase with some samples also being analyzed by 1C.
-------
SECTION 4
PRETEST SITE SURVEYS
INTRODUCTION
The Task Manager, Mr. Mark Antell, had contacted two sites in the Bartow,
Florida phosphate complex at which the program could be conducted. The two
phosphate fertilizer facilities included CF Industries, Inc., Bartow, Florida,
and Agrico Chemical Co., South Pierce, Florida. To determine the feasibility
of utilizing each facility, site surveys were conducted and the following
criteria were evaluated:
• Physical layout of the gypsum ponds
• Access roads for the ROSE van
• Longest unobstructed line of sight available for ROSE system
• Fluoride concentrations in the ponds
• Meteorological patterns which could affect results
• Availability of laboratory facilities for sample recovery
• Possibility of fluoride analyses being done by the facility
• Availability of electrical power.
SITE SURVEYS
CF Industries, Inc.
The layout of the gypsum pond area is illustrated in Figure 1. For the
measurement of HF by the ROSE system a path length of at least 400 meters was
desirable. Therefore, two of the possible sample lines were located at cool-
ing pond No. 1 (lines A and B) while the third possible line of sight was ad-
jacent to cooling pond No. 2 (line C). Pond No. 2 was eliminated from con-
sideration because there was no electric power available and because there
might be interferences in the form of "hot spots" which might be introduced
by plumes from the phosphoric acid plant across the pond from the expected
southeastern winds. The gypsum stacks on the East side of pond No. 1 would
not interfere at either line A or B when the wind was blowing from the
-------
POND FEED
CF INDUSTRIES
PLANT SITE
SETTLING PONDS
POND
r-jSTACK PONDflSTACK
ELEVATION
Figure 1. CF Industries gypsum ponds,
-------
southeast across the pond. Of the two lines of sight, B was preferred (Fig-
ure 2). This was chosen because the path length for the ROSE system was
longer and some onsite electrical power was available. For line B, extra
mobile generators would be required to provide the remaining power not ac-
cessible onsite.
The F~ concentration of pond No. 1, as measured by plant personnel with
an Orion ion specific electrode, was usually between 8,000 and 9,000 ppm.
The pond has a pH of 1.2, although it is commonly between 1.5 and 1.6. In a
previous report by The Research Corporation of New England (TRC), the ambient
HF concentration measured at the edge of the pond was 20-30 ppb and was con-
sidered high enough for the manual sampling methods and the ROSE system.
Meteorological data (i.e., wind speed and wind direction) are measured
at the plant, and this information would be available to GCA. Laboratory
facilities were not available to permit the onsite analysis of the collected
samples.
Agrico Chemical Co.
The plant geography as shown in Figure 4 is amenable to the use of sam-
pling methods in any of three locations. Both the upper and lower gypsum
stacks (B and C) on the east side of the main cooling pond provide a straight
and flat span with no topographic interferences. The effect of the gypsum
stacks on the ambient HF concentrations was unknown at that time. The acces-
sibility of the upper stack to the ROSE van would be a problem and would
prevent its use during the sampling phase. A grassy road (A) on the west
side of the pond could also be utilized. There was, however, a great deal
of brush in this area which may cause obstructions or alter concentration
levels. The road on the lower stock (line C) was chosen for the sampling
effort (Figure 3).
The fluoride concentration in the cooling pond ranged from 9,000 to
14,000 ppm. The laboratory did have space available for the recovery of ;
samples prior to analysis. In addition, Agrico had offered to analyze the
samples for fluoride using ASTM Method D3270, which is an automated colori-
metric method using a Technicon Autoanalyzer with a micro-distillation unit.
The wind speed and direction are measured at the plant and were available
to GCA personnel. No electrical power could be provided at any of sampling
sites and mobile generators are required.
CONCLUSIONS
The physical configurations of the gypsum ponds at both CF Industries and
Agrico Chemical Co. were amenable to the testing program. At CF Industries
ambient HF concentrations at the pond edge were known to be sufficient for both
the ROSE and the wet chemical sampling and analytical methods. The fluoride
concentrations in the Agrico gypsum pond also appeared to be high enough to
provide sufficient HF concentrations for the ambient measurements.
10
-------
Figure 2. Gypsum pond at CF Industries.
Figure 3. Gypsum pond at Agrico,
11
-------
•POSSIBLE
SAMPLING
LINE
LOWER
STACK
ROAD
B
UPPER
STACK
ROAD
C
AGRICO PLANT
SITE
ELEVATION
Figure 4. Agrico gypsum ponds.
12
-------
It was decided that both gypsum ponds should be tested during the prelim-
inary wet chemical field sampling phase. The bulk of this experimental work
should be conducted at CF Industries, where.ambient concentrations are known
and electricity is available. Representative measurements at Agrico using the
wet chemical methods would characterize the HF concentrations at the pond
edge.
13
-------
SECTION 5
LABORATORY PHASE
The purpose of this phase was to select HF collection and analytical
methods, which would facilitate the comparison of optical and chemical methods
in the final field sampling phase. An HF generator was constructed to inves-
tigate the three sampling procedures selected from the literature review. Ion
chromatography (1C) was utilized for fluoride analysis to determine the
collection efficiencies of the three methods and to identify any interference
problems.
HF GENERATION SYSTEM
The design for the HF Generation System and the injection box was sup-
plied by Dr. Jay S. Jacobson of the Boyce Thompson Institute for Plant Research^
(personal communication of Jacobson and Heller, letter of April 10, 1979). An
HF generation system was constructed as illustrated in Figure 5. Air was pumped
through an indicating silica gel drying trap, a tube packed with glass wool,
and a Whatman 42 filter into heated teflon tubing at a flow rate of 1.5 dscfm.
An aqueous HF solution was pumped at 0.05 ml/min into the heated Teflon tubing
(located in the injection box) through which the filtered air flowed. The in-
jection box was kept at 175°F. The fluoride-laden air was then cooled to room
temperature in an ice bath and divided. Sampling took place at two points
downstream of the flow division. The portion of air that was not sampled was
exhausted to a laboratory hood.
Originally, the intent had been to pump the air stream into a section of
PVC tubing with four sampling ports located 3 feet from the duct inlet and
equidistant from each other. The HF analysis yielded values which were lower
than expected from the amount of HF introduced into the system. The split
stream was used to correct the problem.
The amount of fluoride put into the system is dependent on flow rate and
the concentration of the HF solution, i.e., the concentration of solution in
the reservoir times the flow rate (yg HF/ml x ml/min = yg HF/min). Adjustments
of the air flow rate through the system alters the concentration of fluoride
per unit volume of air, but not the amount of fluoride delivered through the
system. The latter is controlled by varying the aqueous HF solution concen-
trations. The concentration of HF in the air stream is determined by the
following equations:
14
-------
FLOWMETER
AIR
COMPRESSOR
SILICA GEL
DRYING TRAP
GLASS WOOL
DRYING TUBE
WHATMAN 42
FILTER
LOW FLOW
PERISTALTIC PUMP
POINT
OF
SAMPLING
THERMOMETER
TEFLON TUBING
INJECTION BOX
HF RESERVOIR
(NALGENE GRADUATED CYLINDER)
ICE BATH
O°C
(COOL AIR TO ROOM
TEMPERATURE)
EXHAUSTED
TO HOOD
POINT OF
SAMPLING
Figure 5. ...HF generation system.
-------
„_,, ri.3 pg HF ml"1 x ml min"1
pg HF/dsft3 = -^
Vm (dsft3)
where Vm dsft3 is at 77°F and 29.92 in.Hg.
_.. 3 yg HF dsft-3
yg HF/dsm = 0<02832 ft3 m-9
HF (ppb)
0.818 yg dsm-3 ppb"1
PRESAMPLING PREPARATION
All impingers, related glassware, and polyethylene sampling bottles
utilized in the laboratory phase were cleaned with an Alconox solution and
rinsed with tapwater and distilled, deionized water. The glassware was air-
dried and capped with parafilm.
The sodium bicarbonate-coated tubes were prepared as outlined in ASTM
D3268. The tubes were cleaned with detergent, alcoholic KOH solution, and
distilled water. While the inner surface was still wet, a 5 percent (by
weight) NaHCOs solution was poured through the tube to coat the internal
surface. Hot, fluoride-free air (prepared by passing air through coiled
copper tubing heated by a heating tape) was blown through the tube to dry the
sodium bicarbonate on the inner wall.
the prefilter and filters were treated with citric acid or sodium
hydroxide respectively, according to ASTM D3266. The filters were immersed
in the appropriate solution (either 0.1 m citric acid in 95 percent ethanol
or 0.5 N NaOH in 95 percent ethanol and 5 percent glycerin) and dried under
an infrared lamp.
All filters and tubes were sealed until sampling occurred. At the
completion of each sampling run, the filters or tubes were resealed until
recovery.
The dry gas meters were calibrated according to procedures in APTD
0576.
SAMPLING PROCEDURES
The three wet chemical sampling procedures, found to be applicable to
this program, are described below. The collection methods are illustrated
in Figure 6.
Double Filter Cassette
The double filter cassette sampling train is a modification of ASTM
D3266, i.e., a double filter cassette is used in place of the AISA Automatic
Tape Sampler. The constituents of the train were a 37 mm Millipore filter
cassette containing a Whatman 42 filter pretreated with a citric acid solution
16
-------
a PREFILTER AND ALKALI TREATED FILTER
25 MM
PLASTIC
FILTER
HOLDER
SHORT TEFLON PROBE
CITRIC ACID TREATED
PREFILTER
NaOH TREATED FILTER
TO VACUUM
b. SODIUM BICARBONATE COATED GLASS TUBE
T
4ft
•7 MM ID GLASS TUBE
INSIDE COATED WITH
SODIUM BICARBONATE
47MM CITRIC ACID—5
TREATED WHATMAN 42
FILTER
-POLYPROPYLENE
FILTER HOLDER
TO VACUUM
C. PREFILTER AND IMPINGER METHOD
d. VACUUM SYSTEM AND SAMPLING TRAINS
POLYPROPYLENE -»,
FILTER HOLDER
WITH 47 MM WHATMAN
42 FILTER tCITRIC
AOD TREATED)
SHORT TEFLON PROBE
TO VACUUM
STANDARD
6REENBUR8-SMITH
IMPINGER WITH
IOO ML O.IN NaOH
FROM
COLLECTION
SYSTEM
THERMOMETER
METERING VALVE
^-MODIFIED OREENBURG-SMITH
WITH IOO ML O.IN NaOH
SILICA GEL
DRYING TUBE
ORIFICE
MANOMETER
VACUUM
PUMP
Figure 6. HF collection methods and sampling train.
-------
back to back with a Whatman 4 filter pretreated with a sodium hydroxide solu-
tion; a modified Greenburg-Smith impinger containing indicating-silica gel;
dry gas meter and an orifice meter; and a leakless lubricating vane pump.
The sampling rate was 0.5 cfm. Leak checks of all sampling trains were
conducted prior to and after each sampling run to determine that a leak rate
of not greater than 0.02 cfm existed. The cassettes were capped to prevent
exposure to the ambient air. After sampling the inlet and outlet were again
plugged. The used filters were placed in clean sample bottles and 10.0 ml of
distilled deionized water and 0.1 ml of l.ON NaOH were added. The bottles
were sealed tightly until analysis.
Sodium Bicarbonate-Coated Tube • •
Sampling with the sodium bicarbonate-coated tube was performed as
described in ASTM D3268. The train consists of a 4-ft glass tube (7 mm ID)
evenly coated with sodium bicarbonate, connected directly to a 47 mm poly-
propylene filter holder containing a citric acid treated Whatman 42 filter.
The tube was followed by the same drying, vacuum and metering equipment as
described for the double filter cassette sampling train.
Both ends of the collecting tube were sealed until sampling took place.
After sampling, the ends were capped until recovery. The air to be sampled
was drawn through the tube at a rate of 0.5 cfm. Each sampling train was
leak checked before and after the sampling run to determine that no leak
greater than 0.02 cfm existed. The collected fluorides were eluted with
8-9 ml of distilled deionized water. One drop of l.ON NaOH was added and the
solution was diluted to 10.0 ml. The samples were stored in clean bottles
and sealed until analysis.
Prefllter and Impinger
Sampling with the prefilter and impinger was performed according to
ASTM D3267. The constituents of the train, as shown in Figure 4, were a
short Teflon probe, a 47 mm Whatman 42 citric acid-treated filter in a poly-
propylene holder, a standard and modified Greenburg-Smith impinger with 100 ml
of 0.1N NaOH, a modified Greenburg-Smith impinger with indicating silica gel,
a dry gas meter, and a leakless lubricated vane pump.
The sampling rate was 1.0 cfm. Leak checks were performed prior to and
after testing to prove that the leak rate was less than 0.02 cfm. At the
completion of each test, the collecting solution was measured and transferred
to a clean sample bottle.
ANALYTICAL PROCEDURES
The samples from the laboratory phase were analyzed for F~ on a Dioriex
Model 14 Ion Chromatograph. This automated ion chromatograph incorporates
the ion-separating capabilities of the ion exchange column with a conducti-
metric detection system.
18
-------
Ion chromatography (1C) is used to identify and quantitate cations and
anions in solution. Conductimetric detection enables the analyst to monitor
the ion separations, but, until recently,.the eluent background conductance
prevented its use. The conductance of the eluent is removed by the appropri-
ate combination of separator and background suppressor columns. The 1C flow
scheme for anion analysis is illustrated in Figure 7.
The principles of 1C can be illustrated using anion analysis as an
example. A sample is injected into the separator column, which consists of
a strong base anion exchange resin in the bicarbonate form. The anions are
distributed between the resin and the NaHC03~Na2C03 eluent. Separation of
the anions depends upon the degree of affinity of the anion for the anionic
exchange groups on the resin. The anions are eluted from the separator
column in the Na+ form. This solution passes through the suppressor column,
which contains a cation exchange resin in the H+ form. The suppressor column
converts the sodium salts of the anions to their corresponding acids. It also
converts the eluent to H2C03, which has a low conductance. The conductivity
of the anions is measured via the peak height, and the anions are identified
by their retention times. The peak heights are converted to concentrations
by comparison with a calibration curve.
