EXPERIMENTAL OBSERVATIONS OF NON-METHANE HYDROCARBONS

AND NOX SOURCES IN URBAN AND RURAL LOCATIONS

By

Henry William Wallace IV

A dissertation submitted in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

WASHINGTON STATE UNIVERSITY College of Engineering and Architecture

July 2013

To the Faculty of Washington State University:

The members of the Committee appointed to examine the dissertation of Henry William Wallace IV find it satisfactory and recommend that it be accepted.

______B. Thomas Jobson, Ph.D, Chair

______Brian K. Lamb, Ph.D

______Timothy M VanReken, Ph.D

______Heping Liu, Ph.D

ii DEDICATION

I dedicate this dissertation and the educational achievement that comes with it to my mother and to the loving memory of my father. I know she is proud of me and that he would be.

iii ATTRIBUTION

This dissertation addresses the sources of nonmethane hydrocarbons NMHCs and NOx in urban and rural environments by developing instrumentation to measure C2 to C12 VOCs and by making experimental observations. It is organized into 6 major chapters: introduction (Chapter 1), instrumentation development and testing for measurement of

Volatile Organic Compounds (VOCs) (Chapter 2), rural observations of NMHCs and

NOx through a forest canopy (Chapter 3), wintertime observations of NMHCs, CO, and

NOx in an urban environment and comparison to vehicle emissions models (Chapter 4 published in Atmospheric Environment as Wallace et al. (2012)), and summertime observations of NMHCs, CO and NOx (Chapter 5), and conclusions (Chapter 6).

Henry W. Wallace IV was the primary author of this dissertation. Dr. B Thomas

Jobson, Dr. Timothy VanReken, Dr. Brian K. Lamb, and Dr. Heping Liu have contributed to this work by editing and providing guidance to the quality of the content.

Dr. B. Thomas Jobson made many intellectual contributions to the design and implementation of the VOC preconcentratioin system during numerous brainstorming sessions throughout the author’s time at WSU. The phase modulation circuit that supplys power to heat the adsorbent tubes was designed and constructed by Fred Schuetze at the

Electronics Shop which is a part of WSU’s Technical Services. The nitrous acid source, which serves as the radical source during the photochemical chamber oxidatioin experiments was designed and constructed by Claudia Toro. Mrs. Toro also assisted with the re-plumbing of the VOC preconcentration system. The photochamber was constructed by a number of people including Dr. B.Thomas Jobson, Jacob McCoskey,

Claudia Toro and myself. The PTR-MS and CO measurements reported in this

iv dissertation were made by Matthew H. Erickson, Jacob McCoskey (Chapter 4 only), and

Dr. B. Thomas Jobson. Figure 3.1 was a modified from a file that was generated by Dr.

Shelley N. Presseley and Figure 3.5 was taken with permission from VanReken et al.,

(Submitted). The jNO2 values used in chapter 3 were provided by Dr. Barry Lefer and Dr.

James Flynn. Output for the vehicle emissions models was provided by Dr. Brian K.

Lamb and Dr. Joseph Vaughn (Chapter 4: MOBILE6.2) and Dr. Rick C. Hardy, Dr.

Jennifer L. Cole, Dr. Sarah M. Strahan, and Dr. Wei Zhang (Chapter 4: MOVES). The organic aerosol mixing ratios presented in Chapter 5 were measured and by Dr. Chen

Song and were downloaded from the ARM database.

v EXPERIMENTAL OBSERVATIONS OF NON-METHANE HYDROCARBONS

AND NOX SOURCES IN URBAN AND RURAL LOCATIONS

ABSTRACT

by Henry William Wallace IV, PhD. Washington State University July 2013

Chair: B. Tom Jobson

Non-methane hydrocarbons (NMHCs) and NO + NO2 (NOx) are precursors for photochemical smog which includes particulate matter and ozone.

NMHCs and NOx have both biogenic and anthropogenic sources. This work characterizes sources of NMHCs and NOx by developing and deploying instrumentation for the quantification of NMHCs emitted from vegetation and spark ignition engines. A two channel volatile organic compound (VOC) preconcentration system was designed and constructed to measure C2 to C12 VOCs with a gas chromatograph-ion trap mass spectrometer GC-ITMS and a gas chromatography-flame ionization detector.

Observations from three field experiments are used to characterize sources of VOC and

NOx, and the impact these pollutants have on local air chemistry. Focus is given to mobile sources emissions as they are an important, often dominant, source of NHMCs and NOx in urban environments. Morning rush hour CO to NOx ratios of 4.2 ± 0.59 mol/mol and 6.9 ± 1.1 mol/mol are reported for Boise, ID and Sacramento, CA, respectively. VOC to CO ratios for both Boise, ID and Sacramento, CA are reported as well. Toluene to CO ratios during the morning rush hour were observed at 2.4 ± 0.04

vi nmol/mol in Boise, ID and 4.7 ± 0.34 nmol/mol in Sacramento, CA while benzene to CO was 1.3 ± 0.02 nmol/mol and 1.1 ± 0.05 nmol/mol in Boise, ID and Sacramento, CA, respectively. A thorough evaluation of the US EPA’s vehicle emissions models

MOBILE6.2 and MOVES is presented. The analysis shows that MOVES represents a significant improvement over MOBILE6.2. The preconcentration system was used quantify VOCs emissions from automobiles in an urban environment. Emissions measured from mobile sources are compared to predicted values from the models and the

MOVES model was found to greatly improve the prediction of CO to NOx when compared to the older MOBILE6.2 model. Biogenic sources of VOCs and NOx are examined in a rural location using a profiling sample inlet to spatially resolve chemical gradients in a forest. Evidence of mixing in the canopy has been explored using the

Leighton relationship and the gradients presented can be analyzed to determine sources and sinks in the canopy.

vii TABLE OF CONTENTS

ABSTRACT ...... vi

TABLE OF CONTENTS ...... viii

LIST OF FIGURES ...... xiii

LIST OF TABLES ...... xxviii

CHAPTER 1: INTRODUCTION ...... 1

1.1. NMHC MEASUREMENT TECHNIQUES ...... 4

1.2. DISSERTATION OVERVIEW...... 6

1.3. REFRENCES ...... 10

CHAPTER 2: INSTRUMENT DESCRIPTION AND DEVELOPMENT ...... 18

2.1. NOxy INSTRUMENT DESCRIPTION ...... 18

2.1.1. Principle of Measurement ...... 18

2.1.1.1. Nitric Oxide (NO) ...... 19

2.1.1.2. Nitrogen Dioxide (NO2) ...... 20

2.1.1.3. Reactive Odd Nitrogen (NOy) ...... 22

2.1.2. Sample Handling ...... 23

2.1.3. Calibration and Zeros ...... 24

2.1.4. Instrument Response ...... 25

2.1.5. Artifacts and Corrections ...... 27

2.2. PRECONCENTRATION GC-ITMS/FID ...... 28

2.2.1. Instrument Description...... 28

2.2.1.2. Adsorbent Traps ...... 38

2.2.1.3. Trap Temperature Control ...... 40

viii 2.2.1.4. Water Trap Temperature Control ...... 41

2.2.1.5. DAQFactory Control Program ...... 42

2.2.2. Principle of Measurement ...... 43

2.2.2.1. FID ...... 43

2.2.2.2. ITMS ...... 48

2.2.3. Sample Handling ...... 50

2.2.3.1. Water Trap ...... 50

2.2.3.2. VOC Pre-concentration ...... 52

2.2.4. Calibration Gases ...... 53

2.2.4.1. GCMS Calibration Gas ...... 54

2.2.4.2. PTR-MS Calibration Gas ...... 54

2.2.4.3. SPECTRA Calibration Gas ...... 55

2.2.4.4. Gasoline Standard ...... 56

2.2.4.5. Zero ...... 56

2.2.5. Method ...... 57

2.3. CHAMBER EXPERIMENT ...... 58

2.3.1. Characterization of Chamber Mixing ...... 58

2.3.2. Exponential Dilution and PTR-MS Intercomparison ...... 61

2.3.3. Gasoline Oxidation ...... 63

2.4. SUMMARY AND FUTURE WORK ...... 66

2.5. REFRENCES ...... 71

2.6. CHAPTER 2 TABLES ...... 80

2.7. CHAPTER 2 FIGURES ...... 90

ix CHAPTER 3: VERTICAL PROFILES THROUGH A FOREST CANOPY DURING

CABINEX ...... 123

3.1. INTRODUCTION ...... 123

3.2. EXPERIMENTAL ...... 125

3.2.1. Site Description ...... 125

3.2.2. Gradient Sampling Methodology ...... 126

3.2.3. Measurement Methods ...... 127

3.2.4. Comparison Between GC-IMS and PTR-MS Data ...... 130

3.3. RESULTS ...... 131

3.3.1. Meteorology ...... 131

3.3.2. Trace Gas Data Time Series ...... 133

3.3.2.1. Ozone ...... 133

3.3.2.2. NO and NO2 ...... 133

3.3.2.3. Volatile Organic Compounds ...... 133

3.4. AIR MASS ORIGINS...... 135

3.4.1. Low and High NOx ...... 136

3.5. VERTICAL GRADIENTS THROUGH A FOREST CANOPY ...... 137

3.5.1. Measurements Overview ...... 137

3.5.2. BVOC and oxidation product gradients ...... 142

3.6. GCMS DATA ...... 144

3.6.1. Comparison of Monoterpene Speciation to Historical Record ...... 144

3.7. HO LOSS FREQUENCY ...... 145

3.8. SUMMARY ...... 147

x 3.9. REFRENCES ...... 149

3.10. CHAPTER 3 TABLES ...... 156

3.11. CHAPTER 3 FIGURES ...... 159

CHAPTER 4: COMPARISON OF WINTERTIME CO TO NOx RATIOS TO MOVES

AND MOBILE6.2 ON-ROAD EMISSIONS INVENTORIES ...... 184

4.1. INTRODUCTION ...... 184

4.2. EXPERIMENTAL ...... 188

4.2.1. Site Location ...... 188

4.2.2. Measurements ...... 189

4.3. Results ...... 191

4.3.1. Time Series of Observations ...... 191

4.3.2. Wind Rose Plots ...... 191

4.3.3. Diel Variation of CO, NOx Mixing Ratios and VMT ...... 192

4.3.4. Emission Ratios Inferred From Measurements...... 194

4.4. CONCLUSIONS ...... 202

4.5. REFERENCES ...... 205

4.6. CHAPTER 4 TABLES ...... 210

4.7. CHAPTER 4 FIGURES ...... 212

CHAPTER 5: CARES FIELD EXPERIMENT ...... 223

5.1. INTRODUCTION ...... 223

5.2. EXPERIMENTAL ...... 228

5.2.1. T0 Site ...... 228

5.2.2. Instrumentation ...... 228

xi 5.3. RESULTS ...... 230

5.3.1. Meteorology ...... 230

5.3.2. Time Series ...... 231

5.4. OH REACTIVITY ...... 232

5.5. EMISSION RATIOS ...... 233

5.5.1. CO to NOx ...... 235

5.5.2. VOCs to CO ...... 237

5.6. INVESTIGATION OF A METRIC FOR PHOTOCHEMICAL ACTIVITY ... 239

5.6.1. Removal by Reaction With OH ...... 239

5.6.2. Photochemical Age...... 241

5.6.3. Photochemistry of Nitrogen Oxides and OA ...... 243

5.6.4. O3, NOz, and OA ...... 245

5.7. SUMMARY ...... 247

5.8. RERENCES ...... 249

5.9. CHAPTER 5 TABLES ...... 260

5.10. CHAPTER 5 FIGURES ...... 262

CHAPTER 6: SUMMARY AND CONCLUSIONS ...... 278

xii LIST OF FIGURES

FIGURE 2.1. COMPARISON OF ROOM TEMPERATURE NO2 AND HONO ABSORPTION CROSS

SECTIONS IN THE REGION OF THE BLUE LED OUTPUT. NO2 CROSS SECTION DATA FROM

BOGUMIL ET AL., 2003 AND HONO FROM STUTZ ET AL., 2000...... 90

FIGURE 2.2. CALIBRATION OF THE NOXY SYSTEM...... 91

FIGURE 2.3. PSD FOR THE NO, CHANNEL1 (BLACK) AND THE NO2 CHANNEL2 (GRAY). THE

SIGNAL GOES TO WHITE NOISE AT 0.3 HZ. THE INSTRUMENT IS NOT CAPABLE OF

RESOLVING ATMOSPHERIC COMPONENTS WITH HIGHER FREQUENCY THAN 0.3 HZ......

...... 92

FIGURE 2.4. ILLUSTRATION OF GC-IFMS/GC-FID ASSEMBLY...... 93

FIGURE 2.5. A PLUMBING SCHEMATIC FOR THE GC-ITMS SYSTEM INCLUDING TWO

SEPARATE CHANNELS, ONE FOR LIGHT HYDROCARBONS (CH1, RED TRACE) AND ONE

FOR HEAVIER VOCS INCLUDING AROMATICS AND BIOGENICS (CH2, BLACK TRACE).

THE FLUSH VALVE IS PLUMBED WITH CHROMATOGRAPHIC GRADE HE THAT IS USED TO

PURGE BOTH CHANNELS (BLUE TRACE)...... 94

FIGURE 2.6. THE COLD TUBE WATER TRAP INLET FOR THE VOC PRE-CONCENTRATION

SYSTEM. THE VALVES ARE INDICATED BY RED CIRCLES AND THE FLOW PATH IS

INDICATED BY BLACK TRACES. AS DRAWN, THE COLD TUBE WATER TRAPS ARE BEING

PURGED WITH 400 SCCM OF DRY N2 AND THERE IS NO FLOW THROUGH THE SAMPLE

FLOW VALVES...... 95

FIGURE 2.7. THE HELIUM FLUSH VALVE IS REPRESENTED BY A RED CIRCLE AND IS

PLUMBED TO THE MULTIPORT FLUSH VALVE. AS DEPICTED, THE HELIUM FLUSH VALVE

xiii IS IN THE “ON” STATE WHICH ALLOWS THE CLEAN CARRIER GAS TO PURGE THE

SYSTEM...... 96

FIGURE 2.8. SHOWS THE COLD BLOCK ASSEMBLY. ALL OF THE MAJOR COMPONENTS ARE

VISIBLE WITH THE EXCEPTION OF THE COLD PROBE FROM THE FLEXI-COOL IMMERSION

COOLER, WHICH IS HIDDEN IN THE COLD BLOCK...... 97

FIGURE 2.9. THIS IS A PLUMBING DIAGRAM OF THE PRE-CONCENTRATION SYSTEM DURING

THE CONDITION WATER TRAP MODE. THE FLOW PATH FOR CHANNEL 1 IS SHOWN BY

RED TRACES, CHANNEL 2 IS BLACK TRACES AND THE HELIUM FLOWS ARE SHOWN BY

BLUE TRACES. ABOVE THE PLUMBING DIAGRAM ARE THE VALVE STATES, EITHER “A”

OR “B” ALONG WITH THE INTERNAL CONNECTIONS FOR THE RESPECTIVE VALVES......

...... 98

FIGURE 2.10. THE PLUMBING AND VALVE POSITIONS DURING SAMPLE MODE. FLOW

THROUGH CHANNEL 1 IS INDICATED WITH RED TRACES. BLACK TRACES INDICATE

CHANNEL 2 FLOW PATHS AND THE BLUE TRACES ARE HELIUM CARRIER AND PURGE.

THE VALVES ABOVE THE PLUMBING DIAGRAM SHOW THE VALVE POSITION ALONG

WITH THE CORRESPONDING INTERNAL TRACES...... 99

FIGURE 2.11. A PLUMBING DIAGRAM FOR COLD/HOT FLUSH MODE. THE COLOR CODING

AND VALVE CONFIGURATION ARE THE SAME AS IN FIGURES 5 AND 6. THE DIFFERENCE

BETWEEN COLD AND HOT FLUSH MODES ARE THE ADSORBENT TUBE TEMPERATURES

ONLY...... 100

FIGURE 2.12. THERMAL DESORPTION MODE PLUMBING. CHANNELS AND VALVE STATES

INDICATED THE SAME AS THE OTHERS IN THIS SECTION...... 101

xiv FIGURE 2.13. A PLUMBING DIAGRAM OF THE SYSTEM IN ANALYSIS/BACK FLUSH MODE.

RED INDICATES CHANNEL 1, BLACK CHANNEL 2 AND BLUE IS FOR HELIUM CARRIER

LINES. THE VALVE STATES ALONG WITH THE INTERNAL CONFIGURATIONS ARE

INDICATED BY THE THREE CIRCLES ABOVE...... 102

FIGURE 2.14. A DRAWING OF THE TENAX GR TRAP USED IN CHANNEL 2 (NOT TO SCALE).

THE TENAX GR AND SILANIZED GLASS WOOL ARE SHOWN IN THE TUBE. THE

THERMOCOUPLE IS ALSO SHOWN PASSING THROUGH THE 1/8” TEE AND THE

THERMOCOUPLE WHICH PASSES THROUGH THE TEE, ONE GLASS WOOL PLUG AND

MEASURES THE TUBE TEMPERATURE AT THE TENAX TUBE INTERFACE...... 103

FIGURE 2.15. ADSORBENT TRAP HEATING CONTROL DIAGRAM...... 104

FIGURE 2.16. THE PROFILES OF 5 HEATING RUNS FOR CHANNEL 1 (GRAYSCALE SQUARES)

AND CHANNEL 2 (BLUE TRIANGLES). THE INITIAL TEMPERATURE RAMP, WHICH STARTS

AT TIME = 3 S IS OUTLINED IN RED...... 104

FIGURE 2.17. SCHEMATIC OF THE PLUMBING USED TO CALIBRATE THE FID ...... 105

FIGURE 2.18. CHROMATOGRAMS OF THE SCOTTY GAS FOR THE 100 (BLUE TRACE), 250

(GREEN TRACE), AND 500 (BLACK TRACE) ΜL SAMPLE LOOPS...... 105

FIGURE 2.19. EARLY ELUTING PEAKS. BLUE IS THE 100 ΜL LOOP, GREEN THE 250 ΜL

LOOP, AND BLACK THE 500 ΜL LOOP...... 106

FIGURE 2.20. LATE ELUTING PEAK SHOWING GOOD CHROMATOGRAPHIC SHAPE. BLUE IS

THE 100 ΜL LOOP, GREEN THE 250 ΜL LOOP, AND BLACK THE 500 ΜL LOOP......

...... 107

FIGURE 2.21. PLOT SHOWING THE LINEARITY OF THE FID ON A PER MOLECULE OF CARBON

FOR THE SCOTT MARRIN, INC. GASES (BLACK CIRCLES), SCOTTY ALKANES AND

xv ALKENES (GREY CIRCLES), PROPYNE (OPEN CIRCLES) AND ACETYLENE (OPEN STARS).

...... 108

FIGURE 2.22. PLOT OF THE FID RESPONSE FOR EACH COMPOUND. THE DASHED LINE IS THE

AVERAGE SLOPE OF THE SCOTT MARRIN, INC. GASES. THE ASTERISK INDICATES THE

SCOTT-MARRIN CALIBRATION STANDARD WHICH HAS A STATED 2% ACCURACY......

...... 109

FIGURE 2.23. CHROMATOGRAMS OF THE SCOTTY GAS DESORBED AT 300 (BLUE), 335

(BLACK), AND 370 (GREEN) OC...... 110

FIGURE 2.24. A TOP VIEW OF THE ITMS SHOWS THE MAIN COMPONENTS OF THE GC-ITMS.

TAKEN FROM THE VARIAN MANUAL ...... 111

FIGURE 2.25. THIS CHROMATOGRAM OF THE CO-ELUTERS ISOPRENE AND TRANS-2-

PENTENE SHOWS THE ANALYTIC POWER OF THE ITMS. THE TRACES SHOWN ARE: THE

TIC (TOP BLUE), M/Z = 70 (MIDDLE BLACK), AND M/Z = 67. WHAT APPEARED TO BE

ONE PEAK IS ACTUALLY TWO AND BY INTEGRATION ONLY THE UNIQUE IONS, THE TWO

CO-ELUTEING COMPOUNDS CAN BE QUANTIFIED...... 112

FIGURE 2.26. A SCHEMATIC OF THE EXPERIMENTAL SETUP FOR THE WATER TRAP

TEMPERATURE TEST...... 112

FIGURE 2.27. INTEGRATED AREAS FOR EACH CONDITION TESTED. A SUMMARY OF THE

CONDITIONS TESTED IS IN TABLE 3...... 113

FIGURE 2.28. CHROMATOGRAMS OF CLEAN NITROGEN GAS DESORBED AT 280 OC (BLACK

TRACE) AND 200 OC (BLUE TRACE). THE INSTRUMENT WAS OPERATED IN EXTERNAL

IONIZATION MODE FOR THIS TEST...... 113

xvi FIGURE 2.29. A CHROMATOGRAM OF THE GCMS CAL GAS TAKEN DURING THE CARES

STUDY...... 114

FIGURE 2.30. POST CARES MULTIPOINT CALIBRATIONS FROM APPROXIMATELY 0.5 PPBV

TO 2.5 PPBV. THE INSTRUMENT WAS OPERATED IN INTERNAL IONIZATION MODE......

...... 114

FIGURE 2.31. A CHROMATOGRAM OF THE PTR-MS CALIBRATION GAS...... 115

FIGURE 2.32. CHROMATOGRAMS OF THE SPECTRA GAS INJECTED DIRECTLY (BLUE) AND BY

THE PRE-CONCENTRATION SYSTEM (BLACK)...... 115

FIGURE 2.33. A CHROMATOGRAM OF THE GASOLINE MIXTURE IN EXTERNAL IONIZATION

MODE. THE LARGEST PEAK AT 23.4 MINUTES IS TOLUENE...... 116

FIGURE 2.34. CHROMATOGRAMS OF ZERO GAS USING AIR FROM THE ZERO AIR GENERATOR

(BLUE TRACE, INTERNAL IONIZATION MODE) AND N2 FROM THE LIQUID NITROGEN GAS

PORT (BLACK TRACE, EXTERNAL IONIZATION MODE). THE DIFFERENCE IN THE

BASELINE SIGNAL IS ATTRIBUTED TO THE BETTER TRAPPING EFFICIENCY OF THE ION

TRAP IN INTERNAL IONIZATION MODE...... 116

FIGURE 2.35. THE CHAMBER MIXING RESULTS FOR EXPONENTIAL DILUTIONS FOR FLOW

RATES (LEFT TO RIGHT): 22.2 L MIN-1, 11.4 L MIN-1, 4.8 L MIN-1. FANS OFF (BLUE),

FANS ON (GREEN), THEORETICAL (BLACK DASHED) ARE SHOWN...... 117

FIGURE 2.36. A TIME SERIES OF THE EXPONENTIAL DILUTION OF BENZENE (GREEN

SQUARES), INDANE (LIGHT BLUE TRIANGLES), TOLUENE (DARK BLUE CIRCLES) AND

1,2,4-TRIMETHYLBENZENE (GRAY DIAMONDS) MEASURED BY THE GC-ITMS......

...... 118

xvii FIGURE 2.37. PTR-MS VS. GC-ITMS DURING THE EXPONENTIAL DILUTION OF BENZENE

(GREEN SQUARES), INDANE (LIGHT BLUE TRIANGLES), TOLUENE (DARK BLUE CIRCLES)

AND 1,2,4-TRIMETHYLBENZENE (GRAY DIAMONDS). THE PTR-MS SHOWS A PROBLEM

WITH THE ZERO AT VALUES BELOW 1 PPBV, POSSIBLY DUE TO A BACKGROUND

CORRECTION PROBLEM...... 119

FIGURE 2.38. EXPERIMENTAL SETUP FOR THE GASOLINE OXIDATION EXPERIMENT......

...... 120

FIGURE 2.39. CROSS SECTIONS OF HONO (TOP) AND NO2 (BOTTOM) WITH THE SPECTRA OF

THE UV LAMPS OVERLAID...... 121

FIGURE 2.40. A PLOT OF THE LN([VOC]T/[VOC]0) VERSUS THE RATE CONSTANT WITH OH.

THE ERROR BARS ARE THE ERRORS ASSOCIATED WITH THE PRECISION OF THE

MEASUREMENTS PROPAGATED THROUGH THE NATURAL LOGARITHM OF A RATIO

(VERTICAL) AND THE STATED ACCURACY OF THE RATE CONSTANT (HORIZONTAL.) THE

LINE IS THE WEIGHTED BEST FIT THROUGH THE POINTS AND THE SLOPE IS EQUAL TO

THE PRODUCT OF THE [OH] AND THE RESIDENCE TIME...... 122

FIGURE 3.1. SCHEMATIC OF THE PROPHET TOWER, PROPHET LABORATORY AND THE

WSU MACL TRAILER SHOWING INLET HEIGHTS FOR THE GRADIENT PROFILE

SAMPLING...... 159

FIGURE 3.2. FLOW DIAGRAM OF THE PROFILING SCHEMATIC...... 160

FIGURE 3.3. PTR-MS VS. GC-ITMS FOR TOTAL MONOTERPENES (TOP) AND ISOPRENE

(BOTTOM) ALONG WITH THE 1:1 LINE (DASHED BLACK)...... 161

FIGURE 3.4. TIME SERIES OF METEOROLOGICAL VALUES FOR THE CABINEX FIELD

CAMPAIGN AS MEASURED AT THE 31 TO 33 M HEIGHT SHOWING RELATIVE HUMIDITY

xviii (BLUE) AND TEMPERATURE (RED), SURFACE AIR PRESSURE AND WIND SPEED......

...... 162

FIGURE 3.5. A PLOT OF AIR MASS RESIDENCE TIME BASED ON HYSPLIT BACK TRAJECTORIES.

...... 163

FIGURE 3.6. A WIND ROSE PLOT FOR THE MEASUREMENT PERIOD. THE WINDS WERE

MEASURED AT THE 33 M ANEMOMETER...... 164

FIGURE 3.7. PLOTS OF O3 AT THE 3 SAMPLE HEIGHTS...... 165

FIGURE 3.8. PLOTS OF NO (RED) AND NO2 (BLACK) FOR THE CAMPAIGN AT THE 34, 20.4,

AND 6 METER HEIGHTS. THE PROFILING EXPERIMENT IS INDICATED BY THE BLUE BOX.

...... 166

FIGURE 3.9. 10 MINUTE AVERAGES OF ISOPRENE MIXING RATIOS SHOWN AT THE 34, 20.4,

AND 6 M HEIGHTS...... 167

FIGURE 3.10. MONOTERPENE MIXING RATIOS (10 MINUTE AVERAGES) AT THE 34, 20.4 AND

6 M HEIGHTS...... 168

FIGURE 3.11. 10 MINUTE AVERAGE MIXING RATIOS OF METHANOL (BLACK) AND ACETONE

(RED) AT 34, 20.4, 6 M HEIGHTS...... 169

FIGURE 3.12. 10 MINUTE AVERAGES OF BENZENE (BLACK CIRCLES) AND TOLUENE (RED

CROSS) AT THE 34, 20.4 AND 6 M HEIGHTS...... 170

FIGURE 3.13. SHOWN TOP IS A LOG HISTOGRAM OF THE NOX DATA WITH TWO CLEAR

MODES. BOTTOM IS THE NOX DATA AT 34 M (BLACK), 20 M (LIGHT GRAY) AND 6 M

(DARK GRAY) ALONG WITH THE 550 PPTV CUTOFF...... 171

O FIGURE 3.14. WIND ROSE PLOTS WITH 15 BINS SORTED BY A NOX CUTOFF OF 550 PPTV.

THE TOP (DARK GRAY) PLOT SHOWS WINDS FOR VALUES LESS THAN 550 PPTV WHILE

xix THE BOTTOM ROSE (LIGHT GRAY) HAS BEEN SORTED BY NOX VALUES OVER 550 PPTV.

...... 172

FIGURE 3.15. A) PLOT OF DIEL AVERAGED METHANOL. B) PLOT OF DIEL AVERAGED

TOLUENE. C) PLOT OF DIEL AVERAGED LIGHT. D) PLOT OF DIEL AVERAGED ISOPRENE.

ALL OF THE DATA WERE SORTED BY THE NOX MIXING RATIO AND THE DATA THAT

WERE TAKEN WHEN NOX WAS GREATER THAN 550 PPVT ARE REPRESENTED BY LIGHT

GRAY TRIANGLES, WHILE THE DATA THAT WERE TAKEN WHEN NOX WAS LESS THAN

550 PPTV ARE DARK GRAY CIRCLES...... 173

FIGURE 3.16. A) DIEL PLOT OF NOX GRADIENTS [NOX,34M]-[NOX,6M] (BLACK SQUARES)

WHEN THE NOX IS LOWER THAN THE 550 PPTV CUTOFF THE ERROR BARS ARE 1

STANDARD DEVIATION. B) HISTOGRAM OF NOX GRADIENTS WHEN NOX IS BELOW THE

550 PPTV CUTOFF. C) DIEL PLOT OF THE NOX GRADIENTS [NOX,34M]-[NOX,6M] (GRAY

CIRCLES) WHEN THE NOX IS GREATER THAN THE 550 PPTV CUTOFF THE ERROR BARS

ARE 1 STANDARD DEVIATION. D) HISTOGRAM OF NOX GRADIENTS WHEN NOX IS

GREATER THAN THE 550 PPTV CUTOFF...... 174

FIGURE 3.17. DIEL AVERAGED JNO2 FOR THE CAMPAIGN AT 34 M (LIGHT GRAY CIRCLES)

AND AT THE GROUND (DARK GRAY TRIANGLES) ...... 175

FIGURE 3.18. DIEL AVERAGES FOR A) O3, B) NO, C) NO2, D) NOX, ARE SHOWN FOR THE

34 M (LIGHT GRAY SQUARES), 20 M (MEDIUM GRAY CIRCLES), AND 6 M (DARK GRAY

TRIANGLES) HEIGHTS. THE VERTICAL DASHED LINES INDICATE THE AVERAGE TIME OF

SUNRISE AND SUNSET DURING CABINEX...... 176

FIGURE 3.19. TOP LEFT PLOTS [NO2]/[NO] CALCULATED VERSUS [NO2]/[NO] OBSERVED

AT THE 34 M HEIGHT. TOP RIGHT IS A LOG HISTOGRAM OF THE RATIO OF NO2 TO NO

xx CALCULATED TO THE RATIO OF NO2 TO NO OBSERVED AT THE 34 M HEIGHT. BOTTOM

LEFT IS NO2]/[NO] CALCULATED VERSUS [NO2]/[NO] OBSERVED AT THE 6 M HEIGHT.

BOTTOM RIGHT SHOWS A LOG HISTOGRAM OF THE RATIO OF NO2 TO NO CALCULATED

TO THE RATIO OF NO2 TO NO OBSERVED AT THE 6 M HEIGHT...... 177

FIGURE 3.20. A) DIEL PLOT OF MONOTERPENES AT 34 M (LIGHT GRAY SQUARES), 20 M

(MEDIUM GRAY CIRCLES), AND 6 M (BLACK TRIANGLES). B) DIEL PLOT OF THE 34 M – 6

M MONOTERPENE GRADIENT. THE ERROR BARS INDICATE 1 STANDARD DEVIATION. C)

DIEL PLOT OF ISOPRENE AT 34 M (LIGHT GRAY SQUARES), 20 M (MEDIUM GRAY

CIRCLES), AND 6 M (BLACK TRIANGLES). D) DIEL PLOT OF THE 34 M – 6 M ISOPRENE

GRADIENT. THE ERROR BARS INDICATE 1 STANDARD DEVIATION...... 178

FIGURE 3.21. A) DIEL PLOT OF MACR+MVK AT 34 M (LIGHT GRAY SQUARES), 20 M

(MEDIUM GRAY CIRCLES), AND 6 M (BLACK TRIANGLES). B) DIEL PLOT OF THE 34 M – 6

M MACR+MVK GRADIENT. THE ERROR BARS INDICATE 1 STANDARD DEVIATION. C)

DIEL PLOT OF HCHO AT 34 M (LIGHT GRAY SQUARES), 20 M (MEDIUM GRAY CIRCLES),

AND 6 M (BLACK TRIANGLES). D) DIEL PLOT OF THE 34 M – 6 M HCHO GRADIENT.

THE ERROR BARS INDICATE 1 STANDARD DEVIATION. E) DIEL PLOT OF CH3OOH AT

34 M (LIGHT GRAY SQUARES), 20 M (MEDIUM GRAY CIRCLES), AND 6 M (BLACK

TRIANGLES). F) DIEL PLOT OF THE 34 M – 6 M CH3OOH GRADIENT. THE ERROR BARS

INDICATE 1 STANDARD DEVIATION...... 179

FIGURE 3.22. A) DIEL PLOT OF ACETONE AT 34 M (LIGHT GRAY SQUARES), 20 M (MEDIUM

GRAY CIRCLES), AND 6 M (BLACK TRIANGLES). B) DIEL PLOT OF THE 34 M – 6 M

ACETONE GRADIENT. THE ERROR BARS INDICATE 1 STANDARD DEVIATION. C) DIEL

PLOT OF METHANOL AT 34 M (LIGHT GRAY SQUARES), 20 M (MEDIUM GRAY CIRCLES),

xxi AND 6 M (BLACK TRIANGLES). D) DIEL PLOT OF THE 34 M – 6 M METHANOL GRADIENT.

THE ERROR BARS INDICATE 1 STANDARD DEVIATION...... 180

FIGURE 3.23. Α-PINENE PLOTTED AGAINST Β-PINENE FOR THIS WORK (DARK GRAY

CIRCLES), WESTBERG 1997 (SOLID BLACK LINE), CABINEX 2009 EMU (DASHED

LINE), AND APEL 2005 (LIGHT GRAY CIRCLES)...... 181

FIGURE 3.24. A PLOT OF THE MEDIAN OH LOSS FREQUENCY AS A FUNCTION OF TIME OF

DAY. THE AREA SHOWN IS THE BY COMPONENT OH LOSS; THE BLUE TRACE IS THE

TOTAL OH LOSS FREQUENCY, THE GREEN TRACE IS THE OH LOSS FOR NO2 PLUS TOTAL

MONOTERPENES AND THE RED IS THE OH LOSS TO MONOTERPENES. BY OVERLAPPING

THE TRACES THE AREA SHOWN CORRESPONDS TO THE INDIVIDUAL COMPONENT......

...... 182

FIGURE 3.25. CONTRIBUTIONS BY PERCENT TO THE HYDROXYL RADICAL LOSS RATE SHOWN

FOR ISOPRENE (BLUE), NO2 (GREEN) AND MONOTERPENES (RED) SPLIT INTO 6 HOUR

SECTIONS. A) 00:00 TO 06:00. B) 06:00 TO 12:00. C) 18:00 TO 24:00 D) 12:00 TO

18:00...... 183

FIGURE 4.1. TIME SERIES OF RELATIVE HUMIDITY (BLUE), TEMPERATURE (RED), CO, NOX,

AND BENZENE MIXING RATIOS FOR THE STUDY PERIOD...... 212

FIGURE 4.2. PLOTS OF HALF HOUR AVERAGED MIXING RATIOS AS A FUNCTION OF WIND

DIRECTION TOGETHER WITH WIND ROSE PLOT SHOWING WIND SPEED AND FRACTIONAL

OCCURRENCE OF WIND FLOW DIRECTION IN 30° INCREMENT BINS. APPROXIMATELY

20% OF THE OBSERVATIONS WERE FROM THE BIN CENTERED ON THE 150° DIRECTION.

...... 213

xxii FIGURE 4.3. DIEL PROFILE OF HALF HOUR AVERAGES OF WEEKDAY VMT AND CO (BLACK

CIRCLES). GREY SHADING FROM 08:15 TO 15:15 INDICATES AVERAGE LENGTH OF

DAYLIGHT PERIOD BETWEEN SUNRISE AND SUNSET FOR THE MEASUREMENT

CAMPAIGN...... 214

FIGURE 4.4. DIEL PROFILE OF HALF HOUR AVERAGED MIXING RATIOS FOR CO (BLACK

CIRCLES), NOX (GREY SQUARES), AND BENZENE (OPEN TRIANGLES)...... 215

FIGURE 4.5. MOBILE6.2 AND MOVES HOURLY WEEKDAY EMISSIONS FOR ADA AND

CANYON COUNTIES. GASOLINE VEHICLE EMISSIONS (OPEN CIRCLES) AND DIESEL

ENGINE VEHICLE (SOLID SQUARES) EMISSIONS ARE DISPLAYED. MOBILE6.2

EMISSIONS ARE GIVEN AS BLACK SYMBOLS MOVES AS BLUE SYMBOLS. LOWER

PANEL DISPLAYS WEEKDAY VMT FOR GASOLINE ENGINES VEHICLES (OPEN CIRCLES)

AND DIESEL ENGINE VEHICLES (SOLID SQUARES). THE GASOLINE VEHICLE VMT ARE

APPROXIMATELY 20 TIMES LARGER THAN DIESEL VMT...... 216

FIGURE 4.6. PLOTS OF LINEAR REGRESSIONS OF AMBIENT MEASUREMENTS FOR THE

EXPERIMENT. THE EMISSION RATIO IS THE SLOPE OF THE REGRESSION. THE GREY

CIRCLES ARE DATA POINTS FROM THE ENTIRE CAMPAIGN, THE BLACK CIRCLES ARE

FROM THE MORNING RUSH HOUR (5-9 AM). LINEAR REGRESSIONS ARE DISPLAYED AS

DASHED LINE THROUGH GREY CIRCLES AND SOLID BLACK LINE THROUGH RUSH HOUR

DATA...... 217

FIGURE 4.7. MEAN RUSH HOUR EMISSION RATIOS BY WIND DIRECTION. WIND DIRECTION

HAS BEEN DIVIDED INTO OCTANTS AND THE DASHED LINES ± 1 STANDARD DEVIATION

FROM THE MEAN. THE ANGULAR UNIT AXIS IS IN DEGREES AND THE RADIAL AXIS IS

MOLE/MOLE. THE PLOTS SHOW LITTLE DIRECTIONAL DEPENDENCE IN THE MOLAR

xxiii RATIOS, ALTHOUGH ΔCO/NOX IS SLIGHTLY LOWER ON AVERAGE WHEN WINDS ARE

FROM THE SOUTH, WHERE AIR FROM THE FREEWAY IS BEING SAMPLED...... 218

FIGURE 4.8. DIEL TREND OF MOLAR RATIOS SHOWING ½ HOUR AVERAGES AS SOLID BLACK

LINE. SHADING INDICATES 1 STANDARD DEVIATION. THERE IS VERY LITTLE CHANGE

THROUGHOUT THE DAY IN THE CO-TO-NOX RATIO INDICATING THAT THERE IS LITTLE

PHOTOCHEMICAL ACTIVITY IN THE DAY...... 219

FIGURE 4.9. HISTOGRAM OF ΔCO/NOX MOLAR RATIO FOR THE TIME PERIODS FROM 08:00

TO 18:00 WITH CO > 200 PPBV. THE EMISSION RATIOS FROM THE MOVES (9.1) AND

MOBILEV6.2 (13.2) MODELS ALONG ARE SHOWN AS THE BLUE AND RED VERTICAL

LINES RESPECTIVELY. THE BLACK VERTICAL LINE SHOWS THE OBSERVED RATIO FROM

THE MORNING RUSH HOUR PERIODS (4.2)...... 220

FIGURE 4.10. HOURLY WEEKDAY MOVES EMISSIONS FOR CO (TOP) AND NOX (MIDDLE)

AND THE CO-TO-NOX MOLAR RATIO COMPARED TO OBSERVATIONS (BOTTOM) WHERE

RED DIAMONDS ARE OBSERVATIONS, BLUE CIRCLES ARE THE URBAN ROAD RATIO,

BLACK CIRCLES ARE THE URBAN INTERSTATE RATIO, AND GREY SQUARES ARE THE

TOTAL RATIO THAT INCLUDES OFF-NETWORK EMISSIONS. THE URBAN ON-ROAD

MOVES EMISSIONS AGREE WELL WITH THE OBSERVATIONS. INCLUSION OF OFF-

NETWORK EMISSIONS SIGNIFICANTLY INCREASES RATIO COMPARED TO

OBSERVATIONS...... 221

FIGURE 4.11. PANEL A) 9-DAY TIME SERIES OF THE HOURLY TOTAL EMISSIONS (BLACK)

AND MOBILE EMISSIONS (RED) OF NOX FOR THE 12 KM X 12 KM GRID CELL OF THE

AIRPACT-3 MODEL CONTAINING THE MEASUREMENT SITE. B) 9-DAY TIME SERIES OF

THE TOTAL AND VEHICLE EMISSIONS OF CO FOR THE SAME GRID CELL. TOTAL CO

xxiv EMISSIONS (BLACK) USED BY AIRPACT-3 AND MOBILE EMISSIONS (RED). C) LOG

HISTOGRAM OF THE CO-TO-NOX MOLAR RATIO FOR THE OBSERVATIONS (GREEN) AND

FOR AIRPACT-3 OUTPUT (BLACK). D) CO VERSUS NOX CORRELATIONS SHOWING

AIRPACT-3 VEHICLE EMISSIONS (BLUE LINE) RATIO OF 18.1±0.09 MOL/MOL,

AIRPACT-3 TOTAL EMISSIONS RATIO (BLACK) WITH A RATIO OF 14.4±0.13, MOVES

EMISSION RATIO 8.9, AND OBSERVED HOURLY AVERAGED RATIO FOR THE ENTIRE

STUDY PERIOD (GREEN) OF 4.6±0.16. THE GREY CIRCLES ARE THE HOURLY AIRPACT-

3 OUTPUT FOR THE MONTH OF JANUARY 2009 AT THE GRID CELL CONTAINING THE

SITE...... 222

FIGURE 5.1. TIME SERIES OF CO, NO, NO2, AND NOY...... 262

FIGURE 5.2. TIME SERIES OF SELECTED VOCS IN PPBV MEASURED WITH THE GC-ITMS.

ISOPRENE (RED), BENZENE, (GRAY), TOLUENE (BLUE), P,M-XYLENE (GREEN), AND

1,2,4-TRIMETHYLBENZENE ARE SHOWN FROM 6/10/2010 TO 7/1/2010...... 263

FIGURE 5.3. TIME SERIES OF ORGANIC AEROSOL, O3, AND NOZ...... 264

FIGURE 5.4. DIEL PLOTS OF OH LOSS FREQUENCY (S-1) FOR THE VOCS MEASURED DURING

CARES...... 265

FIGURE 5.5. PLOT OF CO VERSUS NOX FOR THE MORNING RUSH HOUR (06:00 TO 09:00).

THE LINEAR REGRESSION GAVE A SLOPE OF 6.9 AND A R2 OF 0.68...... 266

FIGURE 5.6. LOG HISTOGRAM OF ΔCO/NOX FOR THE MORNING RUSH HOUR (06:00-09:00)

DURING THE CARES CAMPAIGN...... 266

FIGURE 5.7. A DIEL PLOT OF ΔCO/NOX SHOWING ALL THE DATA (GRAY CIRCLES), THE

MEDIAN (LIGHT BLUE TRIANGLES) AND AVERAGE (DARK BLUE) SQUARES. THE RUSH

HOUR PERIOD (06:00-10:00) IS FLAT...... 267

xxv FIGURE 5.8. FIGURE 8. ALKANE VERSUS CO DURING THE MORNING RUSH HOUR PLOTTED

(GREY CIRCLES) ON A LOG-LOG SCALE ALONG WITH THE LINEAR REGRESSION (BLACK

LINE) THROUGH THE DATA...... 268

FIGURE 5.9. PLOTS OF TOLUENE, BENZENE AND C2-ALKYLBENZENES VERSUS CO DURING

THE MORNING RUSH HOUR ON A LOG-LOG PLOT...... 269

FIGURE 5.10. LOG-LOG PLOTS OF C3- AND C4-ALKYLBENZENES VERSUS CO. THE PLOTS

INCLUDE THE DATA (GRAY CIRCLES) AND THE LINEAR REGRESSION (BLACK TRACE).

...... 270

FIGURE 5.11. FIGURE 11. THE STANDARD DEVIATION OF THE LOG TRANSFORMED VOC

MIXING RATIOS VERSUS THE ESTIMATED LIFETIME ...... 271

FIGURE 5.12. THIS IS A PLOT OF THE TIME EVOLUTION OF AN AIR PARCEL CONTAINING TWO

VOCS “A” AND “C” FOR THE LIMITING CASES OF DILUTION (DASHED LINE) AND

KINETIC REMOVAL (SOLID LINE)...... 272

FIGURE 5.13. PLOTS OF THE LOG TRANSFORMED VOC VERSUS 2,2,4-TRIMETHYLPENTANE.

THE DATA ARE PLOTTED OVER THE DILUTION LINE (BLACK DASHES) AND THE KINETIC

LINE (SOLID BLACK). THE DATA ARE COLOR CODED BY THE RATIO OF NOZ-TO-NOY.

...... 273

FIGURE 5.14. SELECTED PLOTS OF THE LOG TRANSFORMED VOC VERSUS VOC. THE DATA

ARE PLOTTED OVER THE DILUTION LINE (BLACK DASHES) AND THE KINETIC LINE

(SOLID BLACK). THE DATA ARE COLOR CODED BY THE RATIO OF NOZ-TO-NOY......

...... 274

xxvi FIGURE 5.15. DIEL PLOT OF THE NOZ/NOY RATIO. THE DATA (GRAY CIRCLES), ALONG WITH

THE MEAN (BLUE TRIANGLES) AND THE MEDIAN (BLUE SQUARES) VALUES ARE SHOWN.

