FINAL REPORT An Assessment of Nitryl Chloride Formation Chemistry and its Importance in Ozone Non-attainment areas in Texas. AQRP Project

Similar documents
Airborne observations of ammonia emissions from agricultural sources and their implications for ammonium nitrate formation in California

Ongoing EPA efforts to evaluate modeled NO y budgets. Heather Simon, Barron Henderson, Deborah Luecken, Kristen Foley

Using Global and Regional Models to Represent Background Ozone Entering Texas

Examining the impact of heterogeneous nitryl chloride production on air quality across the United States

CHAPTER 1. MEASURES OF ATMOSPHERIC COMPOSITION

Surface-Atmosphere Exchange of Ammonia in a Non-fertilized Grassland and its Implications for PM 2.5

PROBLEMS Sources of CO Sources of tropospheric ozone

AQRP Project Implementation and evaluation of new HONO mechanisms in a 3-D Chemical Transport Model for Spring 2009 in Houston

Continuous measurement of airborne particles and gases

* Author to whom correspondence should be addressed; Tel.:

Lab 4 Major Anions In Atmospheric Aerosol Particles

Sally E. Pusede, Trevor C. VandenBoer, Jennifer G. Murphy, Milos Z. Markovic, Cora J. Young,

EVALUATION OF ORIGINAL AND IMPROVED VERSIONS OF CALPUFF USING THE 1995 SWWYTAF DATA BASE. Technical Report. Prepared by

MODELING CHEMICALLY REACTIVE AIR TOXICS IN THE SAN FRANCISCO BAY AREA USING CAMx

ATOC 3500/CHEM 3151 Week 9, 2016 The Game Changer. Some perspective The British Antarctic Survey The Ozone Hole International Regulations

Implementation and Testing of EQUISOLV II in the CMAQ Modeling System

Photochemical model simulations of air quality for Houston Galveston Brazoria area and analysis of ozone NO x hydrocarbon sensitivity

J4.2 ASSESSMENT OF PM TRANSPORT PATTERNS USING ADVANCED CLUSTERING METHODS AND SIMULATIONS AROUND THE SAN FRANCISCO BAY AREA, CA 3.

Improvement of Meteorological Inputs for Air Quality Study

Emission gas from cooling tower. Cl* + Cl* Cl 2. 1 st : only a fraction of chlorine that is added to cooling tower can be emitted into the atmosphere

EVALUATION OF ATMOSPHERIC PROCESSES FOR OZONE FORMATION FROM VEHICLE EMISSIONS

Comparing Modal and Sectional Approaches in Modeling Particulate Matter in Northern California

ATOC 3500/CHEM 3152 Week 9, March 8, 2016

Inconsistency of ammonium-sulfate aerosol ratios with thermodynamic models in the eastern US: a possible role of organic aerosol

COPYRIGHT 2018 BY QI WANG

DISCOVER-AQ Houston as a case study for understanding spatial and temporal trends in urban particulate matter

Mobile Atmospheric Chemistry Laboratory

Who is polluting the Columbia River Gorge?

Reactive Nitrogen Monitoring

Ammonia Emissions and Nitrogen Deposition in the United States and China

CMAQ Modeling of Atmospheric Mercury

Response to Referee 2

SUPPORTING INFORMATION

Incorporating Space-borne Observations to Improve Biogenic Emission Estimates in Texas (Project )

Bases = Anti-Acids. The process is called neutralization (neither acidic nor basic) O H 3 2H 2

Lawrence Berkeley National Laboratory Lawrence Berkeley National Laboratory

Variability in ammonium nitrate formation and nitric acid depletion with altitude and location over California

Chasing Aerosol Particles Down to Nano Sizes

HIGH RESOLUTION CMAQ APPLICATION FOR THE REGIONAL MUNICIPALITY OF PEEL, ONTARIO, CANADA

An aromatic hydrocarbon study with an extended SAPRC99 mechanism of the CMAQ system: Application for the Houston-Galveston area

The Atmospheric Chemistry and Physics of Ammonia

(for tutoring, homework help, or help with online classes)

Figure S1. Configuration of the CIMS inlet during the KORUS-AQ 2016.