The column system employed for the fluoride analyses consisted of a pre-
column (3 x 150 mm) to remove particulates, strongly retained anions, and
organic species; a separator column (3 x 250 mm) in the HCOs form; and a
supressor column (6 x 250 mm) in the H"*" form to remove the background con-
ductivity of the eluent. The eluent, which was a solution of 0.003 M NaHCOs
and 0.0024 M Na2C03 was pumped through the column at a rate of 150 ml/hr.
The injection loop had a capacity of 100 yl and the sample was introduced
from a 5 ml disposable syringe fitted with a 0.22 ym Millipore filter to
remove particulate matter. A IN I^SOtj solution regenerated the suppressor
column after an 8-hr period.
RESULTS
A series of tests were conducted with each procedure (Table 1). The
sampling period and the approximate HF concentration to be introduced into
the generating system is listed for each trial. The actual HF concentration
is given in Table 2. Sampling runs 1-16 were designed to examine the re-
producibility of the results of each type of sampling device and to provide
information on the sampling period required by each method. Two sampling
trains containing the same sampling method were run simultaneously. Runs
1-6 consisted of two double filter cassette samplers, runs 7-11 utilized
two bicarbonate-coated glass tube trains, and runs 12-16 used two pre-
filter and impinger sampling trains.
A second series of tests were designed to facilitate intercomparisons
of the three methods. Five double filter cassette/sodium bicarbonate-
coated tube runs at 50.1 ppb HF, four runs at 30.8 ppb HF, and three
trials at 18.2 ppb HF were completed. The sampling period was 15 minutes.
In addition, experiments with impinger/bicarbonate-coated tubes and
impinger/double filter cassettes were conducted using 50.1 ppb HF with a
sampling time of 30 minutes.
19
-------
NaHC03
ELUENT
PUMP
SAMPLE
INJECTION
VALVE
R+HC03~
Strong base anion
exchange separator
resin separates
sample anions in
a background of
NaHC03 eluent.
Strong acid sup-
pressor resin re-
moves NaHCOs el-
uent and converts
sample anions to
their acids which
pass unretarded
through the sup-
pressor column.
Conductivity meter
quantifies anion acids
(sample ions) in a
background of dilute
carbonic acid.
SEPARATOR
COLUMN
SUPPRESSOR
COLUMN
(Regenerated
periodically
to remove un-
wanted eluent
ions)
CONDUCTIVITY
METER AND
RECORDER
«^
1
CONDUCTIVITY
CELL
WASTE
Figure 7. Ion chromatography flow scheme (anion analysis illustrated).
20
-------
TABLE 1. TESTING SCHEDULE FOR LABORATORY PHASE
HF range
Sampling concentration
Run No.
1-6
7-11
12-16
17-21
22-30
31-36
37-40
41-43
Sampling methods
Double filter—double filter
NaHC03-coated--NaHC03-coated tube
Impinger — impinger
Double filter — NaHC03~coated tube
Impinger — NaHCOs-coated tube
Impinger — double filter
Double filter — NaHCOs-coated tube
Double filter— NaHCOa-coated tube
period (min)
15
15
30
15
30
30
15
30
(ppb)
50
50
50
50
50
50
30
20
The results of the laboratory sampling phase are presented in Table 2 for
each run. The recovery of the generated HF by each of the three sampling de-
vices is summarized in Table 3. For each amount of HF generated the mean
standard deviation(a) and percent recovery are given for the corresponding
sampling method. The double filter cassette and bicarbonate-coated tube
methods collected 100 percent of the HF generated for each experiment. Re-
plicate experiments indicated that the precision as measured by the relative
standard deviation (RSD, RSD = a/x) was usually less than 8 percent (Table
4) for the double filter cassettes and bicarbonate-coated tubes.
The impinger solutions, however, yielded HF concentrations which were
much higher than the amount of HF generated. It appears that the impinger
solutions or the impinger glassware were contaminated with high concentrations
of fluoride. In these experiments, the impingers were rinsed with deionized
water between runs. To investigate the contamination problem, the impingers
were subjected to a complete washing procedure (including acid rinsing)
between experiements. The HF concentrations found in the impinger solutions
were still higher than the amount of HF generated. This discrepancy between
HF concentrations indicates that some problems exists when the HF is collected
in 0.1N NaOH.
CONCLUSION
The double filter cassettes and sodium bicarbonate-coated tubes were the
most efficient and reproducible of the collection methods which were tested.
It was recommended that these two collection devices be tested further in the
preliminary field phase.
21
-------
TABLE 2. RESULTS OF THE LABORATORY SAMPLING PHASE
N)
NJ
ITT? • L i^l "
HF, ppb determined
tit , ppb
generated Filter-Filter Tube-Tube Filter-Tube
57.5 54
57.5 58
57.3 54
57.3 54
57.3 55
57.3 58
50.1
50.1
50.1
50.1
50.1
50.1
50.1
50.1
30.8
30.8
30.8
30.8
18.2
18.2
18.2
61.5
61.5
61.5
61.5
61. 5U
57. Ob
57.0
58
58
52
57
60
57
50
51
50
51
50
32
30
31
30
19
17
25
59 60
67 63
82 72
62 61
63 70
48
53
42
49
52
28
32
31
29
19
19
18
Impinger-Tube Impinger-Filter Impinger-Impinger
337
210
791
1705
1055
134
374
27
48 35 47
55 165 48
49 953 52
49 1612 50
50 37 49
48 617 50
49
50
56^ I57b
276b
Analyses by Ion Chromatography.
New cleaning process for glasswsare.
-------
TABLE 3. RECOVERY OF HF BY THREE SAMPLING DEVICES
Level of
HF generated
(ppb)
18.2
30.8
50.1
57.4
61.5
x, ppb measured
a
% recovered
x, ppb measured
a
% recovered
x, ppb measured
0
% recovered
x, ppb measured
.a
% recovered
x, ppb measured
a
% recovered
Sampling device
Filter3
20.3
4.0
111.5
30.8
0.96
100.0
49.8
1.43
99.4
56.3
2.59
98.1
Tube3
18.7
0.66
102.6
30.0
2.05
97.4
48.8
2.15
97.4
63.1
3.95
102.6
Impinger
575
1148
163
286
Analysis by Ion Chromatography.
23
-------
TABLE 4. :PRECISION.- LABORATORY PHASE
Group
Filter-filter
Tube-tube
Filter- tube
Filter
Tube
Parameter
Within FI
Within F2
All filters together
Between FX and F2
Within Tj.
Within T2
All tubes together
Between Ti-T2
Within F
Within T
Between F-T
All filters
All tubes
ppb HF generated
18.2
X
20.33
18.7
19.5
a
4.16
0.57
5.14
RSD
20.4
3.04
27.6
30.8
X
30.8
30.0
30.0
a
0.96
1.85
2.65
RSD
3.11
6.08
8.67
50.1
X
50.4
48.8
49.6
49.80
49.4
a
0.55
4.32
4.4
1.40
3.07
RSD
1.09
8.85
8.06
2.81
6.2
57.3
X
55.5
57
56.3
56.3
a
1.97
2.68
2.61
3.3
RSD
3.55
4.71
4.64
5.86
61.5
X
62.8
63.5
63.1
63.1
o
3.30
4.50
3.66
4.73
RSD
5.25
7.08
5.80
7.50
-------
SECTION 6
PRELIMINARY FIELD SAMPLING PHASE
The preliminary field phase was conducted June 13-20, 1979. It
was designed to determine:
1. The compatibility of the selected manual sampling procedures
with the sampling location, i.e., the presence or lack of
interfering substances, sensitivity levels, etc.
2. The range of ambient HF concentrations at the two ponds.
3. If a sampling period of 15 minutes is compatible with the
sensitivity requirements of the analytical techniques.
PRESAMPLING PREPARATION
The impingers, related glassware and polyethylene sampling bottles were
treated as follows:
1. Washed with Alconox solution, rinsed with tap water followed by
distilled deionized water.
2. Scrubbed and rinsed with a 10 percent potassium hydroxide
solution in methanol (alcoholic KOH solution).
3. Rinsed with distilled deionized water.
4. Rinsed with 0.1N HC1.
5. Rinsed with distilled deionized water.
The glassware was then air-dried and capped with parafilm. The required
treated filters and coated glass tubes were prepared as described in the
laboratory phase.
The dry gas meters were calibrated according to procedures in APTD 0576.
The Climatronics Wind Mark III Wind Measuring System was electrically cali-
brated and aligned prior to and after the sampling.
SAMPLING LOCATIONS
The line of sight and associated sampling sites selected for the gypsum
pond at CF Industries are shown in Figure 8. This line of sight was
25
-------
A—i
POND FEED
CF INDUSTRIES
PLANT SITE
EB ffi
SETTLING PONDS
1
POND
1-1 STACK POND I"""! STACK
£ Wj
ELEVATION
Figure 8. Sampling points at CF Industries gypsum ponds.
26
-------
approximately 465 meters in length and was divided into four equal segments.
The sampling sites were situated at the center of each segment.
The line of sight selected for the Agrico gypsum pond was on the east
side lower stack roadway (Figure 9). The line of sight was approximately
600 meters long. It was divided into four equal segments and a sampling site
was situated at the center of each segment. The proposed line of sight on
the western edge of the pond could not be used due to relatively tall shrub-
bery growing in the area between the roadway and the pond. This would inter-
fere with the air flow pattern. The wet chemical sampling probes could have
been elevated above the tops of the shrubbery, however, the ROSE equipment
could not. The wind flow pattern for both sites in past years showed winds
blowing from the southeast.
Sampling Protocol
CF Industries, Inc.—
On the first day, the double filter cassette was utilized at the four
sampling sites to determine the ambient HF concentration. All four sites
were sampled simultaneously. The sampling rate was 0.5 cfm, for a duration
of 15 minutes. Four sets of samples were obtained. The filters were treated
as described in the laboratory phase and were sent to the GCA laboratory for
analysis by 1C. The samples were analyzed the next day, and the results were
transmitted to the field team. The results of these samples (designated Nos.
1-4) are given in Table 5. The values obtained showed that the ambient HF
concentration was at a satisfactory level for the sampling and analytical
methods. Successive samples showed good reproducibility. A gradient along
the line of sight was also shown to be present. The wind was blowing from
the northeast with a speed of 7-10 mph.
The original plan was to sample two sites concurrently with the three
methods, the double filter cassette, bicarbonate-coated tube, and impingers
with the prefilter, being operated simultaneously at each site. Five repli-
cates were to be run. The sampling trains were then to be moved to the next
two sites, and five runs were to be conducted. However, due to electrical
power and equipment constraints, the plan was modified as discussed below.
At each site, the impinger train was operated for 30 minutes at a samp-
ling rate of 1 cfm. Either the double filter cassette or bicarbonate-coated
tube was started simultaneously with the impinger and operated for 15 minutes
with a sampling rate of 0.5 cfm. After 15 minutes, the run was terminated
and the other sampling train was started, and was operated for 15 minutes
with a flow rate of 0.5 cfm. Five replicates were run for each site. The
samples were recovered as described in the laboratory phase, and were returned
to GCA for analysis by 1C. An aliquot of the impinger solution was given to
CF Industries staff.
The windspeed and direction were also obtained (Table 6).
Agrico Chemical Co.—
The revised sampling protocol described for CFI was conducted at the
Agrico Pond. Some samples were analyzed by the Agrico Environmental
27
-------
LOWER
STACK
ROAD
B
UPPER
STACK
ROAD
C
L
AGRICO PLANT
SITE
POND
STACK
POND /\STACK
B^
'STACK
ELEVATION
Figure 9. Sampling points at Agrico Chemical Co.'s gypsum ponds,
28
-------
TABLE 5. PRELIMINARY FIELD PHASE: HF CONCENTRATIONS
Run No.
CFI
1
2
3
4
5
6
7
8
9
10
11
12
rain
13
14
Agrico
1
2
3
4
5
6
7
8
9
10
Date
6/12/79
6/12/79
6/12/79
6/12/79
6/13/79
6/13/79
6/13/79
6/13/79
6/14/79
6/14/79
6/14/79
6/14/79
6/14/79
6/14/79
6/18/79
6/18/79
6/18/79
6/18/79
6/18/79
6/19/79
6/19/79
6/19/79
6/19/79
6/19/79
PPB
Site A Site B
Time F T IF T I
1345 50 29
1500 53 35
1530 65 36
1600
1415 59 41 258 43 42 611
(60) (82)
1500 50 24 19 19 67
(21) (69)
1530 28 23 476 37 36 96
(60) (93)
1605 17 25 18 15
(38)
1000 21 24 (54) 18 17 318
1100
1200
1250
1530
1610
1115 23 21 0 54 32 9
(39) (35) (48)
1200 29 24 0 24 24 194
(27) (27) (30) (50)
1230 38 0 32 23
(26) (52)
1511 27 28 0 39 29
(54)
1545 23 20 22 36 38 29
(33) (55)
1030
1100
1145
1345
1430
HFa
Site C Site D
F T I F T I
71 95
74 . 73
76 79
81
8
(103)
14 17 57 30
32 27 36 17 0
(57)
35 30 208 17 36 154
(51) (34) (64)
14 21 (2) 36 17
21 18 (8) 42 17 0
(103) (54)
35 32 0 19 16 17
(30) (23)
35 25 174 37 24
(28)
39 31 43 39
40 42 46 38
40 37 47 45 409
(50)
3 Ana lysis by ion chromatograph unless noted otherwise, i.e., ( ) *• auto analyzer.
29
-------
TABLE 6. WINDSPEED AND WIND DIRECTION
Date
Agrico
6/18
6/19
CFI
6/12
6/13
6/14
6/14
Runs
A1+B1
A2+B2
A3+B3
A4+B4
A5+B5
C+D1
C+D2
C+D3
C+D4
C+D5
Fl
F2
F3
F4
A+B1
A+B2
A+B3
A+B4
A+B5
C+D1
C+D2
C+D3
C+D4
C+D5
Time
11:15
12:00
12:30
15:11
15:45
10:30
11:00
11:45
13:45
14:30
13:50
15:00
15:30
16:00
Rain
14:15
15:00
15:30
16:05
10:00
11:00
12:00
12:50
Rain
15:30
16:10
WS (Mph)
3
3.5
4
6
5
3 ,
2.5
2.0
5
3.5
10
10
7
8
10
15
10
11
10
9
10
10
14
14
WD (°)
360
Variable
345
360
350
Variable
Variable
Variable
10
35
50
45
95
95
40
40
40
40
40° gusty
40
40
40
25
25
WS Mps
1.3
1.6
1.8
2.7
2.2
1.3
1.1
0.9
2.2
1.6
4.5
4.5
3.1
3.6
4.5
6.7
4.5
4.9
4.5
4.0
4.5
4.5
6.3
6.3
30
-------
Laboratory using a Technicon Autoanalyzer and the semlautomated spectrophoto-
metric procedure (ASTM D 3270). All samples were returned to GCA for analysis
by 1C.