...... 275

FIGURE 5.16. PLOTS OF DIEL AVERAGED O3 AND NOZ (TOP), O3 AND OA (MIDDLE), AND

NOZ AND OA (BOTTOM)...... 276

FIGURE 5.17. PLOT OF OA VERSUS O3. THE RED CIRCLES INDICATE DATA COLLECTED FROM

12:00 TO 20:00...... 277

xxvii LIST OF TABLES

TABLE 2.1. SUMMARY OF CHANNELS LOGGED BY DAQFACTORY...... 80

TABLE 2.2. TEMPERATURE PROGRAM USED FOR THE ABSOLUTE CALIBRATION OF THE FID.

...... 81

TABLE 2.3. SUMMARY OF CALIBRATION GASES USED FOR THE DIRECT CALIBRATION OF THE

FID...... 81

TABLE 2.4. THE TENAX GR TRAP AND COLD TUBE WATER TRAP TEMPERATURES IN OC FOR

THE THREE CONDITIONS TESTED...... 81

TABLE 2.5. A SUMMARY OF THE ARTIFACTS GENERATED FROM THE TENAX GR

DECOMPOSITION...... 82

TABLE 2.6. SHOWS THE COMPOSITION IN PPBV OF THE GCMS CALIBRATION GAS......

...... 83

TABLE 2.7. THE COMPOSITION OF THE PTR-MS CALIBRATION GAS...... 84

TABLE 2.8. SPECTRA GAS COMPOSITION IN PPMV ALONG WITH THE SENSITIVITY (AREA

COUNTS PPBV-1). THESE SENSITIVITIES ARE FOR INTERNAL IONIZATION MODE. THE

REPORTED SENSITIVITIES ARE FROM THE DIRECT INJECTION USING THE SAMPLE LOOP

AND REPRESENT THE ABSOLUTE RESPONSE OF THE SYSTEM TO THE COMPOUNDS THEY

HAVE BEEN SCALED FOR A 300 CM3 SAMPLE CONTAINING 1 PPBV...... 85

TABLE 2.9. TABLE OF COMPOUNDS IDENTIFIED ON THE GC-ITMS, RETENTION TIMES

(MINUTES) AND QUANTITATION IONS (M/Z)...... 87

TABLE 2.10. SUMMARY OF RESULTS FROM THE CHAMBER MIXING EXPERIMENT......

...... 89

xxviii TABLE 2.11. . COMPOUNDS AND RECOMMENDED OH RATE CONSTANTS (CALVERT ET AL.

2002) ...... 89

TABLE 3.1. IONS MONITORED BY PTR-MS FOR GRADIENT PROFILING ...... 156

TABLE 3.2. SUMMARY OF METEOROLOGY DURING CABINEX 2009 COMPARED TO 1984-

1998 JULY-AUGUST DATA FROM PELSTON, MI...... 157

TABLE 3.3. SUMMARY OF MONOTERPENES AT CABINEX...... 157

TABLE 3.4. SUMMARY OF THE 10 MINUTE AVERAGED TRACE GASES MIXING RATIOS (PPBV)

REPORTED FOR EACH HEIGHT...... 158

TABLE 4.1. SUMMARY OF ADA COUNTY DECEMBER AND JANUARY EMISSIONS

INVENTORY...... 210

TABLE 4.2. STATISTICAL SUMMARY OF OBSERVATIONS DURING THE 2-MONTH STUDY

PERIOD...... 210

TABLE 4.3. SLOPES FROM RUSH HOUR LINEAR REGRESSION ANALYSIS OF MEASURED

POLLUTANT RATIOS...... 211

TABLE 4.4. VEHICLE EMISSIONS AND CO/NOX RATIOS FOR MOVES AND MOBILEV6.2

COMPARED TO OBSERVATIONS...... 211

TABLE 5.1. OH+VOC RATE CONSTANTS FOR THE VOCS REPORTED IN THIS WORK ALONG

WITH THE LIFETIME IN HOURS ASSUMING [OH] = 6*106...... 260

TABLE 5.2. VOC-TO-CO EMISSION RATIOS REPORTED BY THIS WORK AND THE WORKS OF

WARNEKE ET AL., (2009) AND LOUGH ET AL., (2005). THE RATIOS ARE REPORTED AS

PPBV PER PPBV CO...... 261

xxix INTRODUCTION

Non methane hydrocarbons (NMHCs) are important trace gases found ubiquitously in the troposphere (Singh and Zimmerman, 1992; Blake et al., 1997; Slemr et al., 2002; Apel et al., 2003). When mixed with NOx and sunlight NMHCs undergo photochemical oxidation which produces photochemical smog which is a combination of ozone, particulate matter and other toxic photochemical products (Altshuller, 1983;

Chameides et al., 1988; Sillman et al., 1995; Odum et al., 1997; Sillman 1999). In urban regions where NOx and volatile organic compounds (VOCS) are copiously emitted this can lead to air pollution problems that harm crops, the environment, and humans (NRC,

1991). From a regulatory point of view, O3 and secondary organic aerosol (SOA) are the two most important secondary pollutants. Ozone and particulate matter (PM) are designated as criteria air pollutants by the United States Environmental Protection

Agency (EPA) and permissible concentration levels are determined by the National

Ambient Air Quality Standards (NAAQS). PM and O3 are also among the most frequently violated NAAQS pollutants (EPA, 2013). It is estimated that 108 million

Americans live in areas that violate the 8-hour standard for O3 and 17.3 million

Americans live in areas that violated either the annual or the 24 hour limit for PM smaller than 2.5 µm (EPA, 2013). Many of the aromatic compounds found in urban environments benzene, toluene o-xylene, p-xylene, m-xylene, toluene, styrene, ethylbenzene, naphthalene, cumene, o-cresol, m-cresol, and p-cresol for example, are considered hazardous air pollutants (HAPs) (Kao, 1994). Aromatic compounds also make up the majority of the SOA precursors in spark ignition emissions (Odum et al., 1997a;

Odum et al., 1997b). Oxidation of NMHCs in the presence of NOx results in production

1 of formaldehyde, a known HAP (Kao, 1994), and peroxyacetylnitrates (PANs) which are lachrymators and are involved in long-range transport of NOx to remote locations (Birch et al., 1984; Brice et al., 1998; Derwent and Jenkin, 1991). The mixing ratios of NMHCs exist over more than 5 orders of magnitude, from tens of ppbvs to sub pptv levels due to variability of the sources and the lifetimes of the particular NMHC (Slemr et al., 2002).

Understanding sources of NMHCs and NOx is important for the development of accurate emission inventories for accurate model modeling of atmospheric chemistry from local to global scales.

Sources of NMHCs include oceans (Bonsang and Lambert, 1985; Bonsang et al.,

1988; Rudolph and Johnen, 1990; Plass-Dülmer et al., 1993), land based vegetation

(Guenther et al., 2006; Helmig et al., 2007; Sakulyanontvittaya et al., 2008) and anthropogenic, including motor vehicle emissions, solvent use, and industrial emissions

(Fujita et al., 1992; Kirchstetter et al., 1996; Jobson et al., 2004). Globally the biosphere is responsible for the vast majority of volatile organic compounds (VOCs) emitted into the atmosphere. Guenther et al. (1995) estimated that biogenic VOC emissions are approximately 1100 TgC yr-1 compared to anthropogenic VOC emissions of 110 -150

TgC yr-1 (Muller et al. 1992). The National Emissions Inventory (NEI) produced by the

US EPA reports a total of 14224 tonnes of anthropogenically emitted VOCs were released from the US in 2012. According to the NEI (2013), in 2012, the largest anthropogenic emissions in the US were due to industrial processes (44%), followed by miscellaneous (30%) and then transportation (22%). Transportation emissions reported in the NEI are from vehicles and they are divided into highway and off highway sources which account for 12% and 10%, respectively. In urban regions free from significant

2 industrial sources mobile emissions are a significant fraction of total VOCs and NOx emissions (Fujita et al., 2012). VOC emissions from mobile sources result from emissions of uncombusted VOCs in exhaust, evaporation during fueling, and also from evaporative emissions due to leaks in fuel tanks and fuel lines (Brooks et al., 1995;

Kirchstetter et al., 1999; Pierson et al., 1999; Lough et al., 2005). Thus, VOC vehicle emissions are a combination of exhaust and vapor headspace. Vehicle emissions measurements made in tunnels show that they include C2 to C12 VOCs including, alkenes, alkenes, alkynes, cycloalkanes, cycloalkenes, aromatics, and oxygenated VOCs (Jobson et al., 2004; Lough et al., 2005; Legreid et al., 2007; Fujita et al., 2012).

Periodically evaluating emission inventories by comparing them to real world measurements of vehicle emissions is important because vehicle emissions change over time as a result of fleet turnover, introduction of stricter emission requirements, and oxygenated fuels programs, while at the same time mobile emission models are continually evolving. Most recently, EPA introduced the MOVES emission model to replace the MOBILE6 emission model. In the early 1990’s a number of studies showed that the older vehicle emissions inventories, based on early versions of the MOBILE model were underestimating CO by a factor of 2-3 (Pierson et al., 1990; Fujita et al.,

1992; Harley et al., 1993) however by 2006 this had switched and it was shown that the models were overestimating CO by a factor of 2 (Parrish, 2006). In the US, road-way emissions of NOx and VOCs have steadily fallen despite increases in vehicle miles travelled (VMT) and growing reliance on diesel fuel (NEI, 2013). This evidence that real world vehicle emissions are changing faster than the models can account for provides motivation for regular evaluation of mobile source emissions. Currently diesel engines

3 account for over half of the mobile emissions of NOx emitted in the United States

(Dallmann and Harley, 2010; Millstein and Harley, 2010; Ban-Weiss et al., 2008)

The need to measure VOCs in urban atmospheres to understand the sources of VOCs and the impact of their photochemical processing on ozone and PM formation motivates the instrumentation developed and used as a part of this dissertation.

These systems were deployed with Washington State University’s Mobile Atmospheric

Chemistry Laboratory (MACL) on several air chemistry field experiments and the results from these field measurements are described in this dissertation.

1.1. NMHC MEASUREMENT TECHNIQUES

A major focus of this dissertation was the construction of a gas chromatography

(GC) based system for the measurement of C2-C12 NMHCs in air. The GC system was designed to measure C2 to C12 compounds emitted from spark ignition vehicles, with a focus on the most important O3 and SOA precursors. NMHC intercomparison experiments (Slemr et al., 2002; Apel et al., 2003; and Rappenglück et al., 2006) demonstrate reasonable agreement amongst methods and research groups for many of these compounds. However, there are only a few groups doing so with great precision and accuracy. NMHC quantification in the atmosphere is challenging because it involves multiple processes: sample handling, preconcentration, desorption/volatilization, separation and detection. The two most common ways to preconcentrate NMHCs are, by trapping them at cryogenic temperatures and by trapping them with solid adsorbents.

Cryogenic trapping is the established standard method (Apel et al., 2003) however, this technique has significant sample handling problems that must be overcome. Water,

4 carbon dioxide and oxygen, which are major components of the troposphere can be collected along with the NMHC at cryogenic temperatures creating problems. To avoid deposition of oxygen, sampling is performed at low pressures. However, water removal and carbon dioxide are problematic. The GC-ITMS system first deployed during the

CABINEX 2009 campaign in rural northern Michigan as part of the research described in this dissertation used a cryogenic trap to preconcentrate hydrocarbons from a 500 cc sample size. This system suffered from significant interference from CO2, which condensed in the cryogenic sample loop along with the hydrocarbons. The impact was splitting of early eluting peaks and high backgrounds of m/z 44. The mass spectrometer used in this research was an ion trap mass spectrometer (ITMS) which has both advantages and disadvantages for trace gas analysis. The ITMS traps and holds ions in orbits within a confined volume before selectively scanning them to the detector. The

ITMS can trap a limited number of ions (20,000) so if it fills with ions from CO2 it loses sensitivity to important analytes. The analytical problems incurred during CABINEX associated with cryogenic trapping of VOCs informed the decision to develop a solid adsorbent based system. Solid adsorbent based systems have the added advantage of being free from cumbersome cryogens. This is not to say cryogen free systems are absolved of problems because they come with their own set of problems, including analyte loss and artifact generation. However, they do not trap either CO2 or O2 improving chromatography, and the water trapping constraints are lessened.

The system described in chapter 2.2 takes a two channel approach to the task of measuring C2 to C12 VOCs by designating a light hydrocarbon channel (C2 to C6) and a

VOC channel (C6 to C12). The design of the light hydrocarbon channel was loosely based

5 on the system described by Tanner et al., (2006) which uses a multi-bed adsorbent to preconcentrate light hydrocarbons. Light hydrocarbons are measured using adsorbent traps by a number of groups. Tanner et al., (2006) use a cold tube water trap at -20 oC to remove water vapor from the sample, a Na2S2O3 ozone scrubber, an Al2O3/KCl Plot column, and a flame ionization detector (FID). Lamanna and Goldstein (1999) use a multi bed adsorbent based preconcentration system to measure both light and heavier hydrocarbons. Their sample handling includes use a Nafion dryer to remove water, an

Acsarite trap to remove CO2 and an ozone trap consisting of KI-impregnated glass wool.

The light hydrocarbon channel is designed with a Carboxen 1000 adsorbent material, an

AL2O3/KCl Plot column, and a flame ionization detector (FID). The VOC channel uses

Tenax GR adsorbent, a DB-624 column, and detection by ITMS. Chapter 2.2 describes the sample handling, water trapping, VOC preconcentration and experiments performed to validate the performance of the instrument. The GC-ITMS system using solid adsorbents was deployed for the CARES 2010 campaign in Sacramento, CA.

1.2. DISSERTATION OVERVIEW

Chapter 2 describes the experimental methods used to measure NMHC and NOx in the field studies. The development and testing of the solid based adsorbent preconcentration system is described in detail. Also described is the NO, NO2, and NOy

(NOxy) instrument that was purchased from Air Quality Design (Golden, CO) as part of the acquisition of the Mobile Atmospheric Chemistry Lab. The NOxy instrument is a research grade, 2 channel chemiluminescence instrument for which I was the instrument mentor. It was used in 3 major field experiments: Treasure Valley PM2.5 Precursor

Study in Jan 2009, CABINEX in July-Aug 2009, and CARES in June 2010.

6 Chapter 3 describes observations of NOx and VOC made at the Program for

Research on: Oxidants Photochemistry, Emissions and Transport (PROPHET) tower during July and August 2009 as a part of the Community Atmosphere Biosphere

Investigation Experiment CABINEX campaign. This site is atmospherically interesting because it lies on the edge of an ozone gradient receiving two distinctly different air masses, clean air from the Canadian north and north west and polluted air from the south and south west (Cooper et al., 2001). Observing the changes in chemistry associated with each airmass type allows us to study the impact of anthropogenic pollution on forested areas. Understanding biogenic VOC sources is important to both rural and urban regions because biogenic sources contribute to VOC emissions in both rural and urban regions.

During this campaign profiles of O3, NOx, VOCs, and photoproducts were taken through the forest canopy at 34, 20.4, and 6 m heights. Gradients were generated from the profiles and then analyzed by airmass origin using NOx as a tracer for anthropogenic influence. This analysis shows that forests set up conditions that favor formation of chemical gradients by inhibiting mixing to the surface and by creating a light gradient. It also allowed us to detect emissions of NOx from soils which are important to the chemistry of remote forested regions. One interesting result during this campaign was that the polluted periods were associated with cloudy conditions which inhibited photochemistry and emissions of isoprene. This chapter has been formatted for submission for a special issue of Atmospheric Chemistry and Physics.

Chapter 4 resulted from a wintertime Particulate Matter (PM) study in Boise, ID. called the Treasure Valley PM2.5 Precursor Study. This is study is an example of an urban environment primarily influenced by emissions from mobile on-road sources in

7 wintertime. The site was located close to I-84 and N. Eagle Road which are the busiest interstate and arterial in the state (Wallace et al., 2012). The site’s proximity to these busy roads along with clear sky wintertime conditions created perfect conditions to observe emissions from the on-road, mobile fleet. These conditions include tight nocturnal inversions that trap morning rush hour emissions which occur before sunrise and when the sun is low in the sky inhibiting mixing form sensible heat flux. In addition, the low photochemical activity during the day prevents perturbation of the emission ratios. Data collected under these ideal conditions were compared to emissions inventories MOBILEv6.2 and MOVES, the EPA’s newest mobile emissions inventory.

This resulted in a manuscript that was published in Atmospheric Environment. CO-to-

NOx emission ratios generated by the two models were compared to the ambient emission ratios observed during the morning rush hour. VOC-to-CO emission ratios for benzene- to-CO, toluene-to-CO, C2-alkylbenzenes-to-CO, and C3-alkylbenzenes-to-CO are also reported.

Chapter 5 describes results from the Carbonaceous Aerosol and Radiative Effects

Study (CARES) field experiment conducted in Sacramento, CA from June 2nd to June

28th, 2010. The CARES experiment was a large multi-institutional field campaign conducted by the US Department of Energy’s Atmospheric Systems Research (ASR) program. This experiment was designed to track the evolution of the Sacramento plume as it is transported up into the forested foothills of the Sierra Nevada during the day and then back into California’s central valley at night (Zaveri et al., 2012). There were two heavily instrumented sites located in Sacramento, CA (T0) and in Cool, CA (T1) which is located in the foothills along with multiple aircraft missions flown with the DOE’s G-1

8 research vessel. The WSU MACL was deployed at the T0 site. Measurements of CO,

NOx, and VOCs were used to characterize emissions and photochemistry at the T0 site.

This chapter is formatted for submission to a special issue of ACP.

9 1.3. REFRENCES

"Air Quality Trends." EPA. Environmental Protection Agency, n.d. Web. 25 June 2013.

Altshuller, A. P. "Review: Natural Volatile Organic Substances And Their Effect On Air

Quality In The United States." Atmospheric Environment (1967) 17.11 (1983): 2131-

2165.

Apel, E.C., Calvert, J.G., Gilpin, T.M., Fehsenfeld, F., and Lonneman, W.A.

"Nonmethane Hydrocarbon Intercomparison Experiment (NOMHICE): Task 4, Ambient

Air." Journal of Geophysical Research: Atmospheres (1984–2012) 108.D9 (2003).

Ban-Weiss, G.A., McLaughlin, J.P., Harley, R.A., Lunden, M.M., Kirchstetter, T.W.,

Kean, A.J., Strawa, A.W., Stevenson, E.D., and Kendall, G.R. "Long-term changes in emissions of nitrogen oxides and particulate matter from on-road gasoline and diesel vehicles." Atmospheric Environment 42.2 (2008): 220-232.

Blake, N.J., Blake, D.R., Chen, T.Y., Collins Jr, J.E., Sachse, G.W., Anderson, B.E., and

Rowland, F.S.. "Distribution and seasonality of selected hydrocarbons and halocarbons over the western Pacific basin during PEM-West A and PEM-West B." Journal of

Geophysical Research 102.D23 (1997): 28315-28.

Bonsang, B., and Lambert, G. "Nonmethane hydrocarbons in an oceanic atmosphere."

Journal of Atmospheric Chemistry 2.3 (1985): 257-271.

10

Bonsang, B., et al. "The marine source of C2-C6 aliphatic hydrocarbons." Journal of

Atmospheric Chemistry 6.1-2 (1988): 3-20.

Brice, K.A., Bottenheim, J.W., Anlauf, K.G., and Wiebe, H.A. "Long‐term measurements of atmospheric peroxyacetylnitrate (PAN) at rural sites in Ontario and Nova Scotia; seasonal variations and long‐range transport." Tellus B 40.5 (1988): 408-425.

Brich, K. A., Penkett, S.A., Atkins, D.H.F., Sandalls, F.J., Bamber, D.J., Tuck, A.F., and

Vaughan, G. "Atmospheric measurements of peroxyacetylnitrate (PAN) in rural, south- east England: Seasonal variations winter photochemistry and long-range transport."

Atmospheric Environment (1967) 18.12 (1984): 2691-2702.

Chameides, W. L., Lindsay, R.W., Richardson, J., and Kiang, C.S. "The role of biogenic hydrocarbons in urban photochemical smog: Atlanta as a case study." Science 241, 4872

(1988): 1473-1475.

Cooper, O. R., Moody, J.L. Thornberry, T.D., Town, M.S., and Carroll, M.A.

"PROPHET 1998 meteorological overview and air-mass classification." Journal of

Geophysical Research 106.D20 (2001): 24289-24.

11 Dallmann, T.R., and Harley, R.A. "Evaluation of mobile source emission trends in the

United States." Journal of Geophysical Research: Atmospheres (1984–2012) 115.D14

(2010).

Derwent, R. G., and Jenkin, M.E. "Hydrocarbons and the long-range transport of ozone and PAN across Europe." Atmospheric Environment. Part A. General Topics 25.8 (1991):

1661-1678.

Fujita, E.M., Croes, B.E., Bennett, C.L., Lawson, D.R., Lurmann, F.W., and Main, H.H.

"Comparison of emission inventory and ambient concentration ratios of CO, NMOG, and

NOx in California's South Coast Air Basin." Journal of the Air & Waste Management

Association 42.3 (1992): 264-276.

Fujita, E.M., Campbell, D.E., Zielinska, B., Chow, J.C., Lindhjem, C.E., DenBleyker, A.,

Bishop, G.A., Schuchmann, B.G., Stedman, D.H., and Lawson, D.R. "Comparison of the

MOVES2010a, MOBILE6. 2, and EMFAC2007 mobile source emission models with on- road traffic tunnel and remote sensing measurements." Journal of the Air & Waste

Management Association 62.10 (2012): 1134-1149.

Guenther, A., Karl, T., Harley, P., Wiedinmyer, C., Palmer, P.I., and Geron, C.

"Estimates of global terrestrial isoprene emissions using MEGAN (Model of Emissions of Gases and Aerosols from Nature)." Atmospheric Chemistry and Physics 6.11 (2006):

3181-3210.

12

Guenther, A., Nicholas Hewitt, C., David Erickson, D., Fall, R., Geron, C., Graedel, T.,

Harley, P., et al. "A global model of natural volatile organic compound emissions."

Journal of Geophysical Research 100.D5 (1995): 8873-8892.

Harley, R.A., Russell, A.G., McRae, G.J., Cass, G.R., and Seinfeld, J.H. "Photochemical modeling of the Southern California air quality study." Environmental Science &

Technology 27.2 (1993): 378-388.

Helmig, D., Ortega, J., Duhl, T., Tanner, D., Guenther, A., Harley, P., Wiedinmyer, C.,

Milford, J., and Sakulyanontvittaya, T. "Sesquiterpene emissions from pine trees-

Identifications, emission rates and flux estimates for the contiguous United States."

Environmental Science & Technology 41.5 (2007): 1545-1553.

Jobson, B. T., Berkowitz, C.M., Kuster, W.C., Goldan, P.D., Williams, E.J., Fesenfeld,

F.C., Apel, E.C., Karl, T., Lonneman, W.A., and Riemer, D. "Hydrocarbon source signatures in Houston, Texas: Influence of the petrochemical industry." Journal of

Geophysical Research: Atmospheres (1984–2012) 109.D24 (2004).

Kao, A.S. "Formation and removal reactions of hazardous air pollutants." Air & Waste

44.5 (1994): 683-696.

13 Kirchstetter, T.W., Singer, B.C., Harley, R.A., Kendall, G.R., and Chan, W. "Impact of oxygenated gasoline use on California light-duty vehicle emissions." Environmental

Science & Technology 30.2 (1996): 661-670.

Lamanna, M.S., and Goldstein, A.H. "In situ measurements of C2‐C10 volatile organic compounds above a Sierra Nevada ponderosa pine plantation." Journal of Geophysical

Research: Atmospheres (1984–2012) 104.D17 (1999): 21247-21262.

Legreid, G., Reimann, S., Steinbacher, M., Staehelin, J., Young, D., and Stemmler, K.

"'Measurements of OVOCs and NMHCs in a Swiss Highway Tunnel for Estimation of

Road Transport Emissions." Environmental Science & Technology 41.20 (2007): 7060-

7066.

Millstein, D. E., and Harley, R.A. "Effects of retrofitting emission control systems on in- use heavy diesel vehicles." Environmental Science & Technology 44.13 (2010): 5042-

5048.

Müller, J.F. "Geographical distribution and seasonal variation of surface emissions and deposition velocities of atmospheric trace gases." Journal of Geophysical Research:

Atmospheres (1984–2012) 97.D4 (1992): 3787-3804.

"National Emissions Inventory (NEI) Air Pollutant Emissions Trends Data." EPA.

Environmental Protection Agency, n.d. Web. 25 June 2013.

14

National Research Council (US). Committee on Tropospheric Ozone Formation and

Measurement. (1991). Rethinking the ozone problem in urban and regional air pollution.

National Academy Press.

Odum, J.R., Jungkamp, T.P.W., Griffin, R.J., Flagan, R.C., and Seinfeld, J.H. "The atmospheric aerosol-forming potential of whole gasoline vapor." Science 276.5309

(1997): 96-99.

Odum, J.R., Jungkamp, T.P.W., Griffin, R.J., Forstner, H.J.L., Flagan, R.C., and Seinfeld,

J.H. "Aromatics, reformulated gasoline, and atmospheric organic aerosol formation."

Environmental Science & Technology 31.7 (1997): 1890-1897.

Parrish, D.D. "Critical evaluation of US on-road vehicle emission inventories."

Atmospheric Environment 40.13 (2006): 2288-2300.

Pierson, W.R., Gertler, A.W., and Bradow, R.J. "Comparison of the SCAQS tunnel study with other on road vehicle emission data." Journal of the Air & Waste Management

Association 40.11 (1990): 1495-1504.

Plass‐Dülmer, C., Khedim, A., Koppmann, R., Johnen, F.J., Rudolph, J., and Kuosa, H.

"Emissions of light nonmethane hydrocarbons from the Atlantic into the atmosphere."

Global Biogeochemical Cycles 7.1 (1993): 211-228.

15 Rappenglück, B., Apel, E., Bauerfeind, M., Bottenheim, J., Brickell, P., Čavolka, P.,

Cech, J., et al. "The first VOC intercomparison exercise within the Global Atmosphere

Watch (GAW)." Atmospheric Environment 40.39 (2006): 7508-7527.

Rudolph, J., and Johnen, F.J. "Measurements of light atmospheric hydrocarbons over the

Atlantic in regions of low biological activity." Journal of Geophysical Research:

Atmospheres (1984–2012) 95.D12 (1990): 20583-20591.

Sakulyanontvittaya, T.T.D., Wiedinmyer, C., Helmig, D., Matsunaga, S., Potosnak, M.,

Milford, J., and Guenther, A. "Monoterpene and sesquiterpene emission estimates for the

United States." Environmental Science & Technology 42.5 (2008): 1623-1629.

Sillman, S., et al. "Photochemistry of ozone formation in Atlanta, GA-Models and measurements." Atmospheric Environment 29.21 (1995): 3055-3066.

Sillman, S. "The relation between ozone, NOx and hydrocarbons in urban and polluted rural environments." Atmospheric Environment 33.12 (1999): 1821-1845.

Singh, H. B., & Zimmerman, P. B. (1992). Atmospheric distribution and sources of nonmethane hydrocarbons. In: Gaseous pollutants: Characterization and cycling, p. 177-

235., 1, 177-235.

Slemr, J., Slemr, F., Partridge, R., D'Souza, H., and Schmidbauer, N. "Accurate measurements of hydrocarbons in the atmosphere (AMOHA): Three European

16 intercomparisons." Journal of Geophysical Research: Atmospheres (1984–2012)

107.D19 (2002): ACH-15.

Tanner, D., Helmig, D., Hueber, J., and Goldan, P. "Gas chromatography system for the automated, unattended, and cryogen-free monitoring of C2 to C6 non-methane hydrocarbons in the remote troposphere." Journal of Chromatography A 1111.1 (2006):

76-88.

Wallace, H. W., Jobson, B.T., Erickson, M.H., McCoskey, J.K., VanReken, T.M., Lamb,

B.K., Vaughan, J.K., et al. "Comparison of wintertime CO to NOx ratios to MOVES and

MOBILE6. 2 on-road emissions inventories." Atmospheric Environment (2012).

Zaveri, R.A., Shaw, W.J., Cziczo, D.J., Schmid, B., Ferrare, R.A., Alexander, M.L.,

Alexandrov. M., et al. "Overview of the 2010 carbonaceous aerosols and radiative effects study (CARES)." Atmospheric Chemistry and Physics 12.16 (2012): 7647-7687.

17 INSTRUMENT DESCRIPTION AND DEVELOPMENT

2.1. NOxy INSTRUMENT DESCRIPTION

The NO, NO2, and NOy mixing ratios presented in this dissertation were measured using a two channel chemiluminescent nitric oxide (NO) detector manufactured by Air

Quality Design (Golden, CO). Detection of NO2 and NOy are achieved using devices that convert NO2 and NOy to NO. This instrument will be referred to as the NOxy instrument.

The NOxy instrument was operated in two modes. The first mode is the standard configuration, where NO and NOx are measured on channel 2 and NOy is measured on channel 1. The second mode is flux mode where NOy measurements are not performed and instead NO is measured on channel 1 and NOx is measured on channel 2. The instrument consists of three rack mount boxes, an inlet case, and a large vacuum pump.

The inlet case houses the photolytic NO2 converter and the molybdenum oxide catalytic converter. The three rack mount boxes are called the control box, ozonizer box and the snooper box. The control box controls the valve states of the instrument and mixes calibration gases, the ozonizer box produces percent levels of ozone for reaction with

NO, and the snooper box houses the photomultiplier tubes (PMTs), the PMT cooler, reaction volumes, and the zero volumes. The pump is external to the rest of the system and not mounted along with the boxes in the rack mount enclosure.

2.1.1. Principle of Measurement

The subsequent sections describe the principles of measurement used by the Air Quality

Design (AQD) NOxy instrument for NO, NO2 and NOy.

18 2.1.1.1. Nitric Oxide (NO)

The AQD NOxy detects NO using a technique known as NO chemiluminescence.

This technique was first used to make atmospheric measurements of NO by Fontijn et al.

(1970). NO is detected via its chemiluminescent reaction with ozone. The reaction produces electronically excited NO2, denoted NO2*, which spontaneously emits a photon that can be measured with a photomultiplier tube. The mechanism of NO chemiluminesce is described by Clyne et al. (1964) and Clough and Thrush (1967):

∗ 푁푂 + 푂3 → 푁푂2 + 푂2 R2.1)

푁푂 + 푂3 → 푁푂2 + 푂2 R2.2)

∗ 푁푂2 → 푁푂2 + ℎ푣 (600 < 휆 < 2800) R2.3)

∗ 푁푂2 + 푀 → 푁푂2 + 푀 R2.4)

Reaction 1 is the reaction of NO with O3 to produce excited state NO2. Reaction 2 is the same reaction only the resulting NO2 is not in an excited state. Reactions 3 and 4 are the fates of the NO2*. Reaction 4 is known as collisional deactivation or quenching. The quenching of excited state NO2 is achieved by collision with an intermediate molecule that takes the extra energy from the NO2*. If an NO2* isn’t quenched, it will chemiluminesce, which is the reason this measurement requires reduced pressure to produce enough light to be detectable. The reaction vessel used in the AQD NOxy system has been designed to maximize the integrated chemiluminescence intensity and has been fully described by Ridley and Grahek, (1990). The AQD NOxy reaction vessel is coupled to the photomultiplier tubes with a red glass filter which transmits photons with wavelength greater than 600 nm. This filter is in place to prevent interference from ozone alkene chemiluminescence which produces light at lower wavelengths. For

19 example, isoprene + O3 chemiluminescence occurs at 490 nm and 550 nm (Guenther and

Hills, 1998). After passing through the red glass filter, a photomultiplier tube converts the light into an electrical current through a cascade of electrons. The resulting current pulses are sent to a preamplifier discriminator which determines which pulses meet a threshold, and then sends a TTL pulse to a counter. The counter counts the TTL pulses and by reading the counts every second, produces a count rate or frequency signal. The mixing ratio of NO in air is linear function of the photon count rate. The instrument response to NO is calibrated using an external compressed gas standard of NO in O2 free nitrogen.

Ridley et al. (1972) and Ridley et al. (1978) mathematically described the chemiluminescence signal for a cylindrical reaction volume of length L and cross sectional area A:

퐶퐼 = 푓(푇, 푃)휙휙푀[푁푂]0(1 − exp(−휔퐿)) (2.1) where 푓(푇, 푃) is the volumetric flow at temperature T and pressure P, [NO]0 is the NO entering the reaction volume, φ is the fraction of excited NO2 produced and φM is the fraction of NO2* that chemiluminesces at T and P, and ωL is the residence time in the volume divided by the chemical reaction time (Ridley and Grahek, 1990). Understanding the theoretical signal is important for understanding the sensitivity of the instrument.

Water, is an effective quencher of NO2* and can cause changes in sensitivity up to 15%

(Ridley and Grahek, 1990). High sample flow and low reaction vessel pressure are needed to maximize instrument sensitivity. In addition, the system was designed to incorporate PMTs with large cross sectional area to maximize sensitivity.

2.1.1.2. Nitrogen Dioxide (NO2)

20 The NOxy system detects NO2 by photolysis-chemiluminesence (PCL), a technique introduced by Kley and McFarland (1980). The NO2 in the sample is photo dissociated with a Blue Light Converter (BLC) from AQD. The BLC uses the same blue light LED that was evaluated by Pollack et al. (2010) for NO2 photolysis efficiency. The

LEDS have a peak emission at 395 nm and a peak width of 20 nm, full with half maximum (FWHM). The BLC used in this work has an internal volume of 17.6 cm3, and is made from sintered PTFE spheres creating a diffusively reflecting cavity. This converter exploits the dynamics of the Leighton relationship by increasing the photolysis rate of NO2 by light with a wavelength less than 420 to 426 nm (Gardener et al. 1987;

Roehl et al. 1994). The reactions forming the Leighton relationship are given below.

푁푂2 + ℎ푣(휆 < 420 푛푚) → 푁푂 + 푂 R2.5

푂 + 푂2 + 푀 → 푂3 + 푀 R2.6

푂3 + 푁푂 → 푁푂2 + 푂2 R2.7

The Leighton relationship can be further perturbed by reactions of NO with peroxy

. . radicals (RO2 ) and hydroperoxy radicals (HO2 ), reactions 8 and 9.

. . 푁푂 + 푅푂2 → 푁푂2 + 푅푂 R2.8

. . 푁푂 + 퐻푂2 → 퐻푂 + 푁푂2 R2.9

The rate of reaction 5, jNO2, is the integral across relevant wavelengths, given by equation

2.2,

휆2 푗 = 퐹(휆)휎(휆, 푇)휙(휆, 푇)푑휆 (2.2) 푁푂2 ∫휆1 where 퐹(휆) is the wavelength dependent actinic flux (photons cm-1s -1 ), σ is the temperature and wavelength dependent absorption cross section (cm2 molecule-1), and φ is the quantum yield (molecule*photon-1) (Sander et al., 2006). While the BLC provides

21 an environment favorable to the photo-dissociation of NO2, the back reaction of ozone or other oxidants with NO can reduce the overall effectiveness of the device. Grouping ozone, hydroperoxide radicals, peroxyl radicals and any other radicals that can convert

NO to NO2 into a category called Ox, Kley and McFarland (1980) derived the effective conversion factor for a photolytic converter, equation 2.3. In equation 2.3 t is the time, j is the jNO2 inside the converter and k is the bulk rate constant of NO with Ox which is defined as the sum of ozone, hydroperoxide radicals, and peroxyl radicals.

푗푡 퐶퐹 = [ ] [1 − exp(−푗푡 − 푘[푂푥]푡)] (2.3) 푗푡+푘[푂푥]푡

Ryerson et al. (2000) and Pollack et al. (2010) note that given the above equation, the best solution to improving the effective conversion factor is to increase j while keeping t small. The BLC used in this work has an internal volume of 17.6 cm3 and the flow rate is approximately 1000 sccm giving the BLC a residence time of approximately

1s, which is ideal. While low residence time prevents the device from achieving the highest possible conversion efficiencies, it minimizes the affect Ox has on the conversion efficiency. The peak intensity of 395 nm light-emitting diodes employed in the BLC is also suited to minimize interference from HONO which has a strong adsorption feature at

385 nm as shown in Figure 2.1.

2.1.1.3. Reactive Odd Nitrogen (NOy)

The term NOy is defined as the sum of all oxides of nitrogen in which the nitrogen has an oxidation state of +2 or larger: NOy = NO + NO2 + HNO3 + HONO+ NO3 + N2O5

+ HNO4 +ClONO2 + Organic Nitrates + Particulate Nitrate (Crosley 1993). Catalytic reduction to nitric oxide of selected species of NOy, particularly NO2, HNO3, and n- propylnitrate with a heated, gold-plated tube in the presence of carbon monoxide was first

22 pioneered by Bollinger et al. (1983). Later the technique was expanded to NOy by Fahey et al. (1985). Kliner et al. (1997) made a thorough investigation of NOy catalytic converters and concluded that the optimum NOy system when using a chemiluminescent detector is a gold catalyst with an H2 reductant. The AQD NOxy system used in this work utilizes a molybdenum oxide catalytic converter temperature controlled at 310 oC, from Thermo Fischer Scientific, Inc. to reduce NOy to NO. Molybdenum oxide catalytic converters have been shown by inter-comparison to be equivalent to reduction with a gold converter using either CO or H2 (Williams et al. 1998).

2.1.2. Sample Handling

With respect to photochemistry, nitric acid is one of the most important components of NOy as it is a major chain termination product in atmospheric chemistry, formed in the reaction between NO2 and OH. There have been a variety of studies examining inlet materials and sample handling requirements to reduce line losses of

HNO3 (Neuman et al. 1999; Apel et al. 1988; Goldan et al. 1983; Talbot et al. 1990). It has been shown that HNO3 can be readily lost by surface adsorption to inlet tubing, even

Teflon. Neuman et al. (1999), conducted a thorough analysis of tubing materials best suited for sampling HNO3 and concluded that PFA Teflon tubing at a temperature

o o between 10 C an 80 C quantitatively transports HNO3. The photolytic NO2 and catalytic NOy converter of the AQD NOxy system are housed in an enclosure which is mounted on the inlet tower. The inlet line to the NOy converter is approximately 30 cm

o and is heated to 30 C to ensure transport of HNO3 to the molybdenum oxide catalytic converter.

23 2.1.3. Calibration and Zeros

Calibration of the AQD NOxy system has been performed in the field and laboratory using two techniques: the first is a standard addition and the second is overflow with zero air and addition of calibration gas. To calculate the conversion efficiency of the BLC photolytic converter, the NO calibration gas (10 ppmv, Scott-

Marrin), diluted in zero air to 10 ppbv, is titrated with O3 to generate NO2. The conversion efficiency of the NOy molybdenum oxide catalytic converter is calculated with nitric acid from a permeation tube (Kintek Inc.). The emission rate when diluted in a known flow of zero air can be converted into a mixing ratio and is used to check the conversion efficiency. In practice the NOy conversion efficiency was not tested during field experiments, but before and after the campaign because they take a long time. We have not noted a drastic change in conversion efficiency of the molybdenum oxide converter. Conversion efficiency of NO2 by the molybdenum oxide catalytic converter is tested during the NO2 calibration of the BLC. Figure 2.2 shows the calibration of the

NOxy system. There are 5 sections to a full calibration: the first is the zero, (NOzero) which is a pre-reactor zero discussed in the next section, the second is calibration (NOcal), the third is titration of to NO to NO2 with the BLC lamps off (NOLamp), the fourth is titration of NO to NO2 with the BLC lamps on (NOxcal) and the fifth is the NOy calibration with HNO3 (NOyHNO3Cal). The sensitivity of each channel to NO is given in equation 2.4. Once the sensitivities and conversion efficiencies have been determined, the signals from the instrument can be converted into mixing ratios.

퐶퐻1 −퐶퐻1 퐶퐻2 −퐶퐻2 퐶퐻1 = ℎ푧(푁푂푐푎푙) ℎ푧(푧푒푟표) ; 퐶퐻2 = ℎ푧(푁푂푐푎푙) ℎ푧(푧푒푟표) (2.4) 푠푒푛푠 푁푂_푝푝푏푣 푠푒푛푠 푁푂_푝푝푏푣

The conversion efficiency of the BLC is calculated with equation 2.5.

24 퐶퐻2ℎ푧(푁푂푥푐푎푙)−퐶퐻2ℎ푧(푁푂퐿푎푚푝) 푁푂2 = (2.5) 퐶퐸 퐶퐻2ℎ푧(푁푂푐푎푙)−퐶퐻2ℎ푧(푁푂퐿푎푚푝)

Finally the NOy conversion efficiency is determined with equation 2.6.

퐶퐻1ℎ푧(푁푂푦퐻푁푂3퐶푎푙)−퐶퐻1푧푒푟표 푁푂푦퐶퐸 = (2.6) 퐻푁푂3푝푝푏푣∗퐶퐻1푠푒푛푠

The instrument zeros are determined in two ways. The most common and fastest way is to use the pre-reactor. The pre reactor is a 1 L volume that serves as a NO + O3 reaction chamber upstream of the actual reaction vessel. The purpose of the zero volume is to convert all the NO to NO2 before the sample enters the reaction vessel to be monitored for chemiluminescence. Any chemiluminescence observed is then attributed to interfering compounds and dark counts. The other technique is to perform zeros by overflowing the inlet with NO, NO2, and NOy free air. One issue with the pre reactor technique is that when the NO concentrations are very high, there may be unreacted NO entering the reaction vessel which can lead to an artificially high zero. However, this is not the case for very low levels of NO. When there is sufficient NO in the system to overload the pre-reactor, the artificially high zeros result in a small loss of signal (<1%) and are not corrected. The issue with using zero air for a zero is that the water vapor is low and generally does not match ambient water vapor concentrations, which results in changes in both background and sensitivity. A combination of both zeroes are used however, the system relies heavily on the pre-reactor zeros. The issue of elevated zeros during high NO events should be investigated.

2.1.4. Instrument Response

This section examines the NOxy instrument’s response. Of interest is the potential capability to perform high sampling rate measurements for eddy covariance flux

25 determination of NO surface fluxes. To capture the smallest eddys using the eddy covariance technique, an instrument must measure temporal variation of atmosphere concentrations at a rate better than 10 Hz. The NOxy instrument has been designed as a

“fast” instrument capable of measuring low mixing ratios of NO, NO2, and NOy, however, this system is just shy of the temporal resolution needed to capture the smallest eddys while making eddy covariance measurements. Response time was explored using a power spectral density analysis common to eddy covariance measurements. This analysis relies on well documented structure of turbulence and energy in the atmosphere.

Turbulent mixing in the atmosphere follows power log scales of -3 for entropy cascades and -5/3 for energy cascades (Gage, 1979; Gage and Nastrom, 1986; Lilly and Peterson,

1983; Sato, 1990). Scalar and wind vectors in the atmosphere with less than 2 km wavelength should exhibit a -5/3 power log relationship, while instrumental noise has a flat slope (Murphy et al., 1989). The power spectrum for a section of data Ci is defined as:

2 푃푆퐷 = |퐹푇(퐶푖푃푖)| (2.7)

Where,

푁−1 1−|푖− | 푃 = 2 (2.8) 푖 (푁+1)/2

N is the number of points in the section and FT is the Fourier transform. Following

Murphy (1989) intervals of 1024 points with a Parzen or triangular window were used to calculate the power spectrum of data collected in “flux” mode during a field campaign in

Xian, China. “Flux” mode involves bypass of the NOy converter so that NO and NOx are continuously measured. The sampling frequency of the PMTs is also increased to 10 Hz.

The analysis is simple; where the power spectrum of the data fits the natural -5/3 power

26 log structure of the atmosphere the instrument is capturing the natural variability of the turbulent atmosphere, and where it rolls over and goes flat is where the instrument noise dominates. This is a robust method to determine the maximum spectral response of an instrument because while an instrument may have a very fast rise time, it may not be able to rapidly resolve small changes in scalar value. A plot of the PSD for a 30 minute section of data during the Xi’an campaign is shown Figure 2.3. The transition from -5/3 log-log slope to instrument noise occurred at 0.3 Hz indicating that the instrument cannot resolve spectral features in the atmosphere that occur at a higher frequency than

0.3 Hz. Operators making flux measurements should bear this in mind when analyzing their data because corrections may need to be made during the analysis.

2.1.5. Artifacts and Corrections

The reflective body of the NO2 photolytic converter is made from sintered PTFE spheres, available from Gigahertz Optik, Inc. One issue with this material is that it is very porous and can lead to NO2 artifacts that change with time. To account for these artifacts, an artifact check sequence was implemented during the CABINEX campaign.

In this procedure, the inlet is overflown with zero air for 2.5 minutes and the NO2 lamps are turned on and off every 30 seconds. An increase in the NO signal indicates photolytic conversion of surface absorbed compounds to NO or NO2. This artifact is then logged and subtracted from the final NO2. During CABINEX the artifact ranged from 12 to 182 pptv and was highest when ambient NO2 was elevated. This artifact should be investigated and thoroughly vetted by future work. As mentioned previously the BLC is not a pure converter, it is more of an equilibrium shifter that pushes NO2 to NO.