RESULTS AND CONCLUSIONS
The results of the preliminary field phase tests are presented in
Table 5. The data indicate that for a majority of runs the filter and tube
results are comparable when the solutions were analyzed by both the 1C and
the spectrophotometric method. Some of the data for the impinger runs,
however, were not comparable to the other results. As demonstrated in
Table 5, the ambient HF concentrations determined by the 1C and autoanalyzer
vary widely, and results for the impinger .solutions are inconclusive. Since
there appears to be some problem with impinger solutions, this method for the
collection of HF was not considered further.
Several conclusions can be drawn from the results of the preliminary
field phase:
1. A sampling period of 15 minutes was adequate for both the
filter and tube collection methods for measuring the HF
concentration at each of the gypsum ponds.
2. The double filter cassette and the bicarbonate-coated
tubes give more consistent results than the impinger
solutions and were selected for use in the formal field
phase.
3. No interferences were observed when either filter or tube
samples were analyzed by either the 1C or autoanalyzer
methods. A previous ROSE study has found that a possible
interferent, SiF^, was not present in the atmosphere above
the gypsum ponds. 0
4. The citrate-treated prefilters, used in the double filter
cassette, were analyzed for fluoride, and the results are
presented in Table 7. The prefilter was intended to
remove particulate matter and was not supposed to remove
any HF. The results indicate that no fluoride was
collected on the citrate prefilter.
TABLE 7. ANALYSIS OF CITRATE FILTERS FOR FLUORIDE
Sample F~ (pg/mL) Blank (yg/mL) net F~ (yg/mL)
1
2
3
4
0.37
0.38
0.36
0.35
0.36
0.36
0.36
0.36
0.01
0.02
0
0
31
-------
5. The precision between the two manual sampling devices is shown
in Table 8. In most of the runs the quantity of ppb collected
by each method is comparable. The data in Table 8 indicate
that both methods can be used for HF sampling since the quan-
tities of HF collected are similar for samples collected
simultaneously.
32
-------
TABLE 8. PRELIMINARY FIELD PHASE: INTERSAMPLING DEVICE PRECISION AT SAME SITE
Group
CFI
Site A
Agrico
Site A
CFI
Site B
Agrico
Site B
Filter
ppb HF
59
50
28
17
21
I 175
X 35
23
29
27
23
S 102
X 25.5
A-J
HO
19
37
18
18
Z 135
X 27.0
54
24
32
39
36
Z 185
x 37.0
Tube
ppb HF
41
24
23
25
24
137
27.5
21
24
28
30
103
25.8
A?
Ht
19
36
15
17
129
25.8
32
24
23
29
38
146
29.2
d
(F-T)
18
26
5
-8
— *}
38
7.6
2
5
-1
-7
-1
-0.3
1
j.
0
1
3
1
6
1.4
22
0
9
10
-2
39
7.8
-------
SECTION 7
FORMAL FIELD PHASE
INTRODUCTION
The objective of the formal field phase of the project was to compare the
results of the simultaneous measurement of ambient HF levels as obtained by
manual wet chemical sampling methods with the EPA ROSE system. Both sampling
systems, were located along the edge of the gypsum ponds at CF Industries and
Agrico Chemical Co.
The protocol for this phase was determined.by the combined results of the
preceding phases. The constraints of the ROSE van and onsite electrical
power were also considered. Sampling at CF Industries was conducted on July
24th and 25th, 1979 while samples were obtained at Agrico on July 26, 1979.
SAMPLING LOCATIONS
The sampling line of sight was adjacent to each pond as shown in Figures
10 and 12. The ROSE van and light source were aligned visually and the dis- .
tance between them was measured with a laser rangefinder. The line of sight
established at each pond was divided into four equal segments. One manual
sample was situated at or as close as possible to the center of each segment.
The locations designated A, B, C and D, were determined by the restraints of
the terrain and positions of the electrical generators.
AT CF Industries, the line of sight for the ROSE system was 3 feet east
of the wet chemical sampling line (Figure II). At Agrico, the line of sight
for the ROSE system is shown in Figure 13. The positions of the sampling
sites were determined by the configuration of the road. The height of the
inlet of each of the manual sampling devices was at the midpoint of the light
beam but did not interfere with the beam,
SAMPLING AND ANALYTICAL PROCEDURES
Manual Wet Chemical Methods
Two collection devices were used with the manual sampling grains; i.e.,
(1) a filter cassette containing a citric acid-treated prefilter followed by
a sodium hydroxide-treated filter (designated F) (Figure 14), and (2) a sodium
bicarbonate-coated pyrex tube (designated T) (Figure 15). The vacuum/metering
system was a Research Appliance Corporation (RAC) meter control console, which
was calibrated according to procedures delineated In EPA publication APTD 0576.
34
-------
u>
COOLING POND
LIGHT SOURCE AND TELESCOPE
RECEIVER
TELESCOPE
OPTICAL SAMPLING PATH
CF INDUSTRIES
Figure 10. Line of sight at CF Industries.
-------
OJ
Figure 11. Sampling line at CF Industries (with light source for
ROSE system in background).
-------
OJ
-J
FF FROM
UPPER POND >
LIGHT SOURCE
TELESCOP
—^:=-.—AH — -
OPTICAL SAMPLING PATH
v
Figure 12. Line of sight at Agrico.
-------
u>
00
Figure 13. Sampling line at Agrico (with ROSE van in foreground)
-------
Figure 14. Sampling train and filter cassette.
Figure 15. Sodium bicarbonate-coated tube.
39
-------
The filters and tubes were prepared as described in Section 4. The sampling
trains are shown in Figure 16. Regardless of the collection device (filter
or tube), the sampling train was used at all four sample locations. The air
was sampled at a rate of 0.5-0.6 acfm for a sampling period of 16 minutes.
Twenty runs were conducted at CF Industries. All odd numbered runs were
executed with the filter cassettes, and the even numbered runs used tubes
for sample collection.. The initial 10 runs were conducted on July 24, 1979
while the remainder of the samples at CF Industries were collected the next
day. Test samples for 18 runs (No. 21-38) were collected at the Agrico gypsum
pond on July 26, 1979. Two sets of samples were collected with the filter
cassette, followed by one run with the bicarbonate-coated tube. This sequence
was repeated six times. The sampling rate and duration was the same as for
the CFI runs.
After completion of the sampling runs the collection devices were removed
and the samples were recovered as follows:
Filter - The sodium hydroxide filter was placed in a 125 ml LPE bottle,
.10 ml of distilled delonized water and 0.1 ml of IN NaOH were
added. The bottle was capped and swirled.
Tube - Two 5 ml portions of distilled deionized water were poured onto
the inner surface of the tube; the tube was swirled and the
liquid collected In a 125 ml LPE bottled. To preserve the sample,
0.1 ml of IN NaOH was added. The bottle was capped and swirled.
Blank filters and tubes were also subjected to the above procedure.
All samples were analyzed the day after they were collected. Analyses
were done at the Agrico Analytical Laboratory using a semi-automated spectro-
photometric procedure with a Technlcon autoanalyzer system. The remaining
aliquots of the CFI samples were brought back to the GCA Laboratory for
analysis by 1C.
Remote Optical Sensing of Emissions (ROSE) System
A schematic of the ROSE optical system is shown in Figure 17. The light
source-telescope system used for the long-path absorption measurements is
shown in Figure 18. The f/5 scope Is of Dall-Kirkham configuration with a
30 cm diameter primary mirror. The light source is a 1000 watt quartz-iodine
lamp (In wavelength regions where the quartz envelope Is opaque, its thermal
emission provides sufficient energy). The light source and telescope system
Is transported to the measurement site In the ROSE van (Figure 19) and in-
stalled in a locally obtained truck which Is driven to a desired location
(Figure 20); a small generator powers the light source if local power is not
available.
The remainder of the ROSE system is permanently installed in an 8.5
meter van. A telescope identical to that described above collects energy
from the remote light source directly, through a port In the side of the van,
as indicated in Figure 17. The receiver telescope focuses energy at the
interferometer aperture, which is adjustable for compatibility with desired
spectral resoltuion. The interferometer and peripheral equipment comprise
40
-------
a. PREF1LTER AND ALKALI TREATED FILTER
IU SHORT TEFLON PROBE
25 MM' kl L 3»
i? Z5V7 ^ U r r ~ ~ :& — CITRIC ACID TREATED
PLMTIC p 7~ «3 DDF«l|Tr»
FILTER U ^t T™*'*'""
HOLDER [I ^NoOH TREATED FILTER
1
TO VACUUM
b. SODIUM BICARBONATE COATED GLASS TUBE
T
4 ft
j
L 1
47MM CITRIC ACID — >&
<-7 MM 10 OLASS TUBE
INSIDE COATED WITH
SODIUM BICARBONATE
i
_^< — POLYPROPYLENE
•—4 *iiT*m uni n*m
TREATED WHATMAN 42 f=TT~
FILTER Y
TO VACUUM
C. VACUUM SYSTEM AND SAMPLING TRAINS
HECK VALVE
THERMOMETER
••-FROM
COLLECTION
SYSTEM
IMPINGER
WITH SILICA CCl
Figure 16. HF sampling trains utilized in the final field phase (a) prefilter and alkali
treated filter (b) sodium bicarbonate coated glass tube (c) vacuum system
for sampling trains.
-------
VAN WALL
Ni
SOURCE TELESCOPE
t
//
't
t
/
/
t
/.
7
t
S
><
<<
A
\\-L
LIGHT ^X
SOURCE-^
/'\
^ v
3
^Wl
RECEIVER
TELESCOPE
INTERFEROMETER
CALIBRATION
CELL
I/
1 .| 4 APERTURE
Figure 17. The ROSE optical system.
-------
CO
Figure 18. The ROSE system light source and telescope.
-------
Figure 19. ROSE van and receiver telescope.
Figure 20. Quartz-iodine light source and telescope.
44
-------
a standard Nicolet Instrument Corporation Model 7199 FT-IR system configured
to fit into the van. Major components of this system consist of a computer
with 40K memory, dual-density disk with 4.8 million, 20-bit word capacity,
teletype, paper tape reader, oscilloscope interactive display unit, and a
high-speed digital plotter. Maximum spectral resolution achievable with this
system is 0.06 cnT1.
The interferometer itself is mounted on the telescope support structure.
All other systems (except the plotter) are arranged in two 19- inch relay racks.
The general layout of the van is shown in Figure 21. Two beamsplitters,
BaF2 and ZnSe, are available for use in the interferometer. A dual-element,
sandwich-type detector is mounted in a liquid nitrogen dewar. For the 1800
to 6000 cm~* region InSb is used, and HgCdTe is used from 600 to 1800 cm"1;
the two regions are scanned separately. During data collection single inter-
ferograms are collected and stored on disk and then averaged at the end of the
data collection period. The inverse Fourier transform of the averaged inter-
fero grams is then calculated by the computer to produce a single spectrum. It
has been found practical to average about 100 inter ferograms; this requires a
data collection time of 16 minutes. (The signal-to-noise of the spectrum is
improved by the square root of the number of inter ferograms collected.)
Sampling Protocol
The manual collection devices were set into the four sampling trains
and initial system readings obtained. The four samplers and the ROSE
system were then started simultaneously. The sampling and spectral ac-
cumulation proceeded for 16 minutes. All systems were stopped simultaneously.
The final readings for the manual samplers were obtained and the spectra
obtained by the ROSE were checked.
CALCULATIONS
Manual Methods
1. The volume of dry gas sampled is converted to standard conditions,
77°F and 29.92 "Hg. (25°C and 760 mm Hg) .
i PM
Vmstd = dry std ft3 = 537! (Y) . (VM) (PB 4- -)
(29.92)(TM)
Where: Y = dry gas meter calibration factor
VM = sample gas volume, ft3
PB = Barometric pressure "Hg
PM = AH, pressure at DGM "^0
TM = Temperature at dry gas meter, °R (°R = °F + 460)
537°R = 77°F +460
2. Vmstd, dry std. m3 = Vmstd (dry std. ft3) x 0.02832
45
-------
B
// // // // // // // // // // // // // // // // // // // // // // // // y/ // // //
•f/.// {/ li /f //
TRACKING
MIRROR
STORAGE
LN2
STORAGE
DOUBLE DOOR
SOURCE UNIT
STORAGE
GN2
STORAGE
INTERFEROMETER
ELECTRONICS
f
I
t
10KVA
GENERATOR
WORK
BENCH
DIGITAL
PLOTTER
/ /{ Jl 'f (1 // // // // // // // // // // // // Jf // // // // // f/ 1 / /I // Ik
I HOIST I
GAS
HANDLING
SYSTEM
RECEIVER
UNIT
DOUBLE DOOR
A
r/ 77 // // //
CAB AREA
Figure 21. Layout of ROSE system in van.
-------
3. The concentration of HF in the ambient air is determined by:
pg HF/ 3 = yg Tl x V. x 20.006(HF)
18.998(F)
Vrastd (m3)
Where: V, = volume of sample, ml
JLi
4. The concentration in ppb is:
Pg HF/m3
0.818 yg/m37ppb
ppb HF =
Where- 0 818 = 20-006 V8/mn°le x JO9 VL/m3
24.45 yL/ymole x 109 ppb
5. The four results from each manual sampling run were averaged
arithmetically and geometrically.
X Arithmetic
X Geometric - {(XA)(XB)(Xc)(XD)}*
ROSE Method
Calibration of the ROSE system field data is normally done by recording
spectra of known amounts of the gas in question (contained in the calibration
cell shown in Figure 17). The transmittance of the gas sample is related to
the cell length and gas concentration by Beer's Law:
, , . -K(v)CL
where T(V) = e
v = wavenuraber (cm )
C = concentration (ppm)
L = path length (meters)
and K(v) = spectral absorption coefficient (ppm meters) .
The quantity K(v) is determined from the calibration spectra and then used
with the field data to determine the average concentration between the
light source and the receiver unit (again using Beer's Law). In filling
the calibration cell either a few torr of the pure gas are admitted to the
evacuated cell and then the total pressure is brought to one atmosphere with
the addition of air or nitrogen, or a premixed sample is admitted to the cell
to a total pressure of one atmosphere. (A total pressure of one atmosphere
is used so that the spectral lines of the gas sample are pressure broadened
to the same extent as in the field data.)
47
-------
Because of Its high reactivity, HF requires a special gas handling system
for filling calibration cells. Such a system was not available at the EPA
laboratory. Therefore another method, which is based on measuring the area
under the absorption curve of the spectral line in question, was used. The
particular advantage of this method is that the area under the absorption curve
is independent of the spectral resolution used, which thus allows the use of
HF data obtained at low resolution**»12 to be used as calibration data. From
the low resolution data, the relationship between the area under an absorption
line and the gas optical depth (product C x L) was determined (private communi-
cation of D. E. Burch and D. A. Gryvnak, Aeronautic Division of Ford Aerospace
and Communications Corporation), and the resulting calibration curve is shown in
Figure 22.
The R(5) line of HF at 4174 cuT1 was selected for the concentration cal-
culation because it provides the most suitable compromise between maximum line
strength and minimum water vapor interference. The signatures of pure H20,
"clean air," and the gypsum pond (typical) are shown in the 4168 to 4178 cm"1
region in Figure 23. It is seen that there are weak H20 lines at approximately
4173.6 and 4173.9 cm"1. The methods used to eliminate the H20 interference and
determine the area under the HF absorption line are described below.
First, the clean air and a gypsum pond spectra were expressed in units of
absorbance [Log T (v)] so that the spectra could be manipulated arithmetically.
The ROSE system software contains an interactive subtraction program that
allows the difference between two spectra (expressed in absorbance) to be .dis-
played on an oscilloscope and allows the optical depth of the background spectrum
to be varied until the desired amount of water vapor interference is subtracted
from the gypsum pond spectrum. In using this method, it was found that the
selection of the amount of water vapor to be subtracted was subjective and not
reliably reproducible. An objective and reliably reproducible method was
achieved simply by multiplying the background spectrum by a factor that would
make the maximum absorbance of the water vapor line at 4176.4 cm"1 equal in
both spectra (Figures 24 and 25). In this way, the optical depth (C x L) of
H20 was made equal in each spectrum. The.background spectrum was then subtracted
from the pond spectrum and the result converted back to transmittance (Figure
26). (Similar sets of data are shown in Figures 27 to 29). All of these data
manipulations were carried out using standard system software.
This substraction method is effective in eliminating H20 interference,
but does increase the noise level in the spectra and makes the determination
of the base line (line of 100% HF transmittance) a potential major source of
error. This problem was handled by taking a laboratory spectrum of the R(5)
line of HF (which was essentially noise-free), expressing the line in assor-
bance, and then multiplying the line by series of factors chosen so that when
the laboratory line was converted back to transmittance, a series of absorp-
tion lines were obtained that spanned the range of transmittances observed in
the field data. A preliminary estimate of the baseline was determined by mea-
suring the maximum HF absorption (difference between pond and clean air spectra
at the center of the R(5) line in Figures 24 and.25) for each spectrum. Then
a laboratory-generated line was selected that gave the best fit between the
general shape of the field line and the estimated baseline. The final result
is indicated in Figures 26 and 29. The area under the absorption curve was
48
-------
HF (R5) N2 BROADENED (« = .Ok ^JJJJJT)
\o
UJ
ID
o
o
O
CO
CQ
-------
a
a
B
t!73.0 t!73.5 «*17lt.O ^171*.5 «H75.0
NflVENUMBERS
Figure 23. (A) Background spectrum, 900-meter path; (B) gypsum pond spectrum,
AGRICO, 630-meter path; and (C) subtracted spectrum.
50
-------
o
o
t
o _
LJ O
CJ <-
is
o J
I-
166 . 0 117b - 0 417^,0 ^17^.0 ^ 17'6 . 0 ^178.0
WRVENUMBERS
Figure 23D. Pure H_0 (not broadened to atmospheric pressure),
51
-------
o
o
o
O)
o
CO
a
r-
o
CO
o
(O
o
a
RGRSML.0%3
07/26/79
IBsSZtSO
38 4189 4170 4171
4172 4173 4174 41.75 4176 4177 4178
UIPVENUMBERS
Figure 24. Original spectrum run~ROSE 43, GCA 32.
52
-------
Peak at 4176.2 cm"1
is adjusted to same
height as peak in
Figure 2*0
BflCKGROUND
(9DJUSTED)
B41774178
UPVENUMBERS
Figure 25. Background spectrum adjusted for constant water vapor;
used to correct run - ROSE 43, GCA 32.
53
-------
o
O
o
o
03* .
0)
O
O
(0 .
0)
O
O
g
UJ
0)
ZQ
?°
«s
r
2°
ao
o
o
•
(0
CD
a
o
03
o
o
o
o
\
BASELINE ERROR RANGE
+6 ppb
BASELINE
-6 ppb
PGRSWL.043 (BflCKGROUND SUBTRflCTED)
-------
o
a
07/86/79
17i04i57
•8 4lil41704171
417Z4173 4174 4175 4176 4177 4178
WRVENUMBERS
Figure 27. Original spectrum; run number ROSE 44, GCA 33.
55
-------
o
o
o
CO
o
CO
n
8s
ex .
O
en
o
ro
BACKGROUND
(flOJUSTED)
(Note:
Peak at 4176.2 cm'1
is adjusted to same
height as peak in
Figure 27)
38 1169 1170 1171
4172 4173 4174
WRVENUMBERS
4175 4176 4177 4178
Figure 28. Adjusted background spectrumj run--ROSE 44, GCA 33.
56
-------
g
•
a .
a
a
a
•
m .
0)
o
a
•
(O
0)
O
O
BASELINE ERROR RANGE
+6 ppb
BASELINE
o
o
•
N
iu°>
U
ZQ
i-o .
zo
cro
QC •
a
a
•
to
CO
o
o
o
o
•
(M
CO
o
o
RGRSPS.044
(BflCKGROUND SUBTRACTED)
-6 ppb
URVENUMBERS
Figure 29. Subtracted spectrum; run—ROSE 44, GCA 33.
57
-------
then measured directly and the concentration determined from the calibration
curve (Figure 22) and the path length. This procedure was carried out for
each field spectrum. A base line error range is indicated in Figures 26 and
29; the magnitude of this error is discussed elsewhere.
Comparison of the HF concentrations measured by the ROSE system (as
described above) with those measured by point sampling methods indicated
general agreement except for the last set of measurements at Agrico. Here,
four of the last five ROSE measurements were significantly higher than the
highest of the point measurements. The point measurements were generally
consistent from site to site, with site D (closest to the van) having the
highest value. The ROSE measurements for these data sets could only be
correct if very high (^400 ppb) HF concentrations existed in the vicinity of
the van. A close inspection of the ROSE spectra for these last five data
sets were appreciably noisier than previous data.
A slightly different data reduction procedure was therefore tried. First,
the ROSE data for each set (two mornings at CFI and a morning and afternoon at
Agrico) were averaged. This had the effect of improving the S/N ratio by a fac-
tor of three. These average data were then reduced as described previously. The
peak transmittance of each of the four averaged spectra was measured, and using
the "known" HF concentrations, an absorption coefficient was calculated. The
four values obtained were 5.14, 5.04, 5.29, and 5.09 x 10~^ (ppm x meters)"1.
The average of these averages was 5.14 x ID"-* (ppm x meters)~ , and this taken
as the value of K for the R(5) line of HF.
Then, all data were reprocessed as above up to the step requiring com-
puter subtraction of pond and background spectra expressed in absorbance.
Instead, the two spectra were plotted and the peak absorbance of the line
center due to HF was measured by subtracting off the H20 interference (Fig-
ure 30). The lines of zero HF absorption were determined visually. The HF
concentration for each line was then calculated from the abosrbance using the
average K value. The results were that the four HF concentrations that were
apparently high were now within the range of the point values. All other HF
concentrations changed by ±3 ppb or less from the previous values.
Sources of Error
ROSE Method--
Potential sources of error in ROSE system measurements have been studied
extensively under both laboratory and field conditions. The laboratory studies
have addressed system reproducibility and calibration error. The reproduci-
bility of the system was tested by collecting separate sets of interferegrams
(50 in each set) on the same gas sample in a cell. Each set was transformed
to give a single spectrum. It was found that for gas samples whose line widths
are comparable 0\0.2 cm"1—gases like CO and HF) with the instrument resolution
(0.125 cm"1), a variation in line intensity of approximately ±1 percent occurred
between the sets of spectra. Calibration accuracy was tested by filling a gas
cell to the same nominal pressure several times and collecting a set of inter-
fere grams after each fill. The error in this case was determined by the accuracy
to which the pressure gauge In the gas handling system could be read—about ±5
percent.
58
-------
Ln
vo
RUN « 3Q
*
BACKGROUND
Absorbance at line center = 0.061
T = 0.869
KCL =0.140
KL = 5.l4x10~3 x 630 = 3.24
C = 43 ppb
WAVENUMBERS
ft 175
Figure 30. Data reduction method - subtracting peak height of background.
-------
Error in field measurements tends to be greater than in laboratory measure-
ments simply because as the light source is moved further and further from the
van, less energy is collected, and the S/N ratio decreases. The most reliable
test of the system for field measurements is to compare from run to run the
spectra obtained for the gases C02 and N20. Since these species have essen-
tially constant concentrations, comparison of spectra from different runs
gives a measure of the overall instrument performance. Figures 31 and 32 com-
pare eight runs from Agrico. The variation in peak absorption of all runs
(strongest and weakest) is 4.6 percent. Figure 33 shows the results for N20,
where the variation is 5.4.percent. Because of the fall-off in detector sen-
sitivity toward shorter wavelength, the S/N ratio is about four times less at
the region of HF absorption than at the regions of C02 and N20 absorption.
This is evidenced in Figures 34 and 35, where another C02 band (located where
the S/N ratio is the same as for HF) Is shown. Here the maximum variation is
14.2 percent. For the HF measurements the noise Is further increased due to
the subtraction of two spectra.. These considerations lead to a maximum error
on any single HF measurement (average of 100 Interferograms) of ±30 percent.
This error is consistent with the observation that on multiple reductions of
the same HF data, the maximum variation In HF concentration on identical runs
was never greater than 30 percent.
Manual Sampling Methods—
The two manual sampling methods for the collection of HF have been studied
in the laboratory and field and the results of these.experiments were presented
in Sections 5 and 6, respectively. The percent recovery of HF for each method
was shown to be about 100 percent (Table 3). The precision for each sampling
device for the laboratory phase is given in Table 4. For both sampling devices
the relative standard deviation was less than 10 percent for HF concentrations
above 18 ppb. The percent recovery and precision shown in Tables 3 and 4,
respectively, are a reflection of the sources of error in both the sampling
devices and the analytical method. In the laboratory phase, ion chromatography
was used to analyze the samples for HF.
The precision obtained In the preliminary field phase is presented in
Table 8. Again, ion chromatography was used to determine HF in the samples.
For the formal field phase, the samples were analyzed by the colorimetric
method using a Technicon Autoanalyzer. The precision and accuracy of the semi-
automated method have been documented by ASTM-Method D3270. A collaborative
study by nine laboratories using the method for the determination of HF in
vegetation gave relative standard deviations ranging from 4 to 13.4 percent for
different types of vegetation. Replicate analyses of standard NaF solutions
by four laboratories had relative standard deviations of 11.4, 3.9, and 3.0
percent for solutions containing 0.28, 1,41 and 2.81 pg F/ml, respectively.
Replicate analyses of standard NaF solutions by four laboratories showed aver-
age recoveries of 101.8, 101.4, and 100.7 percent for solutions containing
0.28, 1.41, and 2.81 yg F/ml, respectively.
60
-------
o
o
LO
O
O
O
•
O
o
»-»
o
o
in
o>
£0
^,«
zKJ
-------
o
o
in
o
1-1
o
o
•
o
o
r-l
o
o
in
O)
to
oc
00
o
QD
O
O
10
O
o
2056
AGRSWL . Oft 1
AGRSWL.OW
WAVENUMBERS
2057
Figure 32. Spectra of C02 at Agrico (absorption at 2056.7 cm'1).
62
-------
o
o
Is
c
(RGRSWL.Otl
AGRSWL.OtS
AGRSWL.0^3
AGRSUIL.OtS
AGRSWL.O'fce
ftGFISWL.Dl*7
RGRSPS.OtB
WAVENUMBERS
Figure 33. Spectra of ^0 at Agrlco.
63
-------
o
o
o _
o
o
•
o
o
UJO
u .
.-•O
cn •
cc
O
O J
o
o
o
CD
AGRSWL .
ftGRSWL.042
RGRSWL.Ot'i
^836
WAVENUMBERS
Figure 34. Spectra of C(>2 - absorption at 4837.25 cm * (Agrico runs 41-44),
-------
a
C3
•
a
a
a
•
o
a
UJO
CJ .