However, in highly oxidative environments, care must be taken to ensure that the

27 calibration of the converter occurred in a representative period. Per equation 11, a calibration of the NO2 converter taken in low Ox would not be valid for one taken in a high Ox environment and vice versa.

In a normal deployment, the converters are located in the air being sampled and there is no need to calculate conversion of NO to NO2. However in the CABINEX study samples were pulled from three heights at flow rates up to 75 SLPM and then sub sampled through the NOxy inlet. The maximum transit time in the tubes was 9 seconds.

This is long enough that NO may be converted to NO2 in the sample tubing by reaction with ambient O3. For CABINEX it was estimated that less than 10% of the NO was reacted at the highest ozone mixing ratios and no corrections were made to the data. The actual conversion may be less since NO2 can be photolyzed in the PFA sample lines running down from the tower to the instrument trailer. The maximum worst case error in

NO2 from the sampling was calculated to be 10%.

2.2. PRECONCENTRATION GC-ITMS/FID

2.2.1. Instrument Description

The gas chromatograph-ion trap mass spectrometer with flame ionization detector

(GC-ITMS/FID) used in this work utilizes a custom built two channel pre-concentration system designed to measure C2 to C12 organic compounds emitted from mobile and biogenic sources, however it will also be able to measure other VOCs with in this volatility range from a variety of sources. This range of compounds is important because it spans the range of VOCs contained in gasoline which impact the chemistry of urban and rural environments (Lough et al., 2005; Gentner et al., 2012). Non methane hydrocarbons (NMHCs) are a major source of anthropogenic O3 which is a primary

28 pollutant and the major source of the hydroxyl radical in the atmosphere (Altshuller,

1983; Sillman et al., 1995). C2 to C6 hydrocarbons, especially the alkenes are important

O3 precursors, while C6 to C12 VOCs include both O3 and secondary organic aerosol

(SOA) precursors. Aromatic compounds in gasoline make up the entirety of SOA precursors, having both sufficient reactivity and yield to produce SOA (Odum et al.,

1997a; Odum et al., 1997b). Intercomparisons of NMHCs have produced a wide variety of results from the participants (Selmr et al., 2002; Apel et al., 2003; Rappenglück et al.,

2006) owing to the wide variety of techniques used to detect NMHCs. Apel et al., (2003) concluded that the techniques that produced the most reliable results used FID detectors and cryotraps packed with glass beads and that great care must be taken when using solid adsorbents to address artifacts and losses. Rappenglück et al. (2006) concluded that only

18 of the 76 compounds tested were accurately identified by 50% or more of the 30 labs.

Thus, the state of NMHC measurements is tenuous and in need of improvement.

The two most common methods of online pre-concentration involve either the use of cryogens or adsorbent traps. Cryogenic trapping of NMHCs is a proven method for concentration of NMHCs (Pollak et al., 1991; Westberg and Zimmerman, 1993; Apel et al., 1994; Apel et al., 1999; Helmig 1999; Plass-Dülmer et al., 2002). However, a number of systems have been designed to be cryogen free employing solid adsorbents to pre- concentrate NMHCs (Holdren et al., 1998; Lamanna and Goldstein, 1999; Tanner et al.,

2006). Solid adsorbent beds have issues with complete recovery and artifact generation

(Harper 2000). Tenax GR the adsorbent used in channel 2 of this system suffers from reaction with O3 and nitrogen oxides to produce acetophenone and benzaldehyde (Neher and Jones 1977; Walling et al., 1986; Pellizzari et al., 1984). The generation of

29 benzaldehyde is particularly problematic because it is a high yield photoproduct of aromatic oxidation by the hydroxyl radical. Reactions with O3 and highly reactive compounds such as terpenes are known to occur on Tenax adsorbent beds (Cao and

Hewitt 1994; Calogriou et al., 1996). Despite the challenges presented with the use of solid adsorbents, there are significant advantages making them preferable for unattended long term use in a variety of locations. The first major advantage is that adsorbent materials are cryogen free. Cryogens are cumbersome and expensive and are not easily available in remote locations. This system is designed for use with Washington State

University’s Mobile Atmospheric Chemistry Laboratory (MACL) which can be deployed in almost any location. The lack of cryogen is highly desirable for use with the MACL and solid adsorbents were chosen for this two channel adsorbent based system. Channel 1 uses Carboxen 1000 (Sigma Aldrich, Co. 60/80 mesh) following the recommendation of

Pollman et al. (2006). Channel 2 uses Tenax GR because it is capable of trapping biogenic VOCs including sequiterpenes (Helmig et al., 1999; Helmig et al., 2004; Helmig et al., 2007.) and also Aromatics and C6 to C12 alkanes (Harper 2000).

Each channel is composed of three main components: a sampling system, a GC, and a detector. The GC is a Varian, Inc. CP 3800 and the mass spectrometer is a Varian

Saturn Ion Trap. The sampling system removes a large fraction of the water vapor from ambient air using a cold tube water traps and traps VOCs from ambient air samples using adsorbent materials. Analyte detection is performed with a FID detector on channel 1 and with the ion trap mass spectrometer (ITMS) using external electron impaction ionization (EI) in channel 2. Both of these techniques will be described thoroughly in section 2.2.2. The system is installed into a custom rack and table system that allows it to

30 be deployed to the field as a unit. Figure 2.4 is a major component schematic, which shows the front of the system. The pre-concentration system with the cold traps and the valves sits above the GC. To the right of GC is the ITMS. There are three rack mount boxes located in the bottom right of the system which control the valves and the temperature zones. The bottom left contains the immersion cooler and the vacuum and sample pumps are located bottom center.

The pre-concentration system has two channels so that it can quantify the entire range of gasoline exhaust emissions (C2 to C12 VOCs). Channel 1 was designed to measure the C2 to C6 light hydrocarbons while channel 2 covers C6 to C12 VOCs including aromatic and biogenic compounds. The difference in volatility between the compounds analyzed by the two channels requires a different approach and necessitates two separate channels. Each channel has been designed with different sample handling parameters, trapping conditions, adsorbent materials, and chromatographic columns.

Channel 1 uses a FID detector, a Carboxen 1000 adsorbent, and an Al2O3/KCL Plot column. Channel 2 uses the ITMS detector, Tenax GR adsorbent, and a DB-624 column.

The three electronics boxes control heating, flow control and the gas flow path of the pre- concentration system for both channels. A plumbing diagram of the pre-concentration system is shown in Figure 2.5. The core of the plumbing is three, high temperature, 2 state, multiport valves (Valco Instruments Company, Inc.). Two of the valves are 12-port

(V1 and V3) and one valve (V2) is a 10-port valve. The valves are labeled in Figure 2.5 and elsewhere as V1, V2, and V3 and will be referred to in this document as the

“injection valve”, the “flush valve” and the “sample valve” respectively. All three valves are controlled by electronic actuators (Valco Instruments Company, Inc.) which are

31 controlled by the DAQFactory data acquisition and control software (Azeotech, Inc.,

Ashland OR). The actuators move the valves back and forth between positions, “A” and

“B.” These positions are not consistent between the 10 and 12 port valves. For example, when the 12-port injection or sample valve is in position “A” port 1 connects to port 2, port 3 to port 4 and so on. When it is in position “B” port 1 connects to port 12 port 11 to port 10 and so on, however the when in state “A” the 10-port flush valve connects port 1 to port 10, port 9 to port 8 and so on. When in position “B” the flush valve connects port

1 to port 2, port 3 to port 4 and so on. The three valves are temperature controlled at 80 oC with temperature controllers (Watlow, Inc., MO, PN: SD6C-HKUA-AARG).

Siltek/Sulfinert® treated stainless steel tubing (Restek Co. Bellefont, PA, PN: 22503),

1/16” O.D. 0.02” I.D. 304 grade, was used throughout the system. The tubing in contact with the analytes was heat traced and temperature controlled at ~80 oC with temperature controllers (Watlow, Inc., MO, PN: SD6C-HKUA-AARG).

There are also three two port valves and four three way valves which, depending on their combination of states, determine the flow of gases through the water trap. All except the helium flush valve are shown in an expanded view of the cold-tube water trap inlet (Figure 2.6). Each channel has a vent trap valve, a flush trap valve, and a sample flow valve, which are clearly visible as red circles. On channel 2 there is an ozone trap consisting of a glass fiber filter coated in Na2S2O3. The ozone trap was used because monoterpenes are highly reactive with O3. On the light hydrocarbons channel, an ascarite trap is placed in line to remove CO2 which can cause chromatographic problems on

Al2O3/KCl Plot column separations. The flush trap valve and vent trap valves are 3-way valves (Parker Hannifin Corp., Bakersfield, NJ, PN: 009-0269-900) that work together to

32 direct flow of either sample air or dry N2 through the cold tube water traps and are controlled in tandem by the control box. The common of the flush and vent valves are connected to the cold tube water traps and the normally opens are connected to the sample inlet (vent trap valve) and the multi-port injection valve (flush trap valve), while the normally closed ports are connected to dry N2 (flush trap valve) and vented to atmosphere (vent valve). The sample flow valves are two port valves (Parker Hannifin

Corp., Bakersfield, NJ, PN: 009-0270-900) which are either open (active) or closed

(inactive). The instrument modes are described later in detail, but to highlight the operation of the cold tube water trap valves, an example is presented. When the instrument is in SAMPLE mode or CONDITION TRAP mode, the vent and flush valve are inactive, connected to the normally open port and the sample flow valve is activated and open allowing a flow of ~100-150 sccm from the instrument inlet to be pulled through the cold tubes. Not depicted in Figure 2.6 is the helium flush valve. The helium flush valve is a two way valve (normally closed) that plumbs helium carrier gas into the multiport flush valve (Figure 2.7). The flow through the flush valve is set with a needle valve to ~ 100 sccm.

The low temperature control in the cold tube water traps and the adsorbent traps is achieved with a “cold block” which is an aluminum block chilled to -50oC with a Flexi-

Cool Immersion Cooler (SSP Scientific, SP industries, Inc., MN: FC55). The aluminum block is a series of three plates that have been machined to accommodate the cold probe and traps when screwed together. Once assembled the cold block is insulated with 1” thick polyisocyanate rigid foam insulation which is glued in place using expanding foam insulation. The insulation is further sealed with silicone room temperature vulcanization

33 and then the entire block is wrapped in duct tape. A 3-D drawing of the cold trap assembly that was used to manufacture the parts for this component is shown in Figure

2.8. The adsorbent traps are mounted vertically and the water traps are mounted horizontally. In this drawing, the top and front insulation has been hidden and the front cold block is transparent to show the water trap tubes, which are 1/4” O.D. 316L grade stainless steel tubes, Sulfinert® coated (Restek, Bellefonte, PA). A more rigorous description of the adsorbent traps and their assembly is present later.

2.2.1.1. Operation Modes

The pre-concentration system operates in six modes. Each mode performs a specific task necessary for the functioning of the pre-concentration system. Table 2.1 contains the valve states and temperatures for each mode. In the DAQFactory program the modes control the valve states alone and do not control the temperatures, with the exception of Safe Mode which also sets the temperatures to safe values for extended idle of the system. This allows the user to switch modes during trouble shooting the flows without demanding extreme temperatures from the heating system. The temperature are set in the master sequence (Master_sequ) which is a DAQFactory sequence that sets the valve states by running subsequences that change valve settings and directly set the temperatures. The heating described in the following sections is for normal running operation.

2.2.1.1.1. Condition Mode

This mode conditions the water traps by bringing them to their sampling temperature (-50 oC for channel 1 and -30oC for channel 2) and drawing ambient air

34 through them at approximately 100 to 150 sccm per channel (Figure 2.9). In practice the channel 1 tube is only able to reach a temperature of -45 oC with humid air flowing through it. The flows through the water traps are controlled by rotameters. The adsorbent traps are isolated from the system with the sample valve in position “A” and the carrier gas flows directly through the injection valve to the columns. While isolated, the adsorbent traps are chilled to their sampling temperatures of -30 oC (channel 1) and -

10 oC (channel 2). The ambient air sample is pulled through all of the plumbing that is used during sample mode except the taps and the plumbing connecting them to the sample valve allowing this to equilibrate with ambient VOCs. Condition mode runs for

15 to 25 minutes. Ample time is needed allow the tubes to come to equilibrium and for the VOCs to pass through, however if the air being sampled is extremely moist, care must be taken to ensure that the tubes do not clog with ice.

2.2.1.1.2. Sample Mode

The plumbing during sample mode and condition trap mode are identical except that for sample mode the sample valve is in position “B” rather than “A.” In sample mode ambient air having passed through the water traps is pulled through the adsorbent traps to trap VOCs from the air stream on the adsorbent. The carrier gas flows from the injectors and directly through the columns. Figure 2.10 shows the plumbing and valve configuration during sample mode. 30 sccm of air is pulled from the dry condition ambient air flow through the cooled adsorbent traps. The sample air flows are controlled by 100 sccm flow controllers (MKS Instruments, Inc.) which are connected to a diaphragm pump (KNF Neuberger, model # 811). The sample flow can be adjusted and the flow is recorded by DAQFactory to determine total sample volume. The sample

35 flow controllers are set to 30 sccm, however the traps provide resistance and it is possible to see flows as low as 25 sccm. If the flow through the traps is less than 25 sccm or unstable, then the trap should be removed and replaced or rebuilt so that ample sample can be collected. Sample mode lasts for 10 minutes so under ideal conditions, the total sample volume is 300 cm3. During a normal run condition trap mode runs for 15 to 25 minutes prior to sample mode so the Tenax-GR adsorbent trap used in channel 2 has equilibrated at -10 °C and the Carboxen 1000 trap in channel 1 has equilibrated at -30 oC before the instrument switches to sample mode. The adsorbent traps should remain at the above temperatures for duration of sample mode.

2.2.1.1.3. Cold/Hot Flush Mode

Cold Flush Mode purges the cold (sample temperature -30 oC and -10 oC for channel 1 and channel 2, respectively) absorbent traps and sample transfer lines with

UHP helium immediately after sample collection. This is done to remove traces of oxygen which can react with and destroy the adsorbent material when it is heated during thermal desorption. The Cold flush mode lasts for 2 minutes and in that time the traps are flushed with 60 mL of helium. After thermal desorption mode, the instrument switches back into Cold Flush Mode for two more minutes and the adsorbent traps are heated to

310 oC and 220 oC channels 1 and 2 respectively. This is the Hot Flush Mode. It is done to clean the traps and to prepare them for the next sample. A plumbing diagram for

Cold/Hot Flush Mode is shown in Figure 2.11.

2.2.1.1.4. Thermal Desorption Mode

36 In Thermal Desorption Mode the valves switch to pass the carrier gas through the traps and to the columns while simultaneously the adsorbent traps are rapidly heated to desorb the collected organics into the carrier gas flow (Figure 2.12). During Thermal desorption mode, the traps are heated to 300oC and 200oC for channels 1 and 2 respectively. The duration is 2 minutes and the carrier flow is increased to 12 ml/min in channel 1 and 7 ml/min in channel 2 using the pressure pulse feature of the GC carrier gas pressure control. The pressure pulse is activated through the GC software methodology. For channel 2 the head pressure of the injection port is increased to 25 psig and for channel 1 it is increased to 30 psig for 2 minutes resulting in an increased flow to more rapidly purge the trap.

2.2.1.1.5. Back Flush Mode

Back Flush Mode prepares the cold tube water traps for the next run by removing

o any water that has frozen in them by heating them to 60 C and pushing clean, dry N2 gas through them at approximately 300 sccm per channel (Figure 2.13). The carrier gas flows from the injector directly onto the column and the adsorbent traps are isolated and the

o adsorbent traps are held at 80 C. The N2 flow is controlled and monitored with two rotameters located on the back of the pre-concentration table. During back flush mode, the flush and vent valves are turned on, enabling N2 to enter and exit the cold tube water traps. The vent valves allow the humid and potentially dirty N2 to exit the system without contaminating any downstream measurements that are attached to a common inlet.

37 2.2.1.1.6. Safe Mode

Safe Mode is enabled when the instrument is going to sit idle without being powered down and the instrument can be put into Safe Mode during certain repairs. It is essentially the same as Back Flush Mode, with the vent and flush valves turned off: the adsorbent traps are isolated from the system, the carrier gas goes straight from the injector to the column through injection valve, and the 30 sccm of air per channel is pulled through the water traps and plumbing to the sample mass flow controllers. Safe mode also sets the water trap temperatures to 60 oC and the adsorbent trap temperatures to 80 oC.

2.2.1.2. Adsorbent Traps

The adsorbent traps are metal tubes filled with a specific mass of adsorbent material held in place by plugs made from silanized glass wool on either end. The channel 1 trap uses 80 mg of Carboxen® 1000 (Sigma-Aldrich, PN: 10478-U) while the channel 2 trap uses 100 mg of Tenax GR® (Scientific Instruments Services, Inc.,

PN:979402). A type K ungrounded thermocouple (OMEGA engineering, Inc., PN:

KMQSS-020U-12) runs through a 1/8” tee (Swagelok, Inc., PN: SS-200-3) to the center of the trap where the adsorbent is located and contacts the metal wall of the tube.

Because the thermocouple contacts the tube wall it is essential that the sheath be ungrounded to prevent damage to the data system. Figure 2.14 shows the trap assembly.

The metal tube has dimensions of 29.21 cm L x 0.305 cm O.D. x 0.262 cm I.D. and are made from grade 600 inconel (Microgroup, Inc. Bethel, MN, PN: 600 F3 0125 x 010 SL)

. Inconel 600 alloy is a nickel/chromium/iron (72% /14% to 17%/6% to 10%) alloy for applications which require resistance to heat and corrosive environments. This material

38 was chosen for its low coefficient of thermal expansion and retention of strength at high temperature relative to the other materials tested. Initially, 0.305 cm O.D. electro- polished 316L stainless steel was used for the traps. The main issue with the 316 stainless steel was that when heated it would expand and this lead to migration of the trap in the cold block over time and more importantly deformation of the trap tubes. Over time the use of copper electrodes and stainless steel lead to corrosion of the copper electrode. The copper-316L stainless steel interfaces lead to build up of copper oxide on the copper electrodes. The copper oxide has much higher electrical resistance than the rest of the system, which resulted in poor transmission of current and hotspots which could be seen as discoloration on the tubes and as blackened copper electrodes. Thus began the search for a more suitable material. The Inconel 600 - copper interface has not suffered from this corrosion. The 316L stainless steel traps were also much thicker and took longer to heat.

The light hydrocarbon trap is made from Carboxen 1000 while the VOC adsorbent material is Tenax GR. Tenax Gr is made from 2,6-diphenylene-oxide polymer resin and 30% graphite, with a 60/80 mesh size. Tenax GR has a low affinity for water, maximum temperature limit of 350 oC, a specific surface area of 24.1, and a density of

0.55 cm3 g-1, and is well suited to trap aromatic VOCs and monoterpenes. Tenax GR fills

3.75 cm of the trap tubes and the glass wool plugs fill another 1 cm each for a total of

5.75 cm of the 29.21 cm tube. Thus, that there is 1.37 cm3 of total dead volume in the trap. The channel 2 trap is almost identical except it houses 80 mg of Carboxen 1000 rather than 100 mg of Tenax GR. Parts for a smaller 0.16 cm O.D. x 0.12 cm I.D. trap have been purchased and they should be tested as they will likely improve the

39 performance of the system by decreasing the dead volume and the heating time, both of which should improve the chromatography.

2.2.1.3. Trap Temperature Control

The adsorbent traps are housed in a cold (-50 oC) aluminum block along with the water traps. Their temperature is controlled by direct electrical heating (AC) and is monitored by type K thermocouples located in the adsorbent bed, which are connected by a LJTick-InAmp (Lab Jack, Inc. Lakewood, CO). The LJTick-InAmp amplifies the thermocouple signal which is then read as an analog input of the U3 LabJack and converted to a temperature using a predefined conversion in DACFactory. A proportional, integral, derivative, (PID) loop running at 10 Hz in DAQFactory controls a

0-5 VDC output which controls the output of a 3 VAC , 20 A, @ 60 Hz transformer using a phase modulation circuit (NU Wave Technologies, Inc., PN: SSRMAN-1P). The

SSRMAN-1P has a compact form factor mounting directly on to a standard, brick type, solid state relay (SSR). The phase modulation circuit is the AC equivalent of the DC pulse width modulation (PWM) circuit. It controls the phase angle of the AC current and delivers power proportional to the 0-5 VDC input. Figure 2.15 shows a schematic of the temperature control of the tubes including the components making up the trap heating system. The main components are the trap tubes, the LJTick-InAmp, the Labjack U3, the

PID loop, the transformer and the phase modulation circuit. This is a novel approach and relatively inexpensive solution to rapid heating because it doesn’t require large batteries or capacitors used in PWM circuits.

The adsorbent trap temperatures are always controlled by the PID loop. The only thing that changes to change the trap temperature is the trap temperature set point. The

40 PID parameters for both channels are P(Gain) = 0.25, I (Integral) = 0.35, and D

(Derivative) = 0.005. Attempts were made to utilize the auto tune feature within the

DAQFactory PID controls to set the PID parameters however, they did not work. The parameters being used were selected through trial and error. Trap heating during thermal desorption mode is rapid and reproducible. During Desorption Mode channel 1 and channel 2 traps are heated from -30 oC and -10 oC to 310 oC and 200 oC, respectively.

This takes approximately 12 seconds for channel 1 and 9 seconds for channel 2. The initial heating is highlighted by a red box in Figure 2.16. which shows the initial temperature ramps for both tubes during 5 separate runs. The heating is very reproducible showing less than 2% deviation during the initial temperature ramp and less that 5% deviation during the temperature hold. Channel 1 shows some oscillations at high temperatures, which is an indication that the PID parameters for channel could be further adjusted. However, this is a minor concern because the compounds should still be desorbed at these high temperatures.

2.2.1.4. Water Trap Temperature Control

The two cold tube water traps are individually temperature controlled with temperature controllers (Watlow, Co., PN: SD6C-HKUA-AARG) which communicate with the DAQFactory control and acquisition software using Modbus protocol on a RS-

485 serial bus. A 4 port USB to serial box (SeaLevel Systems, Inc., PN: 2403) connects the RS-485 serial connections from the controllers to DAQFactory. The temperature of the tubes is read and set using the RS-485 serial communication. This communication protocol is setup and accessed for monitoring/trouble shooting as a device in

DAQFactory under the device configuration button. Temperatures are measured by the

41 Watlow controllers with a type K thermocouple located inside the water traps. The tubes are heated with with 32 AWG polyamide coated heating wire (California Fine Wire Co.,

Stablohm 675) wrapped around the tubes. When fully powered the tubes deliver approximately 90 W of heating to the tubes and surrounding air. During backflush mode, the tubes are heated to 60 oC and are cooled to -45 oC and -30 oC for channels 1 and 2 during condition and sample mode.

2.2.1.5. DAQFactory Control Program

The pre-concentration system was automated with software / hardware interface utilizing DaqFactory (Azeotech) software and a LabJack U-3 and U-12 hardware interface. The Daqfactory control program sets valve states, trap temperatures and auto starts the GC. Autostart of the GC involved the use of a mechanical relay which connected where the mechanical switch that starts GC runs when using the auto sampler.

The instrument can sense when a connection is made across the two wires and this triggers the start of the method. Real time graphs of these variables allowed for monitoring of air flows and trap temperatures and valve states. Several variables were recorded including sample air flows, trap temperatures, trap heating variables, all temperatures and all valves states. These were recorded for later use in analyzing the pre- concentrator performance. The sample flow rates were needed to determine sample size.

Even though the data have different timings, the data are logged in one file by averaging to a fixed 1 s interval. The channels logged are shown in Table 2.1.

42 2.2.2. Principle of Measurement

2.2.2.1. FID

The light hydrocarbon, C2-C6, channel uses a flame ionization detector (FID).

The FID measures ion current created when organic compounds eluting from the GC column are burned in a hydrogen / oxygen flame. The FID is a mass selective detector.

This detector was pioneered by two independent groups at approximately the same time

(Harley, 1958; McWilliams and Dewar, 1958). FIDs have become widely used because they have many advantages including: low background noise (pure H2/O2 flames do not produce ions) linearity over a large dynamic range, and a predictable response to hydrocarbons (Holm 1999). A FID consists of hydrogen and oxygen flame in which entering hydrocarbons are simultaneously “stripped” of hydrogen atoms and the carbon backbones are broken apart or “cracked.” The primary pathway for ion creation is the production of the oxymethylium ion (CHO+) which is produced in the following chemi- ionization reaction (Calcote 1963):

퐶퐻 + 푂 → 퐶퐻푂∗ → 퐶퐻푂+ + 푒− R10

The detector then collects the ions at 100% efficiency and produces a signal (Holm

1990). Holm (1999) states that one ion is produced for every 106 carbon atoms. The reproducibility of the FID signal to carbon atoms in aliphatic compounds lead to the introduction of the concept of effective carbon number response whereby the detector response is proportional to the number of carbon atoms in the compound. There are exceptions to this rule, for example carbonyl groups do not yield an FID response, so acetone with 3 carbon atoms should yield the same response as ethane that has 2 carbon atoms. Sternberg et al. (1962) noted that with the addition of functional groups and

43 increased molecular complexity the response differed from the carbon number of the compound. They proposed that the response to a particular functional group would decrease the observed response, and that this decrease could be predicted. Consequently they developed the concept of the effective carbon number (ECN). The effective carbon number concept was further studied by Scanlon and Willis (1985) who formalized the definition for ECN for a given compound, ECNi as the ratio of the ECN of the reference to a relative molar response factor:

퐸퐶푁푟 퐸퐶푁푖 = 2.9 퐹푟−푚표푙푎푟

The ECN of the reference is 1.0 by definition (Scanlon and Willis 1985). The relative molar response factor is defined as the ratio of the molar response of the reference to the compound:

퐴푟∗푀표푙푒푠𝑖 퐴푟푀푊푟푚𝑖 퐹푟−푚표푙푎푟 = = 2.10 퐴𝑖∗푀표푙푒푠푟 퐴𝑖푀푊𝑖푚푟

Where 퐴푟 and 퐴푖 are the integrated area response, 푀푊푟 and 푀푊푖 are the molecular weights, and 푚푟 and 푚푖 are the masses injected. The subscripts denote either the reference r or the compound i.

The absolute response of the FID used in the system was calibrated by injecting calibration standards directly on to the Al2O3/KCL Plot column. The Plot column is the same as the one described earlier in the system description. Flow through the column was constant at 6 sccm throughout the analysis. The temperature program differed significantly from the standard method described later in this chapter. The method used for the FID calibration was designed to produce rapid, by chromatographic standards, data, completing a run and re-stabilization of the oven in 20 minutes. The oven program

44 is shown in Table 2.2. The FID parameters used did not differ from those described in the instrument description.

To calibrate FID response, a sample loop with a precisely known (±2%) volume was attached to the injection valve. A 500 µL sample loop was used and filled with calibration gas mixtures at a nominal 15 ppmv mixing ratio (Scott Specialty Gases). A low flow of sample gas was used to fill the sample loop, the flow was stopped, and the valve then rotated into injection position where the GC carrier gas at a flow rate of 6 sccm swept the contents directly onto the column. Because the sample loop volume, pressure, and loop temperature determine the number of molecules injected, the requirements on the calibration gas flow were not stringently controlled. Flow of calibration gas was controlled by a needle valve to between ~ 2 and 6 sccm. The main consideration with respect to the calibration gas flow was conservation of the standard gas. During sample load, calibration gas from the cylinder is sent through the sample loop and exhausted to the atmosphere, and the carrier gas from the injector flows through the valve directly to the column. Figure 2.17 is a plumbing diagram of the simple experimental setup, showing both load sample mode and inject sample mode.

The composition of the two standard gases is shown in Table 2.3. Sample loops

(Vici Valco, Inc., PN: SL100CW, SL250CW, and SL500CW), with volumes of 100, 250, and 500 μL respectively, were used to deliver a controlled volume of calibration gas onto the column. The resulting chromatograms from the three sample loops are shown in

Figure 2.18. Each gas was run 5 times for each loop size. The peak areas were integrated and the reproducibility of the runs was very high. The worst relative standard deviation (RSD) was 1.3%, and most RSDs were less than 1%. The loop size had

45 significant impact on the chromatography of the early eluting compounds, while the later eluting compounds were less sensitive to the sample loop volume. The compounds that elute early are poorly retained and so a large injection volume leads to a wider, less

Gaussian peaks, while the compounds that elute later are strongly retained on the column at the initial temperature of 40 oC and are thus refocused into a narrow band. Figure 2.19. shows the early eluting peaks and Figure 2.20 shows the late eluting peaks and their peak shapes. Using the temperature and the pressure in the lab, which were 293.15 K and

96.66 Kpa, the number of moles of gas injected were calculated using the ideal gas law:

푃푉 푛 = 2.11 푅푇

The reason the molar abundance rather than sample volume is used for the calibration of the detector is that the response factor to moles remains invariant with pressure and temperature changes in the lab. The number of moles of each analyte injected is then the total number of moles times the mixing ratio, χ, which is given in Table 2.3 as parts per million volume (ppmv):

푃푉 푛 = ∗ 휒 2.12 푖 푅푇 푖

The number of molecules of each analyte was calculated by multiplying the moles by

Avogadro’s number (6.022 x 1023). Because the FID responds to carbon atoms, the number of moles of carbon was calculated by multiplying 휂I by the carbon number.

Dividing the area counts by the number of molecules gives the absolute response of the detector. Figure 2.18 is a plot the FID response against the number of molecules of carbon. The average response of 2.5 *10-10 counts * molecules carbon -1 was calculated with a linear fit of the Scott Marrin calibration standard shown in Figure 2.21. Figure

2.22 shows the FID response in area counts divided by the number of carbon atoms for

46 each the 9 analytes. The dashed line indicates the average response to the Scott-Marrin standard, which was chosen because it has a 2% accuracy while the Scotty standard has a

10% accuracy. The majority of the gases in the Scotty standard give slightly lower signal than the Scott Marrin average, but this can be accounted for by the uncertainty of the standard. However, acetylene and propyne break this trend. Propyne’s response is too low to be accounted for by the uncertainty and acetylene’s response seems too high, even though it is within the uncertainty of the measurement. For acetylene it is necessary to invoke the effective carbon number when calculating the response. When the effective carbon number was accounted for the compounds calibrated within the uncertainty of the standards. The FID response is similar for these compounds, demonstrating that it is reasonable to use single response factor to calibrate FID sensitivity for the light hydrocarbons (principally C2-C6 alkanes, alkenes) that will be measured on channel 1.

The light hydrocarbon channel has been problematic. Experiments have been performed testing the desorption temperature from 300 to 370 oC with little improvement in recovery of heavy compounds. For this test, the water trap was bypassed and 2.1 sccm of the Scotty standard was diluted in 240 sccm of chromatographic grade ultra-high purity helium. Figure 2.23 shows the small changes in recovery at 300, 335, and 370 oC.

Methane, ethane and ethylene are fully desorbed showing no temperature changes however propane and propylene show strong temperature dependence and the compounds after are not desorbed. Methane is poorly adsorbed but that is expected. Desorption of the heavier compounds at temperatures of 335 oC and 370 oC should occur. Pollman et al.

(2006) reported full desorption from Carboxen 1000 at temperatures of 310 oC. The issue with desorption of heavier compounds is likely due to carrier gas flow problems in the

47 traps. An indication that the there is a flow issue is that the peaks are wide and delayed by almost 2 minutes which is the duration of desorption. Because the EFCs determine the flow through the column by the pressure that they measure at their outlets (the GC inject port), the addition of the adsorbent trap with a large pressure drop across it will result in a much lower flow through than reported by the instrument. During construction of the traps, which involves sucking the adsorbent into the tube, there was much more resistance to flow than the Tenax GR traps. At this point the light hydrocarbons channel cannot be considered fully functional. However, once the flow issue is sorted, the channel will be functional.

2.2.2.2. ITMS

The Varian, Inc. 4000 ITMS is a mass selective detector that utilizes an ion trap to sort ions by mass to charge ratio. The analytes are delivered either directly to the ion trap

(internal ionization mode) or the GC column is plumbing into an external ionization volume (external ionization) and the ions produced are directed into the ion trap by a series of ion lenses. Ionization is done by the standard method of electron impact ionization with 70 eV electrons. The ion trap works by storing ions in stable orbits inside the ion trap and then selectively scanning them to the detector. A turbomolecular pump creates a vacuum of approximately 15 µTorr to allow the trapped ions a sufficient mean free path to produce stable orbits. Figure 2.24 shows a top view of the ITMS and the most important components. The ion trap has some distinct advantages over a standard quadrupole mass spectrometer. The ITMS has a faster full range duty cycle because the ion trap collects and analyzes the full range of m/z for each scan, while the quadrupole has to scan each m/z individually. The high speed collection of a full range of m/z for

48 each scan gives the user added power to quantify analytes. Figure 2.25 shows how the

ITMS allows co-eluting peaks with different mass spectra to be separated. Both isoprene and trans-2-pentene elute at around 5.283 minutes forming what appears to be one peak when looking at the total ion count (TIC) (blue trace). However, with the analytical power of the ion trap they can still be separated and integrated because they have different mass spectra. Isoprene produces an m/z 67 ion, while trans-2-pentene produces a m/z = 70 ion. Integrating the peak area of the respective ion signals for m/z 67 and m/z allows each compound to be quantified. A quadrupole mass spectrometer cannot simultaneously collect ions of two different m/z ratios. In addition to separating co- eluting peaks with different mass spectra, the use of selected ions increases the signal to noise of the analyte. The ITMS has a lower trapping efficiency for lighter ions than heavier ions and this can cause significant differences in ion signal intensity for different compounds. For example, the system is 8.6 times less sensitive to n-pentane than to n- dodecane.

The ITMS is capable of collecting m/z = 0 to m/z = 1000, however in practice an upper and lower limit to ions collected is applied. Trace gases found in air are generally light with masses less than 250, which sets the upper limit for ions collected. The lower limit of the m/z range is set by the fact that the 4 most abundant gases in the earth’s atmosphere are N2, O2, Ar, and CO2 and the ion trap is not housed in a perfect vacuum so the trap is continuously contaminated by the above 4 gases. With a mass of 44, CO2 is the largest of the above four, however it makes up only 0.038% of the atmosphere and many small trace gases have ion fragments with m/z less than 44. After CO2, O2 is the next largest with a m/z of 32, and an abundance of 20.946%. The high atmospheric

49 abundance of O2 means that it must be excluded or else the trap is filled with m/z 32 ions.

The recommended trapping range is m/z 35 to m/z 250. For this range the background ion signal is dominated by m/z 43 and 44, however, it is worth having the ions at m/z 35 to 45 available for compound identification.

2.2.3. Sample Handling

2.2.3.1. Water Trap

The removal of water from both channels is accomplished with a cold tube water trap which has been described briefly in system description and instrument modes. The channel 1 cold tube water trap has to be run at the lowest temperature possible (-42 to -45 o C) to achieve maximum water removal because the Al2O3/KCL Plot column used in channel 1 is very sensitive to water vapor. This column has a solid phase so when water is present in the desorbed sample, it will adsorb to the active sites and drastically change the chromatography. Changes in water vapor produce instability in compound retention times on this column, which presents a major obstacle to compound identification because the detector is a FID. After the cold tube water trap, the sample passes through an Ascarite trap to remove both CO2 and any left over water vapor before passing through the valves and the cold Carboxen 1000 trap (-30 oC). The Tenax GR material has a low affinity for water and the DB-624 column has a gas-liquid interface. Analytes on the DB-624 are absorbed by the active phase, so the chromatography is not nearly as sensitive to water vapor. This meant that the operating parameters for the cold tubes could be optimized to maximize transmission of VOCs. In particular, the transmission of heavier aromatics through the water traps at low temperatures was a concern. Jobson and

McCoskey (2010) showed that losses of C4-alkylbenzenes occurred in this water trap at

50 low mixing ratios, however there were no losses during high ppbv range calibrations. To optimize the operating temperature for the cold tube water traps, humidified calibration gas was sampled. The water trap temperature optimization experimental setup is shown in Figure 2.26. Two streams of clean N2 (750 sccm) from the gas port on a liquid N2 tank were used to produce and approximately 50% relative humidity stream of gas by passing one gas flow through a partially filled horizontal glass tube filled with pure water and mixing the output with the dry gas stream. The humidified air is then mixed in a mixing loop with the PTR-MS calibration gas, approximately 1 ppmv per component (Scott-

Marrin, Inc), to produce approximately 1.3 ppbv mixing ratios. The gas was then pulled through the water trap. The water trap and adsorbent traps were then run at the three temperatures shown in Table 2.4. Other than changing the temperature of the water trap and the adsorbent trap the method was run under standard operating conditions and nothing else was changed during the runs. A minimum of three runs for each temperature was performed. The results for toluene, benzene, alpha-pinene, isoprene, styrene, p- xylene, 1,2,4-trimethylbenzene and 1,2,3,5-tetramethylbenzene are shown in Figure 2.27.

The error bars shown are the standard deviations of the three runs. The largest peak areas for most compounds occurs at -30 oC and -10 oC for the cold tube and the adsorbent tube respectively, suggest greater trapping efficiency. The design goal for the preconcentration system was to operate the Tenax adsorbent at lower temperatures to allow for more efficient trapping of volatile components such as C5 hydrocarbons like isoprene and C1-C3 halocarbons. Operating at lower temperatures runs the risk of water vapor condensation in the Tenax trap so the water trap temperature must be significantly colder than the Tenax. The problem is that cold water trap temperatures may cause loses

51 of lower vapor pressure organics such as monoterpenes and larger monoaromatics. An optimum set of trapping temperatures must be found to provide the greatest analytical range across compound volatilities to maximize the analytical power of the mass spectrometer. Based on this experiment, the recommended trapping temperatures are -30 oC (water trap) and -10 oC Tenax GR adsorbent trap. Further work must be done at lower mixing ratios to test water trap losses and to test trapping ability of less volatile compounds.

2.2.3.2. VOC Pre-concentration

It was noted while performing zeros that artifacts were being generated during desorption on channel 2. The source of the artifact is likely the Tenax GR material. To test the desorption temperature, the GC-ITMS pre-concentration instrument inlet was overflowed with clean, humidified, nitrogen gas for the entire experiment. The water vapor was removed from the sample by flowing through the water trap held at -30oC.

The gas then passed through the adsorbent tubes which were held at -10oC. Prior to desorption, the adsorbent tubes were purged with chromatographic grade helium at a flow of 30 sccm for 2 minutes to remove oxygen from the traps. The m/z trapping range of the ion trap was set to m/z 44 to 500 to allow for identification of any large compounds that might be generated as artifacts. The zeros samples were then desorbed at temperatures of

200, 220, 240, 260, and 280 oC for 2 minutes, and sent to the GC-ITMS for analysis.

Each desorption temperature was run twice. Benzene, toluene, silicic acid, styrene, octamethyl-cyclotetrasiloxane, benzaldehyde, benzofuran, benzonitrile, and cyclopentasiloxane were found in the chromatogram. With the exception of benzene, toluene, and styrene, which have retention times as well as mass spectra to identify them,

52 the reported compounds are a best guess based on potential sources and mass spectral match. The compounds that contain silicon are likely liberated from the silanized glass wool used in the traps, while the aromatic compounds and are from the decomposition of the Tenax GR. In particular benzene is being emitted continuously from the trap during desorption. Benzene is not well retained on the DB-624 column and rather than a peak it has a “mesa” shape that is nearly 2 minutes in duration, matching the desorption time.

Figure 2.28 shows chromatograms from this experiment desorbed at 200 oC and 280 oC.

This experiment was performed in external ionization mode. Based on the results, the optimum desorption temperature for the Tenax GR is 220 oC. Table 2.5 shows a summary of the peak areas. At 220 oC the temperature is sufficient to desorb the heaviest analytes and it also minimizes artifacts. Benzene artifact at this level is equal to approximately 25 pptv, however due to the shape of the benzene generated during desorption a real desorbed peak can still be quantified.

2.2.4. Calibration Gases

Multiple calibration gases from standard mixtures have been run on channel 2.

The calibration gases are referred to as the PTR-MS calibration gas (Scott-Marrin, Inc), the GCMS cal gas (Scott-Marrin, Inc), The Spectra Cal, and the gasoline calibration gas

(in house). The composition of each calibration standard and a sample chromatogram are presented in the following sub sections. To determine ambient mixing ratios, the sensitivities derived from calibration gases were applied to the data. The sample size was first determined from the recorded sample flow controllers and compared to the calibration sample size. If they were within 5% of each other, which they generally were, then no correction was applied, if not, they were corrected for. Then the background was

53 subtracted from the ambient peak size and divided by the sensitivity. The subsequent sections describe the standard gases used.

2.2.4.1. GCMS Calibration Gas

The GCMS calibration contains 32 compounds with a stated NIST traceable precision of ± 5%. Table 2.6. shows the compounds and mixing ratios in the GCMS calibration gas. Figure 2.29. shows a sample chromatogram of the GCMS calibration gas taken during the CARES field experiment. For calibrations, 10.1 sccm of GCMS calibration gas was diluted with 1085 sccm of clean N2 to produce an approximately 3 ppbv level calibration for each of the compounds. After the CARES campaign, multipoint calibrations were performed to determine the linearity of the system. Figure

2.30 shows the results of the span calibration for some selected compounds. Producing reliable low level mixing ratios (<100 pptv) is a challenging task because the gas must be diluted by 1000 times or more. This was part of the reason that the exponential dilutions described later were investigated.

2.2.4.2. PTR-MS Calibration Gas

The composition of the PTR-MS calibration gas which contains 14 compounds is shown in Table 2.7. The PTR-MS calibration gas is prepared by Scott-Marrin, Inc. to an accuracy of ±10% and is not certified. This gas was run on the GC-ITMS and compared to the GC-ITMS. Significant deviations between the two calibration standards were observed. For example, isoprene, benzene, and toluene are 38, 27, and 27 % higher in the

PTR-MS calibration gas than the GC-ITMS gas. 1,2,4-trimethylbenzene was 5% higher and the concentration of p-xylene was found to be 14% higher in the PTR-MS calibration

54 gas than the GCMS calibration gas. Figure 2.31 shows a sample chromatogram of the

PTR-MS calibration standard.

2.2.4.3. SPECTRA Calibration Gas

The Linde SPECTRA gas (Restek Corportation, PN: 34420) is a 56 component gas ± 10%, with a balance of N2. The composition of the trace gases contained in this standard along with the instrument sensitivity and minimum detection limits are shown in

Table 2.8. The ITMS can integrate peaks with a high degree of repeatability when the area counts of a peak are greater than 4000. Based on the sensitivities reported in the absolute calibration using the SPECTRA gas, the minimum detection limits are calculated by dividing 4000 by the sensitivity. The early eluting peaks from chromatograms of the Spectra gas injected directly using a sample loop and also injected with the pre-concentration system is shown in Figure 2.32. The sample loop has an internal volume of 250 µL and the pre-concentration system sampled 300 cm3 of the

Spectra gas diluted to approximately 4.5 ppbv mixing ratios. The sensitivities of the

ITMS in internal and external mode are drastically different, with internal ionization being a factor of 10 more sensitive than external mode. The chromatograms illustrate the difference in the chromatography resulting from direct sample loop injection and adsorbent trapping. The peaks from the pre-concentration system are much wider and the retention times are later than the direct injection. This indicates that the carrier flow through the adsorbent trap is diminished during desorption. The EFCs in the injectors determine flow based on the output pressure and the column dimensions. If there is significant pressure drop before the column then the EFC will report an erroneously high

55 flow through the column. This is an issue that should be investigated further to improve performance of the system.

2.2.4.4. Gasoline Standard

A gaseous mixture of unleaded, ethanol free 87 grade gasoline was prepared by placing 1 ml of gasoline in an evacuated aluminum cylinder, size 150A (29-30L). The cylinder was then pressurized to 900 psig with ultra-high purity nitrogen gas. This mixture was allowed to sit for one week with the bottom heated to 70oC to mix the gases.

This gasoline standard was then mixed at a flow rate of 85 sccm with 3.8 LPM of zero air and 0.8 LPM of N2 for a total flow of approximately 4.7 LPM. This mixture was directed to the photo oxidation chamber and sampled with the GC-ITMS. Figure 2.33 shows a chromatogram of the gasoline. The GC-ITMS was able to resolve 127 distinct peaks the largest 25 of which were identified. Future work will include identification of the rest of the peaks.

2.2.4.5. Zero

Zeros were performed with N2 from the headspace of a liquid N2 tank and with a custom zero air generator which consists of a platinum on alumina catalyst heated to 310 oC. Sample chromatograms of each of these are shown in Figure 2.34. The two zeros are drastically different, mainly due to the fact that the chromatograms are from internal and external mode. The background is much higher in internal mode along with the sensitivity. The method used in internal mode doesn’t turn the ionization on till after the

CO2 peak which elutes at 2 minutes has passed. The peaks in internal mode are due to

Tenax GR and glass wool decomposition.

56 2.2.5. Method

The difference in volatility of the compounds being analyzed by each channel requires the use of two separate columns. The light hydrocarbon, channel 1 (C2 to C6), utilizes a 0.32 x 50 m Al2O3/KCL Plot column (Varian, Inc) and channel 2 uses a 0.32 x

60 m DB-624 column (J & W Scientific). The Plot column uses a solid phase and the

DB-624 has a liquid phase with a film thickness of 1.80 µm. Flow of helium carrier gas onto the columns is controlled by the CP-1177 injectors (Varian). The front injector controls flow through channel 1 and the middle injector controls flow through channel 2.

The CP-1177 employs a type 1 electronic flow controller EFC, which allows them to operate in split/splitless mode as they consist of mass flow controller and a downstream pressure control valve. The flow through the column is determined by the pressure at the column entrance. The injector flows are operated in constant flow mode with flows of

6.0 mL min-1 (channel 1) and 1.0 mL min-1 (channel 2). During the thermal desorption of the analytes onto the column the pressure pulses of 25 psig and 50 pisg with durations of

3 minutes are used for channel 2 and channel 1, respectively. The injectors are operated in split mode with a split ratio of 50 for the first 3 minutes and then are set to splitless for the duration of the run. The greater gas flow in split mode is required to enable the rapid pressure increase of the injector port when performing pressure pulse injections. The temperature program consists of 4 steps: an isothermal hold at 40 oC for 1 minute followed by a temperature ramp of 2 oC min-1 to 80 oC. The temperature ramp is then increased to 4 oC min-1 and held at this rate until 170 oC. Finally, the temperature ramp is increased to 20 oC min-1 and the temperature is brought to 200 oC and held for 15 minutes.

57 o -1 The FID is operated at 300 C with a makeup gas flow rate of 25 mL min , H2 flow of 30 mL min-1, and an air flow of 300 mL min-1. For analysis of ambient VOCs the

ITMS should be operated in internal mode because it has approximately 10 times the sensitivity of internal mode. However during CARES this was not known and the system was run in external mode. The recommended temperatures are: trap temperature 80 oC, transfer line temperature 220 oC and a manifold temperature of 50 oC. The electron multiplier voltage is set during auto-tune and generally ranges from 1500 to 2500 V.

There are 3 ITMS operating parameters used during a run. The first from 0.00 to 3.00 minutes has the trap and ionization off. This is because there is a large peak m/z 44 that elutes early during this period and makes quantative analysis impossible. From 3.00 to

4.50 minutes the trap and ionization (10 µA filament current) are turned on and m/z 50 to

250 are collected. At 4.50 minutes and for the duration of the run the trap and ionization

(10 µA) are turned on and m/z 35 to 250 are collected. The target total ion counts (TIC) for the trap is set to 20,000 when the trap and ionization are turned on. During sample mode approximately 300 ± 5% standard cm3 of air sample is collected. A list of compounds that have been identified along with retention times and quantitation ions is shown in Table 2.9.

2.3. CHAMBER EXPERIMENT

2.3.1. Characterization of Chamber Mixing

The concentration in a well-mixed flow through reactor can be described by equation

2.13:

푆⁄ +휂∗퐶 퐶(푡) = 푉 𝑖푛 (1 − 푒−(휂+푘)푡) + 퐶 (푒−(휂+푘)푡) (2.13) 휂+푘 0

58 Where S is the source term, V is the volume of the chamber, 휂 is the air exchange rate determined from the air flow rate through the chamber divided by the chamber volume,

Cin is the concentration of the analyte in the air entering the chamber, C0 is the concentration at time t=0, Ct is the concentration of the analyte inside the chamber and exiting in the outgoing air flow, and k is a first order loss term that can be related to photochemical destruction in the chamber. If the incoming flow is free of analyte, and there is no source term and no chemical loss, then equation 5 simplifies to equation 6, which describes the dilution of a well-mixed chamber filled with initial analyte concentration Co:

−휂∗푡 퐶(푡) = 퐶표푒 (2.14)

This experiment determines if the chamber can be described as well-mixed for flows ranging from approximately 5 to 22 LPM, yielding residence times ranging from 6.7 to

31 minutes. The chamber is currently equipped with two small, 12 VDC fans constructed from unknown plastic material. One objective of this experiment was to evaluate the performance of the fans under differing flow conditions. Because the fans are made of an unknown material they may emit VOCs or VOCs may stick to them. If they are needed to mix the chamber then they will be used, if not they will be removed. The 148.2 L,

PFA Teflon, photo-chamber described in this study will be used for a number of photo- oxidation and dilution experiments. Understanding mixing in the chamber is required for successful completion of subsequent dilution and photo-oxidation experiments. The chamber is a PFA Teflon cylinder with plexiglass end plates. The dimensions are approximately 57 cm length by 56 cm diameter. Gas enters the chamber through perforations in a 1/4” O.D. PFA Teflon tube approximately 0.25 m long at the bottom of

59 the chamber. The perforations are spaced approximately 1 cm and spiral axially along the tube. Gases vent through an exhaust port and an instrumentation manifold. The exhaust is 8 cm from the top of the chamber on the same plexiglass end as the inlet. Air sample from the middle of the chamber is pulled from a tube that penetrates through the opposite end.

The mixing ratio of nitric oxide inside the chamber was measured using a Thermo

Fischer Scientific, Inc. model 42 NO Chemiluminescent analyzer, and an Air Quality

Design, Inc. chemiluminescent NO detector described in earlier in this document (section

2.1). For each dilution, the chamber was filled to steady state with a mix of nitric oxide and pure air. Flow of pure air was measured with the dry cal flow meter and regulated with the needle valve of a rotameter. NO from a Scott-Marrin, Inc. Riverside CA cylinder 500 ppmv ± 5% was introduced to the flow using a MKS Instruments, Inc.

MASS_FLO® controller. The NO flow was verified using the Alltech, Fischer

Scientific, Inc. Digital Flow Check. The NO and pure air were pre mixed prior to introduction to the chamber in a 50 ml, PFA volume. Once the chamber reached steady state, the NO cal gas was removed and the chamber was purged with zero air. The pure air flows tested were 22.2, 11.4, and 4.8 L min-1. For each flow, a dilution was performed once with the fans on and once with them off. The observed η was calculated using an exponential fit of the data in Igor Pro. The exponential dilutions performed with

NO demonstrate that the chamber is well mixed for flows ranging from 22.2 L min-1 to

4.8 L min-1. η is calculated by diving the flow rate by chamber volume and is referred to as “calculated η”. The flow rate is measured at ± 5% accuracy and the chamber volume is estimated to be accurate to ± 10 %. Figure 2.35 shows the results for the chamber

60 mixing. For the 22.2 L min-1 dilution flow the observed η was 2.45 x 10-3 ± 4.4 x 10-5 s-1 with the fans on and 2.43 x 10-3 s-1 with the fans off and the η calculated from the chamber volume and flow rate was 2.5 x 10-3 s-1 ± 11%. Observed η for 11.4 L min-1 was

1.22 x 10-3 ± 9.8 x 10-6 s-1 with fans on and 1.19 x 10-3 ± 1.1 x 10-5 s-1 with fans off and the calculated η was 1.25 x 10-3 ± 11% s-1. The η observed with a flow of 4.8 L min-1 was

5.40 x 10-4 ± 3.6 x 10-8 s-1with fans on and 5.36 x 10-4 ± 2.6 x 10-8 with fans off and the calculated η was 5.18 x 10-4 ± 11%. The difference between the fans being on versus off was negligible for all flow conditions, thus they were deemed unnecessary and were removed from the chamber for the oxidation experiments. The results are shown in

Figure 2.35. The results for the experiments are summarized in Table 2.10.

2.3.2. Exponential Dilution and PTR-MS Intercomparison

The GC-ITMS system was compared against the proton transfer reactor mass spectrometer (PTR-MS) in an exponential dilution chamber experiment. Both instruments were connected to the chamber manifold so that they could sample gas from it. A solution of benzene, toluene, indane, and 1,2,4-trimethylbenzene was prepared by adding 100 µL of each to a vial and then agitating. This solution was injected into a dynamic dilution system similar to the one described in Faiola et al. (2012). The dilution flow of 3.8 SLPM was N2 from the gas port on liquid nitrogen tank. The syringe pump

-1 (Harvard Apparatus, PN: 70-2209) delivered 1.0 µL hour into the heated flow of N2 to produce approximately 100 ± 3 ppbv of benzene, 92 ± 3 ppbv of toluene, 115 ±2 ppbv of indane, and 97 ± 3 ppbv of 1,2,4-trimethylbenzene. The chamber described in section

2.2.6 was filled with this mixture and allowed to reach steady state. Once the mixing ratios of the four components in the chamber stabilized, the dynamic dilution system was

61 bypassed so that the chamber was purged with zero air. The time series of GC-ITMS results of the exponential dilution are shown in Figure 2.36. The results indicate that the experiment followed what would be predicted for a “well-mixed” chamber from ~100 ppbv down to around ~0.1 ppbv at which time the mixing ratios of the 4 compounds deviate significantly from those predicted by exponential dilution. The purely exponential dilution line is shown in Figure 2.36 as a straight black line and the deviations from this line are apparent. Indane is the first compound to deviate from exponential behavior and benzene is the last. One explanation for this behavior is that at low mixing ratios slow desorption of material from the chamber walls becomes a significant source term (i.e equation 2.13). Benzene is the lightest and most volatile compound so it is not surprising that it is the last to transition from dilution to the wall desorption regime. Toluene and indane are the first to deviate from the exponential dilution trend line and 1,2,4-trimethylbenzene, the heaviest compound behaves more like benzene than indane and toluene. This indicates that there is a chromatographic effect to the chamber and sample lines. As the chamber and sample lines are purged with clean air, the sticky VOCs desorb and are carried to the instrument inlet. While it is unfortunate that the chamber behavior deviates from the predicted exponential decay because low concentrations can no longer be predicted by the chamber mixing equation, it is instrumentally ideal. This setup produces low mixing ratios that continue to decay in a predictable manner providing the ability to sample low mixing ratio values repeatedly.

Figure 2.37 shows the intercomparison with the PTR-MS data. The values agree well until around 1 ppbv, where the PTR-MS appears to suffer from an elevated zero. This is likely caused by a bad zero correction, which would affect the low values more than the

62 higher ones. The chamber described will remain a valuable tool for investigating the performance of the GC-ITMS. Future investigations should start at 1 to 10 ppbv to prevent contamination of the chamber with high levels of VOCs.

2.3.3. Gasoline Oxidation

This section describes the oxidation of gasoline in the photo chamber. It demonstrates that the GC-ITMS can be used in a chamber to determine the level of hydroxyl radical present, something that will be critical for the future experiments using the large flow reactor currently being constructed. The gasoline standard described in section 2.2.4 was added to the chamber at a flow rate of 85 sccm using a MKS instruments, Inc. MassFlo flow controller along with 0.80 LPM of HONO mixture and 3.1 LPM hydrocarbon free air giving a residence time of 34 minutes in the chamber. This mixture was allowed to equilibrate in the chamber overnight and was continuously sampled and analyzed with the ITMS. Once the gases in the chamber reached equilibrium, the UV lights were turned on, producing OH radicals, and the system was again allowed to come to equilibrium.

Figure 2.38 shows the experimental setup used to oxidize gasoline.

The oxidant used in this chamber is the hydroxyl radical which is produced by photolysis of nitrous acid (HONO). HONO was produced following the method of Febo et al.

(1995). HONO is produced when gaseous hydrochloric acid reacts with sodium nitrite.

A gaseous mixture of HCl and N2 are passed through a stirred, heated, sodium nitrite powder before entering the photo-chamber. The reaction that ensues is given below as reaction 3:

퐻퐶푙 + 푁푎푁푂2 → 퐻푂푁푂 + 푁푎퐶푙 R2.11

63 To quantify the HONO, the gas mixture is passed through one of two denuders. The first is coated with NaCl and the second is coated with Na2CO3 + glycerine. The NaCl coated denuder removes only HNO3 while the Na2CO3 denuder removes both HONO and HNO3

(Febo et al. 1995). After passing through the denuders the gas is sent to the NOxy instrument described in 2.1. During the experiment there was 42 ± 3 ppbv of HONO.

The difference between the two denuders is the signal from HONO.

퐻푂푁푂 + ℎ푣 → 퐻푂. + 푁푂 R2.12

The rate that reaction 4 proceeds at is given by equation 7 where j, the photolysis rate constant is the integral across relevant wavelengths, where 퐹(휆) is the wavelength dependent actinic flux (photons*cm-1*s -1), σ is the temperature and wavelength dependent absorption cross section (cm2*molecule-1), and φ is the quantum yield

(molecule*photon-1) (Sander et al. 2006).

휆2 푗 = 퐹(휆)휎(휆, 푇)휙(휆, 푇)푑휆 (2.15) ∫휆1

The chamber was illuminated with 6 x 15 W fluorescent bulbs surrounding on one side and 6x 25 W fluorescent bulbs mounted on the top. The spectra from the bulbs has been measured with a spectroradiometer (Metcon, Gmbh., model: CCD62003) and are plotted with the cross sections of HONO and NO2, in Figure 2.39. The chamber will photolyse

NO2 and thus production of O3 can be expected, although it was not measured in this experiment because O3 + aromatic compounds are slow. The quantum yield of HONO for light with wavelength less than 400 nm is reported to equal one (Cox and Derwent

1977).

The OH reacts with the volatile organics, especially the aromatics, which have high rate constants with OH. Table 10 shows the recommended rate constants at 298 K

64 and 1 atm for selected compounds (Calvert et al., 2002). The generalized reaction of a

VOC with OH is given by the following equation 8.

푉푂퐶 + 푂퐻 → 푝푟표푑푢푐푡푠 (R2.13)

For which the loss rate of the VOC is given by equation 2.16.

−푑 [푉푂퐶] = 푘[푉푂퐶][푂퐻] (2.16) 푑푡

Solving for [VOC] as a function of time and assuming that [OH] is constant, yields:

[푉푂퐶] 퐿푛 ( 푡) = −푘[푂퐻̅̅̅̅]t (2.17) [푉푂퐶]표

Using this relation, the results can be plotted to test that the above model for OH

reactivity is valid, and the average OH concentration can be determined. If this is valid

for compounds with well known OH reactivity, then it can be extrapolated to compounds

with less studied or unknown rate constants.

The natural logs of the ratio of the compounds measured before and after oxidation were plotted against the literature rate constants suggested by Calvert et al.

(2002). Fitting a line through the natural logs of the ratio of the measurements gave a slope of 9.98 x 109 +/- 0.71 x 109 (Figure 2.40). It should be noted that the uncertainty in the slope of the line is 7.2%, which is much better than the stated accuracy to the rate constants. The slope of the line is equal to [푂퐻̅̅̅̅] ∗ 푡 . Dividing by 2040 s, the number of seconds in 34 minutes, the calculated residence time, gives [푂퐻̅̅̅̅] = 4.9 x 106 ± 13% molecules cm-3. The results are shown in Figure 2.40 and the rate constants at 298 K and

1 atm are presented in Table 2.11. Every compound observed in the study obeyed the rate law within the uncertainty of the measurement and the known rate constants. There are a number of compounds in the gasoline sample that have not yet been identified, but

65 as they are identified in future work their rate constants can be quantified with a high degree of precision.

2.4. SUMMARY AND FUTURE WORK

This chapter described the two instruments, (NOxy and GC-ITMS/GC-FID) that

produced a significant portion of the data presented in this manuscript and the photo-

chamber. The principle of measurements along with calibration and generation of mixing

ratios are obtained from them have been described. Both of these instruments have been

integrated and deployed in the MACL and significantly increases the ability of LAR to

quantify trace gases. The NOxy system is able to measure pptv levels of NO, NO2, and

NOy. It has performed reliably in the field and can be deployed as is at any moment,

although the BLC needs to be repaired. Some future work to progress the state of the

NOxy system includes:

 The BLC has experienced a significant loss conversion efficiency. When the

system was deployed in Boise (December 2008 and January 2009) and during the

CABINEX (July 2009) field intensive the conversion efficiency was 60 to 70 %.

Currently the BLC is operating at 25 to 35 % efficiency. There are two likely

causes for this. The first is loss of reflectance due to contamination from soot and

other particles. The second is the loss of LED intensity. The BLC needs to be

opened up and examined. The low conversion efficiency of NO2 significantly

reduces the ability of the system to quantify NO2 in low level environments. This

issue should be addressed before the instrument is deployed again.

66  Investigation into the NOy converter’s conversion efficiency of ammonium nitrate

particles showed that it converts them with 80-85% efficiency. This experiment

should be performed again to see if there is a size dependent relationship.

 The BLC has a known NO2 artifact due to the porosity of the Lambertian

reflective material. A redesign of this converter using a glass cell should be

explored to reduce NO2 artifact.

 Water vapor reduces the sensitivity of NO chemiluminescence detectors by

quenching excited NO2. The dependence of sensitivity on water vapor should be

investigated.

 The issue of elevated zeros when measuring elevated levels (> 250 ppbv) should

be examined.

The GC-ITMS was designed and built as a significant part of this dissertation is a work in progress. The pre-concentration system on the GC-ITMS has been rebuilt and modified multiple times. At this point, the most pressing issue is the light hydrocarbon channel, which has problems with desorption of C3 and above compounds. Based on desorption temperature tests that have been performed the likely problem is low flow through the trap. Once this issue has been resolved the next step will be to determine recoveries of each compound and to determine retention times. Bringing the C2 to C6 light hydrocarbon channel online will allow for the quantification of important O3 precursors emitted from spark ignition engines. A list of recommended tasks for the light hydrocarbon channel include:

 The issue of low flow during desorption is the first task that must be addressed.

Low flow prevents desorption of heavier compounds and leads to poor

67 chromatography (wide peaks) making it impossible to separate the number of

hydrocarbons that the system will face in a polluted urban environment.

 The recovery of the Carboxen 1000 must be determined for C2 to C6

hydrocarbons.

 The retention times of light hydrocarbons needs to be determined along with

assessing the ability of the water trap to produce consistently dry air that does not

lead to changes in retention time on this column.

 The detection limit of this channel should be determined using the exponential

dilution method described for channel 2.

The construction and testing of the VOC Channel has been more successful than the light hydrocarbon channel. There have been a number of successes with this channel listed below:

 It produces linear measurements of VOCs over a large range of mixing ratios with

low (<3%) RSD.

 The GC-ITMS system has been deployed reliably during the CARES field

intensive where it produced a data set that has been useful in describing the air

mass in Sacramento, CA. This includes observations of morning rush hour

emission ratios and an analysis of the variability of mixing ratios. It has also been

used to analyze compounds generated during composting.

 Quantification of important aromatic VOCs in a gasoline mixture. There have also

been an additional 127 peaks found in a blend of gasoline and zero air.

68  Validation of rate constants using relative rate method and determination of the

hydroxyl radical concentration in the photo chamber. This will be extremely

useful in future flow reactor experiments.

 Detection limits ranged from 120 pptv for cis-2-butene to 6 pptv for n-undecane.

In general the aromatic compounds had detection limits of 10 to 20 pptv which

were validated with the exponential dilution experiments.

 The trap heating and control is rapid, reproducible, and can run unattended for

prolonged periods.

Despite the success of this channel in some areas there is still much progress to be made. One of the main objectives of this instrument was to identify and quantify aromatic and biogenic oxidation products to be used as metrics for atmospheric oxidation. One oxidation product that we believe should exist in sufficient quantities in urban air based on sufficient precursors and yield is benzaldehyde. However, this is a Tenax decomposition product and is seen even at the recommended desorption temperature of

200 oC. Injection of diluted gasoline was promising yielded some 127 peaks of which only 25 were identified. Identifying the rest of the peaks in raw gasoline will allow this channel to quantify a significant number of the VOCs found in urban air. The following recommendations for future work are:

 Oxidation products from aromatic compounds need to be identified. The photo

chamber should be used to oxidize large quantities of specific aromatic

compounds to produce oxidation products. The ability of the system to quantify

oxidation products can be determined. One major obstacle to this is that

benzaldehyde is produced from Tenax decomposition products. If benzaldehyde is

69 the predominant oxidation product produced during oxidation of aromatic

compounds, then it may be necessary to explore other adsorbent materials,

particularly the carboxens, which do not produce benzaldehyde. At the same time

these experiments can be used to determine rate constants of VOCs + OH, using

the relative reactivity method.

 The only a small number of peaks have been identified in the gasoline standard. It

would be useful to perform direct injection of the gasoline standard using the 250

µ sample loop. This method produces vastly improved chromatographic

separation which allows for easier identification of compounds. Identifying the

remaining peaks in the gasoline standard will increase the utility of this

instrument in urban environments and for future chamber experiments.

 The exponential dilution experiments validate this technique to produce low

levels of VOCs. The detection limits of the system for multiple compounds

should be experimentally determined so that the detection limit of the system is

known.

70 2.5. REFRENCES

Altshuller, A. P. "Review: Natural volatile organic substances and their effect on air quality in the United States." Atmospheric Environment (1967) 17.11 (1983): 2131-2165.

Apel, E. C., Calvert, J.G., Gilpin, T.M., Fehsenfeld, F.C., Parrish, D.D., and Lonneman,

W.A. "The nonmethane hydrocarbon intercomparison experiment (NOMHICE): task 3."

Journal of Geophysical Research: Atmospheres (1984–2012) 104.D21 (1999): 26069-

26086.

Apel, E. C., Calvert, J.G., Gilpin, T.M., Fehsenfeld, F.C., and Lonneman, W.A.

"Nonmethane hydrocarbon intercomparison experiment (NOMHICE): Task 4, ambient air." Journal of Geophysical Research: Atmospheres (1984–2012) 108.D9 (2003).

Apel, B.R., Povard, V., and Kothny, E.L. “Loss of nitric acid within inlet devices intended to exclude coarse particles during atmospheric sampling.” Atmospheric

Environment 22.11 (1988): 2535-2540. DOI: 10.1016/0004-6981(88)90485-4.

Bollinger, M.J., Sievers, R.E., Fahey, D.W., and Fehsenfeld, F.C. “Conversion of nitrogen dioxide, nitric acid, and n-propylnitrate to nitric oxide by gold-catalyzed reduction with carbon monoxide.” Analytical Chemistry 55. (1983): 1980-1986. Print

Calcote, H. F. "Ion and electron profiles in flames." Symposium (International) on

Combustion. Vol. 9.1. Elsevier, 1963.

71

Calogirou, A., et al. "Decomposition of terpenes by ozone during sampling on Tenax."

Analytical Chemistry 68.9 (1996): 1499-1506.

Cao, X.L., and Hewitt, C.N. "Build-up of artifacts on adsorbents during storage and its effect on passive sampling and gas chromatography-flame ionization detection of low concentrations of volatile organic compounds in air." Journal of Chromatography A

688.1 (1994): 368-374.

Clough, P.N., and Thrush, B.A. “Mechanism of the chemiluminescent reaction between nitric oxide and ozone.” Transactions of the Faraday Society 63. (1967): 915-925. Print.

Clyne, M.A.A., Thrush, B.A., and Wayne, R.P. “kinetics of the chemiluminescent reaction between nitric oxide and ozone.” Transactions of the Faraday Society 60.

(1964): 359-370. Print.

Crosley, D.R. “Issues in the Measurement of Reactive Nitrogen Compounds in the

Atmosphere.” SRI Internal Report MP 94-035, Melno Park, CA. (1993) Print.

Gage, K. S., and Nastrom, G.D. "On the spectrum of atmospheric velocity fluctuations seen by MST/ST radar and their interpretation." Radio Science 20.6 (1985): 1339-1347.

72 Febo, A., Perrino, C., and Allegrini, I. "Measurement of nitrous acid in Milan, Italy, by

DOAS and diffusion denuders." Atmospheric Environment 30.21 (1996): 3599-3609.

Gardner, E.P., Sperry, G.D., and Calvert, J.G. "Primary quantum yields of NO2 photodissociation", Journal of Geophysical Research - Atmospheres 92. 6642-6652

(1987).

Gentner, D. R., Isaacman, G., Worton, D. R., Chan, A. W., Dallmann, T. R., Davis, L et al. "Elucidating secondary organic aerosol from diesel and gasoline vehicles through detailed characterization of organic carbon emissions." Proceedings of the National

Academy of Sciences 109.45 (2012): 18318-18323.

Harley, J. "Flame ionization detector for gas chromatography." Nature 181 (1958): 177-

178.

Helmig, D. "Air analysis by gas chromatography." Journal of Chromatography A 843.1

(1999): 129-146

Helmig, D., Ortega, J., Duhl, T., Tanner, D., Guenther, A., Harley, P., and

Sakulyanontvittaya, T. "Sesquiterpene emissions from pine trees-Identifications, emission rates and flux estimates for the contiguous United States." Environmental

Science & Technology 41.5 (2007): 1545-1553.

73 Detlev, H., et al. "Analytical techniques for sesquiterpene emission rate studies in vegetation enclosure experiments." Atmospheric Environment 38.4 (2004): 557-572.

Fahey, D.W., Eubank, C.S., Hubler, G., and Fehsenfeld, F.C. “Evaluation of a catalytic reduction technique for measurement of total reactive odd-nitrogen NOy in the atmosphere.” Journal of Atmospheric Chemistry 3. (1985): 435-468. Print.

Fontijn, A., Sabadell, A.J., and Ronco, R.J. “Homogenous Chemiluminescent

Measurement of Nitric Oxide with Ozone.” Analytical Chemistry 42. No. 6 (1970): 575-

579. Print.

Goldan, P.D., Kuster, W.C., Albritton, D.L., Fehsenfeld, F.C., Connell, P.S., and Norton,

R.B. “Calibration and tests of the filter-collection method for measuring clean-air, ambient levels of nitric acid.” Atmospheric Environment 17. (1983): 1355-1364. DOI:

10.1016/0004-6981(83)90410-9

Guenther, A.B., and Hills, A.J., “Eddy covariance measurement of isoprene fluxes.”

Journal of Geophysical Research - Atmospheres 103. (1998):13145-13152. DOI:

10.1029/97JD03283

Kley, D., and McFarland M. “Chemiluminescence detector for NO and NO2.”

Atmospheric Technology 12. (1980): 63-69. Print.

74 Kliner, D.A.V, Daube, B.C., Burley, J.D., and Wofsy, S.C. “Laboratory investigation of the catalytic reduction technique for measurement of atmospheric NOy” Journal of

Geophysical Research -Atmospheres 102. D9 (1997): 10759-10776. DOI:

10.1029/96JD03816.

Lilly, D. K., and E. L. Petersen. "Aircraft measurements of atmospheric kinetic energy spectra." Tellus A 35.5 (1983): 379-382.

McWilliam, I. G., and R. A. Dewar. "Flame ionization detector for gas chromatography."

Nature 181 (1958): 760.

Murphy, D. M., et al. "Indicators of transport and vertical motion from correlations between in situ measurements in the Airborne Antarctic Ozone Experiment." Journal of

Geophysical Research: Atmospheres (1984–2012) 94.D9 (1989): 11669-11685.

Nastrom, G. D., Gage, K.S., and Jasperson, W.H. "Kinetic energy spectrum of large-and mesoscale atmospheric processes." Nature 310.5972 (1984): 36-38.

Neher, M.B., and Jones, P.W. "In situ decomposition product isolated from Tenax GC while sampling stack gases." Analytical Chemistry 49.3 (1977): 512-513.

75 Neuman, J.A., Huey, L.G., Ryerson, T.B., and Fahey, D.W. “Study of Inlet Materials for

Sampling Atmospheric Nitric Acid.” Environmental Science and Technology 33

(1999):1133-1136. DOI:10.1021/es980767

Odum, J.R., et al. "The atmospheric aerosol-forming potential of whole gasoline vapor."

Science 276.5309 (1997a): 96-99.

Odum, J.R., et al. "Aromatics, reformulated gasoline, and atmospheric organic aerosol formation." Environmental Science & Technology 31.7 (1997b): 1890-1897.

Pellizzari, E., Demian, B., and Krost., K. "Sampling of organic compounds in the presence of reactive inorganic gases with Tenax GC." Analytical Chemistry 56.4 (1984):

793-798.

Plass-Dülmer, C., et al. "C2-C8 Hydrocarbon measurement and quality control procedures at the Global Atmosphere Watch Observatory Hohenpeissenberg." Journal of

Chromatography A 953.1 (2002): 175-197.

Pollack, A.J., Holdren, M.W., and McClenny, W.A. "Multi-adsorbent preconcentration and gas chromatographic analysis of air toxics with an automated collection/analytical system." Journal of the Air & Waste Management Association 41.9 (1991): 1213-1217.

76 Pollack, I.B., Lerner, B.M., and Ryerson, T.B. “Evaluation of ultraviolet light-emitting diodes for detection of atmospheric NO2 by photolysis – chemiluminescence.” Journal of

Atmospheric Chemistry 65 (2010): 111-125 DOI: 10.1007/s10874-011-9184-3

Pollmann, J., et al. "Evaluation of solid adsorbent materials for cryogen-free trapping— gas chromatographic analysis of atmospheric C2–C6 non-methane hydrocarbons."

Journal of Chromatography A 1134.1 (2006): 1-15.

Rappenglück, B., et al. "The first VOC intercomparison exercise within the Global

Atmosphere Watch (GAW)." Atmospheric Environment 40.39 (2006): 7508-7527.

Ridley, B.A. and Grahek, F.E. “A Small, Low Flow, High Sensitivity Reaction Vessel for

NO Chemiluminescence Detectors.” American Meteorological Society 7 (1990): 307-

311. Print.

Roehl, C.L., Orlando, J.J., Tyndall, G.S., Shetter, R.E, Vasquez, G.J., Cantrell, C.A, and

Calvert, J.G. "Temperature dependence of the quantum yields for the photolysis of NO2 near the dissociation limit." Journal of Physical Chemistry 98. (1994) 7837-7843

Ryerson, T.B., Williams, E.J., Fehsenfeld, F.C. “An efficient photolysis system for fast- response NO2 measurements.” Journal of Geophysical Research – Atmospheres 105.

D21 (2000): 26447-26461. Print.

77 Sander, S.P, et al. "Chemical kinetics and photochemical data for use in atmospheric studies evaluation number 15." (2006).

Sato, K. "Vertical wind disturbances in the troposphere and lower stratosphere observed by the MU radar." Journal of the Atmospheric Sciences 47.23 (1990): 2803-2817.

Scanlon, J.T., and Willis, D.E. "Calculation of flame ionization detector relative response factors using the effective carbon number concept." Journal of Chromatographic Science

23.8 (1985): 333-340.

Slemr, J., et al. "Accurate measurements of hydrocarbons in the atmosphere (AMOHA):

Three European intercomparisons." Journal of Geophysical Research: Atmospheres

(1984–2012) 107.D19 (2002): ACH-15.

Sternberg, J.C., Gallaway, W.S., and Jones, D.T.L. "The mechanism of response of flame ionization detectors." Gas Chromatography (1962): 231-267.

Sillman, S. "The use of NOy, H2O2, and HNO3 as indicators for ozone-NOx-hydrocarbon sensitivity in urban locations." Journal of Geophysical Research 100.D7 (1995): 14175-

14.

78 Talbot, R.W., Vijgen, A.S., and Harriss, R.C. “Measuring tropospheric HNO3: Problems and prospects for nylon filter and mist chamber techniques.” Journal of Geophysical

Research – Atmospheres 95. D6 (1990): 7553-7561. DOI: 10.1029/JD095iD06p07553

Tanner, D., et al. "Gas chromatography system for the automated, unattended, and cryogen-free monitoring of C2 to C6 non-methane hydrocarbons in the remote troposphere." Journal of Chromatography A 1111.1 (2006): 76-88.

Walling, J. F., et al. "Apparent reaction products desorbed from Tenax used to sample ambient air." Atmospheric Environment (1967) 20.1 (1986): 51-57.

Williams E.J., Baumann, K., Roberts, J.M., Bertman, S.B., Norton, R.B., Fehsenfeld,

F.C., Springston, S.R., Nunnermacker, L.J., Newman, L., Olszyna, K., Meagher, J.,

Hartsell, B., Edgerton, E., Pearson, J.R., Rodgers, M.O.

“Intercomparison of ground-based NO y measurement techniques.” Journal of

Geophysical Research 103.D7 (1998): 22261-22280. DOI: 10.1029/98JD00074

Westberg, H., and Zimmerman, P. "Analytical methods used to identify nonmethane organic compounds in ambient atmospheres." Advances in Chemistry Series 232 (1993):

275-275.

79 2.6. CHAPTER 2 TABLES

Table 2.1. Summary of channels logged by DAQFactory. Channel Units Function CH1_HEAT_PV 0-5 VDC 0-5 VDC phase modulation control signal CH2_HEAT_PV 0-5 VDC 0-5 VDC phase modulation control signal CAL_ON_OFF Digital (0,1) calibration valve state CAL_ZERO Digital (0,1) zero air valve state GC_START Digital (0,1) starts GC run when set to 1 He_ON_OFF Digital (0,1) helium flush valve state N2_FLUSH_CH1 Digital (0,1) nitrogen flush valve CH1 N2_FLUSH_CH2 Digital (0,1) nitrogen flush valve CH2 N2_VENT_CH1 Digital (0,1) nitrogen vent valve CH1 N2_VENT_CH2 Digital (0,1) nitrogen vent valve CH2 SAMPLE_CH1 Digital (0,1) sample flow valve CH1 SAMPLE_CH2 Digital (0,1) sample flow valve CH2 CH1_HEAT_ENABLE Digital (0,1) adsorbent trap heater enable CH1 CH2_HEAT_ENABLE Digital (0,1) adsorbent trap heater enable CH2 A1A Digital (0,1) injection valve state A A1B Digital (0,1) injection valve state B A2A Digital (0,1) flush valve state A A2B Digital (0,1) flush valve state B A2A Digital (0,1) sample valve state A A3B Digital (0,1) sample valve state B Trap_Temp_1 oC adsorbent trap temperature CH1 Trap_Temp_2 oC adsorbent trap temperature CH2 INTERNAL TEMP oC ambient temperature UE3 ReadSP1 oC heated zone 1 setpoint ReadSP2 oC heated zone 2 setpoint ReadSP3 oC heated zone 3 setpoint ReadSP4 oC heated zone 4 setpoint ReadSP5 oC heated zone 5 setpoint ReadSP6 oC heated zone 6 setpoint ReadSP7 oC water trap setpoint CH1 ReadSP8 oC water trap setpoint CH2 ReadTemp1 oC heated zone 1 temperature ReadTemp2 oC heated zone 2 temperature ReadTemp3 oC heated zone 3 temperature ReadTemp4 oC heated zone 4 temperature ReadTemp5 oC heated zone 5 temperature ReadTemp6 oC heated zone 6 temperature ReadTemp7 oC water trap temperature CH1 ReadTemp8 oC water trap temperature CH2 CAL_FLOW sccm calibration flow SAMPLE_FLOW_1 sccm channel 1 sample flow SAMPLE_FLOW_2 sccm channel 2 sample flow

80 ZA_FLOW sccm zero air flow

Table 2.2. Temperature program used for the absolute calibration of the FID. Temp Rate Hold Time Total Time oC oC*min-1 minutes minutes 40 0.0 1.0 1.0 100 10.0 0.0 7.0 200 20.0 2.0 14.0

Table 2.3. Summary of calibration gases used for the direct calibration of the FID. Mixing Retention Supplier Ratio Error Time Formula Scott-Marrin, Inc. ppmv +/- minutes ethylene 10.05 2% 2.742 C2H4 2,2-dimethylbutane 10.01 2% 11.175 C6H14 Scott Specialty Gases acetylene 15.2 10% 6.096 C2H2 n-butane 15.2 10% 6.612 C4H10 ethane 15.1 10% 2.311 C2H6 ethylene 15.1 10% 2.742 C2H4 methane 15.1 10% 1.961 CH4 propane 15.3 10% 3.658 C3H8 propylene 15.1 10% 5.306 C3H6 propyne 15.1 10% 9.440 C3H4

Table 2.4. The Tenax GR trap and cold tube water trap temperatures in oC for the three conditions tested.

Condition Tenax GR Cold Tube Temp Temp 1 30 30 2 0 -10 3 -10 -30

81 Table 2.5. A summary of the artifacts generated from the Tenax GR decomposition. Elution Compound Time 280oC 260oC 240oC 220oC 200oC Quant Ions min m/z Area Counts 15.2- Benzene 17.27* 77,78 102553 28193 10539 4894 2203 23.5- Toluene 24.4* 91,92 9002 0 0 0 0 Silicic acid 26.1 193,207,209 9005 0 0 0 0 Styrene 32.45 104,103,78,77 6577 1805 500 0 Octamethyl- Cyclotetrasiloxane 35.7 281,282,265 32799 8445 3213 1235 620 Benzaldehyde 37.7 105,77,50,51 10300 4166 2780 2120 1650 Benzofuran 38.8 118,90,89,63 5917 2200 0 0 0 Benzonitrile 39.6 103,76,50 7749 670 60 0 0 cyclopentasiloxane 42.9 355,356 15400 4689 1684 440 312 * not a variable peak center but a mesa due to long emission time and the compound not being retained

82 Table 2.6. Shows the composition in ppbv of the GCMS calibration gas. Compound ppbv n-pentane 312 i-pentane 315 isoprene 316 2,2,-dimethylbutane 331 dichloromethane 297 2-methylpentane 308 3-methylpentane 301 hexane 295 1,1,1-trichloroethane 313 2-methylhexane 308 3-methylhexane 306 benzene 311 2,2,4-trimethylpentane 311 heptane 253 choloracetone 321 2,3,4-trimethylpentane 306 toluene 309 octane 311 ethylbenzene 309 nonane 306 p-xylene(+ m-xylene 307 o-xylene 308 isopropylbenzene 255 a-pinene 275 3-ethyltoluene 308 4-ethyltoluene 266 decane 290 135-TMB 306 2-ethyltoluene 304 124-TMB 284 123-TMB 256 n-butylbenzene 281

83 Table 2.7. The composition of the PTR-MS calibration gas. Compound ppmv Acetaldehyde 2.0 Acetone 2.0 Acetonitrile 2.0 Benzene 2.0 Isoprene 2.0 Methacrolein 2.0 Methanol 2.0 Methyl Ethyl Keytone 2.0 (1S)-(-)alpha-Pinene 2.0 Styrene 2.0 1,2,3,5- 0.5 Tetramethylbenzene 1,2,4-Trimethylbenzene 2.0 Toluene 2.0 p-Xylene 2.0

84 Table 2.8. Spectra gas composition in ppmv along with the sensitivity (area counts ppbv-1). These sensitivities are for internal ionization mode. The reported sensitivities are from the direct injection using the sample loop and represent the absolute response of the system to the compounds they have been scaled for a 300 cm3 sample containing 1 ppbv.

Compound ppmv Sensitivity MDL ppbv ethylene 1.08 Na na acetylene 1.08 Na na ethane 1.08 Na na propylene 1.04 Na na isobutene 1.03 Na na propane 1.03 Na na 1-butene 1.02 Na na n-butane 1.02 Na na trans-2-butene 1.02 Na na cis-2-butene 1.09 33505 0.119 isopentane 1.1 Na na 1-pentene 1.07 61850 0.065 n-pentane 1.07 49732 0.080 isoprene 1.09 104250 0.038 trans-2-pentene 1.09 101249 0.040 cis-2-pentene 1.1 98606 0.041 2,2-dimethylbutane 1.08 115781 0.035 cyclopentane 1.06 70476 0.057 2,3-dimethylbutane 1.08 74925 0.053 2-methylpentane 1.05 76726 0.052 3-methylpentane 1.08 131522 0.030 1-hexene 1.06 97007 0.041 n-hexane 1.07 103627 0.039 methylcyclopentane 1.09 143386 0.028 2,4-dimethylpentane 1.09 142001 0.028 benzene 1.06 202907 0.020 cyclohexane 1.08 151296 0.026 2-methylhexane 1.08 163177 0.025 2,3-dimethylpentane 1.1 166747 0.024 3-methhexane 1.07 171886 0.023 2,2,4-trimethyl pentane 1.07 277441 0.014

85 n-heptane 1.07 151295 0.026 methylcyclohexane 1.07 217165 0.018 2,3,4-trimethylpentane 1.06 185940 0.022 toluene 1.06 350221 0.011 2-methylheptane 1.06 141494 0.028 3-methylheptane 1.05 205278 0.019 n-octane 1.06 91715 0.044 ethylbenzene 1.05 368819 0.011 p-xylene* 1.07 689257 0.012 m-xylene* 1.07 689257 0.012 styrene 1.06 397001 0.010 o-xylene 1.04 348109 0.011 nonane 1.04 143966 0.028 isopropylbenzene 1.01 267131 0.015 n-propylbenzene 1.08 485495 0.008 m-ethyltoluene 1.06 360444 0.011 p-ethlytoluene 1.02 355468 0.011 1,3,5-TMB 1.05 374064 0.011 o-ethlytoluene 1.03 415742 0.010 1,2,4-TMB 1.03 426937 0.009 n-decane 1.08 253735 0.016 1,2,3-TMB 1.04 413678 0.010 m-diethylbenzene 1.03 463824 0.009 p-diethylbenzene 1.02 469492 0.009 n-undecane 1.06 715152 0.006 n-dodecane 1.05 Na na

86

Table 2.9. Table of compounds identified on the GC-ITMS, retention times (minutes) and quantitation ions (m/z).

# Compound RT Quant Ions 1 trans-2-butene* 3.069 56,55 2 cis-2-butene* 3.724 57,56 4 1-pentene* 4.319 70 5 n-pentane* 4.509 57 6 trans-2-pentene* 5.315 68,67,55,53 7 cis-2-pentene* 4.95 70,55 8 2,2-dimethylbutane* 5.95 71,57,56 9 cyclopentane* 7.683 70,55 10 2,3-dimethylbutane* 7.398 71 11 2-methylpentane* 7.468 55,57 12 3-methylpentane* 8.298 56,57 13 1-hexene* 8.922 64,55,56 14 n-hexane* 9.207 57,56 15 2,4-dimethylpentane* 11.009 85,57,56 16 2-methylhexane* 13.561 85,57,56 17 2,3-dimethylpentane* 13.827 71,56,57 18 3-methhexane* 14.256 71,70,57,56,55 19 2,2,4-trimethyl pentane 15.252 57,56 20 n-heptane* 16.026 71,70, 57 21 methylcyclohexane 18.397 98, 83,55 22 2,3,4-trimethylpentane 20.367 71,70,55 23 2-methylheptane 21.397 57 24 3-methylheptane 22.091 85,84,57,56 25 isoprene 5.663 68,67 26 dimethyl sulfide 6.974 62,45,46,47 27 cyclopentane* 7.684 55, 70 28 3-methyl-furan 11.220 82,81,53 29 MBO 13.565 59,71 30 cyclohexane 14.03 84,69,56 31 benzene 15.181 77,78 32 2-ethyl-furan 18.025 81,96,53 33 methylcyclopentane 11.3 69,56,55 34 2-pentanone 18.942 43,71,87,86 35 toluene 23.633 91,92 36 n-octane 24.332 85, 57

87 37 ethylbenzene 30.301 91,106 38 p-xylene 30.809 91,106 39 m-xylene 30.809 91,106 40 nonane 31.076 57 41 UNK MT1 32.275 93,91,121,136 42 o-xylene 32.453 91,106 43 styrene 32.523 104,103,78,51 44 UNK MT2 33.542 91,92,93,77 45 isopropylbenzene 34.021 105 46 alpha-pinene 34.047 93,92,91 47 UNK MT3 35.041 93,121,91,79,77 48 UNK MT4 35.190 93,91,121,67,79,77 49 n-propylbenzene 35.624 91,120 50 m-ethyltoluene 36.094 105,120,79 51 p-ethlytoluene 36.192 105,79 52 1,3,5-TMB 36.458 105,120 53 n-decane 36.620 57,71 54 beta-pinene 36.656 93, 55 o-ethlytoluene 37.250 105,120,79 56 alpha phellandrene 37.932 77,91,92,93 57 1,2,4-TMB 37.953 105,120 58 3-carene 38.133 91,92,93,79,77 59 alpha-terpinene 38.554 93,91,121,136,75 60 limonene 39.066 67,93,91,94,79 61 tertbutylbenzene 39.385 119,91,134 62 UNK MT6 39.535 91,93,77,79 63 1,2,3-TMB 39.616 105,120 64 Eucalyptol 39.813 93,139,81 65 m-diethylbenzene 40.447 105,120,77,91 66 UNK MT7 40.508 91,92,93,121 67 p-diethylbenzene 40.771 105,134,120,91,77 68 n-undecane 41.397 85,55,71,57 69 UNK MT8 41.957 93.1,91,121,136 70 Terpinolene 42.825 79,93,105,121,136 71 1,6-dimethyl-cyclohexene 44.534 67,109,95,81 72 n-dodecane 45.206 57,71,85,55,83 73 alpha-thujone 45.382 67,68,81,79,95 74 naphthalene 47.160 128 *Retention time is from the direct injection using a sample loop. Retention time during a desorbed run will be lower.