O
•
o
o
Q
•
O
CO
AGRSWL.O«i5
AGRSWL.O'i6
AGRSWL.O«t7
AGRSPS.O«i8
WAVENUMBERS
4838
Figure 35. Spectra of C02 - absorption at 4837.25 cm 1 (Agrico runs 45-48),
-------
RESULTS
Manual Methods
The results of the sampling program are presented in Table 9, for the
autoanalyzer analyses; and in Table 10 for the 1C analyses. The calculations
corresponding to these tables are presented in Appendix B. The samples were
stored for several weeks prior to the 1C analysis because there were some
instrumental problems with the ion chromatograph. Since the holding time for
fluoride in solution is about 7 days, the 1C data were not compared with
the data from the ROSE method. Instead, the data from the autoanalyzer anal-
yses were used in the statistical analysis.
ROSE Method
The results of the data reduction of the ROSE spectra are given in Tables
11 and 12 for CFI and Agrico, respectively. These tables include the data
based on both the peak area and peak height methods,
Weather Conditions
A summary of the wind speed, wind direction, and ambient temperature is
given in Table 13. The relative humidity for the 3 days wa§ 95 percent.
Statistical Analysis arid Discussion of the Data
The final data set, tabulated in the sequence in which the samples were
collected, is presented in Table 14. Five data points have been deleted from
the manual sampling data. These data correspond with the ROSE data which were
omitted because the spectra were either obtained at lower resolution or the
baseline was too noisy.
Graphical representations of the data are presented in the following
Figures:
• Figures 36 to 39 depict the change in the HF concentration with
time at CFI, as measured by the ROSE and manual methods. For
the ROSE method, the data based on peak area (Figures 36 and 38)
and peak height (Figures 37 and 39) are presented. Figures 36
and 37 indicate the concentrations for each sampling site, while
Figures 38 and 39 present the average concentrations of the sam-
pling sites.
• Figures 40 to 43 illustrate the change in the HF concentration
with time at Agrico, as measured by the ROSE and the manual
methods. Figures 40 and 42 are based upon the calculation of the
ROSE data by the peak area method, while Figures 41 and 43 repre-
sent the peak height method for computing the ROSE data. Figures
40 and 41 indicate the concentration for each sampling site while
Figures 42 and 43 present the average concentrations of the sam-
pling points.
66
-------
TABLE 9. HF CONCENTRATION DATA FOR MANUAL SAMPLING AND ANALYSIS BY
SPECTROPHOTOMETRIC METHOD
Concentration
Date
7/24/79
(CFI)
Mean
7/25/79
(CFI)
Mean
7/26/79
(Agrico)
Mean
GCA
run
no.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Sampling
start
time
0905
0936
1008
1043
1124
1200
1224
1246
1307
1330
0805
0830
0940
1005
1027
1050
1115 .
1145
1205
1225
1040
1106
1131
1150
1210
1230
1252
1311
1528
1550
1608
1630
1655
1721
1742
1807
1827
Site
A
26
26
16
27
32
44
35
46
44
42
33.8
30
19
34
14
32
22
52
17
43
27
29.0
20
21
22
34
34
31
35
37
26
30
31
23
32
34
26
25
26
28.6
B
36
27
46
44
32
36
57
24
48
40
39.0
27
15
58
98
34
106
37
7
54
31
46.7
19
29
27
29
26
29
34
-
33
36
27
-
33
30
30
22
26
28.6
C
37
27
52
52
35
-
50
40
43
30
40.7
36
22
43
25
51
-
47
44
60
17
38.3
_
29
27
32
36
34
35
-
37
32
46
27
32
33
30
29
30
32.6
D
40
52
53
27
32
36
50
30
45
40
40.5
44
51
48
29
75
24
42
24
53
24
41.4
31
40
42
43
45
42
46
45
59
50
43
38
52
47
39
34
35
43.0
rithmetic
mean
35
33
42
38
33
39
48
35
45
38
38.6
34
27
46
42
48
51
45
23
53
25
39.4
23
30
29
35
35
34
37
41
39
37
37
29
37
36
31
28
29
33.3
(ppb)
Geometric
mean
34
32
38
36
33
39
47
34
45
38
37.2
34
24
45
32
45
38
44
19
52
24
34.1
23
29
29
34
35
34
37
41
37
36
36
29
36
36
31
27
29
32.6
Collection
device a
F
T
F
T
F
T
F
T
F
• T
F
T
F
T
F
T
F
T
F
T
F
T
F
F
T
F
F
T
F
F
T
F
F
T
F
F
T
F = filter cassette
T - bicarbonate-treated tube
67
-------
TABLE 10. HF CONCENTRATION DATA FOR MANUAL SAMPLING AND
ANALYSIS BY 1C
GCA
run
Uate No.
7/24/79 1
2
3
4
5
6
7
8
9
10
Mean
7/25/79 11
12
13
14
15
16
17
18
19
20
Mean
A
17
31
-
17
9
43
15
35
22
36
25
31
13
26
25
26
17
48
60
19
28
29.3
Site
I) C
37
30
3
52
5
26
28
11
-
36
25.3
44
10
45
92
26
96
37
2
23
24
39.9
25
24
11
46
-
-
9
25
22
23
21
74
19
57
21
6
-
35
46
30
16
33.8
— Arithmetic
L) average
50
47
-
43
-
25
29
24
10
37
33.1
27
75
31
36
66
30
39
30
43
24
40.1
32
39
7
40
7
31
20
24
18
33
25
44
29
40 '
43
31
48
40
34
35
24
37
Geometric
average
30
32
6
36
7
30
18
22
17
32
23
41
21
38
36
23
37
40
20
27
23
31
Collection
device3
F
T
F
T
F
T
F
T
F
T
F
T
F
T
F
T
F
T
F
T
a
F » filter cassette
T = bicarbonate-treated tube
68
-------
TABLE 11. RESULTS OF ROSE MEASUREMENTS—CFI
GCA
run No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
ROSE
run No.
4
6
7
9
10
11
13
14
15
16
17
18
19
20
21
22
23
24
25
26
HF concentration
(ppb)
Peak area Peak height
a
a
a
43
40
41
39
30
43
34
_b
30
28
27
39
27
30
39
36
41
a
a
a
42
39
40
43
30
43
32
_b
27
28
32
39
28
29
38
38
43
aOmitted; spectrum obtained at lower resolution than
rest of data.
Spectrum too noisy; cannot define baseline.
69
-------
TABLE 12. RESULTS OF ROSE MEASUREMENTS--AGRICO
GCA
run No.
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
ROSE
run No.
30
31
32
33
34
35
36 & 37
38
40
41
42
43
44
45
46
47
48
HF
Peak
21
25
37
33
35
40
35
41
40
43
38
62
59
37
46
46
concentration (ppb)
area Peak height
21
23
32
35
36
41
34
38
a _a
39
41
37
46
46
37
41
38
spectrum; cannot define baseline.
70
-------
TABLE 13. WINDSPEED AND WIND DIRECTION DATA AND
AMBIENT TEMPERATURE
Date
7/24/79
(CFI)
Mean
7/25/79
(CFI)
Mean
7/26/79
(Agrlco)
Mean
GCA
run No,
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
30
31
32
33
34
35
36
37
Sampling
period <
(EDT)
0905-0921
0936-0956
1008-1024
1043-1056
1124-1140
1200-1216
1224-1240
1246-1302
1307-1323
1330-1346
0805-0821
0830-0846
0940-0956
1005-1021
1027-1042
1050-1106
1115-1131
1145-1201
1205-1221
1225-1241
1040-1056
1106-1122
1131-1147
1150-1206
1210-1226
,1231-1247
1252-1380
1311-1327
1550-1606
1608-1624
1630-1646
1655-1711
1721-1737
1742-1758
1807-1823
1827-1843
Wind
iirection
(DEC)
125
120
135
150
145
150
150
135
125
135
137
295
125
140
155
140
145
145
140
140
130
140
140
135
120
125
130
125
125
120
105
125
105
110
110
130
140
145
124.3
Wind
speed
(MPH)
11
11
10
11
16
13
12
12
12
14
12.2
12
12
13
14
13
12
12
12
12
12
12.4
9
11
11
10
10
11
10
9
7
7
8
7
8
5
5
7
8.4
Ambient
Temp.
(°F)
79
82
83
83
83
85
89
89
91
89
85
81
81
82
82
83
84
84
84
85
86
83
85
88
89
89
90
91
92
93
95
95
95
95
95
95
93
93
92
71
-------
TABLE 14. HF CONCENTRATION DATA GROUPED.IN SEQUENCE OBTAINED
Concentration (ppb)
Manual sampling
Date
7/24/79
(CFI)
Mean
7/25/79
(CFI)
Mean
7/26
(Agrico)
Mean
Grand
mean
GCA
run
No.
4
5
6
7
8
9
10
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
30
31
32
33
34
35
36
37
Sampling
start
time
1043
1124
1200
1224
1246
1307
1330
0830
0940
1005
1027
1050
1115
1145
1205
1225
1040
1106
1131
1150
1210
1231
1252
1311
1550
1608
1630
1635
1722
1742
1809
1827
t
A
27
32
44
35
46
44
42
38.6
19
34
14
32
22
52
17
43
27
28.9
20
21
22
34
34
31
35
37
30
31
23
32
34
26
25
26
28.8
31.0
Site
B
44
32
36
57
24
48
40
40.1
15
58
98
34
106
37
7
54
31
48.8
19
29
27
29
26
29
34
-
36
27
-
33
30
30
22
26
28.4
39.1
C
52
35
-
50
40
43
30
41.7
22
43
25
51
-
47
44
60
17
38.6
_
29
27
32
36
34
35
-
32
46
27
32
33
30
29
30
32.3
36.1
D
27
32
36
50 '
30
45
40
37.1
51
48
29
75
24
42
24
53
24
41.1
31
40
42
43
45
42
46
45
50
43
38
52
47
39
34
35
42.0
40.7
Arith-
metic
mean
38
33
39
48
35
45
38
39.4
27
46
42
48
51
45 .
23
53
25
40.0
23
30
29
35
35
34
37
41
37
37
29
37
36
31
28
29
33.0
36.4
Geo-
metric
mean
36
33
39
47
34
45
38
38.0
24
45
32
45
38
44
19
52
24
34.1
23
29
29
34
35
34
37
41
36
36
29
36
36
31
27
29
32.3
34.0
Collec-
tion
device3
T
F
T
F
T
F
T
T
F
T
F
T
F
T
F
T
F
T
F
F
T
F
F
T
F
T
F
F
T
F
F
T
ROSE method
Peak
area
43
40
41
39
30
43
34
38.6
30
28
27
39
27
30
39
36
41
33.0
21
25
37
33
35
40
35
41
40
43
38
62
59
37
'46
46
39.9
37.6
Peak
height
42
39
40
43
30
43
32
38.4
27
28
32
39
28
29
38
38
43
33.6
21
23
32
35
36
41
34
38
39
41
37
46
46
37
41
38
36.6
36.1
F = Filter cassette.
T = Bicarbonate - treated tube.
72
-------
7/24/79
U)
6O
SO
40
_a
O
U
U
i
A
§
8O
10
0839 EOT
A $ O
© -
A ROSE
SITE KEY:
Q'ROSE
• •A
X»B
1152 EOT
* o
1326 EOT
i
1294367
SAMPLING PERIOD
IO
7/25/79
60
SO
^40
ex
o.
O
SSO
UJ
U
20
IO
X(98) X(IO6)
A(75)
234 3678
SAMPLING PERIOD
1214 EOT
9 10
Figure 36. Changes in concentration with time at CFI for each manual sampling site
and for the ROSE data based on peak area.
-------
TO
6O
SO
4O
7/24/79
111
u
20
10
A
9
O859 tOT
A Q O
A ROSE
1152 EOT
SITE K€Y =
Q'ROSE
• * A
X» B
O.C
A« o
1326 EOT
801
1294 9 6
SAMPLING PERIOD
10
17/25/79
X(98)
X(IO6)
6O
5O
40
90
UJ
U
I 20
10
A
3 4 5 6 T
SAMPLING PERIOD
10
Figure 37. Changes in concentration with time at CFI for each manual sampling site
and for the ROSE data, based on peak height.
-------
TO
60
SO
40
O
K 90
O
I
80
10
7/24/79
AVERAGE OF MANUAL METHODS
0859 EOT
1132 EOT
*
1326 EOT
*
3 49 6 7
SAMPLING PERIOD
• 9 K>
sop
7/25/79
60
50
4O
<50
8C
Ul
O
§ to
10
AVERAGE OF MANUAL METHODS
0810 EOT
t
-ROSE
1019 EOT
1214 EOT
849*7
SAMPLING PERIOD
IO
Figure 38. Changes in concentration with time at CFI for the average of the manual sampling
sites and for the ROSE data (peak area method).
-------
7/24/79
O
8
60
SO
40
30
20
10
-ROSE
-AVERAGE OF MANUAL METHODS
O859 EOT
1192 EOT
1326 EOT
I
7/25/79
60
^ 40
a.
o.
O
$30
Ul
O
8
10
AVERAGE OF MANUAL METHODS
0810 EOT
t
1019 EOT
12(4 EOT
3 4 S 6 7
SAMPLING PERIOD
IO
3 4 S « 7
SAMPLING PERIOD
» IO
CHANGES IN CONCENTRATION WITH TIME AT CFI
Figure 39. Changes in concentration with time at CFI for the averages of the manual sampling
sites and for the ROSE data (peak height method).
-------
7O
6O
50
40
30
; Z
I |
; § 20
10
7/26/79
Ar
A
A
SITE KEY =
0 = ROSE
• =A
X =B
O =C
1038 EOT
1311 EOT 1948 EOT
1830 EOT
123456
7 8 M 10 II 12
SAMPLING PERIOD
13 14 15 16 17
Figure 40. Changes in concentration with time at Agrico for each manual sampling
site and for the ROSE data (peak area method).