88 Table 2.10. Summary of results from the chamber mixing experiment. Chamber Volumetric mixing Fan Observed η Calculated η Volume Flow ratio Power NO L L x min-1 ppbv state s-1 s-1 148 22.2 250 on 2.45E-03 ± 4.4E-5 2.50E-03 ± 11% 148 22.2 250 off 2.43E-03 ± 2.5E-5 2.50E-03 ± 11% 148 11.4 250 on 1.22E-03 ± 9.8E-6 1.25E-03 ± 11% 148 11.4 250 off 1.19E-03 ± 1.1E-5 1.25E-03 ± 11% 148 4.8 125 on 5.40E-04 ± 3.6E-8 5.18E-04 ± 11% 148 4.8 125 off 5.37E-04 ± 2.6E-8 5.18E-04 ± 11%

Table 2.11. Compounds and recommended OH rate constants (Calvert et al. 2002) k, cm3 molecules -1 s - Compound Name 1 err % 1,2,3- trimethylbenzene 3.27E-11 25 1,2,4- trimethylbenzene 3.25E-11 25 1,3,5- trimethylbenzene 5.67E-11 20 benzene 1.39E-12 20 ethylbenzene 7.00E-12 25 isopropylbenzene 6.30E-12 30 m-ethyltoluene 1.86E-11 35 p-ethyltoluene 1.18E-11 35 n-propylbenzene 5.80E-12 25 napthalene 2.30E-11 25 o-ethyltoluene 1.19E-11 35 p,m-xylene 1.80E-11 20 toluene 5.63E-12 20

89 2.7. CHAPTER 2 FIGURES

Figure 2.1. Comparison of room temperature NO2 and HONO absorption cross sections in the region of the blue LED output. NO2 cross section data from Bogumil et al., 2003 and HONO from Stutz et al., 2000.

90

Figure 2.2. Calibration of the NOxy system.

91

Figure 2.3. PSD for the NO, channel1 (black) and the NO2 Channel2 (gray). The signal goes to white noise at 0.3 Hz. The instrument is not capable of resolving atmospheric components with higher frequency than 0.3 Hz.

92

Figure 2.4. Illustration of GC-IFMS/GC-FID assembly.

93

Figure 2.5. A plumbing schematic for the GC-ITMS system including two separate channels, one for light hydrocarbons (CH1, red trace) and one for heavier VOCs including aromatics and biogenics (CH2, black trace). The flush valve is plumbed with chromatographic grade He that is used to purge both channels (blue trace).

94

Figure 2.6. The cold tube water trap inlet for the VOC pre-concentration system. The valves are indicated by red circles and the flow path is indicated by black traces. As drawn, the cold tube water traps are being purged with 400 sccm of dry N2 and there is no flow through the sample flow valves.

95

Figure 2.7. The helium flush valve is represented by a red circle and is plumbed to the multiport flush valve. As depicted, the helium flush valve is in the “on” state which allows the clean carrier gas to purge the system.

96

Figure 2.8. Shows the cold block assembly. All of the major components are visible with the exception of the cold probe from the Flexi-Cool immersion cooler, which is hidden in the cold block.

97

Figure 2.9. This is a plumbing diagram of the pre-concentration system during the

Condition Water Trap mode. The flow path for channel 1 is shown by red traces, channel

2 is black traces and the helium flows are shown by blue traces. Above the plumbing diagram are the valve states, either “A” or “B” along with the internal connections for the respective valves.

98

Figure 2.10. The plumbing and valve positions during sample mode. Flow through channel 1 is indicated with red traces. Black traces indicate channel 2 flow paths and the blue traces are helium carrier and purge. The valves above the plumbing diagram show the valve position along with the corresponding internal traces.

99

Figure 2.11. A plumbing diagram for Cold/Hot Flush Mode. The color coding and valve configuration are the same as in figures 5 and 6. The difference between cold and hot flush modes are the adsorbent tube temperatures only.

100

Figure 2.12. Thermal Desorption Mode plumbing. Channels and valve states indicated the same as the others in this section.

101

Figure 2.13. A plumbing diagram of the system in Analysis/Back Flush Mode. Red indicates channel 1, black channel 2 and blue is for helium carrier lines. The valve states along with the internal configurations are indicated by the three circles above.

102

Figure 2.14. A drawing of the Tenax GR trap used in channel 2 (not to scale). The Tenax

GR and silanized glass wool are shown in the tube. The thermocouple is also shown passing through the 1/8” tee and the thermocouple which passes through the tee, one glass wool plug and measures the tube temperature at the Tenax tube interface.

103

Figure 2.15. Adsorbent trap heating control diagram.

Figure 2.16. The profiles of 5 heating runs for channel 1 (grayscale squares) and channel

2 (blue triangles). The initial temperature ramp, which starts at time = 3 s is outlined in red.

104

Figure 2.17. Schematic of the plumbing used to calibrate the FID

Figure 2.18. Chromatograms of the Scotty gas for the 100 (blue trace), 250 (green trace), and 500 (black trace) μL sample loops.

105

Figure 2.19. Early eluting peaks. Blue is the 100 μL loop, green the 250 μL loop, and black the 500 μL loop.

106

Figure 2.20. Late eluting peak showing good chromatographic shape. Blue is the 100 μL loop, green the 250 μL loop, and black the 500 μL loop.

107

Figure 2.21. Plot showing the linearity of the FID on a per molecule of carbon for the

Scott Marrin, Inc. gases (black circles), Scotty alkanes and alkenes (grey circles), propyne (open circles) and acetylene (open stars).

108

Figure 2.22. Plot of the FID response for each compound. The dashed line is the average slope of the Scott Marrin, Inc. gases. The asterisk indicates the Scott-Marrin calibration standard which has a stated 2% accuracy.

109

Figure 2.23. Chromatograms of the Scotty gas desorbed at 300 (blue), 335 (black), and

370 (green) oC.

110

Figure 2.24. A top view of the ITMS shows the main components of the GC-ITMS.

Taken from the Varian manual

111

Figure 2.25. This chromatogram of the co-eluters isoprene and trans-2-pentene shows the analytic power of the ITMS. The traces shown are: The TIC (top blue), m/z = 70 (middle black), and m/z = 67. What appeared to be one peak is actually two and by integration only the unique ions, the two co-eluting compounds can be quantified.

Figure 2.26. A schematic of the experimental setup for the water trap temperature test.

112

Figure 2.27. Integrated areas for each condition tested. A summary of the conditions tested is in Table 3.

Figure 2.28. Chromatograms of clean nitrogen gas desorbed at 280 oC (black trace) and

200 oC (blue trace). The instrument was operated in external ionization mode for this test.

113

Figure 2.29. A chromatogram of the GCMS calibration gas taken during the CARES study.

Figure 2.30. Post CARES multipoint calibrations from approximately 0.5 ppbv to 2.5 ppbv. The instrument was operated in internal ionization mode.

114

Figure 2.31. A chromatogram of the PTR-MS calibration gas.

Figure 2.32. Chromatograms of the Spectra gas injected directly (blue) and by the pre- concentration system (black).

115

Figure 2.33. A chromatogram of the gasoline mixture in external ionization mode. The largest peak at 23.4 minutes is toluene.

Figure 2.34. Chromatograms of zero gas using air from the zero air generator (blue trace, internal ionization mode) and N2 from the liquid nitrogen gas port (black trace, external ionization mode). The difference in the baseline signal is attributed to the better trapping efficiency of the ion trap in internal ionization mode.

116

Figure 2.35. The chamber mixing results for exponential dilutions for flow rates (left to right): 22.2 L min-1, 11.4 L min-1, 4.8 L min-1. Fans off (blue), fans on (green), theoretical (black dashed) are shown.

117

Figure 2.36. A time series of the exponential dilution of benzene (green squares), indane

(light blue triangles), toluene (dark blue circles) and 1,2,4-trimethylbenzene (gray diamonds) measured by the GC-ITMS.

118

Figure 2.37. PTR-MS vs. GC-ITMS during the exponential dilution of benzene (green squares), indane (light blue triangles), toluene (dark blue circles) and 1,2,4- trimethylbenzene (gray diamonds). The PTR-MS shows a problem with the zero at values below 1 ppbv, possibly due to a background correction problem.

119

Figure 2.38. Experimental setup for the gasoline oxidation experiment.

The hydroxyl radical used in the gasoline oxidation was produced by photolysis of

HONO. In the first part of the experiment the lights in the photo-chamber were off while the gases equilibrated. Once stable VOC measurements were observed in the chamber,

Reaction 4, the photolysis of HONO was commenced by turning the chamber lights on.

120 .

Figure 2.39. Cross sections of HONO (top) and NO2 (bottom) with the spectra of the UV lamps overlaid.

121

Figure 2.40. A plot of the Ln([VOC]t/[VOC]0) versus the rate constant with OH. The error bars are the errors associated with the precision of the measurements propagated through the natural logarithm of a ratio (vertical) and the stated accuracy of the rate constant (horizontal.) The line is the weighted best fit through the points and the slope is equal to the product of the [OH] and the residence time.

122 VERTICAL PROFILES THROUGH A FOREST CANOPY DURING CABINEX

3.1. INTRODUCTION

The Community Atmosphere Biosphere Interaction Experiment (CABINEX) was a multi-investigator field project conducted in the summer of 2009 at the University of

Michigan Biological station in northern Michigan. The purpose of the experiment was to investigate the influence of biogenic emissions of reactive hydrocarbons on HOx cycling and aerosol formation and growth in and above the forest canopy. The biosphere is responsible for the vast majority of volatile organic compounds (VOCs) emitted into the atmosphere. Guenther et al. (1995) estimated that global biogenic VOC emissions are approximately 1100 TgC yr-1 compared to anthropogenic VOC emissions of 110 -150

TgC yr-1 (Muller et al. 1992). Forests cover approximately one third of the earth’s land mass and account for two thirds of the total BVOC emissions (Goldstein and Gabally

2007; Guenther et al. 1995). Oxidation of reactive hydrocarbons is thought to be a major source of secondary organic aerosol (SOA) and contributes to ozone formation in polluted environments. Major biogenic SOA precursors include monoterpenes (C10H16) and sequiterpenes (C15H24) compounds, while isoprene emissions can have an important role in ozone formation. Over 5000 terpenes including monoterpenes, sequiterpenes

(C15) , diterpenes (C20), and higher molecular weight terpenes have been identified, however only the monoterpenes and sequiterpenes have volatilities to exist in the gas phase under atmospheric conditions (Geron et al., 2000). Oxidation of these compounds in the atmosphere by ozone, hydroxyl radical, and nitrate radical produce a wide variety of photoproducts. Goldstein and Gabally (2007) estimate the myriad of VOCs, which includes compounds directly emitted in the atmosphere and also their oxidation products

123 to be between 10,000-100,000 VOCs. Forests are also an important source of tropospheric NOx (NO+NO2). Emissions of NOx from microbial processes in soils account for ~15% of the global NOx emissions (Holland et al., 1999; Hudman et al.,

2010). In addition to emitting numerous VOCs, forests present complex terrain that complicates mixing and sets up meteorological and photochemical conditions that favor formation of chemical gradients (Rizzo et al., 2010;Steiner et al ., 2011; Wolfe et al.,

2011). Given the significant emission of NOx and VOCs, remote forests are important for shaping oxidation that occurs in the atmosphere, and also have climate change implications through their contribution to secondary organic aerosol (SOA).

Understanding emission rates from forests, mixing through forests, chemical gradients that exist in forests and the impact of transported pollution to forests is important for modeling oxidant and aerosol formation from biogenic emissions.

The CABINEX field experiment was conducted at the Program for Research on

Oxidants: Photochemistry, Emissions, and Transport (PROPHET) site. The site has a long record of BVOC measurements including monoterpenes, isoprene and sesquiterpenes made during several summer field experiments dating from 1997

(Westberg et al., 2001; Barker Jr., et al. 2001; Apel et al., 2002; Presley et al., 2005;

Presley et al., 2006; Ortega et al., 2007; Kim et al., 2009). Ortega et al. (2007) made emission measurements of the five most common trees at the PROPHET site during July and August, 2003 by taking samples from branch enclosures. In addition to this history of

BVOC measurements, a number of interesting observations about the chemistry of forests have been made. Di Carlo et al. (2004) demonstrated that there was missing OH reactivity as the measurements of volatile organic compounds could not account for the

124 measured OH loss frequency. The missing OH reactivity was shown to have a temperature dependence that was identical to a monoterpene emission curve, suggesting missing monoterpenes, or improperly measured monoterpenes. Kanawade et al. (2011) have shown that isoprene emissions suppress growth of aerosols at the PROPHET site.

The purpose of this manuscript is to report vertical profiles of ozone, VOCs, NO, and

NO2 made during the CABINEX field experiment as part of the first field deployment of

LAR’s Mobile Atmospheric Chemistry Lab (MACL).

3.2. EXPERIMENTAL

3.2.1. Site Description

The PROPHET tower is located at the University of Michigan Biological Station, an approximately 10,000 acre land parcel in a rural area near Pelston, Michigan. The site is located on a high-outwash plain deposited by glacial drift at the northern tip of

Michigan’s Lower Peninsula (45o30’N, 84o24’W) (Caroll, 2001.) The soils are described a “mostly highly drained spodosols” and is surrounded by a heavily forested area classified as a transition forest (Pearsall, 1995; Schmid et al., 2003). Transition forests are found south of the northern boreal forests and north of the deciduous forests. The composition of the forest around the PROPHET tower includes big-tooth aspen (Populus grandidentata 28.9%), red oak (Quercus rubra 26.6%), red maple (Acer rubra 19.1%), white pine (Pinus strobes 11.0%), and Betula papyrifera (8.7%) (Ortega et al. 2007). The canopy height is approximately 22 m with an over story LAI of 3.8 m2m-2 (Ortega et al.

2007; Pratt et al. 2012). The surrounding area was heavily burned in 1880 and then repeatedly logged until 1923 and after approximately 80 years of growth is in transition from big tooth aspen to white pine (Schmid et al., 2003; Bergen and Dronova, 2007).

125 Emissions are expected to switch from isoprene dominated to terpene dominated as the succession to white pine proceeds. This will drastically change the forest chemistry and aerosol formation (Bergen and Dronova, 2007). The site is semi remote with no major pollution sources nearby and the closest (<50 km) towns have fewer than 50,000 inhabitants. Major population centers of Detroit (5,200,000), Toronto (2,600,000),

Chicago (9,700,000) are located over 350 km away (US D.O.C, 2013). The nearest town is Pellston, located 5.5 km to the west, with a population of less than 1000. State highway 31 runs north-south through Pellston and there is a small airport in Pellston with fewer than 20 flights per day (Carroll, 2001.).

3.2.2. Gradient Sampling Methodology

Measurements were made at the PROPHET tower site at the University of

Michigan Biological Station from July 1 to August 8, 2009. The PROPHET tower is a

31.4 m scaffold tower. A glass manifold for air sampling extends to a height of 35-m above the ground and was used for instruments house in the PROPHET lab at the foot of the tower. For this study, 3 PFA sampling lines were used as inlets for the gas phase instrumentation housed in the Mobile Atmospheric Chemistry Lab (MACL) positioned at the base of the PROPHET tower. The MACL is an environmentally controlled mobile laboratory owned and operated by WSU’s Laboratory for Atmospheric Research. The inlets consisted of 5/8” OD PFA tubing and sampled from the tower at heights of 34-m and 20.4 m with the third inlet protruding from the MACL roof at a height of 6-m. Air flow rates through the tubing were 82, 64, and 58 SLPM through the 34 m, 20.4 m, and

6m respectively. The sample inlet heights were chosen to be above, in, and below the forest canopy to measure concentration gradients as shown in Fig. 3.1. Gas phase

126 instruments sub-sampled from these three inlets. Ozone was continuously measured from each inlet by either Dasibi or TECO ozone monitors. A proton transfer reaction mass spectrometer (PTR-MS) and a two channel NOx instrument alternatively sampled from each sample line by means of a 3-way PTFE Teflon solenoid valve (Teqcom) as illustrated in Fig. 3.2. The instruments sampled from each height for 10 minutes before the valves automatically switched, providing a profile every 30 minutes. The GC-MS system sampled from the 6-m inlet with some samples collected from the 20.4 m height.

3.2.3. Measurement Methods

Gradient profiles of VOCs were measured with a PTR-MS. The ions monitored are shown in Table 3.1. The PTR-MS mass spectrum is interpreted as an M+1 mass spectrum where M is the compound’s molecular weight. Monoterpene isomers cannot be distinguished by this method so the total monoterpene abundance is inferred from the abundance of the m/z 137 ion. Other major biogenic emissions from this forest include isoprene (m/z 69), methanol (m/z 33), acetaldehyde (m/z 45) and acetone (m/z 59) (Karl et al., 2003). The PTR-MS was operated with a -30 °C cold trap to remove water vapor as described by Jobson and McCoskey (2010). This allowed the PTR-MS drift tube to be operated dry and at a lower electric field (80 Td) and made measurements of formaldehyde (HCHO) and the radical chain termination product (CH3OOH) possible.

The PTR-MS was calibrated by diluting a multiple component compressed gas standard

(Scott-Marrin, stated accuracy ±10%) with humid zero air to 20 ppbv mixing ratio that was then sampled through the cold trap. The standard included isoprene, a-pinene, methacrolein and other compounds as noted in the Table. A permeation device (Kintek)

127 was used to calibrate formaldehyde response. An automated sequence was used to regularly zero and calibrate the instrument every 4 hours.

The NO, NO2 analyzer was operated in “Flux” configuration, where the NO and

NO2 were measured on separate channels and the NOy converter was bypassed. The instrument was zeroed using the pre-reactors for one minute every 30 minutes. After the pre-reactor zero, both channels measured NO for 1 minute before the Blue Light

Converter (BLC) (Air Quality Design, Inc., Golden, CO) was turned on for 28 minutes.

The two channels were operated in NO mode to allow for comparison during data analysis. NO and NO2 were calibrated every 23 hours using NO calibration gas 5.03 ppmv (Scott-Marrin, Inc., stated accuracy ± 2% EPA) that was diluted in 2 SLPM to approximately 10 ppbv. The conversion efficiency of the BLC was determined using gas phase titration with photolytically generated O3. The average conversion efficiency of the BLC during the campaign was 65 %. NO2 artifact was measured every 4 hours by placing the instrument on zero air for 5 minutes and turning the BLC on and off every 30 s. The NO2 artifact, which was less than 200 pptv was interpolated and subtracted from the NO2. Conversion of NO to NO2 in the sampling lines was not corrected for. Air was sampled through the 34 m inlet line at a flow rate of 82 SLPM giving a residence time in the tube of 6.2 seconds. The worst case scenario, which occurred when O3 was at the maximum of 53.7 ppbv, resulted in an underestimation of NO by 13%. During median O3 conditions (29.6 ppbv) NO would be under reported by 7%. However, PFA Teflon is not opaque in the ultraviolet range and it was impossible to know what the jNO2 inside the sample tube was so NO was not corrected for ozone losses. The overall uncertainty of

NO and NO2 was estimated at 15% and 20% respectively.

128 Additional VOCs were quantified using a Varian Gas Chromatography Ion Trap

Mass Spectrometer (GC-ITMS) and a VOC pre-concentration system. VOCs from a 400 to 1100 mL air sampling were cryogenically trapped by pulling ambient air at a flow rate of approximately 20 to 60 sccm through an open tubular trap immersed in liquid nitrogen. The sample size was determined by measuring the pressure change of a reference volume (8.324 L) as it was filled after evacuation. Prior to sampling, O3 was removed with a heated chemical trap consisting of a 25 mm Teflon filter coated in sodium thiosulfate and water vapor was removed with a cold trap described by Jobson and McCoskey (2010). The ozone filter was replaced every 2 days. The trap consisted of

0.020” ID Siltek coated tubing (Restek) and was resitively heated to 130 °C to desorb the analytes into the GC carrier gas flow. The pre-concentration system has been fully described by Faiola et al. (2012). Separation was performed on an HP-624 column

(0.320 mm, 1.80 µm, 60 m). A pressure pulse of 25 psig for one minute was applied to the sample loop to sweep the analytes onto the column. After the pressure pulse, the flow was held constant at 1 ml min-1. The initial oven temperature was 30 oC and was held isothermally for 5 minutes. After the initial isothermal section, the temperature was ramped at 2 oC min-1 to 110 oC and then taken to 240 oC at 50 oC min -1. The total analysis time was 47.6. The ion trap was operated in external ionization mode with an ionization current of 25 µA and a dynode voltage of 1200V. The target ion count was

20,000 and the maximum ionization time was 65,000 µs. The mass to charge range monitored by the ion trap was 45 to 300 m/z and the ion trap was turned on at 1.00 minutes into the run. Identification of monoterpenes was done by comparison of chromatogram peaks to known retention times and mass spectrum of monoterpene test

129 mixtures prepared in the lab. Calibration was based on an external compressed gas standard containing α-pinene (Scott-Marrin, stated accuracy ±10%). One problem encountered with the analysis was the collection of CO2 in the trap and its influence on the chromatography and performance of the ion trap in the early part of the chromatogram. The ions from the desorbed CO2 fill the ion trap reducing analyte sensitivity to isoprene and caused large baseline rise in first few minutes of the chromatogram. This lead to peak splitting and made integration of peaks eluting during this period, such as isoprene, difficult.

3.2.4. Comparison Between GC-IMS and PTR-MS Data

Significant differences between the monoterpene and isoprene mixing ratios reported by the PTR-MS and GC-ITMS were observed during this campaign. Figure 3.3 shows plots of total monoterpenes and isoprene measured by the two techniques. For monoterpenes, the linear correlation is acceptable having an r2 of 0.88 and a slope of 1.70 when the intercept is forced through zero. This indicates that there is a calibration issue, but that the two instruments are both responding to monoterpenes. The monoterpene response in the PTR-MS was calibrated to the m/z 137 ion signal produced by α-pinene.

+ Monoterpenes fragment upon protonation by H3O in the PTR-MS. Sensitivities for other monoterpenes will differ from α-pinene due to differences in the degree of fragmentation. According to the GC-ITMs results, α-pinene was 78% of the total monoterpene abundance. Given the abundance of α-pinene there should not be such a large discrepancy between the PTR-MS and GC-ITMS results. Variability may be due to sample handling problems with the GC-MS such as reactions with O3 and monoterpenes in the cryogenic trap, variability trapping efficiencies of monoterpenes in the cryotrap,

130 and/or differences in calibration and variable relative abundances of the monoterpenes.

The two techniques compare even more poorly when measuring isoprene. When the intercept is held at zero the slope is 0.99 which is a good result, however the r2 is 0.49.

Often the PTR-MS reports consistently higher isoprene than the GC-ITMS. One problem with the GC-MS analysis was the trapping of CO2 which caused lower sensitivities because CO2 filled the ion trap, and impacted the chromatography and the ability to integrate the isoprene peak which appeared on the “hump” of the CO2 induced background signal. While cryogenic trapping is a well-established approach for VOC sample pre-concentration from air, several problems associated with sample handling

(CO2, O3) and interfacing to an ion trap motivated the development of the adsorbent based pre-concentration described in chapter 2.2. In this environment there is good reason to believe that the PTR-MS measurements are of higher fidelity than the GC-

ITMS measurements

3.3. RESULTS

3.3.1. Meteorology

The meteorology of the site has been extensively documented by Cooper et al.

(2001). The site receives air from two dominant origins: clean, continental-polar air characterized flows from the north and north west and polluted maritime-subtropical air characterized by flow from the south-east through south-west (Cooper et al., 2001). The site lies on the edge of a summer ozone gradient and experiences transitions between clean continental air and aged polluted air. This makes the site atmospherically interesting because both the pristine and polluted conditions can be observed (Cooper et al., 2001; Vukovich, 1994.)

131 The temperature, pressure, wind speed and relative humidity measured at the 31 to 33 m height are shown for the campaign in Figure 3.4. Compared to the 1984 to 1998 averages presented in Cooper et al. (2001), the summer of 2009 was much different: The

CABINEX measurement period, July 1- August 8 2009 was exceptionally cool and cloudy (Bryan et al., 2012). Only two of four the dominant flow regimes mentioned in

Cooper et al. (2001) were observed. The south easterly flow and north easterly flows were not as prominent as reported by Cooper. The vast majority of the air coming in from northern flow was from the north west (Figure 3.5). A detailed explanation of the

Hysplit back trajectory analysis used to produce Figure 3.5 can be found in VanReken et al., (Submitted 2013). Figure 3.6 displays a wind rose plot showing that the dominant wind flow was from the northwest, with approximately half of the air received coming from 270o to 360o. The wind speeds were comparable to the 1984 to 1998 average at

Travers City, MI and Saul Ste. Marie, MI. July and August 2009 brought 3.81 cm and

13.9 cm of precipitation respectively for a total of 17.7 cm which is higher than the 1984 to 1998 average of 14.3 cm. The pressure during the observation period ranged from a maximum of 994.2 mb to a minimum of 973.8 mb and was significantly lower at an average 984.7 mb (4.8 mb stdev) than the 1984 to 1998 average of 993.4 mb (4.4 mb stdev). The average surface air pressure for CABINEX was 2 standard deviations lower than the average reported from 1984 to 1998, which is consistent with the passage of multiple low pressure fronts which brought rain from the northwest, cooler temperatures, and clean air from Canada. Despite having drastically different meteorology than the previously studied periods, the site still experienced transitions between clean air from

132 the northerly flows and polluted air from southerly flows. Table 3.2 summarizes the meteorological conditions observed during the campaign.

3.3.2. Trace Gas Data Time Series

3.3.2.1. Ozone

Figure 3.7 displays the ozone mixing ratio measured simultaneously from the 3 measurement heights. Ozone mixing ratios above the canopy were typically significantly higher than those below. At the 34-m height ozone mixing ratios varied between 53.7 and 12.2 ppbv. No evidence of particularly photochemically polluted air was observed.

3.3.2.2. NO and NO2

Mixing ratios of NO and NO2 measured over the course of the experiment are shown in Figure 3.8. Shown are 10 minute averages. NO mixing ratios peaked during the day when photolysis of NO2 was at a maximum and NO2 mixing ratios were highest during the mornings and evenings when the PBL height was lowest. The profiling portion of the campaign is indicated by a blue box.

3.3.2.3. Volatile Organic Compounds

Mixing ratios of selected compounds measured by the PTR-MS are shown in

Figures 3.9-3.12 to illustrate variation over the course of the profiling section of the field experiment. Isoprene mixing ratios displayed a pronounced daily variation, Figure 3.9, shows isoprene with maximum values occurring in the day and lowest values at night.

Isoprene is emitted during the day when photosynthesis is occurring and thus its emissions are light dependent. Low concentrations of isoprene were measured during

133 periods of overcast skies and rain. Monoterpene emissions do not generally display a light dependent emission and typically night time mixing ratios are highest because of lower mixed layer heights. This was evident in the CABINEX data, shown in Fig. 3.10, with highest values of monoterpenes occurring at night.

Methanol and acetone are also emitted by forests and emissions of these compounds have been measured at the PROPHET tower site by Karl et al. (2003). These compounds have much longer lifetimes than isoprene and monoterpenes and can be transported into the free troposphere to impact chemistry in regions far distant from their surface sources. The temporal variation of these compounds in Fig. 3.11 shows a tight correlation between the two compounds which is in good agreement with Karl et al.

(2003) suggesting that they have a common source, vegetation. Karl et al. (2003) report temperature dependence of methanol following the form: 퐸푓 ∗ exp (훽(푇 − 푇푠)) and that based on fluxes during spring and summer the ratio of acetone to methanol should be

0.18. We observed an emission ratio of acetone to methanol of 0.3, which agrees with the fall observations of Karl et al (2003) which report a ratio between 0.22 and 0.7 . One reason this might be is the relatively cloudy, cool, and rainy conditions experienced at the

PROPHET site during the summer of 2009, which are more typical of fall.

Figure 3.12 shows mixing ratios of benzene and toluene. These compounds are emitted from anthropogenic sources such as vehicle emissions. In general mixing ratios of these compounds were low and no obvious impact from local traffic was evident.

Variations in abundance were associated with mesoscale meteorology and transport form polluted regions from large urban centers to the south. Air mass origin is discussed in section 3.4.

134 3.4. AIR MASS ORIGINS

The NOx (NO + NO2) concentration measured at the site was influenced by air mass origin. Plotting a log histogram of the NOx values yields two distinct modes, which can be attributed to the two dominant flows experienced at the site. Figure 3.13 shows a log histogram of NOx and all of the NOx data plotted against time of day. The plot of

NOx against time of day shows a distinct baseline NOx and also elevated values. The two different modes of NOx were attributed to the influence of the 2 air mass origins that characterize flow to the site: clean continental-polar air, and polluted southerly air. A

NOx value of 550 pptv measured at the 34 m height was selected as a threshold for anthropogenic influence. This value was aggressively low selected to isolate the

“pristine” periods and minimize anthropogenic influence on the “pristine” data. Data above this threshold was classified as having been recently been exposed to anthropogenic emissions and those below 550 were said to be pristine. This criterion was then used to sort the other data into these air mass influence categories.

Sorting wind direction by these criteria yields the wind rose plots illustrated in

Fig. 3.14 which is consistent with the trajectory analysis. When the NOx is less than 550 pptv the winds were predominantly from the west and when the NOx was greater than

550 pptv the winds were from the south. Analysis of other anthropogenic tracers such as benzene and toluene were consistent with the 550 pptv cutoff as there was significantly more benzene and toluene when the NOx was elevated.

135 3.4.1. Gradients Sorted by NOx Cutoff

The case for cutoff point for anthropogenic influence of 550 pptv of NOx was discussed in the meteorology section of this paper. This section furthers the analysis of the data using this cutoff point. There were a total of 667 profile points, of which 388 were below the NOx cutoff and 279 were above it. Anthropogenic tracers toluene are shown along with isoprene, methanol and light flux having all been sorted by the NOx cutoff point (Figure 3.15). The diel plots of benzene and toluene show elevated levels when the NOx is above 550 pptv. The polluted periods were associated with cloudy days and as a result figure 3.15 C shows diminished light flux when NOx was above the cutoff period. Isoprene emissions which are light dependent are diminished during the polluted period. High NOx periods were also associated with lower temperatures, less photochemistry and fewer emissions of BVOCs. Sorting NOx by the cutoff gives the plots in Figure 3.16. This data supports the claim that elevated NOx is the result of transported pollution to the site from the industrialized parts of Michigan and the greater

Ohio River Valley. When NOx is above the cutoff value, the diel plots of the NOx gradients during the night, when the atmosphere has been shown to be stable 85-95% of the time (Steiner et al., 2011), show that there was a significant positive gradient ranging from 200 to 400 pptv. This means that the NOx is coming in from aloft. The histogram of NOx gradients (n=279) for the high NOx data shows a bias towards positive large gradients. Negative gradients do occur, however the number of negative gradients smaller than -300 pptv occurs less than 5% of the time. When the NOx is below the cutoff, the gradients during the night time when atmospheric stability is most likely to occur are small, negative and statistically significant. This is evidence that soils are emitting NOx,

136 and that this NOx builds up under the canopy at night under stable conditions. The low

NOx histogram (N=388) shows that most of the gradients are small (<|-200 pptv |) and occur frequently. Gradients of -50 pptv or smaller occur 31% of the time. These results indicate that soil NOx is very important to the chemistry of remote forested areas contributing a significant amount of the NOx.

3.5. VERTICAL GRADIENTS THROUGH A FOREST CANOPY

3.5.1. Measurements Overview

The analysis presented in this chapter is based on NO, NO2, O3, formaldehyde

(HCHO), methacrolein and methyl vinyl ketone (MACR+MVK), methyl hydroperoxide

(CH3OOH), monoterpenes, isoprene, benzene, and toluene. NO and NO2 are unique because they are both transported to the site as anthropogenic pollution and emitted biogenically by the soils as the result to of microbial activity. Ozone is a photochemical pollutant that may be produced locally or also transported to the site in an anthropogenic plume. HCHO, CH3OOH, and MACR+MCK are photo-products that are thought to be locally produced through the oxidation of isoprene and other VOCs. Isoprene oxidation is important to this forest because BVOC emissions are isoprene dominated. Methyl hydroperoxides have been shown to be very important reactive constituents of the atmosphere, acting as reservoirs for major oxidants (Zhang et al., 2012; Madronich and

Calvert 1999.) When a methyl hydroperoxyl radical is formed in the atmosphere NO and

HO2 compete for it. When it reacts with NO HCHO is the resulting product and when it reacts with HO2 methyl hydroperoxide is formed. Another source of methyl hydroperoxide is ozonolysis of alkenes such as isoprene, monoterpenes and sesquiterpenes. Monoterpenes and isoprene are directly emitted by the forest although

137 their emission mechanisms differ. Some monoterpenes emissions are temperature dependent only while others are light and temperature dependent like isoprene (Di Carlo et al., 2004). Benzene and toluene are anthropogenic tracers and because of the site’s relatively remote location are associated with transport from the industrialized regions of

Michigan and the Ohio River Valley. The average, median, maximum, minimum, and standard deviation for all of these trace gases at each height has been included as Table

3.3. The highest average and maximum mixing ratios are found above the canopy for O3,

NO, NO2, HCHO, MACR+MVK, CH3OOH and toluene. On average monoterpenes and isoprene are found in highest abundance below the canopy although the maximum isoprene observation occurred at the 34 m height. Toluene was found in highest average abundance at the 34 m height, however the maximum 10 minute average occurred at the

20 m height. Benzene did not show any significant difference between the heights.

The forest canopy greatly increases the complexity of mixing and chemistry at the

PROPHET site and sets up conditions favorable for the development of chemical gradients. The formation of gradients between the understory and the canopy are the result of photochemical oxidation, emissions, mixing, actinic flux gradients, deposition and transport (Wolfe et al. 2011). Emissions of BVOCs occur in the canopy and at the ground, in the understory. Anthropogenic VOCS are transported in from aloft and evidence suggests that NOx is both transported in from aloft and also emitted from the soil. NOx is important for this environment because it recycles HO2 to HO and in so doing produces O3. A significant gradient in ultra violet photons exists between the ground and above the canopy (Figure 3.17). Between the hours 10:00 and 18:00 average JNO2 above the canopy is 6.8 times higher than JNO2 at the ground. Light absorbed at the canopy

138 also inhibits mixing from occurring at the ground level as a large percentage of the sensible heat flux occurs at the canopy (Bryan et al., 2012). This creates stability in the understory and inhibits mixing allowing for trace gases to accumulate during the daytime.

The above canopy jNO2 is an excellent proxy for stability observed above the canopy.

Detailed descriptions of stability and mixing during CABINEX have been published by

Bryan et al. (2012) and Steiner et al. (2011). Using an Obukhov length stability classification, Steiner et al. (2011) demonstrated that stable conditions occurred above and at the forest canopy 80 to 90% during the night time (20:00-07:00), unstable conditions occurred 80 to 90% of the time during the daytime (09:30-18:00). During mornings (07:00-09:30) and evening (18:00-20:00) they show that the atmosphere was in transition between stability classes. Examination of the gradients that exist in this forest and the conditions that generate them elucidates the dynamics of chemistry, mixing, and transport in this transition forest.

The diel averages of NO, NO2, NOx and O3 at 6 m, 20 m, and 34 m show significant differences (Figure 3.18). During the night (23:00-06:00) O3 shows a continuous vertical gradient. Average O3 mixing ratios during this time are 30 ppbv, 26 ppbv and 17 ppbv for the 34 m, 20 m, and 6 m heights, respectively. Strong production of O3 was not observed as daytime O3 increased on average only 5 ppbv over the observed night time O3 values. This small variation is likely driven by dry deposition of

O3 to the forest canopy. During the well mixed portion of the day (09:30-17:00 EDT) O3 mixing ratios at the 34 m and 20 m sampling heights were almost identical, while O3 mixing ratios were approximately 6 ppbv lower at the 6 m height. This indicates that the forest canopy and air above it are somewhat well mixed, while the air below the canopy

139 in the forest understory is not. The lower O3 mixing ratios at 6-m are attributed to loss by dry deposition.

NO shows a different trend than O3. During the night NO within and above the canopy is essentially titrated away by O3 to NO2. However NO at the 6 m height is observed to exist at mixing ratios of ~5 pptv. While O3 mixing ratios are lower at this height, 17 ppbv, they are high enough to titrate the NO. One likely explanation for NO existing at detectable levels at the 6 m height is that NO is continuously emitted by the soils. At sunrise NO mixing ratios shoot up to their maximum which occurs at 09:00.

After 09:00 NO drops steadily until sunset as the PBL height grows. The presence of a morning NO peak is the result of photolysis of NO2 and PBL evolution throughout the day. As with O3, NO at the 34 m and 20 m heights are very similar. NO below the canopy is lower due to the JNO2 gradient presented in Figure 3.17. Both NOx and NO2 steadily climb through the night from a nightly low which occurs at sunset to a peak that occurs with the NO peak at 09:00. After 09:00 NO2 and NOx fall as the PBL height increases.

Just after 18:00 the PBL height falls and concentrations begin to rise again. Only one small difference between NOx and NO2 exists as NOx during CABINEX was dominated by NO2. NO2 at the 6 m height is higher than the 34 m and 20 m heights while NOx is essentially the same during the day.

The relative amounts of NO, NO2, and O3 in the daytime can be simply described by a set of reactions that rapidly come into photo stationary equilibrium.

푁푂 + 푂3 → 푁푂2 + 푂2 R3.1

푁푂2 + ℎ푣 → 푁푂 + 푂 R3.2

푂 + 푂2 + 푀 → 푂3 + 푀 R3.3

140 A 1D model of O3, jNO2, NO, and NO2 was run to determine the time it would take for air to come to equilibrium. The model results indicated that during the daytime above the canopy photo stationary equilibrium would is achieved within 5 to 10 minutes and below the canopy this takes 15 to 30 minutes due to the lower jNO2 values. This simple model also indicated that the NO2 to NO ratio at the 6 m height would be about 5 times higher due to the lower jNO2 observed at the 6 m height. The model indicated that with narrow range of O3 and NOx observed, the time it takes to reach equilibrium was most sensitive to jNO2. The jNO2 gradient between the ground and the 34 m inlet along with diminished daytime mixing between the 6 m and 34 m inlet produces conditions where observations deviate from photo stationary equilibrium.

The photo stationary equilibrium ratio of NO2 to NO was calculated using O3 and jNO2 values from above and below the canopy and compared to the measured NO2 to NO ratio. Figure 3.19 shows the results from this analysis at both the 6 m and 34 m heights.

The deviations from photo stationary equilibrium are consistent with the established forest canopy air exchange that involving coherent structures or “bursts” and “sweeps”

(Brunet and Irvine, 2000; Collineau and Brunet, 1993a; Raupach et al., 1996; Steiner et al. 2011, Bryan et al., 2012). A “burst” is a slow upward discharge of air from the canopy to the atmosphere above while a “sweep” is a rapid downward penetration of the forest canopy by air from the atmosphere above. “Bursts” and “sweeps” are thought to be important means of moving scalars through the canopy and may help explain the presence of chemical gradients (Steiner et al., 2011; Thomas and Foken, 2007). The ratio of NO2 to NO at 34 m height approaches the calculated photo stationary equilibrium much more closely that the NO2 to NO ratio at the 6 m height. This is consistent with the

141 shorter time to reach photo stationary equilibrium at the 34 m height. Deviations from the photo stationary state at the 34 meter height are consistent with air from below the forest

“bursting” up from the understory resulting in a measured NO2 to NO is higher than the predicted. This is readily apparent in the log histogram of the calculated ratio to the observed ratio (Figure 3.19b). Most of the calculated NO2 to NO ratios are lower than the measured NO2 to NO ratio at the 34 m height, which is consistent with air from the understory slowly “bursting” up. Below the canopy at the 6 m height, the observations and model results don’t agree. The plot of the modeled NO2 to NO versus observed NO2 to NO shows a scatter of values. This can be attributed to two processes: the first is the emission of NO from the soil which lowers the observed NO2 to NO ratio. The second is the “sweeps” of air that has a low NO2 to NO ratio from above the canopy. The impact of “sweeps” on the NO2 to NO ratios is shown in the log histogram of the ratio of calculated to measured NO2 to NO ratios at the 6 m height, which show a mode centered at 10.5.

3.5.2. BVOC and oxidation product gradients

The diel profiles and gradients of total monoterpenes and isoprene were quantified during CABINEX (Figure 20). The diel maximum monoterpene mixing ratios were observed during the night, while the maximum diel isoprene ratios are observed at around 16:00, where the minimum diel monoterpene mixing ratios are observed. The minimum diel isoprene mixing ratios were observed at sunrise, but were low throughout the night. Both isoprene and monoterpenes are emitted from the plants however; monoterpene emissions are temperature dependent, while isoprene emissions are light dependent (Guenther et al., 2006; Di Carlo et al., 2004). Isoprene above the canopy is

142 also being removed by OH and producing HCHO and MACR+MVK more rapidly than below the canopy. The daytime monoterpene gradients are negligible and become negative, down to -0.3 ppbv during the night. The opposite for the isoprene gradients are true: they are negative during the day, going as low as -0.6 ppbv and go to zero at night.

One confounding issue is that during the day both isoprene and monoterpenes are co- emitted, however there is an isoprene gradient and there is no apparent monoterpene gradient. The answer is that during the day monoterpene concentrations fall to around 0.1 ppbv and the PTR-MS , which has a detection limit of ~20 pptv would have a hard time resolving the small 10 to 30 pptv gradients that are likely to exist. Another important observation is that gradients exist during the day when the atmosphere should be unstable. This is because the forest inhibits mixing by changing the height at which the sensible heat flux enters the atmosphere and by inhibiting advection of pollutants (Steiner et al, 2011; Bryan et al., 2012). While there is instability at the canopy level and above it, below the atmosphere is stable and isoprene can build up until a “sweep” brings air aloft into the understory, or a “burst” removes it. Isoprene above the canopy is subject to both chemical removal and dilution, which leads to lower mixing ratios above the canopy.

Oxidation products HCHO, MACR+MVK and CH3OOH were also measured with the

PTR-MS during the canopy profiling experiment (Figure 3.21). All of the oxidation products showed the highest diel average mixing ratios during the day, although the time that this maximum occurred differed. MACR+MVK mixing ratios peak at around 10:00,

HCHO mixing ratios peak at around 17:00, and the CH3OOH mixing ratios peak at around 16:00. The gradients for all three are very similar and always positive. These products are generated above the forest canopy and even though there is mixing during

143 the day the forest inhibits thorough mixing of the oxidation products down through the canopy. During the night when atmospherically stable conditions exist, and photochemical production halts, the gradients grow. Without new sources, the development of these strong gradients at night is the result of depositional losses within the forest canopy. Diel profiles and gradients of methanol and acetone, which have been reported as abundant VOCs at the PROPHET site by Karl et al. (2003) are shown in

Figure 3.22. The emissions of methanol and acetone are well correlated with an r2 of 0.6.

3.6. GCMS DATA

3.6.1. Comparison of Monoterpene Speciation to Historical Record

The GC-ITMS cryogen trap system was able to identify 7 monoterpenes during the CABINEX 2009 field intensive including α-pinene, β-pinene, camphene, α- phellandrene, 3-carene, limonene, and one unidentified monoterpene. α-pinene was by far the most abundant monoterpene accounting for 76.8 % of all observed monoterpenes, with β-pinene as the second most abundant monoterpene at 11.6%. Table 3.4 shows the percent abundance, maximum, minimum, median and standard deviation. Monoterpenes have been measured at this site for a number of summers. One confounding issue is that there does not seem to be any consistency in the observed ratio of α-pinene to β-pinene.

Figure 3.23 shows the α-pinene to β-pinene ratio for 4 independent measurements of monoterpenes. There is not good agreement between the works. This work observed an

α-pinene to β-pinene ratio of 5.0 at the 6 m height. The maximum α-pinene to β-pinene ratio was 20.4, the minimum was 2.6 and the standard deviation was 2.9. Measurements made concurrently by the group at Eastern Michigan University at the 20 m height, observed a α-pinene to β-pinene ratio of 1.7. Our results appear to be in agreement with

144 unpublished measurements made in 1997 by WSU’s H. Westberg which observed a α- pinene to β-pinene ratio of 4.2. Apel et al., (2005) have observed the most drastically different α-pinene to β-pinene ratio. Their work indicated that β-pinene and not α-pinene is the dominant monoterpene at the PROPHET site. These results suggest that either the monoterpene emissions are extremely variable, or different systems are not performing identically. The large difference between the two ratio observed during CABINEX indicate that a monoterpene intercomparison is necessary.

3.7. OH LOSS FREQUENCY

The hydroxyl loss frequencies have been calculated by determining the product of the rate constant times the concentration of measured BVOCs. The concentration of trace gas ί in molecules cm-3 was determined by using the ideal gas law:

푛 푃 [푖] = 휒 ∗ ∗ 10−9 = 휒 ∗ 10−9 3.1 푖 푉 푖 푅푇

Where 휒푖 is the mixing ratio of ί in ppbv, P is the pressure in kPa, R is the gas constant and equal to 8314 cm3 kPa K−1 mol−1, and T is the temperature in Kelvin. The rate constants for limonene, α-pinene, β-pinene and isoprene were calculated using the

Arrhenius equation:

−퐵𝑖/푇 푘푖 = 퐴푖 ∗ 푒 3.2

Where the constants A and B are the Arrhenius constants and T is the temperature in

Kelvin. The remaining monoterpenes did not have Arrhenius constants in the literature, and the rate constants at 298 K were used. Monoterpenes are the smallest contribution to

OH loss, and limonene, α-pinene and β-pinene account for 89.4% of the total monoterpene mass, therefore the errors created by using fixed rate constants are minimal.

Isoprene has been shown to be the dominant BVOC emitted by the forest surrounding the

145 PROPHET (Guenter et al., 1995). Figure 3.24 shows the hydroxyl loss frequency s-1 by

NO2, Isoprene, and total monoterpenes. The area shown per color corresponds to the OH loss rate. Hydroxyl loss frequency peaks at 16:00 EDT at ~2.5 s-1 and reaches a minimum of ~0.5 s-1 during the night. This work shows that Isoprene is the dominant BVOC hydroxyl radical sink during the day, when isoprene emissions are high. However, during the night to early morning when isoprene emission are zero (00:00-06:00) NO2 is the major sink for hydroxyl radicals. Monoterpenes are a small contribution to the hydroxyl loss frequency.

Figure 3.25 shows the percent contribution to the hydroxyl radical loss frequency for isoprene, NO2 and total monoterpenes. During the morning period (00:00 to 06:00)

NO2 accounts for 58% of the OH loss frequency, while monoterpenes account for 12%, and Isoprene 31%. Once the sun comes up and the biosphere begins to emit isoprene

(06:00 to 12:00), isoprene becomes the major sink for HO radicals accounting for 62% of the OH loss. This transitional period includes NOx and monoterpenes trapped in the nocturnal boundary layer and also sees its’ breakdown and growth of the PBL height. As the PBL height grows, NO2 and monoterpenes are diluted and their total and fractional contributions to HO loss shrink. During the day both NO2 and monoterpene mixing ratios fall due to mixing, however isoprene mixing ratios are at their maximum. That the isoprene emissions reach a maximum during the day when sensible heat flux also at a maximum and atmospheric stability is a testament to the strength of their source: From

12:00 to 18:00 isoprene accounts for 95% of the hydroxyl radical loss with NO2 and monoterpenes contributing 4% and 1% respectively. 18:00 to 24:00 includes another transitional period where the nocturnal boundary layer reforms, isoprene emission shut

146 down and the contribution to OH loss from NO2 and monoterpenes grows. Isoprene is still the dominant OH sink at 83% with NO2 following at 14% and monoterpenes making up the final 3%.

3.8. SUMMARY

Profiles of NOx, VOCs, and O3 at 6 m, 20 m, and 34 m heights made during the

2009 CABINEX campaign are reported for a transitional forest located at the PROPHET tower at the UMBS in northern Michigan. The campaign occurred during an unusually cold and rainy summer; however the site experienced transitions between pristine and polluted conditions that make it relevant to understanding the influence of anthropogenic pollution on forested areas. A 550 pptv cutoff for NOx was chosen as a division between pristine influence and transported pollution. The winds during the clean periods were from the west and north while during the polluted period the winds were primarily from the south. Sorting the data by this cutoff point showed that there is a soil NOx source that produces a NOx gradient during the day. Examining two anthropogenic tracers benzene and toluene by this definition, showed that they are found in higher abundance during the polluted period. The profiles showed that the forest canopy sets up conditions favorable for the formation of chemical gradients. Examining the NO2 to NO ratio showed evidence that exchange of air between the above canopy and below canopy leads to perturbation of the photo-stationary equilibrium. Examination of the profiles and chemical gradients that exist as a result of the light gradient and the forest canopy’s inhibition of mixing indicate that mixing of air in a forest canopy is a complex process. Daytime gradients of isoprene show that isoprene is able to build up under the canopy because the canopy inhibits mixing to the ground level. This is consistent with works that point to “bursts” and

147 “sweeps” as the means of mixing within a canopy. Monoterpene gradients occur at the night time although they probably exist during the day as well but our instrumentation is not sensitive enough to quantify small (30 pptv or less) gradients. Oxidation product gradients for MACR+MVK, CH3OOH, and HCHO show that the largest gradients occur during the night although they are present during the day. These large gradients are the result of depositional loss. The ratio of α-pinene to β-pinene observed during the campaign was compared to previous measurements and to measurements made by EMU during CABINEX. The ratio of α-pinene and β-pinene shows very little consistency.

Either there is a great deal of variability in the monoterpene emissions or the different measurements are not producing the same results. The OH loss rate has also been examined and isoprene is the largest OH sink, although during the night NO2 is the largest sink. Discrepancies between the PTR-MS and the GC-ITMS demand further investigation through intercomparison. The two methods obtain different results.

148 3.9. REFRENCES

Apel, E. C., et al. "Measurement and interpretation of isoprene fluxes and isoprene, methacrolein, and methyl vinyl ketone mixing ratios at the PROPHET site during the

1998 Intensive." Journal of geophysical research 107.D3 (2002): 4034.

Atkinson, R., and Arey, J. "Gas-phase tropospheric chemistry of biogenic volatile organic compounds: a review." Atmospheric Environment 37 (2003): 197-219.

Barket, D.J., et al. "Intercomparison of automated methodologies for determination of ambient isoprene during the PROPHET 1998 summer campaign." Journal of

Geophysical Research: Atmospheres (1984–2012) 106.D20 (2001): 24301-24313.

Bergen, K.M., and Dronova, I. "Observing succession on aspen-dominated landscapes using a remote sensing-ecosystem approach." Landscape Ecology 22.9 (2007): 1395-

1410.

Brunet, Y., and Irvine, M.R. "The control of coherent eddies in vegetation canopies: streamwise structure spacing, canopy shear scale and atmospheric stability." Boundary-

Layer Meteorology 94.1 (2000): 139-163.

Calvert, J.G., et al. “The mechanisms of atmospheric oxidation of the alkenes.” Vol. 5.

New York: Oxford University Press, 2000.

149 Carroll, M.A., Bertman, S.B., and Shepson, P.B. "Overview of the Program for Research on Oxidants: Photochemistry, Emissions, and Transport (PROPHET) summer 1998 measurements intensive." Journal of Geophysical Research: Atmospheres (1984–2012)

106.D20 (2001): 24275-24288

Collineau, S., and Brunet, Y. "Detection of turbulent coherent motions in a forest canopy part I: Wavelet analysis." Boundary-Layer Meteorology 65.4 (1993a): 357-379.

Collineau, S., and Brunet, Y. "Detection of turbulent coherent motions in a forest canopy part II: Time-scales and conditional averages." Boundary-Layer Meteorology 66.1-2

(1993b): 49-73.

Cooper, O.R., et al. "PROPHET 1998 meteorological overview and air-mass classification." Journal of Geophysical Research 106.D20 (2001): 24289-24.

Di Carlo, P., et al. "Missing OH reactivity in a forest: Evidence for unknown reactive biogenic VOCs." Science 304.5671 (2004): 722-725.

Faiola, C.L., et al. "Quantification of biogenic volatile organic compounds with a flame ionization detector using the effective carbon number concept." Meas. Tech. Discuss 5

(2012): 2415-2447.

150 Gill, K.J., and Hites, R.A. "Rate constants for the gas-phase reactions of the hydroxyl radical with isoprene, α-and β-pinene, and limonene as a function of temperature." The

Journal of Physical Chemistry A 106.11 (2002): 2538-2544.

Goldstein, A.H., and Galbally, I.E. "Known and unexplored organic constituents in the earth's atmosphere." Environmental Science & Technology 41.5 (2007): 1514-1521.

Guenther, A., et al. "Estimates of global terrestrial isoprene emissions using MEGAN

(Model of Emissions of Gases and Aerosols from Nature)." Atmospheric Chemistry and

Physics 6.11 (2006): 3181-3210.

Guenther, A., et al. "A global model of natural volatile organic compound emissions."

Journal of Geophysical Research 100.D5 (1995): 8873-8892.

Hansen, M. C., et al. "Global percent tree cover at a spatial resolution of 500 meters: First results of the MODIS vegetation continuous fields algorithm." Earth Interactions 7.10

(2003): 1-15.

Holland, Elisabeth A., et al. "Contemporary and pre-industrial global reactive nitrogen budgets." Biogeochemistry 46.1-3 (1999): 7-43.

151 Holzinger, R., et al. "Observations of oxidation products above a forest imply biogenic emissions of very reactive compounds." Atmospheric Chemistry and Physics 5.1 (2005):

67-75.

Hudman, R. C., et al. "Steps towards a mechanistic model of global soil nitric oxide emissions: implementation and space based-constraints." Atmospheric Chemistry and

Physics 12.16 (2012): 7779-7795.

Karl, T., Guenther, A., Spirig, C., Hansel, A., and Fall, R. (2003). Seasonal variation of biogenic VOC emissions above a mixed hardwood forest in northern Michigan.

Geophysical Research Letters, 30(23).

Kleindienst, Tadeusz E., Geoffrey W. Harris, and Pitts, J.N. "Rates and temperature dependences of the reaction of hydroxyl radical with isoprene, its oxidation products, and selected terpenes." Environmental Science & Technology 16.12 (1982): 844-846.

Jobson, B. T., and McCoskey, J.K. "Sample drying to improve HCHO measurements by

PTR-MS instruments: laboratory and field measurements." Atmospheric Chemistry and

Physics 10 (2010): 1821-1835.

Madronich, S., and Calvert, J.G. "Permutation reactions of organic peroxy radicals in the troposphere." Journal of Geophysical Research: Atmospheres (1984–2012) 95.D5

(1990): 5697-5715.

152

Mao, J., et al. "Insights into hydroxyl measurements and atmospheric oxidation in a

California forest." Atmospheric Chemistry and Physics 12.17 (2012): 8009-8020.

Ortega, John, et al. "Flux estimates and OH reaction potential of reactive biogenic volatile organic compounds (BVOCs) from a mixed northern hardwood forest."

Atmospheric Environment 41.26 (2007): 5479-5495.

Pearsall, Douglas R. Landscape ecosystems of the University of Michigan biological station: Ecosystem diversity and ground-cover diversity. No. DOE/OR/00033--T708.

Michigan Univ., Ann Arbor, MI (United States), 1995.

Pressley, Shelley, et al. "Long-term isoprene flux measurements above a northern hardwood forest." Journal of Geophysical Research: Atmospheres (1984–2012) 110.D7

(2005).

Pressley, S.N., et al. "Relationships among canopy scale energy fluxes and isoprene flux derived from long-term, seasonal eddy covariance measurements over a hardwood forest." Agricultural and Forest Meteorology 136.3 (2006): 188-202.

Raupach, M.R. "Conditional statistics of Reynolds stress in rough-wall and smooth-wall turbulent boundary layers." Journal of Fluid Mechanics 108 (1981): 363-382.

153 Rizzo, L.V., et al. "Aerosol properties, in-canopy gradients, turbulent fluxes and VOC concentrations at a pristine forest site in Amazonia." Atmospheric Environment 44.4

(2010): 503-511.

Schmid, H.P., et al. "Ecosystem-atmosphere exchange of carbon dioxide over a mixed hardwood forest in northern lower Michigan." Journal of Geophysical Research 108.D14

(2003): 4417.

Thomas, C., and Foken, T. "Flux contribution of coherent structures and its implications for the exchange of energy and matter in a tall spruce canopy." Boundary-Layer

Meteorology 123.2 (2007): 317-337.

United States Department of Commerce. American Fact Finder. United States Census

Burea. 2010. Web. May 2013.

Vukovich, Fred M. "Regional-scale boundary layer ozone variations in the eastern United

States and their association with meteorological variations." Atmospheric Environment

29.17 (1995): 2259-2273.

Westberg, H., et al. "Measurement of isoprene fluxes at the PROPHET site." Journal of

Geophysical Research: Atmospheres (1984–2012) 106.D20 (2001): 24347-24358.

154 Wolfe, G. M., et al. "The Chemistry of Atmosphere-Forest Exchange(CAFE) Model- Part

2: Application to BEARPEX-2007 observations." Atmospheric Chemistry and Physics

11.3 (2011): 1269-1294.

Zhang, X., et al. "Methyl hydroperoxide (CH3OOH) in urban, suburban and rural atmosphere: ambient concentration, budget, and contribution to the atmospheric oxidizing capacity." Atmos. Chem. Phys 12 (2012): 8951-8962.

155 3.10. CHAPTER 3 TABLES

Table 3.1. Ions monitored by PTR-MS for gradient profiling m/z Compound Description 31 Formaldehyde photoproduct 33 Methanol* biogenic emission 42 Acetonitrile* wood smoke tracer 45 Acetaldehyde* biogenic emission 49 CH3OOH chain termination photoproduct 59 Acetone* biogenic emission 69 Isoprene* biogenic emission 71 Methacrolein* + methyl vinyl isoprene photoproducts ketone 75 Hydroxyl acetone oxidation product 79 Benzene* anthropogenic 93 Toluene * anthropogenic 107 C2-alkylbenzenes* anthropogenic 121 C3-alkylbenzenes* anthropogenic 137 Monoterpenes* biogenic emissions 139 Nopinione O3 + b-pinene reaction product * indicates compound present in compressed gas standard, specifically p-xylene for C2- alkylbenzene response, 1,2,3-trimethylbenzene for C3-alknylbenzene response, and α- pinene for monoterpene response.

156

Table 3.2. Summary of meteorology during CABINEX 2009 compared to 1984-1998

July-August data from Pellston, MI.

CABINEX 1984-1998 July-August Pellston* Traverse City, MI†* Average daily maximum 21.4 ± 2.8 25.5 ± 4.0 temperature (oC) Average daily minimum 11.9 ± 2.8 12.4 ± 4.3 temperature (oC) Average pressure (mb) 980.4 ± 4.8 993.4 ± 4.4 Total Precipitation 17.7 cm 14.3 ± 4 Average wind speed @ 33- 2.9 ± 2.7 3.2 ± 1.2† m (m/s) Average wind speed @ 6-m 0.5 ± 0.4 NA (m/s) Max wind speed @ 33-m 9.5 NA (m/s) Max wind speed @ 6-m 8.5 NA (m/s) * data from Copper et al., 2001

Table 3.3. Summary of monoterpenes (ppbv) at CABINEX. Monoterpene Abundance % Mean Maximum Minimum Median σ α-pinene 76.8 0.08 0.64 BDL 0.064 0.08 β-pinene 11.6 0.01 0.12 BDL 0.005 0.014 Camphene 6.4 0.007 0.056 BDL 0.004 0.009 α-phellandrene 2.1 0.002 0.016 BDL 0.001 0.003 3-Carene 2 0.002 0.04 BDL 0.001 0.003 Limonene 1 0.001 0.06 BDL BDL 0.005 UNK MT1 0.1 0.0001 0.02 BDL BDL 0.001 Isoprene NA 0.67 6.0 BDL 0.36 0.87

157 Table 3.4. Summary of the 10 minute averaged trace gases mixing ratios (ppbv) reported for each height.

Compound / height Average Median Max Min Ozone 34 m 31.2 ± 7.7 29.6 53.7 12.2 20 m 29.2 ± 8.5 28.2 53.4 7 6 m 22.7 ± 9.5 22.7 45.2 BDL NO 34 m 0.047 ± 0.083 0.015 0.781 BDL 20 m 0.031 ± 0.067 0.009 0.640 BDL 6 m 0.019 ±0.032 0.008 0.276 BDL

NO2 34 m 0.712 ± 0.739 0.420 4.885 BDL 20 m 0.670 ± 0.653 0.413 3.995 0.039 6 m 0.706 ± 0.963 0.467 4.971 0.057 Isoprene 34 m 0.77 ± 0.75 0.51 4.84 BDL 20 m 0.93 ± 0.89 0.62 4.46 BDL 6 m 0.93 ± 0.91 0.63 4.46 BDL Monoterpenes 34 m 0.19 ± 0.15 0.15 1.15 BDL 20 m 0.21 ± 0.16 0.16 1.38 BDL 6 m 0.26 ± 0.26 0.18 3.67 BDL Methanol 34 m 3.96 ± 1.56 3.48 9.04 0.95 20 m 3.84 ± 1.59 3.49 9.18 0.71 6 m 3.25 ± 1.55 3.14 8.24 BDL Acetone 34 m 1.98 ± 0.51 1.88 3.58 0.52 20 m 1.93 ± 0.51 1.83 3.42 0.84 6 m 1.80 ± 0.62 1.71 4.22 0.05 Formaldehyde 34m 0.08 ± 0.31 0.73 3.42 BDL 20m 0.72 ± 0.31 0.68 4.22 0.04 6m 0.58 ± 0.32 0.55 1.87 BDL MACR+MVK 34 m 0.44 ± 0.29 0.35 1.96 BDL 20m 0.42 ± 0.26 0.35 1.47 BDL 6m 0.35 ± 0.24 0.29 1.63 BDL

CH3OOH 34m 0.19 ± 0.09 0.18 0.55 BDL

CH3OOH 20m 0.17 ± 0.10 0.15 0.58 BDL

CH3OOH 6m 0.14 ± 0.11 0.11 0.52 BDL toluene 34m 0.09 ± 0.05 0.08 0.28 BDL toluene 20m 0.08 ± 0.05 0.07 0.30 BDL toluene 6m 0.06 ± 0.04 0.05 0.24 BDL benzene 34m 0.05 ± 0.03 0.05 0.18 BDL benzene 20m 0.05 ± 0.04 0.05 0.19 BDL benzene 6m 0.05 ± 0.03 0.04 0.16 BDL

158 3.11. CHAPTER 3 FIGURES

Figure 3.1. Schematic of the PROPHET tower, PROPHET laboratory and the WSU

MACL trailer showing inlet heights for the gradient profile sampling. Schematic courtesy of Shelley Pressley.

159

Figure 3.2. Flow diagram of the profiling schematic.

160

Figure 3.3. PTR-MS vs. GC-ITMS for total monoterpenes (top) and isoprene (bottom) along with the 1:1 line (dashed black).

161

Figure 3.4. Time series of meteorological values for the CABINEX field campaign as measured at the 31 to 33 m height showing relative humidity (blue) and temperature

(red), surface air pressure and wind speed.

162

Figure 3.5. A plot of air mass residence time based on hysplit back trajectories

(VanReken et al., Submitted).

163

Figure 3.6. A wind rose plot for the measurement period. The winds were measured at the 33 m anemometer.

164

Figure 3.7. Plots of O3 at the 3 sample heights.

165

Figure 3.8. Plots of NO (red) and NO2 (black) for the campaign at the 34, 20.4, and 6 meter heights. The profiling experiment is indicated by the blue box.

166

Figure 3.9. 10 minute averages of isoprene mixing ratios shown at the 34, 20.4, and 6 m heights.

167

Figure 3.10. Monoterpene mixing ratios (10 minute averages) at the 34, 20.4 and 6 m heights.

168

Figure 3.11. 10 minute average mixing ratios of methanol (black) and acetone (red) at

34, 20.4, 6 m heights.

169

Figure 3.12. 10 minute averages of benzene (black circles) and toluene (red cross) at the

34, 20.4 and 6 m heights.

170

Figure 3.13. Shown top is a log histogram of the NOx data with two clear modes.

Bottom is the NOx data at 34 m (black), 20 m (light gray) and 6 m (dark gray) along with the 550 pptv cutoff.

171

o Figure 3.14. Wind rose plots with 15 bins sorted by a NOx cutoff of 550 pptv. The top

(dark gray) plot shows winds for values less than 550 pptv while the bottom rose (light gray) has been sorted by NOx values over 550 pptv.

172

Figure 3.15. A) Plot of diel averaged methanol. B) Plot of diel averaged toluene. C) plot of diel averaged light. D) Plot of diel averaged isoprene. All of the data were sorted by the NOx mixing ratio and the data that were taken when NOx was greater than 550 ppvt are represented by light gray triangles, while the data that were taken when NOx was less than 550 pptv are dark gray circles.

173

Figure 3.16. A) diel plot of NOx gradients [NOx,34m]-[NOx,6m] (black squares) when the

NOx is lower than the 550 pptv cutoff the error bars are 1 standard deviation. B)

Histogram of NOx gradients when NOx is below the 550 pptv cutoff. C) Diel plot of the

NOx gradients [NOx,34m]-[NOx,6m] (gray circles) when the NOx is greater than the 550 pptv cutoff the error bars are 1 standard deviation. D) Histogram of NOx gradients when

NOx is greater than the 550 pptv cutoff.

174

Figure 3.17. Diel averaged JNO2 for the campaign at 34 m (light gray circles) and at the ground (dark gray triangles). Data from the CABINEX archives.

175

Figure 3.18. diel averages for A) O3, B) NO, C) NO2, D) NOx, are shown for the 34 m

(light gray squares), 20 m (medium gray circles), and 6 m (dark gray triangles) heights.

The vertical dashed lines indicate the average time of sunrise and sunset during

CABINEX.

176

Figure 3.19. Top left plots [NO2]/[NO] calculated versus [NO2]/[NO] observed at the 34 m height. Top right is a log histogram of the ratio of NO2 to NO calculated to the ratio of

NO2 to NO observed at the 34 m height. Bottom left is NO2]/[NO] calculated versus

[NO2]/[NO] observed at the 6 m height. Bottom right shows a log histogram of the ratio of NO2 to NO calculated to the ratio of NO2 to NO observed at the 6 m height.

177

Figure 3.20. A) Diel plot of monoterpenes at 34 m (light gray squares), 20 m (medium gray circles), and 6 m (black triangles). B) Diel plot of the 34 m – 6 m monoterpene gradient. The error bars indicate 1 standard deviation. C) Diel plot of isoprene at 34 m

(light gray squares), 20 m (medium gray circles), and 6 m (black triangles). D) Diel plot of the 34 m – 6 m isoprene gradient. The error bars indicate 1 standard deviation.

178

Figure 3.21. A) Diel plot of MACR+MVK at 34 m (light gray squares), 20 m (medium gray circles), and 6 m (black triangles). B) Diel plot of the 34 m – 6 m MACR+MVK gradient. The error bars indicate 1 standard deviation. C) Diel plot of HCHO at 34 m

(light gray squares), 20 m (medium gray circles), and 6 m (black triangles). D) Diel plot of the 34 m – 6 m HCHO gradient. The error bars indicate 1 standard deviation. E) Diel plot of CH3OOH at 34 m (light gray squares), 20 m (medium gray circles), and 6 m

(black triangles). F) Diel plot of the 34 m – 6 m CH3OOH gradient. The error bars indicate 1 standard deviation.

179

Figure 3.22. A) Diel plot of acetone at 34 m (light gray squares), 20 m (medium gray circles), and 6 m (black triangles). B) Diel plot of the 34 m – 6 m acetone gradient. The error bars indicate 1 standard deviation. C) Diel plot of methanol at 34 m (light gray squares), 20 m (medium gray circles), and 6 m (black triangles). D) Diel plot of the 34 m

– 6 m methanol gradient. The error bars indicate 1 standard deviation.

180

Figure 3.23. α-pinene plotted against β-pinene for this work (dark gray circles),

Westberg 1997 (solid black line), CABINEX 2009 EMU (dashed line), and Apel 2005

(light gray circles).

181

Figure 3.24. A plot of the median OH loss frequency as a function of time of day. The area shown is the by component OH loss; the blue trace is the total OH loss frequency, the green trace is the OH loss for NO2 plus total monoterpenes and the red is the OH loss to monoterpenes. By overlapping the traces the area shown corresponds to the individual component.

182

Figure 3.25. Contributions by percent to the hydroxyl radical loss rate shown for isoprene (blue), NO2 (green) and monoterpenes (red) split into 6 hour sections. A) 00:00 to 06:00. B) 06:00 to 12:00. C) 18:00 to 24:00 D) 12:00 to 18:00.

183 COMPARISON OF WINTERTIME CO TO NOx RATIOS TO

MOVES AND MOBILE6.2 ON-ROAD EMISSIONS INVENTORIES

(Published in Atmospheric Environment 2012)

4.1. INTRODUCTION

Mobile emissions are an important source of NO and NO2 (NOx), CO, and volatile organic compounds (VOCs) in urban areas and their photochemistry causes ozone and particulate matter (PM) formation in summer. NOx may also play a role in wintertime PM production through heterogeneous conversion of NO2 to HNO3 (Silvia et al., 2007; Kleffmann et al., 1998). Quantifying vehicle emission rates and parameterizing emissions for emission inventories are essential components of air quality management.

Periodically evaluating emission inventories by comparing them to real world measurements of vehicle emissions is important because vehicle emissions change over time as a result of fleet turnover, introduction of stricter emission requirements and oxygenated fuels programs. In the US, road-way emissions of NOx have steadily grown with increase in vehicle miles travelled (VMT) and increased reliance on diesel fuel, and now account for over half of the anthropogenic NOx emissions in the United States, while

CO emissions from vehicles have fallen (Parrish, 2006; Harley et al., 2010; Millstein,

2010; Ban-Weiss et al., 2008). Measured ratios of CO-to-NOx in urban areas across the

United States have decreased by 6.6 ± 0.3 % per year from 1994 to 2003, but this decrease is not reflected in vehicle emission inventories (Parrish, 2006).

Evaluating the mobile emission inventory is difficult owing to the complexity of the emissions process and the large number and wide variety of vehicles involved. Some important factors affecting emissions of CO, NOx, and VOCs include engine type and

184 age, composition of fuel including the increasing use of oxygenated fuels, engine load, efficiency of exhaust treatment devices such as 3-way catalytic converters, and ambient temperature (Gentner et al., 2009; Rubin et al., 2006; Lough et al., 2005; Kirchsetter et al., 1999; Singer et al., 1999). While NOx and CO emissions result from the combustion process, emission of VOCs can come from both evaporative emissions of fuel and the exhaust. Both evaporative and exhaust emissions are temperature dependent so these emissions vary seasonally and by region (Gentner et al., 2009).

A number of studies have shown that diesel vehicles are currently contributing a large fraction of total anthropogenic NOx (Ban-Weiss et al., 2008a); Dallman and Harley,

2010). Ban-Weiss et al. (2008b) conclude that diesel engines are the largest source of anthropogenic NOx in the United States. Dallmann and Harley (2010) estimated from fuel-based emissions calculations that on-road diesel vehicles account for 68% of the on- road NOx emissions in the US. Parrish (2006) concluded that the US Environmental

Protection Agency’s (EPA) road-way vehicles emissions model MOBILE6.2 captures the general downward trend in CO emissions over time, but that CO emissions were still over predicted by a factor of 2. The trend of increasing NOx emissions has not been captured by MOBILE6.2, leading to an overall greater CO-to-NOx ratio than actually emitted by the vehicle fleet (Parrish, 2006; Singer et al., 2009).

The US EPA released a new on-road vehicle emissions model in December 2009.

The new model MOter Vehicle Emission Simulator (MOVES) replaces the MOBILE6.2 vehicle emissions model. MOVES differs from MOBILE6.2 is several substantial ways.

An important improvement is that MOVES emission factors are derived from extensive in-use data sets rather than the heavy reliance on new-vehicle factory certification data

185 employed by MOBILE6.2. Examples of major sources of in-use data include a detailed analysis of 70,000 vehicles in Arizona’s inspection and maintenance (I/M) program, a study of PM emissions from nearly 500 gasoline cars and trucks in Kansas City,

Missouri, and a study of 400 in-use heavy duty vehicles over a range of speeds and driving conditions as well as diesel crank case emissions and extended idling emissions.

MOVES generates emission rates based on driving patterns to account for acceleration, braking, cruising and average speed whereas MOBILE6.2 emission rates were based on aggregate driving patterns and average speed only (EPA FAQ sheet, 2010).

One approach for evaluating the fidelity of vehicle emissions models is to measure in traffic tunnels (Rogak et al., 1998; Fraser et al., 1998; Kirchstetter et al.,

1999a; McGaughey et al., 2004; Stemmler et al., 2005; Grieshop et al., 2006; Dallmann et al., 2010). Tunnel measurements offer an efficient means of characterizing mobile emissions, because they are free from photochemistry and other urban emission sources as the emissions are captured within the tunnel. Emissions from individual automobiles can be obtained along with the vehicle type (Kirchstetter et al., 1999). However, tunnel studies do not capture some components of the vehicle emission process, in particular cold starts, and can be biased towards particular engine speeds and loads. Another approach is to make ambient measurements in a city during morning rush hour to establish emission ratios from roadway sources. Ambient data has the potential to capture all components of the emission process from cold starts and vehicle engine loads in commuter traffic conditions. However, the site location is critical to what parts of the emissions are observed. To observe all components of an area’s emission a site should be placed in a location that is free from local sources and receives well mixed air that is

186 representative of the city. The drawback is that ambient data also include emission from other sources such as area sources, off road engine emissions, and point sources.

Contributions from these other sources are difficult to account for, potentially confounding the source contribution from vehicles. Another potential problem, especially during the summer, is that the morning rush hour can occur during a period in the day when surface heating is beginning to cause rapid growth of the mixed layer and surface emissions are being mixed with polluted air from the residual layer, and photochemistry can alter the observed emission ratios.

In this paper we report ambient CO-to-NOx ratios from wintertime urban measurements made in the Treasure Valley region of Idaho and compare the observations to CO-to-NOx ratios predicted by MOBILE6.2 and MOVES. Treasure Valley, Idaho contains several cities, the largest being Boise in Ada County, and has experienced rapid population growth to the current population of approximately 600,000 people, and with it the attendant air pollution issues. In particular the valley experiences wintertime stagnation events that cause elevated PM2.5 concentrations. According to a 2008 emissions inventory developed for this region that used the MOBILE6.2 vehicle emission model, on-road emissions are the largest sources of CO and NOx in the valley during winter. CO and NOx emissions from area sources, point sources, non-road, and roadway sources are given in Table 1 for Ada County in December and January. Emissions of CO are dominated by just two sources, on-road emissions from vehicles (66.2%) and emissions from residential wood combustion (25%). Most wood combustion occurs during weekends and after 4 PM according to surveys used in the compilation of the inventory (Environ/ERG, 2010). Thus during the weekdays CO emissions are dominated

187 by roadway sources. For NOx, roadway sources (53.7%) and areas sources from combustion of fuels for building heating (30.7%) account for most of the emissions.

Urban area emissions of CO and NOx are thus dominated by mobile sources during winter making comparisons between vehicle emission models and ambient data tractable.

During these months in the Treasure Valley the morning rush hour occurs largely before sunrise, so emissions are not photochemically processed and occur within a relatively shallow mixed layer that is not yet increasing in height due to sensible heat flux.

Wintertime conditions in this region provide an excellent opportunity to compare ambient morning rush hour CO-to-NOx and VOC-to-CO ratios with MOBILE6.2 and MOVES emission inventories. The ambient data are also compared to the AIRPACT-3 chemical forecast model that provides hourly forecasts of various pollutants for the Pacific

Northwest (Chen et al., 2008). The forecast model is based on a 2005 vehicle emission inventory produced by MOBILE6.2.

4.2. EXPERIMENTAL

4.2.1. Site Location

Measurements were made from December 1, 2008 to January 31, 2009 in

Meridian, Idaho. Meridian is a suburban area located approximately 11 km west of the

Boise city center. The measurement site was co-located with a state air monitoring site

(St. Lukes) run by the Idaho Departmental of Environmental quality on a large vacant lot behind a hospital (latitude 43.6008 N, longitude 116.3484 W). The site was approximately 300 m north of an east-west oriented interstate highway (I-84), with a major urban arterial road (Eagle Road) to west of the site. The intersection of Eagle

Road and I-84 is the busiest intersection in Idaho. An automated traffic reporter (ATR)

188 was located close to this location on the interstate highway. Light duty vehicle (6 ft to

22.9 ft, 1.8m to 7.0 m length) traffic counts during the morning rush hour (5-9 am) range between 2,000 and 6,000 vehicles per hour, while those of heavy duty vehicles (> 7.0 m length) range between 400 and 1,000 vehicles per hour. The monitoring site is surrounded by a mix of light industrial and commercial spaces with limited residential development in the immediate area. The fetches to the west north-west and south east included mixed commercial and residential areas and were the dominant air flow directions. The site was selected to reflect regional conditions in the valley, and was located alongside the IDEQ’s Meridian Hospital site, which was situated mid-point between the high population areas of Boise in Ada County to the east and the smaller towns of Nampa and Caldwell in Canyon County to the west. Air flow through the valley is impacted by drainage flows from the surrounding mountains and induces a prevailing southeast air flow.

4.2.2. Measurements

NO, NO2, and NOy (NOxy) were measured using a two-channel chemiluminescence NO detector (Air Quality Design). NOy was measured on one channel by conversion to NO with a molybdenum oxide catalytic converter. NO2 and

NO were measured on the other channel. NO2 was converted to NO using a blue light photolytic converter. The flow rates through both channels were regulated at 1 slpm ±

2% with mass flow controllers, producing a residence time in the photolytic converter of approximately one second. Calibrations were performed every 23 hours, and zeros were performed every 120 seconds. NO calibration gas of 5.03 ppmv ± 1%, NIST traceable,

(Scott-Marrin, Inc.) was diluted in dry zero air to provide a 25 ± 5.4% ppbV NO

189 calibration level. Calibration of the NO2 converter efficiency was performed using gas phase titration of NO to NO2. A nitric acid permeation tube (KinTek, Laboratories) was used to calibrate the conversion efficiency of the molybdenum oxide catalytic converter.

The conversion efficiency of the NOy converter remained consistently between 96 and

98% throughout the campaign. CO was measured with a VUV fluorescence instrument

(Aerolaser GmbH). Standard addition calibrations were performed every 4 hours using

10.01 ppmv standard calibration gas (± 1%, NIST traceable, Scott-Marrin, Inc.). Volatile organic compounds (VOCs) were measured with a Proton Transfer Reaction Mass

Spectrometer (PTR-MS, Ionicon Analytik GmbH). The instrument was operated at 120

Townsend (Td) drift field condition (2 mbar drift pressure, 60 °C drift temperature, and a drift voltage of 535 volts). The PTR-MS was calibrated and zeroed on an automated, regular schedule using a 14 VOC component compressed gas standard (Scott-Marrin,

Inc.). The standard was diluted to 19.8 ppbv with humid zero air produced by scrubbing ambient air with a Pt catalyst (1% Pt on alumina spheres) at 260 °C. Twenty-five organic ions were monitored with dwell times ranging from 1 to 5 seconds resulting in a 1 minute measurement cycle. The CO and PTR-MS instruments sub-sampled from the main inlet line that consisted of a 1/2” PFA tube with a flow rate of 30 slpm pulling from an inlet height of approximately 10 m. The NOxy inlet responsible for conversion of NO2 and

NOy was located at a height of approximately 6 m. After conversion, the sample reaches the mass flow controller where the pressure drops to 15 torr and then flows to the detector at 1 slpm in 1/4” PFA tubes. Temperature, pressure, wind speed, wind direction, relative humidity, and precipitation were measured at a 10-m height with a using a Vaisala WXT

510 sensor.

190 4.3. Results

4.3.1. Time Series of Observations

Time series of the 5-minute averages for relative humidity, temperature, CO, NOx and benzene mixing ratios for the December to January measurement period are presented in Figure 4.1. A variety of pollution levels were observed, ranging from relatively clean air in mid-January to very polluted urban air. The variability in pollutant levels reflects air flow in the valley and the measurement site’s location at the western edge of the suburban sprawl of Boise. The average temperature was -0.6 oC with maximum and minimum temperatures of 13.8 oC and -13.1 oC, respectively. Morning fog was a frequent occurrence. The maximum CO mixing ratio was1460 ppbv and the maximum NOx mixing ratio was 546 ppbv. Aromatic VOCs, as measured by the PTR-

MS, were well correlated with CO and NOx mixing ratios, reflecting the fact that these

VOC species are dominated by vehicle emissions. The maximum benzene mixing ratio as a 5-minute average was 3.5 ppbv; this observation was not matched by corresponding increases in CO and NOx. There were several additional instances of short duration periods of elevated benzene and toluene mixing ratios suggesting non-mobile sources of

VOCs impacted the site. Table 2 summarizes the conditions and mixing ratios of CO,

NOx, and aromatic hydrocarbons measured at the site.

4.3.2. Wind Rose Plots

Wind rose plots in Figure 4.2 show the distribution of wind speeds and trace gas mixing ratios as a function of wind direction for half-hour averages. The plots reveal that elevated CO and NOx data were collected when winds were from the south (13%), south-

191 southeast (20%), and east-southeast sectors (15%). These sectors are impacted by nearby roadway emissions. About 15% of the data was collected under west-northwest wind conditions. The fetch in this direction includes busy urban arterial roads and a mix of residential and commercial development. Typically, the highest mixing ratios of CO,

NOx, and VOCs were observed from southerly quadrants, suggesting the impact of the highway and major roads to the south. While wind flow from the eastern sectors was infrequent, occurring less than 5% of the time, we note that elevated levels of NOx, and

VOCs were occasionally observed.

4.3.3. Diel Variation of CO, NOx Mixing Ratios and VMT

Figure 4.3 shows the half-hour average CO mixing ratio as a function of the time of day along with sunrise and sunset times for December 31, the midpoint of the measurement period. Also shown are the typical wintertime weekday vehicle miles travelled (VMT) for gasoline powered vehicles in Ada County. The VMT data are in one-hour time steps and were determined using all available ATR data sets available within the county (Environ/ERG, 2010). During the morning commute between 05:00 and 09:00, VMT for spark ignition vehicles increases by a factor of 9. This large change in the roadway emissions source corresponds to approximately a factor of 3 increase in the average CO mixing ratio measured at the site. Morning rush hour emissions mostly occur before sunrise within a relatively shallow mixed layer of approximately 300 m depth as measured by an aerosol lidar located on site. Average CO mixing ratios decrease after 09:30 as the VMT decreases slightly and as daytime surface heating causes boundary layer growth and the dilution of surface emissions. CO mixing ratios increase again at sunset as VMT steadily increases to a maximum of ~1 x 106 miles hr-1 and as the

192 mixed layer height decreases. CO mixing ratios decrease overnight as VMT rapidly decreases. The diel variation of CO mixing ratios is qualitatively consistent with the large changes in source strength from mobile emissions and daily growth of the boundary layer into which the emissions are mixed. Figure 4.4 shows the correspondence in the daily pattern of mixing ratios for CO, NOx, and benzene suggesting mobile emissions are significant sources for each of these species.

The hourly wintertime weekday emissions for CO and NOx for both spark ignition engine (gasoline) and compression engine (diesel) vehicles for Ada and Canyon Counties are shown in Figure 4.5. These emissions were calculated using the MOBILE6.2 for the week of February 11, 2008 and the MOVES model for the month of January, 2008.

These periods are representative of typical wintertime conditions during the field study.

The figure shows that CO emissions were dominated by gasoline vehicles with morning and evening rush hour maxima reflecting local maxima in VMT at 08:00 and 18:00.

Diesel engine vehicles account for only ~5% of VMT and do not display distinct rush hour maxima. The diesel VMT are dominated by the heavy duty diesel vehicle category

(HDDV: 8,500 to 60,000 lbs gross vehicle weight; 3,860 to 27,200 kg).

Differences between the MOVES and MOBILE6.2 emission inventories are readily apparent. For gasoline vehicles MOVES and MOBILE6.2 predict similar hourly CO emissions. For the time period between the VMT local maxima, 08:00 and 18:00, hourly

MOVES CO emissions are on average 13% larger than MOBILE6.2. For the same time period, MOVES hourly NOx emissions from gasoline vehicles are on average 45% greater than MOBILE6.2. MOVES diesel emissions between 08:00 and 18:00 are on average 2.4 times greater than MOBILE6.2. This large difference is in part due to an

193 updating of the HDDV fleet inventory from a review of local motor vehicle registration data that was not available for the MOBILE6.2 emissions estimates. The HDDV fleet age distribution was older than that used in MOBILE6.2 and VMT were doubled. The

MOVES model predicts 40% of total NOx emissions from 08:00 to 18:00 were from diesel engine vehicles. The diesel engine NOx emissions are almost entirely (99.5%) due to the HDDV vehicle class category. Interestingly, MOVES predicts NO2 emissions from HDDV are 5% of their total molar NOx emissions. At night diesel NOx emissions can contribute up to 74% of total vehicle NOx emissions. Increased NOx emissions from diesel vehicles help to bring MOVES into better agreement with observations by lowering the CO-to-NOx. MOVES emissions highlight the affect that HDDV emissions could have on regional air quality.

4.3.4. Emission Ratios Inferred From Measurements

Vehicle emission ratios were calculated from the ambient data during the morning rush hour from 05:00 to 09:00 by plotting the CO versus NOx mixing ratios and determining the slope (Parrish, 2006; Fujita et al. 1992; Luke et al, 2010). This method is unaffected by pollutants with a significant regional background concentration such as

CO, as the background only affects the intercept of the graph and not the slope. Days displaying a minimum of 200 ppbv increase in CO during the morning rush hour were selected for analysis to ensure a well-defined trend line over a reasonably large dynamic range. This selection criterion netted eight days with well-defined rush hour increases in

CO and NOx. Results of the regression analysis are shown in Table 3. The average CO- to-NOx morning rush hour molar ratio, as determined from the slope of the regression analysis from the eight different days was 4.2 ± 0.6. This value is significantly lower

194 than the measured summertime average of 7.9 ± 0.1 for U.S. cities reported by Parrish et al. (2006) for the year 2000. A more recent ratio based on summertime 2006 measurements in Houston, TX yielded a morning rush hour ratio of 5.8 ± 0.9 (Luke et al., 2010), comparable with our ratio and suggesting a continued decline in urban CO emissions and / or an increase in NOx emissions in the US. Rush hour correlations between aromatic VOCs and CO were also investigated with the results summarized in

Table 3. The average benzene-to-CO molar ratio was 1.3x10-3 ± 13%, the average

-3 toluene-to-CO ratio was 2.4x10 ± 16%, and the average C2-alkybenzene-to-CO ratio was 2.3x10-3 ± 18%. The benzene-to-CO and toluene-to-CO values compare reasonably well to those reported by Lough et al. (2005) using measurements conducted in a tunnel study.

Figure 4.6 illustrates correlations between CO and NOx and aromatic VOCs and

CO for all the morning rush hour periods together with all the data from the two-month study period. The CO versus NOx regression over the entire campaign data yielded an intercept of 110 ppbv and this value was used as the nominal background CO value in subsequent analyses. The linear regression fits determined from the selected rush hour periods were not significantly different than the regression fits through all the data and other morning rush hour periods. This similarity suggests the ambient CO-to-NOx ratio was determined primarily by roadway emissions throughout the day.

Figure 4.7 displays the rush hour ΔCO-to-NOx and selected VOC-to-ΔCO molar ratios as a function of wind direction. Ratios were determined by averaging together the individual five-minute averages for each wind sector to yield the wind sector mean values shown in the figure. Each octant of the wind compass is represented by an average ratio.

195 The 110 ppbv CO background was subtracted from the CO data to yield a ΔCO value.

Backgrounds for the other data were considered negligible; for clean air periods the VOC and NOx data was often near the instrument detection limits. There was no clear dependence of the ΔCO/NOx ratio as a function of wind direction. Most of the data were collected from the south-southeast (SSE) and west-northwest (WNW) wind sectors. The fetch to the SSE direction includes interstate highway emissions and a mixed commercial residential area and yielded an average ΔCO/NOx ratio of 4.2. The fetch to the WNW direction is impacted by emissions from residences and major urban arterial roads and displays a slightly larger ratio of 6.0. The slightly lower ratio from the southerly wind sectors may reflect the impact of the interstate highway emissions. The VOC-to-CO ratios for benzene, toluene, and C2-alkylbenzenes also did not display pronounced differences due to wind direction, but the southerly sector data that is impacted by interstate highway emissions displayed slightly lower ratios than other sectors.

The diel trend of the ΔCO/NOx ratio and selected VOC-to-ΔCO ratios are shown in Figure 4.8. Shown are half-hour averages with ±1 standard deviation range. During summer, NOx is typically quickly removed relative to CO by the reaction of NO2 with hydroxyl radical. Summer time ΔCO/NOx ratios are typically elevated during the daylight hours compared to the morning rush hour by a factor of 1.5 (Fujita et al, 1992).

The diel trend of the ΔCO/NOx ratio, illustrated in Figure 4.8, did not exhibit large changes during the day that would be characteristic of photochemical removal of NOx.

The lowest half-hour average ΔCO/NOx ratios were observed in the morning between 5 am and noon. The average of the individual half-hour ratios for ΔCO/NOx was 4.6 ± 2.3, close to the selected rush hour data emission ratio of 4.2 ± 0.6. The benzene/ΔCO ratio

196 and toluene/ΔCO, ratio data also shown in Figure 4.8, displayed no significant diel trend, though both ratios were lowest during the morning rush hour. If photochemical removal were a significant loss process then the toluene/ΔCO ratio would be expected to be much lower during the day compared to morning periods. The diel trends suggest that photochemical removal did not significantly affect the ΔCO/NOx ratio during the day and that the observed ratios reflect the mix of emission sources for these species.

The statistical distribution of the ΔCO/NOx ratio for weekday high VMT periods,

(08:00-18:00) is shown as a histogram plot in Figure 4.9, filtered for CO > 200 ppbv.

These data are obviously impacted by vehicle emissions since the CO mixing ratios are significantly larger than background values. Vehicle emissions are the only significant source of CO during the daytime hours of the weekday, and thus these filtered data provide the best comparison to the vehicle emission models. The median value of the

ΔCO/NOx ratio from these data is 4.95, slightly largely than the ratio derived from the rush hour correlation between CO and NOx (4.2). Also shown in the figure are the predicted CO-to-NOx ratios from MOBILE6.2 (13.2) and MOVES (9.1) for the same time period. The bulk of the observations, 60%, have a ΔCO/NOx ratio between 3.5 and

5.5. Approximately 97% of the observations are less than the MOVES ratio and 99% are less than the MOBILE6.2 ratio. The CO-to-NOx ratio determined from the emission inventory in Table 1 (9.5) is also much higher than observed.

The large differences between the observed ratio and those predicted by the emission models suggests that: 1) there are other sources with low CO-to-NOx emission ratios that are mixed with vehicle emissions resulting in the observed range, or 2) that the vehicle emission models are significantly over-predicting CO, as suggested by Parrish et

197 al., (2006), or under predicting NOx. Both explanations are plausible. The observed CO- to-NOx ratio did not display a significant wind direction dependence, suggesting uniformly distributed area sources that would mix with mobile emissions to yield the range of ratios observed. The most important NOx area sources are industrial, commercial, and residential combustion (Table 1). These emission sources have a CO- to-NOx ratio < 1 except for wood combustion which has a ratio of 62. Surveys conducted by the state as part of the emission inventory development process reveal that residential wood combustion primarily occurs on weekends and after 4 PM (Environ/ERG, 2010).