-------
oo
OL
Q.
<
cr
UJ
u
z
o
o
70
60
5O
40
30
20
10
7/26/79
A
A
1036 EOT
A
Ar
A
O
x
ROSE
1311 EOT
I .
1548 EOT
_t ,
A
SITE KEY =
D = ROSE
• = A
X =B
O =C
A =D
ROSE
7 8 M 10 II 12
SAMPLING PERIOD
1830 EOT
13 14 15 16 17
Figure 41. Changes in concentration with time at Agrico for each manual sampling
site and for the ROSE data (peak height method).
-------
VO
7O
60h
50
1 40|-
o.
O
< 30
Ui
U
§ 20
10
7/26/79
1038 EOT
I
ROSE
AVERAGE OF MANUAL METHODS
1311 EOT 1548 EOT
-t-^v. t ,
7 8 M IO II 12 13 14 15
SAMPLING PERIOD
1830 EOT
16 17
Figure 42.
Changes in concentration with time at Agrico for the averages of the manual sampling
sites and for the ROSE data (peak area method) .
-------
7O
oo
o
Z
O
60-
50
4O-
3O
Ul
o
I 2<>
JO
7/26/79
1038 EOT
ROSE
AVERAGE OF MANUAL METHODS
1311 EOT
1548 EOT
_J ,
1 VV
8 M IO II 12
SAMPLING PERIOD
1830 EOT
13 14 15 16 17
Figure 43. Changes in concentration with time at Agrico for the averages of the manual sampling
sites and for the ROSE data (peak height method).
-------
• Figure 44 plots the HF concentration measured by the ROSE method
(based on peak area) against the HF concentration obtained by the
manual sampling method at CFI (7/24/79). Figure 45 makes a sim-
ilar comparison, however, the ROSE data is based on the peak
height method for calculating the HF concentrations.
• Figures 46 and 47 compare the ROSE data for the peak area and
peak height methods, respectively, with the manual sampling re-
sults at CFI (7/25/79).
• Figures 48 and 49 plots the ROSE data for the peak area and peak
height methods, respectively, against the HF concentrations ob-
tained by the manual sampling efforts at Agrico.
• Figures 50 and 51 compares the overall data sets for both sampling
methods at the two gypsum ponds. The ROSE data is based upon the
peak area method and peak height method in Figures 50 and 51,
respectively.
The ROSE system measures the average concentration of HF molecules in the
30 cm diameter cylinder extending through the atmosphere from the source to
the receiver telescope. The point sampling systems measure the point concen-
trations of HF. Both the.optical and point method measurements were averaged
over 16-minute time intervals for each sampling period. If the HF concentra-
tion were relatively uniform along the sampling path, fairly small differences
would be expected between values obtained at the four sampling sites during a
given sampling period, and reasonable agreement between an average of the sam-
pling site values and a ROSE measurement would be expected. If, on the other
handj there were appreciable concentration gradients along the path, then the
two methods could give widely differeing results without either being
"incorrect." The situations that best illustrate this are: (1) a spatially
small but high concentration HF pocket could slowly traverse the area of a
single point monitor (the result would be a high reading at one site, but no
appreciable affect on the ROSE data); and (2) an extended pocket of high HF
concentration that.slowly passed through the optical path but missed the
point monitors (obvious results).
Inspection of data shown In Figures 36 and 37, with the above in mind,
show the following:
(I) During the sampling at CFI on 7/24 the HF concentration spread
between sampling sites is at maximum about ±35 percent of the
average value for a sampling period. The agreement between the
ROSE and average point values are all within the estimated ROSE
error, and furthermore, each method follows the same up and down
trends.
(2) During the sampling at CFI on.7/25 the fluctuations between point
measurement sites during a given sampling period were much greater
than on 7/24, and the highs and lows vary between sites for different
81
-------
60
o
o
x
»-
UJ
5
UJ
or
UJ
0.
UJ
(/)
o
a
a
a
Z
O
Z
UJ
u
Z
o
o
SO
40
30
20
10
X- FILTER
• = TUBE
0 10 20 30 40 50 60
CONCENTRATION (ppb)-MANUAL (ARITHMETIC AVERAGE AVERAGE OF SITES)
Figure 44. Comparison of HF concentrations measured by two techniques at CFI
7/24/79). (Peak area method for ROSE data).
82
-------
o
o
X
»-
UJ
2
z
o
UJ
X
UJ
a
UJ
c/»
o
a:
i
o
a
a
O
Z
UJ
O
z
o
a
X- FILTER
•= TUBE
'0 10 20 30 40 50 60
CONCENTRATION (ppb)-MANUAL (ARITHMETIC AVERAGE AVERAGE OF SITES)
Figure 45. Comparison of HF concentrations measured by two techniques at
CFI (7724/79). (Peak height method for ROSE data).
83
-------
o
o
I
t-
Ul
UJ
o:
o
a:
i
A
a
a
Z
O
-------
X- FILTER
• = TUBE
'0 10 20 30 40 50 60
CONCENTRATION (ppb)-MANUAL (ARITHMETIC AVERAGE AVERAGE OF SITES)
Figure 47. Comparison of HF concentrations measured by two techniques at
CFI site (7/25/79). (Peak height method for ROSE data).
85
-------
70
o
o
I
h-
UJ
2
UJ
tr
UJ
o.
UJ
O
IT
I
a
a
a
Z
O
<
-------
60
o
o
i
t-
UJ
h-
z
o
UJ
X
UJ
a.
ui
V)
o
o:
i
Z
g
<
a;
»-
z
UJ
o
z
o
o
50
40
30
20
10
X- FILTER
•= TUBE
• X
0 10 20 30 40 SO 60
CONCENTRATION (ppb)-MANUAL (ARITHMETIC AVERAGE AVERAGE OF SITES)
Figure 49. Comparison of HF concentrations measured by two techniques at
Agrico site (7/26/79). (Peak height method for ROSE data).
87
-------
70
60 -
o
? 50
H
Ul
S
UJ
IT
UJ
0.
UJ
CO
o
« 30
a
a
Z
O
20
z
UJ
o
z
o
o
10
X- FILTER
•= TUBE
X(2)
0 10 20 30 40 50 60
CONCENTRATION (ppb)-MANUAL (ARITHMETIC AVERAGE AVERAGE OF SITES)
Figure 50. Composite comparison of HF concentrations measured by two
techniques. (Peak area method for ROSE data).
88
-------
X- FILTER
• = TUBE
'0 10 20 30 40 50 60
CONCENTRATION (ppb)-MANUAL (ARITHMETIC AVERAGE AVERAGE OF SITES)
Figure 51. Composite comparison of HF concentrations measured by two
techniques. (Peak height method for ROSE data).
89
-------
periods. The point measurements thus indicate widely varying HF
concentrations. As might be expected, in this case the agreement
between the two methods is not as good as on 7/24.
The analysis of the measurements at Agrico shows an appreciable differ-
ence from CFI in that a true HF gradient along the sampling path is indicated
by the point measurement (Figures 40 and 41). Sites A and B are generally
the lowest and Site D always the highest (by an appreciable amount) in HF
concentration. The readings at D can be explaiend by the fact that the site
was next to a small stream of liquid leaking from an upper gypsum pond. Also,
it should be rioted that the truck holding the ROSE light source and telescope
was in an area surrounded by pond runoff. The data fall into two categories:
(1) During the first eight sampling periods (1038 to 1311 hours)
the agreement between the methods is excellent.
(2) In the eight late afternoon sampling periods (1548 to 1830
hours) the ROSE data gave HF concentrations always a slightly
higher than the average of the point sampling data.
The original ROSE data for the last six runs at Agrico were reprocessed
a number of times to eliminate the possibility that human error could have
caused the difference between the point and ROSE values. Also, the C02 and
N20 concentrations measured by the ROSE system were compared for the last six
runs. No mistakes in data reduction were found for the HF measurements. The
C02 and N£0 concentrations were all within ±5 percent of each other (respec-
tively) when measured in spectral regions of maximum S/N. In the spectral
region where the signal to noise was about the same for C02 as for HF, the
apparent C02 concentration varied about ±15 percent. The internal checks on
the spectral data provided by (X^ and N20 show that there is nothing abnormal
about the spectra measured by the ROSE system during the last six sampling
periods at Agrico. The consistently higher values obtained by the ROSE system
during the warm afternoon may have been due to the pond runoff in the vicinity
of the light source.
To determine whether any statistical differences existed between the sam-
pling sites at CFI and Agrico, and analysis of variance was calculated for the
data obtained each sampling day (no data were deleted for this analysis). The
analyses are presented in Tables 15 to 17. The results of these analyses in-
dicated no differences among the sites at CFI, however, a significant differ-
ence existed among the sampling sites at Agrico.
Statistical analyses based on the differences between the manual sampling
data and the ROSE values are presented in Tables 18 and 19 for all of the data
and for each day of Campling. The difference between the arithmetic mean of
the manual methods (A) and the ROSE data (R) was computed from the data in
Table 14. Tables 18 and 19 represent the statistical analyses based on the
ROSE data computed by the peak area and peak height methods, respectively.
90
-------
TABLE 15. COMPARISON OF MANUAL SAMPLING DATA AT CFI
(7/24/79)
Date
7/24/79
(CFI)
Source df
Sites 3
Times 9
Error 26
Total 38
F 5%, 3/26
9/26
Start
time
0859
0931
1007
1037
1119
1152
1227
1243
1304
1326
SS
305
935
2374
3616
= 2.98
= 2.27
Site
A
26
26
16
27
32
44
35
46
44
42
338
.5
.5
.6
.0
B
36
27
46
44
32
36
57
24
48
40
390
MSV
101.8
935.7
2374.8
C
37
27
52
52
35
-
50
40
43
30
366
F
1.11
1.63
D
40
52
53
27
32
36
50
30
45
40
405
Total
139
132
167
150
131
.116
192
140
180
152
1499
Conclusions:
a. No significant difference among the sampling
locations.
b. No significant difference among the sampling
times.
91
-------
TABLE 16. COMPARISON OF MANUAL SAMPLING DATA AT CFI
(7/25/79)
Site
Date
7/25/79
(CFI)
Source
Sites
Time
Error
Total
Cf* ovl-
o tax L
time
0830
0942
1010
1027
1049
1115
1133
1154
1214
df
3 1
8 4
23 10
34 16
A
19
34
14
32
22
52
17
43
27
360
SS
,835.8
,220.5
,548.1
,604.4
B
15
58
98
34
106
37
7
54
31
440
MSV
611.9
527.6
458.6
C
22
43
25
51
-
47
44
60
17
309
F
1.34
1.15
D
51
48
29
75
24
42
24
53
24
370
Total
107
183
166
192
152
178
92
210
99
1379
F 5%, 3/23 = 3.03
8/73 =2.37
Conclusions:
No significant differences among the sampling sites
and among the sampling times.
92
-------
TABLE 17. COMPARISON OF MANUAL SAMPLING DATA AT
AGRICO (7/26/79)
Date
7/25/79
Agrico
Source
Sites
Error
Total
Start
time
1038
1106
1130
1149
1209
1230
1251
1311
1548
1607
1632
1704
1732
1746
1809
1830
df SS
3 1885.
56 1506.
59 3391.
Site
A
20
21
22
34
34
31
35
37
30
31
23
32
34
26
25
26
461
4
5
9
B
19
29
27
29
26
29
34
-
36
27
-
33
30
30
22
26
397
MSV
628.5
26.90
C
29
27
32
36
34
35
-
32
46
27
32
33
30
29
30
452
F
23.3
D Total
31
40
42
43
45
42
46
45
50
43
38
52
47
39
34
35
672 1982
F 5%, 3/56 = 2.76
Conclusions:
There is a significant difference among the
sampling sites.
93
-------
TABLE 18. STATISTICAL ANALYSES BASED ON DIFFERENCE VALUES FOR MANUAL
SAMPLING DATA AND ROSE DATA (PEAK AREA METHOD)
Concentrations in ppb
_ _ _ ^
n d(1) Sd(2) (A)( } (peak area) (R/A) (5) Range of (R/A)
All. data 32 -1.2 11.9 36.4 37.6 1.09 ± 0.34 0.53 - 1.70
CFI 7 +0.8 5.8 39.4 38.6 1.00 ± 0.18 0.86 - 1.33
7/24
CFI 9 +7.0 14.6 40.0 33.0 0.93 ± 0.45 0.53 - 1.70
7/25
Agrico 16 -6.8 9.3 33.0 39.8 1.21 ± 0.29 0.83 - 1.68
7/26
_
d (average of differences) = Z(A-R)/n
d (standard deviation of differences) = [I(d-d)2/n-lF
(3) _
(A) = mean of data for manual methods
_
R = mean of ROSE values (peak area method)
(5) -
(R/A) = E(R/A)/n
94
-------
TABLE 19. STATISTICAL ANALYSES BASED ON DIFFERENCE VALUES FOR MANUAL
SAMPLING DATA AND ROSE DATA (PEAK HEIGHT METHOD)
Concentrations in ppb
n d(1) Sd(2) A(3) (peak height) (R/5)(5) (R/A)
All data 32 +0.3 9.7 36.4 36.1 1.04 ± 0.28 0.55 •*• 1.72
CFI 7 +1.0 4.8 39.4 38.4 0.9810.13 0.84+1.18
7/24
CFI 9 +6.4 14.5 40.0 33.6 0.94 ± 0.44 0.55 •»• 1.72
7/25
Agrico 16 -3.6 5.6 33.0 36.6 1.11 ± 0.18 0.92 •»• 1.46
7/26
_
d (average of differences) = Z(A-R)/n
(2)Sd (standard deviation of differences) = [Z(d-!)2/n-lp
(3) _
(A) = mean of data for manual methods
R = mean of ROSE values (peak height method)
(5)
(R/A) = E(R/A)/n
95
-------
The mean HF concentration determined by the manual sampling methods for
all of the data was 36.4 ppb compared to a concentration of 37.6 ppb calcu-
lated by the peak area method of the ROSE system. The peak height method
(Table 19) gave an overall average of 36.1 ppb HF. The standard deviation
of the difference was 11.9 and 9.7 ppb for Tables 18 and 19, respectively,
with the variation at CFI on 7/25/79 contributing significantly to raise the
overall standard deviation.
Theoretically, random errors should give an overall d that is zero or
very close to zero. The small d, based on both methods for computing the
ROSE data, indicates a good correlation between the ROSE system and the
manual sampling methods. In addition, the overall average of the quotient
(R/A) should be about one if the data from the manual sampling methods and
the ROSE system is comparable for any given run. The values for (R/A)approach
one in both tables.
96
-------
REFERENCES
1. Herget, W. F., and J. D. Brasher. Applied Optics U5, 3404-3420 (1979).
2. Jacobson, J. S., and L. H, Weinstein, Journal of Occupational Medicine, _ ,
_19, 79-87 (1977).
3. American Society for Testing and Materials, Annual Book of ASTM Standards,
D3266-D3270, 1978.
4. Intersociety Committee, Methods of Air Sampling and Analysis, American
Public Health Association, 1972.
5. Kommers, F. J. W. Determination of Fluorides. Research Institute for
Environmental Hygiene, Delft, Netherlands. N79-16435/6WP. 1976.
6. Israel, G. W. Evaluation and Comparison of Three Atmospheric Fluoride
Monitors Under Field Conditions, Atmospheric Environment 8:159-166
(1974).
7. Okita, T., K. Kaneda, T. Yanaka, and R. Sugar. Determination of Gaseous
and Particulates Chloride and Fluoride in the Atmosphere. Atmospheric
Environmental 8:927-936 (1974).
8. Glabisz, U., and Z. Trojanowski. Metody Oznaczania Nieorganicznych
Zwiazkow Fluoru W Gazach Odlotowych Z Instalacji Kwasu I Nawozow
Fosforowych Oraz W Powietrzu atmosferycznym. Prace Naukowe Akademii
Ekonomicznej We Wroclawiv No. 9/1137:253-257 (1976).
9. Jacobson, Jay S., and L. 1. Heller. Evaluation of Probes for Source
Sampling of Hydrogen Fluoride. APCA Journal Vol. 26 (H):1065-68 1976.
10. Boscak, V., N. E. Boune, and N. Ostojic. Measurement of Fluoride Emis-
sions from Gypsum Ponds. Draft Final Report. TRC - The Research
Corporation. Prepared for U.S. Environmental Protection Agency. Con-
tract No. 69-01-4145.
11. Randall, C. M. Line-by-Line Calculations of Hot Gas Spectra Including
HF and HC1. SAMSO Report TR-75-288, Air Force Systems Command, Los
Angeles, California, December 1975.
12. Meredith, R. E., and F. G. Smith. Broadening of HF Lines by H2, D2,
and N£. J. Chem. Phys. 60, pp. 3388-3391, May 1974.
97
-------
BIBLIOGRAPHY
Ferraro, J., and J. Basil (editors). Fourier Transform Infrared Spectroscopy,
Applications to Chemical Systems Vol. 2, Academic Press, N.Y. 1979.
Chapter 2, Hanst, Phillip, Trace Gas Analysis; Chapter 3, Herget, William
R., Air Pollution: Ground Based Sensing of Source Emissions.
98
-------
APPENDIX A - PROJECT PARTICIPANTS
The following personnel from GCA, EPA, CF Industries and Agrico Chemical
Company participated in the performance of this task.
GCA Technology Division
Dan Bause
John Fitzgerald
Mark McCabe
Kenneth McGregor
Dan Montanaro
Howard F. Schiff
Verne Shortell
EPA Division of Stationary Source Enforcement, Washington, D.'C.
Mark Antell
EPA-RTP
William F. Herget
CF Industries
Steve Martin
William Schimming
Agrico Chemical Company
Ed Germain
Maurice Johnson
Harold Long
Charles Kinsey
99
-------
APPENDIX B - LABORATORY RESULTS AND CALCULATIONS
1. PRELIMINARY LABORATORY PHASE
a. Laboratory Results
b. Calculations
2. PRELIMINARY FIELD PHASE
a. Laboratory Results
b. Calculations
3. FORMAL FIELD PHASE
a. Laboratory Results
b. Calculations
100
-------
APPENDIX B-l
LABORATORY PHASE - LAB RESULTS
101
-------
c^^
T3/1
73
03
0. 3
/. ^
-- I--
102
-------
u
• ........-,.. ..y
A/A
(Lome. F-
3V I*)
o.
'/ear^
4-88
103
- <5"-/ V-
T
"T
T-
.1
t
i
__i—
I
-------
F" analysis trom H:>i-zs*
-------
37^ /T
J.IO
/.
/.
X
x
3^33
3.SY
Z-I3-T
/•3O
O.JQ
0. f /
fc-H-Zl
<*,(&>
-------
fo
.3V T>S"
3 /
C*v*»JKYweft>'A<•'«. pK.u
ZD
3fk'*- /.'-/6 zv
^TV O /* //
3fbcl fL-lb-F
0
Atcricvti-3
a-
O.S's
K.-H-T
rib
_- /"X?
s-
J
/
i S /
(
I -HO
O.Sfc
O.'i\
6,
,c, J3,
v
.. i .. _. _i
-------
**
380 y
380^
H. 1 1
A 36
i 3813 \R-dtf-7& 0.93
ii' ' i i
/.a/
I- CO
107
-------
C-ic'/-> Wf-
-<±
i.oa
r
$
108
-------
C
I
•""1
0
I
"T
i .—_._...__. __„
I 109
-------
*, r . tit*1'ay I . *VWi//j -.1 '
•4$ .»-a3,.fc*
*ir '-*1 -
.lS\.ar.i&
-------
-------
APPENDIX B-2
PRELIMINARY FIELD PHASE
(a) Lab Results
112
-------
(ott-U
M
fl3
H*
fts
fit
HI
flfc
61
(ill
0 1 2
fill:
on '
no
1 i !;
Tie
in
no
IM;
0,70
•O..S.3
f.os
0-73
Ci,SS
0.18/G,?
;'^t One;,;;*
' .I-H
wiju
r.-.v fr-M'>
/- ^ M"
0.^
Vio\ ^^Ov
6.G1
3.0^/
6.M
0-33L
c.^
0-4?
IM
rjn
ii1 i •:••'
,0^
ran
I&B
110/1
ricA
111
rizn
IBS
IMA
'6
nsfl
HW
riU
i nn
rm
IiU
I l-V- &
.1.1M i;
0-S1
0.11
0,18
0,47
0,37
0,27
0,9?
6.57
C.f/9
Mi'
07)
4 -T
III 'T l| in i||lll 11 III ' Ml
113
-------
Cone. Oy/n / J
V; IB
./'I
Wf
Hi;
'
-------
I
T3
Ts
Tl
T'c
Ti
V
"V <
T14
T15
TK-.
17
Tib
o,7lo
0,42
O.il
0,53
0,48
o/l!
03?
0 ^;
T30
131
TO
T»
TM
ra
TIC
"01
TVZ
Si
TS4'
TS7
TO
'i to
TU
742
VI
'/ ••)
115
Vb'
0,70
0,4^
0,5(0
u.ft
0,7t>/O.Jt
0,4,3
io/
2/,3
'•57
O.t?
0'/ V
-------
V I
• . /\
V -0 ,
vci
v ic
V II
v :r
(i.
I .j
i
<:
-Z3//-/7
Al
0,17
C;^
C.M;
o.r/
*%.6V
I .(
,\ .•
O.bL
1.00
o.-n
116
-------
li/i
us
126
14.1
1ST;
Itfc
H-1 r
S 5
HF
cere,
0.3?
0,3..?
o,vS"
as?
0,37
O.Sf
0.50
0,U
6.56
on
Ml
Co?
0,41
04
DHB
0,15
A/0
0,5'.'
u
K.IOB
117
-------
APPENDIX B-2
PRELIMINARY FIELD PHASE
(b) Calculations
118
-------
cp
r
MM
v£>
6,
tH
73*3
m
A
Ci7MW£ Fil1&&
5.
4.
.3 .
.6
55/5"
If
*Q
Afjfatl Majnlft*
'1
t*M
1,0J 4
.>.c .•" '
•'If
*5Q
I//
£
non
till
v
16.1
-------
ro
o
! ,
,557
I'
'i
l/m
Uag
-Til
Xfc
n.n
r/5
T/7
'M Tj
fltf-*l
S'-Sflt
3/-S7S
£-V//
9-277
3I.&L
6 -9*2
7.ni
74GL
2?-3/3.
J5V/
37
337
'«? 5
36:05 j
3/tfc
V
\S6.35I
7 cm
6-937
.09
.46$ J< .
< .
•*>
./&(* .9
'•5?
f.v^
;/«j
-
fl
^ !
...
.50
.60
SC
,,,
*
tc
JO. I
r-^1-"
.— if- -771
37.^7
13?:^ .
A- W6
~/T7^~/^2fn~7r-
3/. 33 f/77 :/f^>3
1.74?
6-C60
.
I/7T75
?- J3.^5
1,577
i.iiJ^^>!33.^
-------
AH | ft?
7/zr
/./c
/JO
(>pb
l.O
to
/ o
/O
1.0
/.O
itJ&l
/.*
/.I
?.C
J.O
6&.K'J
/
.CM
.049
3Q&S
/./6
//a
/.id
v
i.10-
.6*3*
-------
ao/cif-c
N)
ro
J44
,430
$44
.444
444
.&(?•
.50%
aao
.65'
•W \
•044
.4®
.344
O04
-10
£>
JT_.1
£1
rf: „ •
*3/53
JflOJ
^.-?/
-20.7
-------
NJ
OJ
..: y*
ff
••*#
ftM
fit
pb I Mac fb-lbc.
ppb
33.7?
=£^=5
rt-T
,< .51
/
-------
GCA/TECHNOLOGY DIVISION
BURLINGTON ROAD, BEDFORD, MASSACHUSETTS 01730 / PHONE: 617-275-9000
JOB NO £ """
SHEET OF_
BY
A 4
T
cf-t . r
/Y A /*' o c>
o
o
A ^/ V ^ <
A yiy c9 r; /^
/i /t// o- v-i ^o ti
T /^O'dl V f/f~ "$?O'/I
A /'' /.o/ //!•"
& '** *- o o
T ft L-K O Yfl(
10 A /'^ a,6>t> ?,;G
7 ^ v,^ tf-r/^ V^//
/V ? o J c? £3>^ <2
r
-------
Ml
JOB MA
PROJECT
SUBJECT
GCA/TECHNOIQGY DIVISION ««A
BURLINGTON ROAD, BEDFORD, MASSACHUSETTS 01730 / PHONE, 617-275-9000
— -T/^1
J^ ,
8MFET OP
BY
BATE
CH'K. BY
DATE CH'K.
-— ' -"*""~~: §'K. CH'K BY
> '*
q(> *
O
O
O
,^
0
y
-------
GCA/TECHNOLOGY DIVISION ••* SHEET—OF-
BURLINGTON ROAD, BEDFORD, MASSACHUSETTS 01730 /. PHONE: 617-275-9000 BY~
JOB NO DATE.
A
A ,<30~) <" /^<
g ,0^^ ?v
T 2-/ -^
126
PROJECT CH'K. BY_
SUBJECT /\ .S\ DATE CH'K..
O
T /<*''//' '<^l V-6
//I
in
-------
APPENDIX B-3
FORMAL FIELD PHASE
(a) Lab Results
127
-------
P../
7.5-^f
iW<:
y.mr
128
-------
../*=.•».?..<:
X,
,C»_¥.?
f / X.V
in ni
15^7
,03?
129
-------
7S~
S3
y&3
'?/
8O(
069
. f *
99 I
/
20
FS/
r'rao
'\F5t
, \F.&^
* $
T ,'
9-'
<5 S
,70
= C .Hj^-*-*->
,6 &
P.
i¥
Ft,
r
y.63
^A*L..
.>
-------
131
-------
HF
ir/p
ample
5Q r
/Fl
/Fa .
Ft i
X FT
"
PI
Fiz
FI3
•••'F15"
Fit
'•pn
AT3
•flto
1.13
3.35
i-S'8
Ull.t7
Lie?
.'Sfc
ibZ
Z.025,17
5-53
net pno
431
42f
430
17
n
rz
132
rt
T7
IB
n
TIC
Til
/.w
.97
^.0?
z.y
,13
1.13
0 ,«?/o.
-------
HF
iij"
-^ i %. \/
Tl3v
TK '
TIT '
Tib
TH
TIE
Til
T30
Cfai
TM
T23
T25
rat
T37
T2£
nn
T3C>
Til
T33
T33
T3f
T2c-
IJJ
Tjt
TJ7
T3^
I.04//.04 A
. /*
, |-°l i
1.04/1.03 i
7, l.23/l,2t . j
/ '
•f l.oz '
/ I.OS"
I >
sl. C.-77/Q,t2
7 f,2t
< J.jf (l.
ol?3^B//c.
/ 1.12
^ MS"
V 1,03/I.OS1
^ i.ts-
J o.w
/ 1.31
V 1,05
^ |.«?/IX7
^ 1.34
/
17 , 2.U/3.0?
^ l.fl
• I.2C
\y i ^7
|.Zi
/ i.zt
V ^7/0,12
/ 2 .S3
Tfl v
^
T40 ^
TxJ )
^;
; 0,7t
1.17
&T3~~~\£~8I fc>
. .••. — - J^in^ |^f f ^*"
o,z?fe
0.7Q
P,?3
l,o/
0..13
0,37
I.I3//./3
JXL.
1,511
C-2S/Q.2S
/
V /'°X , , I,
N^) mtettankf4«f\
rWft./^5^.VitVofSppm J
435 .
1,30
0.43
o,«/o.a«"
/ v
o,#?
-
133
-------
APPENDIX B-3
FORMAL FIELD PHASE
(b) Calculations
134
-------
01
i, Of^|.£k:
1 1
v!