For this study we assumed that wood-burning activity did not significantly impact the measured weekday ratios. A simple calculation was performed to determine how large of an area combustion source would be needed to match the observed median CO-to-NOx ratio when mixed with the predicted roadway emissions from MOVES. Updating the on- road source in Table 1 with MOVES-based emissions (1353 tonnes NOx and 7962 tonnes

CO for December and January) significantly increases the relative importance of vehicle emissions relative to other categories. The resulting CO-to-NOx ratio based on Table 1 that includes MOVES and excludes residential wood combustion is 6.6, a better match to observations but still high. Increasing the emissions from industrial, commercial, and residential combustion sources (excluding wood) by approximately 2.5 times yields an overall CO-to-NOx molar ratio of 5.1, similar to observations. It is unlikely that area source emissions could be so severely under-predicted. A more plausible explanation reveals itself from inspection of the MOVES inventory. A significant component of the mobile emissions predicted by MOVES arises from off-network emissions involving engine cold starts and other emissions when the vehicle is not on the road. In MOVES

198 the off-network emissions can be separated from the running emissions. Figure 4.10 displays the hourly weekday Ada County emissions of CO and NOx along with the resulting CO-to-NOx molar ratio predicted by MOVES and the observed ambient ratio for weekdays when CO was greater than 200 ppbv. The MOVES emissions are categorized into urban road, rural road, and off-network emissions. There are separate road types within these categories, the most important distinction being the difference between interstate highway and other roads types as the highway has a large fraction of heavy duty diesel vehicle traffic. At the county level the CO emission inventory is significantly affected by off-network emissions accounting for, on average, 65% of the hourly emitted

CO and 23% of the NOx. The off-network emissions dramatically affect the CO-to-NOx ratio. During the morning rush hour the CO-to-NOx molar ratio from MOVES peaks at

11. The average ratio for the 08:00 to 18:00 time period is 8.4 ± 0.7. These ratios are significantly higher than the observed ratio shown in the lower panel of Figure 4.10. The observed ratio is in much better agreement with MOVES prediction for urban roadway emissions. Urban roads (excluding interstate highway emissions) yield an average CO- to-NOx ratio of 4.5 ± 0.3 for the morning rush hour period (05:00 to 09:00), and 4.3 ± 0.3 for the 08:00 to 18:00 time period. These values compare very well to the observations where a ratio of 4.7 ± 0.5 was measured for the morning rush hour and a ratio of 5.2 ± 0.5 for the 08:00 to 18:00 time period. The emissions ratio from the urban interstate highway category predicted by MOVES is significantly smaller, with a ratio of 2.4 ± 0.5 for the morning rush hour period and 2.5 ± 0.3 for the 08:00 to 18:00 period. Roadway emissions from the nearby highway and major urban roads are likely a dominant source as the measurements site’s fetch to the south is a low density urban area where significant

199 off-network emissions would not be expected to occur. At the county level adding together highway and urban road emissions yields a CO-to-NOx ratio of 3.8 ± 0.3 for the

08:00 to 18:00 time period, 27% lower than observed. The observed ratio may reflect a combination of road and highway emissions plus some contribution from off-network emissions. The off-network emissions shown in Figure 4.10 are for the entire county.

The areas with the highest off-network emissions will be located to the east and north of the measurement site, areas with the highest population densities. During the night,

MOVES predicts a low CO-to-NOx for interstate highway emissions due to the greater fraction of heavy duty diesel engine traffic on the roads at that time. At night the observed ratio is in much better agreement with general urban roadway emissions that with highway emissions. MOVES appears to capture the on-road emission well, however due to site location we do not believe that we are able to effectively evaluate off-network emissions.

The importance of off-network emissions for CO and NOx is striking and emphasizes the complexity in comparing ambient surface measurements to emission inventory predictions. Off-network emissions may be an explanation for previously reported poor agreement between emission models and observations if surface sites did not capture residential and business cold starts. Comparison between ambient measurements and chemical transport models is even more complex due to the inherent mismatch between the spatial and temporal averaging of models and measurements.

However comparison of species ratios, such as CO-to-NOx ratios, is reasonable since this removes the effects of dispersion that impacts comparison of concentrations. In wintertime when photochemistry is slow, comparison between measurements and models

200 may be a good way to evaluate emission inventories. We compared our observed CO-to-

NOx ratio to those predicted by a chemical forecast model that predicts hourly concentrations of wide range of species for the Pacific Northwest region of the US. The

Air Indicator Report for Public Awareness and Community Tracking or AIRPACT-3 is an air quality forecasting system that generates public daily air quality forecasts for the

Pacific Northwest. Within the AIRPACT-3 forecasting system, the chemical transport model (CTM) is the Community Multi-scale Air Quality Model (CMAQ), and the emissions are processed using the Sparse Matrix Operator Kernel Emissions (SMOKE) model version 2.1. The CMAQ CTM uses a grid 12-km cells with 21 vertical layers, and a time step of one hour. On-road mobile emissions are generated with SMOKE using emission factors from EPA MOBILE6.2 The emissions are generated at 30 oC and then adjusted to ambient temperatures for evaporative effects and cold starts (Chen et al,

2008). AIRPACT-3 mobile emissions of CO and NOx exhibit a temporal profile with clear morning and evening rush hour peaks during the weekday which are absent during weekends as shown in Figure 11. The contribution of on-road emissions to total emissions used in AIRPACT-3 is 76% for CO and 46% for NOx. These values are similar to the more recent 2008 emission inventory compiled by IDEQ shown in Table 1 where on-road CO is 66% of total CO emissions, and on-road NOx is 54% of total NOx emissions. The measured CO-to-NOx ratios were compared to model emission ratios and model output. Model data where compared to one-hour averages of the ambient data utilizing the entire measurement period. Hourly averages did not affect the value of

ΔCO/NOx. Within the AIRPACT-3 grid domain for this area, the CO-to-NOx molar emission ratio from on-road sources was 18.1 ± 0.09, and from all sources was 14.4 ±

201 0.13, as shown in Figure 4.11. Model output concentration data yielded an average ratio of 13.8 ± 0.2, similar to the total emissions ratio. The AIRPACT-3 CO-to-NOx ratios are

3 times higher than measured and the variability of the ratio much lower than measured as shown in the histograms of the ΔCO/NOx ratio in Figure 4.11c. The CO-to-NOx emission ratio for on-road emissions used in AIRPACT-3 was 4.7 times higher than the morning rush hour measurements. The nature of our measurement site location means that it would not be significantly impacted by off-network emissions and thus should display a lower CO-to-NOx ratio than the AIRPACT-3 grid as was the case. The on- road sources for AIRPACT-3 include off-network emissions but these can’t be easily separated from the compiled MOBILE 6.2 emissions inventory to make an assessment of off-network contributions. A summary of the emission ratios from MOVES,

MOBILEv6.2, AIRPACT-3 and the observations can be found in Table 4.

Implementation of the MOVES emission inventory in the model domain will significantly impact the CO-to-NOx ratios due to increases in roadway NOx emissions.

Future comparisons that use ambient data to evaluate the vehicle emissions inventory will have to be designed to quantify the contribution of off-network emissions.

4.4. CONCLUSIONS

Wintertime measurements of CO, NOx, and selected VOCs were made at the St.

Lukes monitoring site in Meridian, Idaho from December 1, 2008 to January 31, 2009.

The site was located in a light industrial commercial area close to a major highway and busy urban arterial roads and thus significantly impacted by roadway vehicle emissions.

Morning rush hour occurred during dark thus ambient measurements of these gases should reflect unprocessed vehicle emissions and be comparable to emission model

202 predictions. The measured CO-to-NOx molar ratio during the morning rush hour periods

(5-9 am) with elevated CO was 4.2 ± 0.6. The average CO-to-NOx ratio during the weekday between the hours of 08:00 and 18:00 when VMT is highest and for periods of elevated CO (> 200 ppbv) was 5.2 ± 0.5. These ratios were compared to the those predicted from the US EPA vehicle emission model MOVES. MOVES provides detailed accounting of emissions from different road types including off-network emissions from residential and business cold starts. The observed ratios were in excellent agreement with the urban roadway emission ratios predicted by MOVES for Ada county Idaho. The

MOVES model predicted a ratio of 4.5 ± 0.3 during rush hour and 4.3 ± 0.3 for the 08:00 to 18:00 time period for urban roadway emissions. MOVES predicted a highway emissions ratio of 2.4 and 2.5 for the same time periods respectively. Off-network emissions are a significant fraction of the hourly emitted CO and NOx from vehicles in the county and have dramatic impact on the CO-to-NOx ratio. Including emissions from urban and rural roads, interstate highway, and off-network emissions, MOVES predicts a

CO-to-NOx ratio of 8.4 ± 0.7 for the 08:00 to 18:00 time period, and a ratio that peaks at

11 during the morning rush hour. We have concluded that the nature of the site location on the western edge of suburban sprawl of the Boise metropolitan area and the prevailing wind direction meant that the site was not significantly impacted by off-network emissions from residential and business cold starts. The reasonable agreement between observations and MOVES roadway emission ratios suggests the on-road component of

MOVES is a significant improvement from older MOBILE6.2 vehicle emissions model.

At the county level MOBILE6.2 predicted a CO-to-NOx ratio of 11.8 compared to 8.4 for

MOVES. The lower CO-to-NOx ratio compared to MOBILE6.2 is the result of increased

203 NOx emissions. We believe that the emission inventory using MOVES more accurately reflects real emissions for the Treasure Valley region. MOVES predicts that heavy duty diesel vehicles accounted for 40% of total roadway NOx emissions in Ada County. These vehicles account for only 5% of vehicle miles travelled but clearly will have an important impact on air quality in this region.

Our study suggests that off-network vehicle emissions deserve more study as they are a significant source of roadway CO and NOx emissions and make a significant contribution to the overall emission inventory in the region. Off-network emissions may be the cause of previous discrepancies between MOBILE6.2 predictions of CO-to-NOx ratios and ambient measurements. The more detailed nature of the emissions allocation in MOVES compared to MOBILE6.2 will allow for more refined assessments through comparison to observations. In this study wind rose plots revealed only small variations in the CO-to-NOx ratio as a function of wind direction indicating a uniform emission source and no clear evidence of the impact of off-network emissions from wind sectors with large residential areas. The CO-to-NOx ratio also did not display a significant time of day dependence further suggesting more uniformity in the hourly emissions ratio that predicted by MOVES. Quantifying real world off-network emissions presents itself as the next challenge in verifying vehicle emission inventories.

204 4.5. REFERENCES

Ban-Weiss, G.A., McLaughlin, J.P., Harley, R.A., Lunden, M.M., Kirchstetter, T.W.,

Kean, A.J., Strawa, A.W., Stevenson, E.D., Kendall, G.R. “Long-term changes in emissions of nitrogen oxides and particulate matter from on-road gasoline and diesel vehicles.” Atmospheric Environment 42, (2008a) 220-232.

Ban-Weiss, G.A., McLaughlin, J.P., Harley, R.A., Kean, A.J., Grosjean, E., Grosjean, D.

“Carbonyl and nitrogen dioxide emissions from gasoline- and diesel-powered motor vehicles.” Environmental Science and Technology 42, (2008b) 3944-3950.

CRC. “Review of the 2009 Draft Motor Vehicle Emissions Simulator (MOVES) model,

Final Report CRC Project E-68a.” (2010).

Dallmann, T.R., Harley, R.A. “Evaluation of MOBILE6.2 source emission trends in the

United States.” Journal of Geophysical Research 115, (2010) D14305.

Environ/ERG. 2008, 2015, and 2023 Emissions Inventories for the Treasure Valley

Airshed, Final Report, August 2010.: ENVIRON International Corporation, 773 San

Marin Drive, Suite 2115, Novato, CA 94998, and Eastern Research Group, 10860 Gold

Center Drive, Suite 275, Rancho Cordova, CA, 95670. (2010)

205 Fraser, M.P., Cass, G.R., Simoneit, B.R.T. “Gas-phase and particle-phase organic compounds emitted from motor vehicle traffic in a Los Angeles roadway tunnel.”

Environmental Science and Technology 32, (1998) 2051-2060.

Fujita, E.M., Croes, B.,E., Bennett, C.L., Lawson, D.R., Lurmann, F.W., Main, H.H.

“Comparison of Emission Inventory and Ambient Concentration ratios of CO, NMOG, and NOx in California’s South Coast Air Basin.” Journal of Air and Waste Management

42, (1992) 264-276.

Gentner, D.R., Harley R.A., Miller, A.M., Glodstein, A.H. “Diurnal and Seasonal

Variability of Gasoline-Related Volatile Organic Compound Emissions in Riverside

California.” Environmental Science and Technology 43, (2009) 4247-4252.

Grieshop, A.P., Lipsky, E.M., Pekney, N.J., Takahama, S., Robinson, A.L. “Fine particle emission factors from vehicles in a highway tunnel: Effects of fleet composition and season.” Atmospheric Environment 40, (2006) S287-S298.

Grivicke, R. “Wintertime Air Quality and Boundary Layer Measurements in the Boise,

ID Treasure Valley” (2011) in progress.

Harley, R.A., Coulter-Burke, S.C., Yeung, T.S. “Relating Liquid Fuel and Headspace

Vapor Composition for California Reformulated Gasoline Samples Containing Ethanol.”

Environmental Science and Technology 34, (2000) 4088-4094.

206 Harley, R.A., Kirchsetetter, T.W., Singer, B.C., Kendall, G.R., Chan, W. “Impact of

Oxygenated Gasoline Use on California Light-Duty Vehicle Emissions.” Environmental

Science and Technology 30, (1996) 661-670.

Kirchstetter, T.W., Singer, B.C., Harley, R.A., Kendall, G.R., Hesson, J.M. “Impact of

California Reformulated Gasoline on Motor Vehicle Emissions. 2. Volatile Organic

Compound Speciation and Reactivity.” Environmental Science and Technology 33,

(1999a) 329-336.

Kirchstetter, T.W., Singer, B.C., Harley, R.A., Kendall, G.R., Hesson, J.M. “Impact of

California Reformulated Gasoline on Motor Vehicle Emissions. 1. Mass Emission

Rates.” Environmental Science and Technology 33, (1999b) 329-336.

Kleffmann, J. Becker, K.H., Weisen, P. “Heterogeneous NO2 conversion process on acid surfaces: Possible Atmospheric Implication.” Atmospheric Environment 32, (1998)

2721-2729.

Lough, G.C., Schauer, J.J., Lonneman, W.A., Allen, M.K. “Summer and Winter

Nonmethane Hydrocarbon Emissions from On-Road Motor Vehicles in the Midwestern

United States.” Journal of Air and Waste Management Association 55, (2005) 629-646.

Luke, W.T., Kelly, P., Lefer, B.L., Flynn, J., Rappengluck, B., Luechner, M., Dibb, J.E.,

Ziemba, L.D., Anderson, C.H., Buhr, M. “Measurements of primary trace gases and

207 NO(y) composition in Houston, Texas.” Atmospheric Environment 44, (2010) 4068-

4080.

McGaughe, G.R., Desai, D.R., Allen, D.T. Selia, R.L., Lonnemann, W.A., Fraser, M.P.,

Harley, R.A., Pollack, A.K., Ivy, J.M., Price, J.H. “Analysis of motor vehicle emissions in a Houston tunnel during the Texas Air Quality Study 2000: Findings from EPA’s

Particulate Matter Supersites Program.” Atmospheric Environment 38, (2004) 3363-3372.

Millstein, D.E., Harley, R.A. “Effects of Retrofitting Emission Control Systems on In-

Use Heavy Diesel Vehicles.” Environmental Science and Technology 44, (2010) 5042-

5048.

Mwaniki, G.R., Pressley, S.N, Rosenkrance, C., Wallace, H.W., Erickson, M.H.,

Grivicke, R., McCoskey, J.K., Jobson, B.T., Lamb, B.K., VanReken, T.M., Submitted

2011. “Wintertime particulate nitrate formation in Idaho’s Treasure Valley.” Atmospheric

Environment.

National Research Council, 2004. Air Quality Managemant in the United States, National

Research Counsil, Washington D.C.

Parrish, D.D. “Critical evaluation of US on-road vehicle emission inventories.”

Atmospheric Environment 40, (2006) 2288-2300.

208 Rogak, S.N., Pott, U., Dann, T., Wang, D. “Gaseous emissions from vehicles in a traffic tunnel in Vancouver, British Columbia.” Journal of the Air and Waste Management

Association 48, (1998) 604-615.

Rubin, J.I., Kean, A.J., Harley, R.A., Millet, D.B., Goldstein, A.H. “Temperature dependence of volatile organic compound evaporative emissions from motor vehicles.”

Journal of Geophysical Research 111, (2006) D03305.

Silva, P.J., Vawdrey, E.L., Corbett, M., Erup, M. “Fine particle concentrations and composition during wintertime inversions in Logan, Utah, USA.” Atmospheric

Environment 41, (2007) 5410-5422.

Singer B.C., Kirchstetter, T.W., Harley, R.A. Kendall, G.R., Hesson, J.M., “A Fuel Based

Approach to Estimating Motor Vehicle Cold-Start Emissions.” Journal of Air and Waste

Management Association 49, (1999) 125-135.

Stemmler, K., Bugmann, S., Buchmann, B., Reimann, S., Staehlin, J. “Large decrease of

VOC emission of Switzerland’s car fleet during the past decade: Results from a highway tunnel study.” Atmospheric Environment 39, (2005) 1009-1018.

209

4.6. CHAPTER 4 TABLES

Table 4.1. Summary of Ada County December and January Emissions Inventory. Source category NOx CO % NOx % CO CO / NOx tonnes tonnes emissions emissions molar ratio Area source 428 3396 30.7 27.4 8.5 Industrial combustion 51 37 Commercial combustion 105 78 Residential combustion 215 90 Wood burning 57 3120 Point Source 56 28 4 0.2 0.5 Non Road 162 772 11.6 6.2 5.1 On Road 736 8085 53.7 66.2 11.8 (MOBILE6.2) Total 1394 12408 100 100 9.5

Table 4.2. Statistical summary of observations during the 2-month study period. Measurement Mean Median Max Min Units NO 13.0 3.0 402 0.020 ppbv NO2 11.6 9.2 144 0.035 ppbv NOy 31.5 18.7 520 0.030 ppbv CO 222 168 1460 78 ppbv benzene 0.27 0.19 3.53 0.1 ppbv toluene 0.41 0.24 5.55 0.1 ppbv C2-alkylbenzenes 0.35 0.20 5.03 0.1 ppbv temperature -0.6 -1.1 13.8 -13.1 °C relative humidity 76 79 99 30 %

210

Table 4.3. Slopes from rush hour linear regression analysis of measured pollutant ratios.

C2- Date CO/NOx benzene/CO toluene/CO AKB/CO (mol/mol) ± (mmol/mol) ± (mmol/mol) ± (mmol/mol) ± 12/7/2008 4.2 0.28 1.2 0.061 2.1 0.11 2.1 0.031 12/9/2008 4.7 0.22 1.1 0.082 2.1 0.0073 1.9 0.058 12/10/2008 5 0.37 NA NA NA NA NA NA 12/12/2008 3.6 0.19 1.4 0.059 2.2 0.11 2.1 0.096 1/20/2009 3.2 0.16 1.2 0.046 2.3 0.059 2.2 0.081 1/23/2009 4.2 0.39 1.6 0.10 3.3 0.20 3.2 0.41 1/27/2009 4.5 0.33 1.4 0.074 2.4 0.16 2.4 0.014 1/29/2009 3.8 0.15 1.4 0.053 2.5 0.081 2.2 0.079 Average 4.2 0.59 1.3 0.017 2.4 0.040 2.3 0.42 All Rush Hour 4.1 0.33 1.4 0.0052 2.6 0.0043 2.2 0.026 All DATA 4.6 0.02 1.5 4.0e-3 2.8 8.8e-3 2.4 0.0078

Table 4.4. Vehicle emissions and CO/NOx ratios for MOVES and MOBILEv6.2 compared to observations. CO NOX Molar MOVES WEEKDAY1 tonnes/day tonnes/day CO/NOx Off-Network Total 120.4 4.9 26.6 Urban Interstate Total 13.8 6.2 2.5 Urban Roads Total 40.1 9.7 4.6 Rural Interstate Total 2.1 1.3 1.8 Rural Roads Total 7.7 2.1 4.1 TOTAL 184.0 24.0 8.4 MOBILEv6.2 Winter months (NDJF) 133 12.1 11.7 AIRPACT-3 Mobile Emissions 18.1 Total Emissions 14.4 OBSERVED rush hour correlation (05:00-09:00) 4.2 day time average, CO>200 ppbv (08:00-18:00) 5.2 All data correlation 4.6 1 MOVES and MOBILE emissions for Ada County, ID.

211 4.7. CHAPTER 4 FIGURES

Figure 4.1. Time series of relative humidity (blue), temperature (red), CO, NOx, and benzene mixing ratios for the study period.

212

Figure 4.2. Plots of half hour averaged mixing ratios as a function of wind direction together with wind rose plot showing wind speed and fractional occurrence of wind flow direction in 30° increment bins. Approximately 20% of the observations were from the bin centered on the 150° direction.

213

Figure 4.3. Diel profile of half hour averages of weekday VMT and CO (black circles).

Grey shading from 08:15 to 15:15 indicates average length of daylight period between sunrise and sunset for the measurement campaign.

214

Figure 4.4. Diel profile of half hour averaged mixing ratios for CO (black circles), NOx

(grey squares), and benzene (open triangles).

215

Figure 4.5. MOBILE6.2 and MOVES hourly weekday emissions for Ada and Canyon counties. Gasoline vehicle emissions (open circles) and diesel engine vehicle (solid squares) emissions are displayed. MOBILE6.2 emissions are given as black symbols

MOVES as blue symbols. Lower panel displays weekday VMT for gasoline engines vehicles (open circles) and diesel engine vehicles (solid squares). The gasoline vehicle

VMT are approximately 20 times larger than diesel VMT.

216

Figure 4.6. Plots of linear regressions of ambient measurements for the experiment. The emission ratio is the slope of the regression. The grey circles are data points from the entire campaign, the black circles are from the morning rush hour (5-9 am). Linear regressions are displayed as dashed line through grey circles and solid black line through rush hour data.

217

Figure 4.7. Mean rush hour emission ratios by wind direction. Wind direction has been divided into octants and the dashed lines ± 1 standard deviation from the mean. The angular unit axis is in degrees and the radial axis is mole/mole. The plots show little directional dependence in the molar ratios, although ΔCO/NOx is slightly lower on average when winds are from the south, where air from the freeway is being sampled.

218

Figure 4.8. Diel trend of molar ratios showing ½ hour averages as solid black line.

Shading indicates 1 standard deviation. There is very little change throughout the day in the CO-to-NOx ratio indicating that there is little photochemical activity in the day.

219

Figure 4.9. Histogram of ΔCO/NOx molar ratio for the time periods from 08:00 to 18:00 with CO > 200 ppbv. The emission ratios from the MOVES (9.1) and MOBILEv6.2

(13.2) models along are shown as the blue and red vertical lines respectively. The black vertical line shows the observed ratio from the morning rush hour periods (4.2).

220

Figure 4.10. Hourly weekday MOVES emissions for CO (top) and NOx (middle) and the CO-to-NOx molar ratio compared to observations (bottom) where red diamonds are observations, blue circles are the urban road ratio, black circles are the urban interstate ratio, and grey squares are the total ratio that includes off-network emissions. The urban on-road MOVES emissions agree well with the observations. Inclusion of off-network emissions significantly increases ratio compared to observations.

221

Figure 4.11. Panel A) 9-day time series of the hourly total emissions (black) and mobile emissions (red) of NOx for the 12 km x 12 km grid cell of the AIRPACT-3 model containing the measurement site. B) 9-day time series of the total and vehicle emissions of CO for the same grid cell. Total CO emissions (black) used by AIRPACT-3 and mobile emissions (red). C) Log histogram of the CO-to-NOx molar ratio for the observations (green) and for AIRPACT-3 output (black). D) CO versus NOx correlations showing AIRPACT-3 vehicle emissions (blue line) ratio of 18.1±0.09 mol/mol,

AIRPACT-3 total emissions ratio (black) with a ratio of 14.4±0.13, MOVES emission ratio 8.9, and observed hourly averaged ratio for the entire study period (green) of

4.6±0.16. The grey circles are the hourly AIRPACT-3 output for the month of January

2009 at the grid cell containing the site.

222 CARES FIELD EXPERIMENT

5.1. INTRODUCTION

The Carbonaceous Aerosols and Radiative Effects Study (CARES) was a project directed by the United States Department of Energy (DOE) Atmospheric Radiation

Measurement (ARM) program, which ran from June 2-28 2010. The objective of this study was to measure aerosol formation and radiative properties in California’s Central

Valley when anthropogenic pollution interacts with biogenic or natural sources. The mix of anthropogenic pollution with biogenic volatile organic compounds (VOCs) has been shown to produce higher than expected aerosol formation (Volkamer et al., 2006; de

Gouw et al., 2005). This is important because this region experiences violations of the

National Ambient Air Quality Standards (NAAQS) for particulate matter (PM) and ozone. Both O3 and PM have negative health effects and can damage ecosystems and crops. The link between respiratory and cardiovascular illnesses in humans and aerosols has been well established (e.g. Thurston et al. 1994; Schwartz et al. 1996; Harrison and

Yin, 2000; Pope 2000; Davidson et al., 2005; Pope and Dockery, 2006; Valavanidis et al.

2008). Aerosols are also important for understanding the radiative forcing of climate through their direct and indirect effects. The direct effect is the scattering and absorption of long and shortwave radiation while the two indirect effects are the ability of aerosols to change cloud albedos and cloud lifetimes.

Aerosols are broken into two source categories: primary and secondary. Primary aerosols are directly emitted to the atmosphere from fossil fuel combustion, ocean spray, volcanoes, wind driven process, biological processes and biomass burning. Secondary aerosols are created through chemical processing in the atmosphere. Organic aerosol

223 (OA) is a major component of submicron particulate matter (PM) accounting for 20-90% of aerosol mass and much of this mass is thought to be secondary (Kanakidou et al.,

2005; Murphy et al., 2006; Zhang et al., 2007; Jimenez et al., 2009). Secondary aerosols are formed through condensation and chemical processing, either homogenous or heterogeneous (Seinfeld and Pandis, 2006; Hallquist et al., 2009, Jimenez et al., 2009).

The composition of atmospheric aerosols is highly complex; including mixtures of organic and inorganic material that can exist in the liquid phase, solid phase or a combination of the two (Zaveri et al., 2012; Seinfeld and Pandis, 1998; Hallquist, et al.,

2009). Studies have shown that models based on chamber experiments using Raoult’s law calculations grossly under-predict aerosol concentrations in urban regions where both biogenic and anthropogenic precursors are present (Volkamer et al., 2006, de Gouw et al.,

2005.) Efforts to improve modeling of SOA formation, by adding the missing precursors intermediate volatility organic compounds (IVOCs) and semi-volatile organic compounds, which are emitted by vehicles, have improved agreement with observations

(Donahue et al., 2006, Robinson et al., 2007). However, there continues to be evidence that the atmospheric community still lacks a fundamental understanding of PM formation and evolution. Numerous studies have demonstrated that oxidation of carbonaceous compounds leads to secondary aerosol formation (Kroll et al., 2006; Donahue et al.,

2012; Getner et al., 2012; Jimenez et al., 2009), but the vast number of VOCs and potential SOA precursors complicate the problem of understanding SOA formation.

Understanding VOC composition and oxidation is critical to solving the discrepancies between SOA models and prediction, which necessitates real world observations of pollution. The atmospheric community is unable to precisely model the

224 concentration, optical and microphysical properties of aerosols in the atmosphere. The lack of precise models makes the radiative forcing from aerosols the largest source of error in current global climate models (Forster et al., 2007). The potential for SOA to affect climate and human health, combined with our inability to accurately predict aerosols make this topic an important area of research.

The CARES field campaign involved careful observation of the time evolution of aerosol chemical and physical properties in California’s Central Valley during transport from urban Sacramento, CA to the surrounding foothills of the Sierra Nevada range and during return flow back in to the valley at night (Zaveri et al., 2012). Two heavily instrumented sites were located at the American River College in Sacramento, CA (T0) and in Cool, CA (T1) which is located in the foothills of the Sierra Nevada mountain range. The T0 site was chosen as a source site for anthropogenic emissions (Zaveri et al., 2012) and the Washington State University (WSU) Mobile Atmospheric Chemistry

Laboratory (MACL) was deployed at this site. This work aims to characterize the trace gas composition and photochemistry at the T0 site. Specifically, we calculated emission ratios to characterize mobile source emissions and identified a metric for determining the extent of photochemical activity. To accomplish this, we present observations of CO,

NO, NO2, NOy, O3, and VOCs.

One of the fundamental problems with understanding and predicting the evolution of pollution in the atmosphere is that emission rates of pollution, especially mobile sources are difficult to estimate. One important tool to understand sources of pollution is the use of emission ratios. Emission ratios are useful because they allow us to relate trace gases with poorly cataloged and poorly understood emissions to trace gases that are better

225 understood. For trace gases with no significant background mixing ratio, emission ratios are invariant under dilution, meaning that the source signature remains unchanged if the trace gases are not being reacted away. Emission ratios also allow for direct comparison to emissions inventories without intensive modeling efforts and they provide an important historical record to track ever changing emissions. Measurements of VOCs,

CO, and NOx made near mobile source emissions can under specific conditions be used to report emissions ratios from mobile sources. The data set collected allowed us to observe and report emission ratios made for CO, NOx, and VOCs. We report morning rush hour emission ratios for CO-to-NOx and VOC-to-CO following the methods described by Fujita, et al., 1992; Parrish 2006; Luke et al., 2010; and Wallace et al.,

2012,. The emission ratios reported in this work are the best estimate of the on-road mobile source emissions for the Sacramento, CA area. There are surprisingly few observations of emission ratios of VOC-to-CO from mobile on-road sources in the literature which is unfortunate because emissions from mobile sources change over time

(Kirchstetter et al., 1999; Parrish 2006, Dallmann and Harley, 2010).

Understanding the extent of photochemistry at the T0 site is important because photochemical oxidation of VOCs in the presence of NOx results in production of O3 and

SOA. The major atmospheric oxidant is the hydroxyl radical, which occurs in very low abundances, from 2 to 20 x 106 molecules cm-3, and is notoriously difficult to measure.

Because there are only a handful of groups in the world with the capability to measure this important radical, it is important to develop methods to examine the extent of photochemical activity in other ways. This is a challenging task in an environment with continuous emissions of VOCs and NOx. This work examines variability of the VOC

226 mixing ratios observed at the T0 site. The findings indicate that variability of VOCs was driven by emissions and dilution rather than chemical loss. VOC mixing ratios did not exhibit evidence of photochemical processing. In locations where the mixing ratios of a trace gas are driven by emission and dilution, identifying a photoproduct is necessary to characterize the photochemistry. Significant photochemical activity was observed at the

T0 site, however little evidence for photochemical activity at the T0 site was found in the

VOC mixing ratios. The best evidence for photochemical activity was best observed by focusing analysis on the few photoproducts available in the dataset. NOz (NOy -NOx) was found to be the most reliable photoproduct showing correlation with both organic aerosol (OA) and O3. Based on the observations made at the T0 site p,m-xylene, toluene and the trimethylbenzene isomers are suggested for further investigation using a flow through photo-chamber to determine a quantifiable photoproduct that can be used to determine airmass age. Observations of VOCs and determination of photochemical processing are important understanding the sources of SOA and the relationship between

SOA and photochemical aging in real atmospheres.

This work describes the gas phase measurements of CO, NOx and VOCs made at the T0 site of the U.S. DOE’s CARES field intensive experiment. These measurements contribute to the CARES science objectives by characterizing the gas phase SOA precursors at the T0 site. Morning rush hour emission ratios of CO-to-NOx and VOC-to-

CO which represent our best measure of the signature of mobile on-road sources are reported. The T0 site is shown to be VOC limited with NO2 the largest, daytime hydroxyl radical sink. As a result, the NOz-to-NOy ratio is the best metric for atmospheric oxidation in this data set.

227 5.2. EXPERIMENTAL

5.2.1. T0 Site

Atmospheric measurements presented in this work were made on the campus of the American River College in Sacramento, CA. The WSU MACL was deployed just north of the main parking lot from June 2-28, 2009. The parking lot was not a major source of pollution during measurements because classes were out and the parking lot was sparsely used. The T0 site has been described extensively by Zaveri et al (2012).

The site is located approximately 14 km northeast of the urban core in the suburban sprawl. To the west of the site lies I-80 (2.5 km), and surrounding the site are busy arterials and residential/commercial areas. The dominant source of pollution (CO, NOx, and VOCs) was from mobile on-road emissions. There were large oak trees 50 m to the north and west of the MACL.

5.2.2. Instrumentation

All of the analytical instrumentation was housed in the environmentally controlled

MACL. NO, NO2, and NOy (NO + NO2 + HONO + PAN + particulate nitrate +…) were measured with a high sensitivity chemiluminescence NO monitor (Air Quality Design,

Inc., Golden, CO) which has been fully described in Chapter 2.1. The NOxy inlet which houses the photolytic Blue Light Converter (BLC) (Air Quality Design, Inc., Golden,

CO) and the molybdenum oxide catalytic NOy converter (Thermo Scientific) was mounted at a height of 6 m. Calibrations of NO and NO2 were performed (100 ppbv level) automatically every 23 hours. Pre-reactor zeros were performed every 120 s and zero air zeros were performed with calibrations. No difference between the pre-reactor

228 zeros and the zero air zeros was observed. NO2 artifact was measured every 4 hours and found to be smaller than 1% of the ambient NO2, so no corrections were made. NO,

NO2, and NOy data are reported every 120 s. The detection limit of NO, NO2, and NOy defined by 3 standard deviations above the zero was 10, 20, 20 pptv respectively. 24

VOCs ranging from C6 to C12 were quantified during this field experiment. VOCs reported in this work were monitored with a GC-ITMS (Varian, Inc. 4000-MS) with a custom VOC pre-concentration system, fully described in chapter 2.2. The system was operated in external ionization mode with a target ion count of 20,000 and a filament current of 25 µA. Calibrations were performed manually every two to three days. Zeros were performed before and after calibrations. Data points represent 600 s samples and the timestamp reported is at the end of the 10 minute sample collection period. Detection limits are compound specific ranging from 1 to 50 pptv. VOC, CO and O3 were sampled from the main inlet line which consisted of 12.7 mm OD (1/2”) PFA Teflon tube mounted at a height of 9-m on the MACL crank up tower. Air flow through the tubing was 30 SLPM. O3 was monitored with a model 3000 UV absorption O3 monitor (Dasibi

Environmental Corp, Glendale, CA) and CO measured with a model 5002 vacuum ultra violet resonance florescence (Aero-Laser, GmbH., Garmisch-Partenkirchen, Germany).

The operating parameters for the CO instrument follow: cell pressure 7.35 torr, cell temperature 24 C, monochrometer pressure 3.6 bar, monochrometer temperature 24 C, monochrometer N2 flow 35 sccm, lamp gas flow 35 sccm, lamp gas temperature 36 C.

Standard addition calibrations and zeros were performed every 4 h using 10.01 ppmv CO standard (1%, NIST traceable, Scott-Marrin, Inc.). The O3 and CO data reported are 5 minute averages. Organic aerosol (OA) was measured by Pacific Northwest National

229 Laboratory personnel using a High Resolution Time-of-Flight Aerosol Mass

Spectrometer (HR-ToF-AMS) (Aerodyne Research Inc., Boston, MA) which is described by Fierz et al. (2007).

5.3. RESULTS

5.3.1. Meteorology

During the summer months the Sacramento urban plume acts a flow reactor as thermally driven upslope winds carry pollutants from the urban core northeast into the

Sierra Nevada foot hills (Bao et al., 2009; Zaveri et al. 2012). Dillon et al. (2002) applied a Lagrangian analysis of pollution transported from the urban centers in California’s

Central Valley to the biogenic volatile organic compound (BVOC) rich foothills of the

Sierra Nevada Range. At night the flow switches from upslope to downslope (Van Ooy and Caroll, 1995 Bao et al., 2009) and the mix of aged pollutants, aerosols, and BVOCs can flow back into the valley creating a natural reactor (Zaveri, 2012). To examine this hypothesis, two heavily instrumented ground sites were placed in the both the urban core of Sacramento, CA and 40 km northeast in the foothills of Cool, CA. Fast et al. (2012) have described the meteorological conditions observed during the CARES campaign.

The planetary boundary layer height affects dilution of pollutants by limiting the volume into which pollutants are mixed. To measure the PBL height during the campaign, radiosondes from both sites were launched 5 times per day at (08:00, 11:00, 13:00, 15:00, and 17:00 Pacific Daylight Savings Time) when the DOE G-1 aircraft was flying and once a day (13:00) when it was not. The PBL generally started at a height of 100 to 200 m at 06:00 and steadily increased to the daily maximum of 0.5 to 2.5 km which occurred around 16:00.

230 5.3.2. Time Series

nd Time series of the 5-min averages for CO, NO, NO2, and NOy from June 2 to

June 28Th 2010 CARES measurement campaign are presented in Figure 5.1. The average

CO mixing ratio was 152 ppbv with maximum and minimum mixing ratios of 846 and 88 ppbv, respectively. NOy was observed to have an average mixing ratio of 8.4 ppbv with maximum and minimums of 80 ppbv and 0.16 ppbv respectively. CO, NOx, and NOy exhibit covariance that is indicative of a common source, which will be discussed in more detail in the section on emission ratios. Variability in the CO, NOx, and NOy data can be attributed to variations in source strength and the evolution of the PBL height discussed in the meteorology section. This pattern of emissions and dilution is typical for summertime measurements made in urban environments (Fujita et al., 1992; Stephens et al., 2008; Luke et al., 2010; Lefer et al., 2010). The highest median mixing ratios for CO and NOx, which are primary pollutants, occurred in the early morning (08:00 PDST) and again during the evening (21:00) which reflects the effects of urban emissions during morning and evening rush hours coupled with minimal morning mixing.

A time series of selected VOCs, isoprene, toluene, benzene, 1,2,4- trimethylbenzene, and p,m-xylene are shown in Figure 5.2. The mixing ratios of VOCs were generally low (<2 ppbv) and a summary of the trace gases is presented in Table 1.

The GC-ITMS data were not collected prior to 6/10/2010 due to instrumentation failure, and there was a gap in data from 6/12/2010 20:00 until 6/13/2010 19:00 due to a heater malfunction on the temperature controlled inlet line. Many of the VOCs exhibited covariance with CO and NOx all of which are known tracers of automobile exhaust

(Lough et al., 2005; Fujita et al., 2011; Getner et al., 2012; Wallace et al., 2012). The

231 time series of the photochemical products O3, NOz and organic aerosol are displayed in

Figure 5.3.

5.4. OH REACTIVITY

This section summarizes the likely fate of an OH radical by calculating OH loss frequencies for the VOCs and NOx. The OH loss frequency of VOCs can be used to determine the potential of an air mass to form SOA and O3. The rate constants at 298 K along with the lifetime (assuming [OH] = 6x106) of the anthropogenic compounds measured during CARES are presented in Table 5.1. At the specified hydroxyl radical concentration, 1,3,5-trimethylbenzene has the shortest lifetime at 0.8 hrs while benzene has the longest lifetime at 33.3 hrs. Rate constants were calculated from Arrhenius constants where available and if not, the recommended rate at or near 298 K was used

(Calvert et al., 2002). To calculate the rate constants for NO and NO2, both of which are termolecular reactions in the falloff region at 298 K and 1 atm, the following equation from Troe (1979) and Troe ( 1983) was used.

푘0[푀] 퐴 푘 = 푘 [푀] ∗ 퐹푐 5.1 1+( 0 ) 푘∞

Where,

푘0[푀] 2 −1 퐴 = (1 + (log10( )) ) 5.2 푘∞

The low and high pressure limits were calculated using values found in Finlayson-Pitts and Pitts (1999). A diel plot of the OH loss frequency for all of the VOCs measured during CARES is shown in Figure 5.4 on both a linear and log scale. The log scale makes the miniscule contributions from benzene and C4- alkylbenzenes visible. This plot shows that NO2 is the largest OH sink with the largest VOC sink due to isoprene,

232 followed by toluene, C2-alkylbenzenes, and C3-alkylbenzenes. Monoterpenes were observed infrequently at the site and the median value was below detection limit, so they have been omitted from the analysis. This helps to explain why NOz/NOy ratios are good photochemical tracers in this environment. OH is most likely to react with NO2 at the T0 site converting it to NOz. The reaction between OH and NO2 is likely the major radical chain termination at this location. Photolysis rates of HNO3 at the surface are low so the likely fate of HNO3 is loss through deposition. OH + NO2 during the day is 6 times more likely to occur than OH + VOCs. Not all VOCs were measured by the GC-ITMS. The

C2-C6 light hydrocarbons were missing but the photochemistry at the T0 site is still likely

VOC limited as there needs to be more VOC reactivity than NOx reactivity.

The failure of the light hydrocarbon channel left out the C2 to C6 alkenes, alkanes and aldehydes, which are important O3 precursors but not SOA precursors. Aromatic

VOCs have been shown to make up the SOA precursors in gasoline (Odum et al., 1997a).

The importance of a SOA precursor depends on the loss rate and the SOA yield which are both high for aromatic compounds. Odum et al. (1997b) showed that reformulated gasoline with lower aromatic content reduced the aerosol formation. This means that the

T0 plume is ripe with NOx and when transported to the surrounding forested regions and mixed with BVOCs could potentially result in rapid production of particles and O3.

5.5. EMISSION RATIOS

In the absence of a high degree of certainty about emissions inventories, emissions ratios are a useful means to extract information about emissions and sources of pollutants.

They can be examined for directional, temporal variability and can be compared to emissions inventories. Ratios of trace gases can be calculated using ambient data and so

233 they can be collected during a variety of experiments that are not specifically designed to study emissions, making them a versatile tool for evaluating emissions models. In particular, on-road emissions are an important source of pollution, contributing a significant portion of the NOx, CO and VOCs emitted by cities (Lough et al., 2005;

Parrish et al., 2006; Wallace et al., 2012). Due to pollution controls such as oxygenated fuels, improvements in catalytic converters, fleet turnover and mandatory emissions testing programs, emissions have fallen despite increased vehicle miles traveled (Parrish,

2006; Dallmann and Harley, 2010). Thus, emissions inventories are a moving target and should be regularly evaluated. Examination of emission ratios of ambient data has both advantages and disadvantages when compared to more specific emissions studies such as tunnel studies, on-road monitoring, garage studies or examination of the wealth of emissions testing data. One advantage of examining ambient emissions ratios is that they have the potential to capture the entire emissions of a city, including starting emissions and point sources. However, understanding site location is critical to the interpretation of the ambient data and the sources. The advantage of more specific emissions studies is that they yield high fidelity, specific information about emissions. For example, tunnel measurements guarantee that emissions are from vehicles and give vehicle-specific emission ratios and emission factors. However, tunnel studies often have fixed speeds and grades resulting in emissions from “warm” vehicles with fixed engine loads, and thus are not always representative of emission from actual on-road vehicles and miss starting emissions.

This section presents emission ratios of CO-to-NOx and VOCs-to-CO that are derived from ambient observations collected during the morning rush hour (06:00-10:00).

234 Extraction of emission ratios from morning rush hour gives the best measurements of mobile sources for a number of reasons: there is low photochemical activity, the mobile source increases significantly as vehicle miles traveled (VMT) increase significantly, and the PBL height is low leading to high concentrations (Fujita et al., 1992; Parrish, 2006;

Wallace et al., 2012).

5.5.1. CO to NOx

Fujita et al. (1992) first used morning rush hour CO-to-NOx to evaluate emissions inventories in California’s south coast air basin. Parrish’s (2006) summertime analysis of

CO-to-NOx ratios shows that CO-to-NOx ratios have steadily decreased since 1996 and reported a factor of 2 overestimation of CO in the National Emissions Inventory (NEI).

CO-to-NOx ratios are sensitive to a number of factors, such as temperature, vehicle maintenance, fleet composition and vehicle age, fuel composition, and prevalence of diesel vehicles which are copious emitters of NOx. This work adds another data point to a growing collection of reported CO-to-NOx observations. We report a morning time rush hour (06:00-10:00) CO-to-NOx of 6.9 ± 0.15 (Figure 5.5). This value was

2 determined by a linear regression of the CO versus NOx scatter plot and the R for this is

0.68 indicating that the site has received air from numerous sources and is representative of the area (Parrish, 2006). The linear regression has a positive y-intercept indicating the regional background value for CO. Figure 5.6 shows a log histogram of ΔCO/NOx that has a mode centered at 5.9. The ΔCO value is determined by subtracting the background

CO concentration from the CO data. NOx is not background corrected because no significant NOx background was observed. The two techniques yield different emission ratios, indicating that some bias has been introduced in subtracting the CO background or

235 the un-weighted linear regression is biased by high values. While the two techniques are not in perfect agreement, they are close and both values are in reasonable agreement with the linear extrapolation of the Los Angeles CO-to-NOx historical record ~(6.5) present in

Parrish (2006). The agreement with the extrapolated results of Parrish (2006) indicates that the site receives CO and NOx that have the emission signature of mobile, on-road emissions. This CO-to-NOx (6.9) is higher than the measurements of Luke et al., (2009)

(5.9) and Wallace et al., (2012) (4.6), although the center of the histogram (5.9) is in better agreement. Measurements made by Wallace et al., (2012) were located within 300 m from busy roads and interstate 84, so the measurements captured “warm” running emissions. Measurements made by Luke et al, (2009) were made at the University of

Houston’s Moody towers, which are around 1 km away from I-45 and are surrounded by residential housing. Observations in this work were collected farther away from major roads and the site was surrounded by a residential and commercial area. The reason that a higher CO-to-NOx ratio was observed at the T0 site than in Boise is likely due to decreased impact from diesel vehicles.

Figure 5.7 shows a diel plot of the ΔCO/NOx ratio. During the morning rush hour the ΔCO/NOx ratio is flat due to fresh emissions and low photochemical activity, however at 10:00 the ΔCO/NOx ratio begins to ramp up. This increase is due to photochemical conversion of NOx to NOz. The ratio increases to a maximum average value of ~16 at 15:00 and then decreases to a value of ~9 at 00:00 the next day. It isn’t until the morning rush hour that the ΔCO/NOx is driven down to the morning rush hour minimum.

236 There are a number of ways that NOx can be converted to NOz. During the day, the most important is the conversion of NO2 to HNO3 by the hydroxyl radical. The perturbation of the CO-to-NOx ratio during the daytime hours provides evidence of photochemical activity at the site. This summertime observation contrasts the CO-to-

NOx ratio observed during the Boise study presented in Wallace et al. (2012), which did not show any evidence of daytime photochemical activity. The elevated CO-to-NOx ratios observed at the CARES T0 site were attributed to photochemical activity, and suggests that measurements of oxides of nitrogen may help characterize photochemical processing at the site.

5.5.2. VOCs to CO

VOCs are emitted by on-road vehicles through two paths. The first is from un-combusted fuel in the exhaust and the second is evaporative emissions from cracks and leaks in the fuel delivery and holding system (Brooks et al.,

1995; Pierson et al., 1999; Lough et al., 2005; Kirchstetter et al., 1999). Urban concentrations of VOCs resulting from vehicle emissions should look like a combination of gasoline vapor and exhaust. VOC-to-CO emission ratios are reported for this work for the morning rush hour. The evaporative emissions have temperature dependence, so the use of morning rush hour emission ratios may bias the data towards the exhaust component as they exclude the hotter parts of the day when VMT fall off slightly and evaporative emissions increase. The emission ratios reported are the slope of the linear regression fit through the plot of each VOC versus CO. Table 5.2 presents the emission ratios for the VOCs-to-

CO with the statistical error of the fit reported and the R2 along with the average

237 summertime emission ratios from Lough et al., (2005) and Warneke et al., (2009).

Warneke et al., (2009) compares VOC -to-CO ratios collected in Boston/New

York (2004 and 2006) and Los Angeles (2004) to the US EPA’s 1999 national emissions inventory. The spread of the data shows that VOC emissions from cities are highly variable. The VOC-to-CO emission ratios reported here generally agree with the values reported by Warneke et al., (2009) and Lough et al., (2005) to within a factor of 2 with some exceptions (e.g. see toluene in Lough et al., 2005 data).

A plot of measured alkanes-to-CO is shown on a log-log scale in Figure

5.8. The reason that the data are shown on a log-log plot is to visually demonstrate that the linear fit is not biased by elevated values and that the slope reported is representative of the entire morning rush hour data set. Hexane and decane have a R2 of 0.43 and 0.17 respectively, while the rest of the alkanes have

2 R greater than 0.68. Figure 5.9 shows benzene, toluene and speciated C2- alkylbenzenes plotted on a log-log scale versus CO. The linear regressions are representative of the low value data and do not show strong bias from elevated values. All of the R2 values for the VOC’s shown in Figure 5.8 are better than

0.72. Figure 5.10 shows the emission ratios for C3- and C4-alkylbenzenes versus

CO plotted on a log-log scale. Of the VOC’s reported these compounds were found in the lowest abundances and were frequently below or near the detection limit of the system. Isopropylbenzene, 1,2,3-trimethylbenzene, 1,3,5- trimethylbenzene, and 2-ethyltoluene suffer from digitization of the data below 10 pptv, which impacts the ability of the data to capture the covariance. The

238 majority of isopropylbenzene mixing ratios were 1, 2, or 3 pptv and as a result the

R2 is 0.47. This digitization was user imposed when the numbers were truncated at 1 pptv.

5.6. INVESTIGATION OF A METRIC FOR PHOTOCHEMICAL ACTIVITY

In this section we examine the variability of the VOC data to determine if evidence of photochemical activity can be observed in the VOC data. In general, a pollutant may have the following fates: it can be removed chemically (including reactions associated with photochemical activity), it can be transported to or away from the measurement platform, it can be diluted, or it can be deposited to a particle or surface.

Transport of pollutants during CARES was thoroughly investigated by Fast et al., (2012).

This section will focus on fates associated with dilution and chemical loss by comparing the data to simple analytical models that account for dilution and chemical removal in a

Lagrangian parcel of air isolated from sources.

5.6.1. Removal by Reaction With OH

VOCs are chemically removed from the atmosphere by reaction with oxidants such as the hydroxyl radical (OH), the nitrate radical (NO3), and ozone (O3). The hydroxyl radical is the most important oxidant for removing VOC’s during the day in mid-latitude summertime conditions because it is produced and regenerated via photochemical activity. Thus, chemical loss of VOCs during daytime in summer will be dominated by photochemical reactions specifically. Removal of a hypothetical VOC, Xi, by the hydroxyl radical can be described by the following reaction.

. 푋푖 + 푂퐻 → 푃푟표푑푢푐푡푠 R5.1

239 The rate of loss of Xi follows the differential equation given below.

푑[푋 ] 𝑖 = −푘 [푂퐻][푋 ] 5.3 푑푡 푖 푖

Assuming that the hydroxyl radical concentration is time independent, integrating equation 1 yields a solution for Xi as a function of time of the following form.

[푋푖]푡 = [푋0]퐸푥푝(−푘푖 ∗ [푂퐻] ∗ 푡) 5.4

Which can be rearranged to

[푋푖] 1 ln ( 푡) = − 푡 5.5 [푋푖]0 휏 where the lifetime is defined as 휏 = 1/푘푖[푂퐻]. Jobson et al. (1998) demonstrated that, in the absence of mixing, the variance of the ln-transformed data, Slnx, should be inversely related to the lifetime provided there is only an initial injection of [X0 ] and the compound has a sufficiently short lifetime. Thus, the following relationship can be established for a compound where mixing ratios are solely being driven by chemical loss:

1 푆 = 휎 5.6 푙푛푥 휏 푡

The proportionality constant, 휎푡, is the standard deviation of the time distribution.

Experimental investigation of the relationship described by equation 5.6using the VOC data collected during the CARES investigation shows no relationship between the variance of the log transformed mixing ratios and the lifetime (Figure 5.11). To estimate the lifetime, an OH concentration of 6 x 106 was assumed. The lack of a relationship between Slnx and the lifetime indicates that the variability of the VOCs observed at the T0 site was due to factors other than chemical loss. Consequently, variability in VOC mixing ratios was not an adequate measure for identifying the extent of photochemical activity at the T0 site.

240 5.6.2. Photochemical Age.

The oxidation of VOCs by the hydroxyl radical leads to the production of both ozone and SOA. Because the hydroxyl radical is extremely difficult to measure, various metrics to determine the extent which OH has acted on an air parcel have been developed. These methods determine airmass age by determining the amount of processing that has taken place by the hydroxyl radical. One such metric is the ratio of

NOz-to-NOy, which has been correlated with ozone production (Trainer et al. 1993;

Ryerson et al. 1998; Ryerson et al. 2003) and with SOA formation (Miakawa et al.,

2008). Another method to determine photochemical airmass age involves observation of the evolution of VOCs. Plotting the log transformed concentration of two chemical compounds with differing rate constants which are exposed to the same [OH] will yield a slope equal to the ratio of the rate constants. By assuming a hydroxyl radical concentration, the kinetic line can be used as a metric for airmass age. For purely kinetic loss of two VOCs, as time increases, the ratio progresses down the kinetic line. This type of analysis has been applied previously to ambient data to determine the airmass age

(Rudolph and Johnen, 1990; Parrish, et al., 1992; Jobson et al., 1994; McKeen and Liu,

1993; Honrath et al., 2008). Figure 5.12 shows the two limiting cases of dilution into background air free of analytes and kinetic loss discussed in McKeen and Liu (1993).

The dashed line in figure shows dilution and the solid black line shows kinetic loss.

When the plume is free from fresh emissions the data will fall between the kinetic and dilution lines and an airmass age can be determined from the data. The condition that the plume is free of continued emissions of VOCs is explicitly not met by the T0 site which is fumigated with fresh emissions.

241 Plots of log transformed mixing ratios for selected VOCs against ln transformed

2,2,4-trimethylpentane are shown in Figure 5.13. The compound 2,2,4-trimethylpentane was selected because it occurs in abundance well above the detection limit of the system

(the maximum mixing ratio was 550 pptv and the average was 61 pptv) and the rate constant with OH was sufficiently low (3.34*10-12) relative to some of the aromatics which are an order of magnitude more reactive. Choosing compounds with vastly different rate constants with OH would give the best possibility to determine if OH chemistry was observable using this method. The data are shaded by the ratio of NOz-to-

NOy which is a strong indication of photochemical activity. Coincidentally, it is also correlated to boundary layer height as the highest NOz-to-NOy ratios occur at around

16:00 when the boundary layer is near its maximum. The ln-lm data for all of the graphs are scatted about the clean air dilution line. Based on the analysis presented in the model above this is consistent with the VOC mixing ratios being driven by emissions and dilution. That the lowest mixing ratios tend to correspond with the highest NOz to NOy ratios is explained by maximum dilution occurring when the maximum NOz to NOy ratios occur. The slope of ln(toluene) vs. ln(2,2,4-trimethylpentane) was 0.65 and the slope of ln(1,2,4-trimethylbenzene) vs. ln(2,2,4-trimethylpentane) was 0.75, which indicates that there is a mechanism other than dilution driving the mixing ratios. It is likely that there are some fresh emissions with a different emission ratio of toluene and 1,2,4- trimethybenzene that are causing the deviations from the dilution line. The remaining ln- ln plots shown in Figure 5.13, ln(p,m-xylene) vs. ln(2,2,4-trimethylpentane), ln(o-xylene) vs. ln(2,2,4-trimethylpentane) and ln(3-methylpentane) vs. ln(2,2,4-trimethylpentane) lie along the 1:1 dilution line.

242 The ln-ln plots in Figure 5.14 were selected because they include compounds with interesting relationships or a history of use in this type of analysis (Jobson et al., 2004).

For instance, toluene and benzene have been used in a number of studies to determine airmass age, p,m-xylene and o-xylene are isomers with a very high emission correlation

2 (R =0.98), and 1,2,4-trimethyl benzene is the most abundant C3- alkylbenzene and has a reasonably high rate constant with OH. By comparing the log transformation of 1,2,4- trimethylbenzene versus p,m-xylene, which has a comparable OH reaction rate and also benzene which a much lower reaction rate, it is clear that there is no evidence of relative loss driven by OH oxidation chemistry. As with Figure 5.13, the plots shown in Figure

5.14 indicate that the ratio of the VOCs is driven by dilution as most of the data fall along the 1:1 dilution line. The log plots of p,m-xylene versus o-xylene shows the least scatter of all of the plots. This is attributed to the high degree of co-emission that these isomers exhibit. Figures 5.13 and 5.14 are consistent with dilution and emissions as the drivers of

VOC variability. They demonstrate that sampling at the T0 site resulted in observations of fresh VOC emissions, which validates the site location.

5.6.3. Photochemistry of Nitrogen Oxides and OA

The overall reaction for NOx, VOCs and light in the atmosphere can be generally summarized as:

푁푂푥 + 푉푂퐶 + ℎ푣 → 푂3 + 퐻푁푂3 + 푃퐴푁 + 푃푎푟푡푖푐푙푒푠 + 퐶푂2 + 퐻퐶퐻푂 … R5.2

The NOy reported here and elsewhere is defined functionally as reactive oxides of nitrogen which can be converted to NO with a molybdenum oxide catalytic converter at

o 310 C. This includes NO + NO2 + HNO3 + HONO+ NO3 + N2O5 + HNO4 +ClONO2 + organic nitrates + particulate nitrate. NOz is defined as NOy –NOx. Examining the

243 summarized reaction of VOCs and NOx in the presence of sunlight reveals that NOz and,

O3, and particles are all products, so there should be some relationship between measured concentrations in urban air. In an urban environment HONO, PAN and HNO3 can make up significant fractions of NOy. NOz-to-NOy has been used as a tracer for photochemical activity (Trainer et al., 1993; Miyakawa et al., 2008). There are far too many pathways to describe all the mechanisms involved in producing NOz but it is worth noting some of the major photochemical pathways to convert NOx to NOz:

푀 . 푁푂2 + 퐻푂 → 퐻푁푂3 R5.3

푀 푁푂 + 퐻푂. → 퐻푂푁푂 R5.4

푀 . ∗ 푅푂2 + 푁푂 → 푅푂푁푂2 → 푅푂푁푂2 R5.5

Figure 5.15 shows the diel mean and median NOz-to-NOy ratio. There are very small differences between the mean and median. This ratio experiences a minimum during the morning rush hour as fresh emission of NOx from mobile sources mean that the majority of NOy is NOx. Starting at 08:00 photochemistry takes over. Fresh emissions of NOx are still being emitted, however NOz is being produced and accumulating in the atmosphere so the ratio increases linearly to the maximum average value of 0.6 at 16:00.

The values then decrease to the nighttime values of around 0.2. NOz-to-NOy grows slowly overnight from 00:00 to 04:00. From 00:00 to 04:00 there are still emissions of

NOx yet the ratio of NOz-to-NOy is slightly increasing. This indicates that the nitrate radical, an important night time oxidant for the Sacramento area is present because overnight production of NOz is occurring. It can react with organics in much the same way that the hydroxyl radical does, by either H abstraction or addition to carbon double bonds. The rates of reaction are sufficiently fast to make it atmospherically relevant

244 (Smith et al., 1995; Geyer et al., 2001; Stutz et al., 2004; Atkinson and Arey, 2007).

Night time oxidation of NOx produces NOz through the following mechanisms all of which involve nitrate:

푁푂 + 푂3 → 푁푂2 + 푂2 R5.6

. 푁푂2 + 푂3 → 푁푂3 + 푂3 R5.7

푀 . 푁푂3 + 푁푂2 → 푁2푂5 R5.8

푁2푂5 + 퐻2푂 → 2퐻푁푂3 R5.9

This will only happen if there is O3 around to keep NO low and produce NO2 because if

NO occurs in large abundances it will destroy NO3.

푁푂3 + 푁푂 → 2푁푂2 R5.10

Based on the reactivities, both O3 and NOx occur in sufficient mixing ratios during the

. night time to produce both NO3 and N2O5. As illustrated below, oxides of nitrogen provide the best tracers of oxidation (both during the daytime and the night time) of all the data examined in this work.

5.6.4. O3, NOz, and OA

Both O3 and OA are products of the general oxidation reaction (R2), as such there should be some relationship for their mixing ratios. Miyakawa et al., (2008) have demonstrated positive correlation between NOz-to-NOy ratios and OA mass. Figure 5.16 shows the diel, average mixing ratio of O3, OA, and NOz. Ozone, NOz, and OA exhibit some degree of covariance. However, OA and NOz lag O3 during the initial daytime production and their average maxima peaks much earlier at 12:00 than O3 which peaks at

16:00. OA and NOz also fall off earlier than O3. OA is the only compound to shoot back up briefly at 20:00. The increase of OA at 20:00, around sunset, can be explained by

245 examining the behavior of compounds such as CO and NOx. CO and NOx also increase during this time and are commonly used as tracers for primary emissions sources. OA has both a component that is photochemical and also a primary component. The diel plots show the tightest correlation between OA and NOz. It is interesting that both OA and NOz have delayed morning production when compared to O3. It is probable this results from the breakup of the nocturnal boundary layer resulting in higher O3 aloft mixing into the

PBL. Setyan et al., (2012) found nitrate to make up only 3.1% of the aerosol mass

+ + observed at the T1 site. However they found a much higher slope of NO to NO2 than is

+ + observed for pure NH4NO3. An elevated NO to NO2 has been found to indicate the presence of organonitrates, which can explain the relationship between NOz and OA.

The NOy measurement includes nitrate in particles, so incorporation of nitrate in the growth of OA may be important and it also may explain the reason why NOz correlates well with OA. NOz appears to be useful for understanding OA and could potentially be used by air quality managers to help predict OA. The correlation of OA and NOz deserves more study. A plot of OA versus O3, with the highest photochemically active period (12:00-20:00) indicated by red dots in Figure 5.17 shows that the majority of the elevated NOz, OA, and O3 are observed during this period. Despite the differences in observed diel averaged NOz, O3, and OA, there is a relationship between the three. For

O3 and NOz this relationship is the strongest from 12:00 to 20:00. When compared to O3 and NOz, OA has a component that is elevated and is not well correlated with O3 and

NOz. The portion of OA that is not correlated with O3 and NOz is due to primary emissions of OA. OA has both a primary and secondary component that are indistinguishable from each other. Oxidized organic aerosol (OOA) which has been

246 presented by Jimenez et al., (2009) was not available as a data product, but comparing

OOA and NOz could yield a better correlation since there is some evidence that the OOA component is composed of secondary OA (Lanz et al., 2007; Ulbrich et al., 2009;

Volkamer et al., 2006; Zhang et al., 2005). NOz is most highly correlated with OA during the daytime and it is possible that NOz, which is easily measured, can be used as a tracer to inform modelers and air quality managers as to how much OA there is in this region.

5.7. SUMMARY

This work presented the trace gases observed at the T0 site during the CARES campaign. This was the maiden voyage for the GC-ITMS VOC pre-concentration system described in chapter 2.2. The system operated well producing a valuable data set that cataloged VOC-to-CO ratios from mobile sources, quantified important VOC mixing ratios and helped to define the T0 site as VOC limited. It is a robust, cryogen free platform capable of observing VOCs in the atmosphere. The VOC-to- CO emission ratios measured were in good agreement with those of Warneke et al., (2009). The addition of this system adds a vital measurement platform to the WSU MACL. Future work with this dataset includes comparison to the MOVES emissions model. The major findings of this study include:

 Morning rush hour emission ratios of CO-to-NOx along with VOC-to-CO ratios

are presented. CO-to-NOx and VOC-to-CO ratios measured at the T0 site in

Sacramento, CA during the CARES field campaign agree with those presented

previously by Parrish (2006).

247  This work demonstrated that the variability of VOCs observed was due to sources

and dilution rather than photochemical loss, so VOC variability could not be used

to characterize photochemical activity at the site. To observe the chemistry of the

VOCs in the manner used in this analysis the site would have to be farther away

from the site and have a path that was reasonably free from emissions of fresh

VOCs. For areas littered with urban regions like California this condition is

unlikely to be met and a photoproduct is needed to calculate airmass age. The

chamber and oxidant source described in chapter 2.2 will be a useful tool to

identify a probable oxidation product for future field work.

 Based on the work here, the most abundant and most reactive VOCs are p,m-

xylene, toluene and 1,3,5-TMB.

 Even though the VOC data did not provide a dataset that could be used to

determine airmass age, NOz showed potential for prediction of OA. This study

found that NOz has the strongest correlation with OA and may be useful to predict

organic aerosol as it likely correlates with the SOA component of OA. The

relationship between NOz production and OA deserves further study.

 The growth of the NOz-to-NOy ratio provides evidence of nighttime production of

NOz and indicates that the nitrate radical is being produced and is a potentially

important oxidant during the night.

248 5.8. RERENCES

Atkinson, R., & Arey, J. “Mechanisms of the gas-phase reactions of aromatic hydrocarbons and PAHs with OH and NO 3 radicals.” Polycyclic Aromatic Compounds,

27(1), (2007) 15-40.

Bao, J. W., Michelson, S. A., Persson, P. O. G., Djalalova, I. V., & Wilczak, J. M.

“Observed and WRF-simulated low-level winds in a high-ozone episode during the

Central California Ozone Study.” Journal of Applied Meteorology and Climatology,

47(9), (2008) 2372-2394.

Brooks, D. J., Peltier, R. J., Baldus, S. L., Reuter, R. M., Bandy III, W. J., & Sprik, T. L.

“Real-World Hot Soak Evaporative Emissions—A Pilot Study. SAE Paper

951007.”Auto/Oil Air Quality Improvement Research Program, 3, (1995) 1-20.

Calvert, J. G., Atkinson, R., Becker, K. H., Kamens, R. M., Seinfeld, J. H., Wallington,

T. J., & Yarwood, G. “The mechanisms of atmospheric oxidation of aromatic hydrocarbons” (2002) 346-397 New York: Oxford University Press.

Finlayson-Pitts, B. J., & Pitts Jr, J. N. “Chemistry of the upper and lower atmosphere:

Theory, experiments, and applications.” Academic press (1999).

249 Forster, P., Ramaswamy, V., Artaxo, P., Berntsen, T., Betts, R., Fahey, D. W., et al.,

“Changes in atmospheric constituents and in radiative forcing.” Climate Change, 20

(2007)

Fujita, E. M., Croes, B. E., Bennett, C. L., Lawson, D. R., Lurmann, F. W., & Main, H.

H. “Comparison of emission inventory and ambient concentration ratios of CO, NMOG, and NOx in California's South Coast Air Basin.” Journal of the Air & Waste

Management Association, 42(3), (1992) 264-276.

Fujita, E. M., Campbell, D. E., Zielinska, B., Chow, J. C., Lindhjem, C. E., DenBleyker,

A., et al., “Comparison of the MOVES2010a, MOBILE6. 2, and EMFAC2007 mobile source emission models with on-road traffic tunnel and remote sensing measurements.

Journal of the Air & Waste Management Association, 62(10), (2012) 1134-1149.

Dallmann, T. R., & Harley, R. A. “Evaluation of mobile source emission trends in the

United States.” Journal of Geophysical Research: Atmospheres (1984–2012), 115(D14)

(2010).

Davidson, C. I., Phalen, R. F., & Solomon, P. A. “Airborne particulate matter and human health: A review.” Aerosol Science and Technology, 39(8), (2005) 737-749.

De Gouw, J. A., Middlebrook, A. M., Warneke, C., Goldan, P. D., Kuster, W. C.,

Roberts, J. M., et al., “Budget of organic carbon in a polluted atmosphere: Results from

250 the New England Air Quality Study in 2002.” Journal of geophysical research,

110(D16), (2005) D16305.

Dillon, M. B., Lamanna, M. S., Schade, G. W., Goldstein, A. H., & Cohen, R. C.

“Chemical evolution of the Sacramento urban plume: Transport and oxidation.” Journal of Geophysical Research: Atmospheres (1984–2012), 107(D5), (2002) ACH-3.

Donahue, N. M., Robinson, A. L., Stanier, C. O., & Pandis, S. N. “Coupled partitioning, dilution, and chemical aging of semivolatile organics.” Environmental Science &

Technology, 40(8), (2006) 2635-2643.

Donahue, N. M., Henry, K. M., Mentel, T. F., Kiendler-Scharr, A., Spindler, C., Bohn,

B., et al., “Aging of biogenic secondary organic aerosol via gas-phase OH radical reactions.” Proceedings of the National Academy of Sciences, 109(34), (2012) 13503-

13508.

Fierz, M., Vernooij, M. G., & Burtscher, H. (2007) “An improved low-flow thermodenuder.” Journal of Aerosol Science, 38(11), (2007) 1163-1168.

Kroll, J. H., Ng, N. L., Murphy, S. M., Flagan, R. C., & Seinfeld, J. H. “Secondary organic aerosol formation from isoprene photooxidation.” Environmental Science &

Technology, 40(6), (2006) 1869-1877.

251 Lefer, B., Rappenglück, B., Flynn, J., & Haman, C. “Photochemical and meteorological relationships during the Texas-II Radical and Aerosol Measurement Project (TRAMP).”

Atmospheric Environment, 44(33), (2010) 4005-4013.

Lough, G. C., Schauer, J. J., Lonneman, W. A., & Allen, M. K. “Summer and winter nonmethane hydrocarbon emissions from on-road motor vehicles in the Midwestern

United States.” Journal of the Air & Waste Management Association, 55(5), (2005) 629-

646.

Luke, W. T., Kelley, P., Lefer, B. L., Flynn, J., Rappenglück, B., Leuchner, M., Buhr, M., et al., “Measurements of primary trace gases and NOY composition in Houston, Texas.”

Atmospheric Environment, 44(33), (2010) 4068-4080.

Gentner, D. R., Isaacman, G., Worton, D. R., Chan, A. W., Dallmann, T. R., Davis, L.,

Goldstein, A. H., et al., “Elucidating secondary organic aerosol from diesel and gasoline vehicles through detailed characterization of organic carbon emissions.”

Proceedings of the National Academy of Sciences, 109(45), (2012) 18318-18323.

Geyer, A., Alicke, B., Konrad, S., Schmitz, T., Stutz, J., & Platt, U. “Chemistry and oxidation capacity of the nitrate radical in the continental boundary layer near Berlin.”

Journal of Geophysical Research: Atmospheres (1984–2012), 106(D8), (2001) 8013-

8025.

252 Hallquist, M., Wenger, J. C., Baltensperger, U., Rudich, Y., Simpson, D., Claeys, M., ...

& Wildt, J. “The formation, properties and impact of secondary organic aerosol: current and emerging issues.” Atmospheric Chemistry and Physics, 9(14), (2009) 5155-5236.

Harrison, R. M., & Yin, J. ‘Particulate matter in the atmosphere: which particle properties are important for its effects on health?” Science of the Total Environment, 249(1), (2000)

85-101.

Honrath, R. E., Helmig, D., Owen, R. C., Parrish, D. D., & Tanner, D. M. “Nonmethane hydrocarbons at Pico Mountain, Azores: 2. Eventspecific analyses of the impacts of mixing and photochemistry on hydrocarbon ratios.” Journal of Geophysical Research:

Atmospheres (1984–2012), 113(D20) (2008).

Jimenez, J. L., Canagaratna, M. R., Donahue, N. M., Prevot, A. S. H., Zhang, Q., Kroll, J.

H., et al., “Evolution of organic aerosols in the atmosphere.” Science, 326(5959), (2009)

1525-1529.

Jobson, B. T., Wu, Z., Niki, H., & Barrie, L. A. “Seasonal trends of isoprene, C2–C5 alkanes, and acetylene at a remote boreal site in Canada.” Journal of Geophysical

Research: Atmospheres (1984–2012), 99(D1), (1994) 1589-1599.

253 Kanakidou, M., Seinfeld, J. H., Pandis, S. N., Barnes, I., Dentener, F. J., Facchini, M. C., et al., “Organic aerosol and global climate modeling: a review.” Atmospheric Chemistry and Physics, 5(4), (2005) 1053-1123.

Kirchstetter, T. W., Singer, B. C., Harley, R. A., Kendall, G. R., & Hesson, J. M. “Impact of California reformulated gasoline on motor vehicle emissions. 2. Volatile organic compound speciation and reactivity.” Environmental Science & Technology, 33(2),

(1999) 329-336.

Kleinman, L. I., Springston, S. R., Daum, P. H., Lee, Y. N., Nunnermacker, L. J., Senum,

G. I., et al., “The time evolution of aerosol composition over the Mexico City plateau.”

Atmospheric Chemistry Physics, 8(6), (1999) 1559-1575.

Lough, G. C., Schauer, J. J., Lonneman, W. A., & Allen, M. K. “Summer and winter nonmethane hydrocarbon emissions from on-road motor vehicles in the Midwestern

United States.” Journal of the Air & Waste Management Association, 55(5), (2005) 629-

646.

McKeen, S. A., & Liu, S. C. “Hydrocarbon ratios and photochemical history of air masses.” Geophysical Research Letters, 20(21), (1993) 2363-2366.

254 Miyakawa, T., Takegawa, N., & Kondo, Y. “Photochemical evolution of submicron aerosol chemical composition in the Tokyo megacity region in summer.” Journal of

Geophysical Research: Atmospheres (1984–2012), 113(D14) (2008).

Murphy, D. M., Cziczo, D. J., Froyd, K. D., Hudson, P. K., Matthew, B. M.,

Middlebrook, A. M., et al., “Single-particle mass spectrometry of tropospheric aerosol particles.” Journal of Geophysical Research: Atmospheres (1984–2012), 111(D23)

(2006).

Odum, J. R., Jungkamp, T. P. W., Griffin, R. J., Flagan, R. C., & Seinfeld, J. H. “The atmospheric aerosol-forming potential of whole gasoline vapor.” Science, 276(5309),

(1997a) 96-99.

Odum, J. R., Jungkamp, T. P. W., Griffin, R. J., Forstner, H. J. L., Flagan, R. C., &

Seinfeld, J. H. “Aromatics, reformulated gasoline, and atmospheric organic aerosol formation.” Environmental Science & Technology, 31(7), (1997b) 1890-1897.

Parrish, D. D., Hahn, C. J., Williams, E. J., Norton, R. B., Fehsenfeld, F. C., Singh, H. B., et al., “Indications of photochemical histories of Pacific air masses from measurements of atmospheric trace species at Point Arena, California.” Journal of Geophysical Research:

Atmospheres (1984–2012), 97(D14), (1992) 15883-15901.

255 Pierson, W. R., Schorran, D. E., Fujita, E. M., Sagebiel, J. C., Lawson, D. R., & Tanner,

R. L. “Assessment of nontailpipe hydrocarbon emissions from motor vehicles.” Journal of the Air & Waste Management Association, 49(5), (1999) 498-519.

POPE III, C. A. “What do epidemiologic findings tell us about health effects of environmental aerosols?” Journal of Aerosol Medicine, 13(4), (2000) 335-354.

Pope III, C. A., & Dockery, D. W. “Health effects of fine particulate air pollution: lines that connect.” Journal of the Air & Waste Management Association, 56(6), (2006) 709-

742.

Robinson, A. L., Donahue, N. M., Shrivastava, M. K., Weitkamp, E. A., Sage, A. M.,

Grieshop, A. P., et al., “Rethinking organic aerosols: Semivolatile emissions and photochemical aging.” Science, 315(5816), (2007) 1259-1262.

Rudolph, J., & Johnen, F. J. “Measurements of light atmospheric hydrocarbons over the

Atlantic in regions of low biological activity.” Journal of Geophysical Research:

Atmospheres (1984–2012), 95(D12), (1990) 20583-20591.

Ryerson, T. B., Buhr, M. P., Frost, G. J., Goldan, P. D., Holloway, J. S., Hübler, G., et al., “Emissions lifetimes and ozone formation in power plant plumes.” Journal of

Geophysical Research: Atmospheres (1984–2012), 103(D17), (1998) 22569-22583.

256 Ryerson, T. B., Trainer, M., Angevine, W. M., Brock, C. A., Dissly, R. W., Fehsenfeld,

F. C., et al., “Effect of petrochemical industrial emissions of reactive alkenes and NOx on tropospheric ozone formation in Houston, Texas.” Journal of Geophysical Research:

Atmospheres (1984–2012), 108(D8) (2003).

Schwartz, J., Dockery, D. W., & Neas, L. M. “Is daily mortality associated specifically with fine particles?” Journal of the Air & Waste Management Association, 46(10), (1996)

927-939.

Seinfeld, J. H., & Pandis, S. N. (2006). Atmospheric chemistry and physics: from air pollution to climate change. Wiley-Interscience, Hoboken, NJ.

Setyan, A., Zhang, Q., Merkel, M., Knighton, W. B., Sun, Y., Song, C., et al., (2012).

“Characterization of submicron particles influenced by mixed biogenic and anthropogenic emissions using high-resolution aerosol mass spectrometry: results from

CARES.” Atmospheric Chemistry and Physics, 12(17), 8131-8156.

Smith, N., Plane, J., Nien, C. F., & Solomon, P. A. “Nighttime radical chemistry in the

San Joaquin Valley.” Atmospheric Environment, 29(21), (1995) 2887-2897.

Stephens, S., Madronich, S., Wu, F., Olson, J. B., Ramos, R., Retama, A., & Munoz, R.

“Weekly patterns of México City’s surface concentrations of CO, NOx, PM10 and O3 during 1986–2007.” Atmos. Chem. Phys, 8, (2008). 5313-5325.

257

Stutz, J., Alicke, B., Ackermann, R., Geyer, A., White, A., & Williams, E. “Vertical profiles of NO3, N2O5, O3, and NOx in the nocturnal boundary layer: 1. Observations during the Texas Air Quality Study 2000.” Journal of Geophysical Research, 109(D12),

(2004) D12306.

Thurston, G. D., Ito, K., Hayes, C. G., Bates, D. V., & Lippmann, M. “Respiratory hospital admissions and summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols.” Environmental Research, 65(2), (1994) 271-

290.

Trainer, M., Parrish, D. D., Buhr, M. P., Norton, R. B., Fehsenfeld, F. C., Anlauf, K. G., et al., “Correlation of ozone with NOy in photochemically aged air.” Journal of

Geophysical Research: Atmospheres (1984–2012), 98(D2), (1993) 2917-2925.

Troe, J. “Predictive possibilities of unimolecular rate theory.” Journal of Physical

Chemistry, 83(1), (1979) 114-126.

Troe, J. “Theory of Thermal Unimolecular Reactions in the Falloff Range. I. Strong

Collision Rate Constants.” Berichte der Bunsengesellschaft für physikalische Chemie,

87(2), (1983) 161-169.

258 Van Ooy, D. J., & Carroll, J. J. “The spatial variation of ozone climatology on the western slope of the Sierra Nevada.” Atmospheric Environment, 29(11), (1995) 1319-

1330.

Valavanidis, A., Fiotakis, K., & Vlachogianni, T. “Airborne particulate matter and human health: toxicological assessment and importance of size and composition of particles for oxidative damage and carcinogenic mechanisms.” Journal of Environmental Science and

Health, Part C, 26(4), (2008) 339-362.

Volkamer, R., Jimenez, J. L., San Martini, F., Dzepina, K., Zhang, Q., Salcedo, D.,

“Secondary organic aerosol formation from anthropogenic air pollution: Rapid and higher than expected.” Geophysical Research Letters, 33(17) (2006).

Zhang, Q., Jimenez, J. L., Canagaratna, M. R., Allan, J. D., Coe, H., Ulbrich, I.,

“Ubiquity and dominance of oxygenated species in organic aerosols in anthropogenically-influenced Northern Hemisphere midlatitudes.” Geophysical Research

Letters, 34(13), (2007) L13801.

Zaveri, R. A., Shaw, W. J., Cziczo, D. J., Schmid, B., Ferrare, R. A., Alexander, M. L., et al., “Overview of the 2010 Carbonaceous Aerosols and Radiative Effects

Study(CARES).” Atmospheric Chemistry and Physics, 12(16), (2012) 7647-7687.

259 5.9. CHAPTER 5 TABLES

Table 5.1. OH+VOC rate constants for the VOCs reported in this work along with the lifetime in hours assuming [OH] = 6*106.

k @ 298K: Lifetime for (cm3 molecule-1 reaction with Compound s-1) Error OH (h) hexane 5.55E-12 3.6% 8.3 nonane 1.05E-11 3.8% 4.4 decane 1.12E-11 5.4% 4.1 2,2,4-trimethylpentane 3.34E-12 3.9% 11.9 3-methylpentane* 5.62E-12 1.8% 8.2 benzene 1.39E-12 10% 33.3 toluene 5.63E-12 20% 8.2 ethylbenzene 7.00E-12 25% 6.6 p,m-xylene* 1.80E-11 30% 2.6 o-xylene 1.36E-11 25% 3.4 1,3,5-TMB 5.67E-11 25% 0.8 1,2,3-TMB 3.27E-11 25% 1.4 1,2,4-TMB 3.25E-11 25% 1.4 2-ethyltoluene 1.19E-11 35% 3.9 3-ethyltoluene 1.86E-11 35% 2.5 4-ethyltoluene 1.18E-11 35% 3.9 n-butylbenzene 4.50E-12 35% 10.3 isopropylbenzene 6.30E-12 30% 7.3 *p-xylene and m-xylene elute at the same time and have the same mass spectra so they can’t be distinguished.

260

Table 5.2. VOC-to-CO emission ratios reported by this work and the works of Warneke et al., (2009) and Lough et al., (2005). The ratios are reported as ppbv per ppbv CO.

Lough et Boston/New Los Angeles Boston/New Boise Compound This Work +/- r2 al., (2005) York 2004* 2002* York 2006* 2008/2009 hexane 7.00E-04 1.00E-04 0.43 2.44E-03 1.07E-03 1.13E-03 nonane 1.40E-03 1.50E-05 0.68 1.59E-04 decane 7.10E-05 2.50E-05 0.17 7.66E-05 1.00E-05 2,2,4-triMEpentane 9.40E-03 6.90E-05 0.77 2.92E-03 4.76E-03 3-methylpentane 1.30E-03 9.20E-05 0.77 1.75E-03 1.28E-03 benzene 1.10E-03 4.60E-05 0.92 3.68E-03 6.17E-04 9.50E-04 1.09E-03 1.50E-03 toluene 4.70E-03 3.40E-04 0.77 6.59E-05 2.62E-03 3.51E-03 3.79E-03 2.60E-03 ethylbenzene 6.00E-04 4.70E-05 0.72 1.14E-04 8.10E-05 2.20E-04 p,m-xylene 2.10E-03 2.00E-04 0.73 2.40E-03 1.16E-03 6.40E-04 o-xylene 6.70E-04 5.10E-05 0.76 8.57E-04 4.59E-04 4.50E-04 1,3,5-TMB 1.30E-04 1.10E-05 0.71 3.78E-04 9.10E-05 1,2,3-TMB 1.20E-04 9.80E-06 0.72 6.67E-05 6.90E-05 1,2,4-TMB 6.00E-04 4.90E-05 0.75 3.88E-04 3.50E-04 2.50E-04 2-ethyltoluene 1.20E-04 8.80E-06 0.76 1.24E-04 1.00E-04 3-ethyltoluene 3.60E-04 3.90E-05 0.59 1.11E-03 4-ethyltoluene 2.70E-04 3.10E-05 0.57 6.52E-04 isopropylbenzene 2.20E-05 3.10E-06 0.47 1.18E-04 2.50E-05

*Warneke et al., 2009

261 5.10. CHAPTER 5 FIGURES

Figure 5.1. Time series of CO, NO, NO2, and NOy.

262

Figure 5.2. Time series of selected VOCs in ppbv measured with the GC-ITMS. Isoprene (red), benzene, (gray), toluene (blue), p,m-xylene (green), and 1,2,4-trimethylbenzene are shown from 6/10/2010 to 7/1/2010.

263

Figure 5.3. Time series of organic aerosol, O3, and NOz.

264

Figure 5.4. Diel plots of OH loss frequency (s-1) for the VOCs measured during CARES.

265

Figure 5.5. Plot of CO versus NOx for the morning rush hour (06:00 to 09:00). The linear regression gave a slope of 6.9 and a R2 of 0.68.

Figure 5.6. Log histogram of ΔCO/NOx for the morning rush hour (06:00-09:00) during the CARES campaign.

266

Figure 5.7. A diel plot of ΔCO/NOx showing all the data (gray circles), the median (light blue triangles) and average (dark blue) squares. The rush hour period (06:00-10:00) is flat.

267

Figure 5.8. Figure 8. Alkane versus CO during the morning rush hour plotted (grey circles) on a Log-Log scale along with the linear regression (black line) through the data.

268

Figure 5.9. Plots of toluene, benzene and C2-alkylbenzenes versus CO during the morning rush hour on a Log-Log plot.

269

Figure 5.10. Log-Log plots of C3- and C4-alkylbenzenes versus CO. The plots include the data (gray circles) and the linear regression (black trace).

270

Figure 5.11. The standard deviation of the log transformed VOC mixing ratios versus the estimated lifetime

271

Figure 5.12. This is a plot of the time evolution of an air parcel containing two VOCs

“A” and “C” for the limiting cases of dilution (dashed line) and kinetic removal (solid line).

272

Figure 5.13. Plots of the ln transformed VOC versus 2,2,4-trimethylpentane. The data are plotted over the dilution line (black dashes) and the kinetic line (solid black). The data are color coded by the ratio of NOz-to-NOy.

273

Figure 5.14. Selected plots of the ln transformed VOC versus VOC. The data are plotted over the dilution line (black dashes) and the kinetic line (solid black). The data are color coded by the ratio of NOz-to-NOy.

274

Figure 5.15. Diel plot of the NOz/NOy ratio. The data (gray circles), along with the mean

(blue triangles) and the median (blue squares) values are shown.

275

Figure 5.16. Plots of diel averaged O3 and NOz (top), O3 and OA (middle), and NOz and

OA (bottom).

276

Figure 5.17. Plot of OA versus O3. The red circles indicate data collected from 12:00 to

20:00.

277 SUMMARY AND CONCLUSIONS

Chapter 2 is a description of the instrumentation used to produce a significant portion of the data presented in this dissertation. Descriptions of the NOxy system, the

VOC preconcentration and a photochemical chamber including principles of measurement, calibration and operation are presented. Both the GCMS and the NOxy instruments have been integrated and deployed in the MACL and significantly increases the ability of LAR to quantify trace gases. The NOxy system is able to measure pptv levels of NO, NO2, and NOy. It has performed reliably in the field and can be deployed as is. The GC-ITMS preconcentration system was designed and built as a significant part of this dissertation represents a major improvement over the cryogenic version deployed during CABINEX however, it remains work in progress. One of the main objectives of this instrument was to identify and quantify aromatic and biogenic oxidation products to be used as metrics for atmospheric oxidation. Analysis of the VOC data collected during

CARES demonstrates that an oxidation product is needed to quantify airmass age in urban environments. The issues that have been encountered with this system are described and recommendations for future progress are suggested. The photochamber is well mixed for flows ranging from 22.4 to 4.8 LPM and is a useful laboratory tool allowing for investigation of oxidation of VOCs by the hydroxyl radical under laboratory conditions. Aside from getting the light hydrocarbon channel working, using the chamber to generate oxidation products will be the next step to improve the capabilities of the system. The chamber is also useful for producing low mixing ratios using exponential dilution.

278 The future direction of this system includes fixing the light hydrocarbons channel, identifying oxidation products in channel 2. This channel is malfunctioning due to a lack of flow through the adsorbent tubes during desorption mode. This results in poor desorption of analytes and bad chromatography. The ability of the system to quantify oxidation products is hindered by a benzaldehyde artifact from the Tenax. While the

Tenax is a good adsorbent because it traps a wide range of VOCs and has a low affinity for water vapor, the inability to quantify benzaldehyde which is the most abundant aromatic oxidation product is reason to explore other adsorbent materials.

Chapter 3 describes findings from the CABINEX field campaign made during

July 2009 at the PROPHET tower in northern Michigan. Profiles of NOx, VOCs, and O3 at 6 m, 20 m, and 34 m heights are reported. The contrast between clean and polluted conditions at the PROPHET tower shows the impact of transported pollution on the chemistry of a forest. The site is located in a heavily forested region and thus was primarily impacted by local emissions of VOCs and NOx from soils. Profiles of VOCs,

NOx and oxidation products including HCHO and CH3OOH were reported. This study confirmed that forests produce gradients of VOC, NOx, O3 and oxidation products. An analysis of gradients during clean conditions showed that emissions of NOx from soils accumulate under the canopy during the night. These gradients are useful to help understand mixing through the forest canopy. Examination of the Leighton relationship above and below the canopy showed evidence of air mixing between the understory and below the canopy. The cryogenic VOC preconcentration system used with the ITMS for this study suffered significant problems associated with trapping CO2 and helped to motivate the development of the adsorbent based preconcentration system.

279

The results from a comparison between data collected during January 2008 and

December 2009 in Meridian, ID and the EPA’s mobile emissions models MOVES and

MOBILE6.2 are presented here as chapter 4 and were published as Wallace et al. (2012).

Emission ratios of CO to NOx and VOCs to CO were inferred from morning rush hour data and compared to mobile emissions models. Major findings of this work are that

MOVES represents a major improvement over MOBILE6.2 for running emissions, diesel vehicles are significant sources of NOx in this region, wintertime studies provide ideal conditions to study emissions, and starting emissions need more study.

Chapter 5 described the results of measurements made in Sacramento, CA during the CARES campaign in June of 2010. The WSU MACL was deployed with the adsorbent based VOC preconcentration system, which collected data on one channel (C2-

C12) due to a malfunction on the light hydrocarbon channel (C2-C6). Morning time emission ratios of CO to NOx and VOCs to CO inferred from the data are compared to values published by Lough et al. (2005) and Warneke et al. (2009). The variability of the

VOC data was driven by emission and dilution rather than chemistry. To see evidence of photochemical oxidation with the ITMS, a photoproduct is needed. Significant evidence of photochemistry was observed using the NOx , NOy data. The best correlation with OA was from NOz. This is because NO2 was the largest measured OH sink in this environment.

280