•/r
: 1
OU
: lit
I/ >
rr» -ni> •
, /a£«:>,.rtMf,
s
c
b
fS
f3~
r'5
f/o-
F//
FiS
t
c
R
B
C
9-
.Xf
2/5j
7.3//,
i
j
3.n\\
7.716!
i /P. 65! /5.f£'
; 5Z7? | ^.33:
5. 03
3f.fl
3.3?
fjo
Ail
1..1
31.51
I i
: 3/.^ ;
i
. •vt* J2--
35.05
V/.05
fo.to
34./1
nsi
•J5./3
: 5o. /
-------
fHJTO
CO
¥\
W
\\ t\
'• c
i P
; \3 . A
; J e
' ! e
16 . ft
c£ i il
R
t
C
P
j C
D
g.OfO
.a//
//.aa/! ^.'d\ ntf 51.1C
UK; 35-3^; 35.0?; ¥J.?7
^.J7o,
75 =
/f.^!(
7.lofti
7.7/7
fit
<5"/: ^/./J
ns/
'W3 t"c).?7. fa*
y. 75 a/. ?? x. ti
¥.15
11.51
at.n
fv.?j
ufficoo
UIS \ 3t./0.
31. IS
?•<# .sin"': ?7.w:
6.11 tt.fy-
Lit 31.13'- IS.'
-------
P&RIOO
= Vie
/. i \ /
A/ ft
6
C
D
^^w r^
e
c
D
.&]
C
fU
^
^7
£53
..07..
7tt
yy.v,.
(•fw v U.'
r*^=r
1.063 \~
=*fc
IS'13
f.000
5.303
6.10
n./o
33-5?
J3.7J
£J.3*
aj.05
33.19-
».fr
35.
5. oy
37.'
If. 8-3
¥!.&
21.37
So.//
39.96
-------
7/* Ay
'JL\ A
a
c
3)
f\ ^">?
S '. c
c r v
l\-\l
1.11'j .&}{
5.^ 2/Jo
^ '' 1>-I^ 1/< / :•/ •'
y.
V .JC-7
f-
-o.7*
i .......
36.
138
-------
^U
vo
i
ft
6
e
ffl
A
ass"
4fr
^m
n.i6 I 11.01 \
n.ff 3?.
-------
iff -'
.fa*
, c
c :
10
U-
-f/d .l-o? 5.5'3?' aj-i'; . 2\.r.l
•<./
\*6
ft.C~. yr./f 5A5
;'.;,/7 j././?. ^-'.^
<•'• i5 ; yy, /»
'//
;. t ?
:.?/?
; t*.ffb.= na
^'•l- , S7.T3 : #.££
.ytt \ 7.s>5
,23£ I 7.323
ja.^l i7.0
37. ?/
-n
0 V
11 ll.fc £5.-)?
s/.ss
37 5/
l./l f.,'i //.i?
-------
flrff
. c
i "*>
i
! A
: 6
;_«L>_i_.ui«-j_.
T37 -»11
/< i »»
//
D
*
c
.5*7
•3/0
.PY?
-7;'l ; 11.0^ /|.63
/i.'/i-:: 7;.;7 I »,>-/->
n:r:i
3-3531 /3.
Iff*
a/.
3-333
59.73;
* Hl.ll
= 25.0?
3o.j'o
a/. JO
J5.0
15? 7r.7$" ' K.s? :
. -'^j8?; a?.?;
I //.# n.^ :5;.f7
I .3o3 ; /.3/ .' /.^(J
: ?.2.? ; 3^.^5 : /^".^
5. /5
5:55;
Ait^f
-------
I ^Lt -'
N3
^~ ' . .fl£T : Ik
i c*] c
: -I
as
; 31
i c.
i 3
J
v^'-^'W ®
-.„:•- . /tx&-
&/
P~:'^/f» ; •tfi' " •'
.5./V' 5". 1:."
-J^o d. ',/
l.l
I (,.323[
U.57.75
^cr.ffi
X7//
^-35
5^05
V.95
-------
D*fe
ft-
U)
6
C
^:
!
I ^./5«i
rfe*
ftt
wl
aiooj
*1H*
-------
APPENDIX C
ROSE RESULTS
144
-------
ftREA
KTg
&Y
SogTEAtT
5.1.
~Tvw» 'is
/ i
ROSE
Vai i 3/ai •
CF 1
3
t
S
I- "I)
8 f.VJ
r.«0 |i
•0 '
-------
APPENDIX D
SAMPLING EQUIPMENT CALIBRATIONS
I. Dry Gas Meters for Laboratory Phase
II. Dry Gas Meters and RAC Control Box for Preliminary
Field Phase
III. RAC Control Boxes for Final Field Phase
146
-------
APPENDIX D-l
DRY GAS METERS FOR LABORATORY PHASE
147
-------
T«St
/?/>
30.23
H
i
t
*mW-jfJ*Al.'
-------
-------
APPENDIX D-2
DRY GAS METERS AND RAC CONTROL BOX
FOR PRELIMINARY FIELD PHASE
150
-------
.1 _ -.. ._f7A'.T!?._.
it t *j . . _^t
-------
jJ! G M CMtW'ofJ
....... 4..
I
152
I _...).
-------
-------
APPENDIX D-3
RAC CONTROL BOXES FOR FINAL FIELD
SAMPLING PHASE
154
-------
-.. . Box No..
T'jl
.Ha rone trie pro-.iucc, P( v!L_fn. Hg Dry gas rrtftcr Ho.
7I T ;"~ T~/'
Or1f!i-
Calculations
AH
0.5
1.0
?.o
4.0
6.0
T.o~"
i U
AN
T3T6
0.0368
0.0737
0.147
0.294
0.43)
0.588
Y
*w ^b (td * 4
y /p AM \ L
46°
0.0317
(tw + 400) a
y • Ratio of accuracy of wet test .*••.- tor to dry test meter. Tolerance • i 0.01
Orifice preciurc differential that oiv(^s 0.75 cfm of air at 70° T and 29.92
"
inches of mcrcory, in. HjO. "c'crancc • * 0.15
155
-------
I •-! , i-—' •j»7£4tft
-------
-------
i • .<«:• ;-. cjr f,f .,.,
-------
*.J.\
E
'
p»a.
i
i-
xy/1 /<•/
27
W.T
Met
'..2JK
r"'
! i-
I <•
";
).
-------
n
A/"-)
Date
cA/?7
.Barometric pressure,
Orifice
manometer
setting,
AH,
in. H20
0.5
Gas volume
wet test
meter
Gas volume
dry gas ^
meter
Box No..
in. Hg Dry gas meter No.
Temperature
Wet test
Meter
tw»
°F
tfl-.O
Dry gas meter
Inlet
Outlet
Average
95-. 4 2? i/
Time
0,
I min
i:
JL
1.0
•r.
nn.'i.
0-003.
2.0
10
/o, v^-'i
ti'i.u
, 429
4.0
10
I'LL.
/CO.
6.0
10
8.0
10
10 A \ b
Average
(Logo
JLool
QiQW
o.i^i
Calculations
AH
0.5
1.0
2.0
4.0
6.0
8.0
AU
on
TJTe
0.0368
0.0737
0.147
0.294
0.431
0.588
Y
^ Pb (td + 460)
pb * ife) (*« + 46°)
AH*
0.0317 fiH f(tw + 4CO) ol2
Pb (td + 4GO) [ Vw
Y • Ratio of accuracy of wet test meter to dry test meter. Tolerance « * 0.01
« Orifice pressure differential that gives 0.75 cfm of air at 70° f and 29.92
inches of mercury, in. HgO. Tolerance • * 0.15
160
-------
,
TTemp.
4 «
.
4*-
P'G,.
FIMA
\06/
.1. '-/
w.r
W.T
w.r
.f "•- -f'-^i
.5 Toy
•P.4.T,
^51 *j
ff*
/
2i^
£9
^ .-w.ah.
V/' •!-.
3")
SS"
X.
9
•o.
4,
/Oh
JQ.2.
^3 2., 7H
/054
^A
\
3.
79,
Pf
_.
/.o
./.O.
no
fit
0.
-------
/£
s&c.
D.a.M
2
i«
is
MA */
1 ••(
:o
,
"7
•'; T
.
,-/ C-73
W$
/;•/.
BAR...
"is-
°F
w.r
w.r
w.r
"P. 4.17
IM.\
/7
••+--
T
.jj..
I;'/' ./f
\<
162
ai
73
lO
-=/-
767.
/67
//$:
///, 7
/LL.
119
119
lAI
CSV
w.r./vr
-------
Date
--> / £-_5 / •/• i
.Barometric press
Orifice
manometer
setting,
Ail.
in. H20
0.5
1.0
2.0
4.0
6.0
8.0
Gas volume
wet test
meter
ft3
5
5
10
10
10
10
ure, Ph • ' in. Ho
Gas volume
dry gas
meter
X23^
£.11$
iV.&f
10. J'/rt
/0.3/G
/o- tf?
Box
Dry gas meter
Temperature
Wet test
Meter
°F
k?
LI
69
62
h S
& 0
Dry gas meter
Inlet
°F
98J
101. f
12 Q
I/%J
US
Hf.f
Outlet
°F
n
yt.v
9?
SL^.p
Average
td,
°F
§^-/
•?0.z.
/03-f
loi.l
/oo,Y
fS-J- \ 99- /
Average
Nn ^J>_J u
flrt.
Time
o,
min
/3.I?
7-21
/3,lf
/ ' i0^}
^J lJ*2
JL%L
Y
•?£3
*\l(£)
AHp
/-8fV
/. 830
' J / ^ *r"x /
/•H /.?63
^rl /. f-J^r
SJ
/•9/0
r
,00 f-
'001
•005
'OOf
-ooz
•one
oo9
.035^
,o/y
.09?
JOG
•o'f'T
Ll^L
Calculations
AM
0.5
1.0
2.0
4.0
6.0
8.0
AH
13.6
0.0368
0.0737
0.147
0.294
0.431
0.588
Y
\ Pb (td t 460)
Vd(Pb * faTg) (tw * 460)
AHg
0.0317 AH f(tw + 460} fl]2
Pb (t
-------
s f
D.&.M
a
0
2
0
12.
Q
J,
r
6
A
,5
-9,/J
.9-1
&ZJ37.
ft 9. 7V
"f"
9 'JO -
flMAJ
%
&£:_L3- L
7M
•p. a.
W, T.
Av@~
26..
'j?
-ZsM
W.f
=,/ 9
164
A/
fo
7?
t£L
We
#9
//•
f(9
..
f/T-
...2.?...
.__.
7?
r?
..£?
f.£
f&
/"X
67
fj.
9o..
/ae -
-------
wm-w.
I. REPORT NO.
EPA-340/1-80-019
REPORT DOCUMENTATION
PACE
4. ttend SuMIIU
Correlation of Remote and Wet Chemical Sampling Techniques for
Hydrogen Fluoride from Gypsum Ponds
7. Author<«> H. F. Schiff, D. Bause, J. Fitzgerald, M. McCabe, D. Mon-
tanaro, and V. Shortell
1. Recipient**
J. Report Date
June 1981
WO*
GC.A-TK-an-7fi-G
». Performing Onjenlutlon Name and AddreM
GCA Corporation
GCA/Technology Division
Bedford, MA 01730
10. Prolect/Taek/Worh UnH No.
Task OrcW SQ
11. ContrecMQ or QrantfQ) No.
(068-01-4143
(6)
IS. Sponsoring Organliatlon Name and Addrett
Division of Stationary Source Enforcement, Headquarters
Office of Enforcement
U.S. EPA
1*. Typo of Report 4 Pwtad Covered
Final
March 1979 to June 198
14.
Wfl <5lri n c/frin
Ik. Supplementary Net**
Project Officer
Mark R. Antell
401 M Street SW
Washington, D.C.
(202) 755-8137
1C AMrect (Limit: 200 word*)
For several years, the Environmental Protection Agency (EPA) has used
the Remote Optical Sensing of Emissions (ROSE) system to characterize the
gaseous pollutants emitted by a variety of point and extended area sources.
The purpose of this program was to extend the data base of this versatile
and promising pollutant sensor by comparing the data generated by the ROSE
system with data generated by standard techniques for the sampling and anal-
ysis of hydrogen fluoride. The program was divided into five phases in-
cluding a literature review, pretest survey, sampling and analytical trials
In the laboratory, preliminary field phase, and the final, collaborative
field phase. The field sampling efforts were conducted along gypsum ponds
at two phosphate fertilizer facilities. For the formal sampling phase, both
tliu double filter cassette and sodium bicarbonate-coated tube were used for
the point sampHnj;. The point sampling effort was conducted simultaneously
with the operation of the ROSE system. A sampling period of 15 minutes was
compatible with the sensitivity requirements of the analytical methods. The
fluoride collected by the wet chemical methods was analyzed colorlmetrically
using a semlautomatcd method with lanthanum-alizarin complexone for the
colorimetric reagent. Two data reduction methods, a peak area and peak
height procedure, were used to compute the HF concentrations from the spectra
obtained by the ROSE system. In 32 Independent tests of comparable ambient
HF concentrations, the overall average HF concentration was 37.6 ppb (ROSE
system, peak area method), 36.1 ppb (ROSE system, peak height method) and
36.4 ppb (wet chemical techniques). The standard deviation between the ROSE
system data and the manual sampling results was 11.9 ppb and 9.7 ppb for the
peak area and peak height computation procedures, respectively.
It. Document Analycl* i. Descriptor*
Keywords;
Hydrogen fluoride
Sampling methods
Air pollution
b. Identlflera/Opan-Cnded Term»
Remote sampling
Correlation
a. COSATI FleM/Qroup
Availability Statement
Unlimited
19. Security CI*M (Thl« Report)
Unclassified
JO. Security Claw CThlt Peg*)
Unclassified
21. No. of
165
22. Price
(tee ANSI-U9.il)
S«« fnttructlont on ft*v*r*«
165
OPTIONAL PONM 272 (4-77)
(Formerly NTIS-3S)
Department of Commerce
------- |