The Importance of Ammonia in Modeling Atmospheric Transport and Deposition of Air Pollution. Organization of Talk:

NO X emissions, isoprene oxidation pathways, and implications for surface ozone in the Southeast United States

CHAPTER 8. AEROSOLS 8.1 SOURCES AND SINKS OF AEROSOLS

NOAA s Air Quality Forecasting Activities. Steve Fine NOAA Air Quality Program

Source apportionment of fine particulate matter over the Eastern U.S. Part I. Source sensitivity simulations using CMAQ with the Brute Force method

Measuring Total Reactive N and its Composition

New Science Implementation in CMAQ-Hg: Test over a Continental United States Domain

Responsibilities of Harvard Atmospheric Chemistry Modeling Group

SCICHEM: A Puff Model with Chemistry. Part 2: Ozone and Particulate Matter

Review of the IMPROVE Equation for Estimating Ambient Light Extinction

CONTENTS 1 MEASURES OF ATMOSPHERIC COMPOSITION

Measurements of Ozone. Why is Ozone Important?

Tananyag fejlesztés idegen nyelven

REGIONAL AIR QUALITY FORECASTING OVER GREECE WITHIN PROMOTE

Air Monitoring. Semi-continuous determination of ambient air quality

Change of aerosol and precipitation in the mid troposphere over central Japan caused by Miyake volcano effluents

Effects of NO x control and plume mixing on nighttime chemical processing of plumes from coal-fired power plants

Supplementary Information

Wet plus Dry Deposition of Atmospheric Hg in the SE U.S. NADP Technical Meeting Sept , 2007 Boulder, CO, USA

Conceptual Model for Ozone in the Austin-Round Rock Metropolitan Statistical Area

FRAPPÉ/DISCOVER-AQ (July/August 2014) in perspective of multi-year ozone analysis

J1.7 IMPACT OF THE ON-ROAD AND MOBILE SOURCES ON THE BENZENE AND TOLUENE EMISSIONS AND CONCENTRATIONS IN THE HOUSTON-GALVESTON AREA

Figure 1. A terrain map of Texas and Mexico as well as some major cites and points of interest to the BRAVO study.

The Effect of Future Climate Change on Aerosols: Biogenic SOA and Inorganics

1.07 A FOUR MODEL INTERCOMPARISON CONCERNING CHEMICAL MECHANISMS AND NUMERICAL INTEGRATION METHODS

OVERVIEW OF CMAQ 5.0 AND CAMX 5.4 5/17/2012 1

Elements and Their Oxides

1. Introduction. JOURNAL OF GEOPHYSICAL RESEARCH: ATMOSPHERES, VOL. 118, , doi: /jgrd.50637, 2013

ATOC 3500/CHEM 3151 Air Pollution Chemistry Lecture 1

Bloomington in particular, the local soil contains CaCO 3, which donates a carbonate (CO 2-3 ) ion to help

Final Exam: Monday March 17 3:00-6:00 pm (here in Center 113) Slides from Review Sessions are posted on course website:

Thermodynamic characterization of Mexico City aerosol during MILAGRO 2006

Abstract. 1 Introduction

Know and Respond AQ Alert Service. Paul Willis SCOTTISH AIR QUALITY DATABASE AND WEBSITE ANNUAL SEMINAR Stirling 30 th March 2011

FINAL REPORT. Interactions between Organic Aerosol and NOy: Influence on Oxidant Production. AQRP Project

Chemical mass closure of atmospheric aerosol collected over Athens, Greece.

CH 221 Chapter Four Part II Concept Guide

H. M. Allen et al. Correspondence to: J. L. Fry

Air Quality Modelling for Health Impacts Studies

Highlights of last lecture

Regional Production, Quarterly report on the daily analyses and forecasts activities, and verification of the MATCH performances

TM4-ECPL model : Oceanic Sources for Oxygenated VOC and Aerosols

Transition from high- to low-no x control of night-time oxidation in the southeastern US

Project Summary. Sanford Sillman. is a way to evaluate the sensitivity to its two main precursors, nitrogen oxides (NO x

Analysis of Data from the 2009 SOOT Experiment

P2.11 THE LAKE SHADOW EFFECT OF LAKE BREEZE CIRCULATIONS AND RECENT EXAMPLES FROM GOES VISIBLE SATELLITE IMAGERY. Frank S. Dempsey

ClO + O -> Cl + O 2 Net: O 3 + O -> O 2 + O 2

Sources and Properties of Atmospheric Aerosol in Texas: DISCOVER-AQ Measurements and Validation

Evaluation of lower/middle tropospheric ozone from air quality models using TES and ozonesondes

Global Change and Air Pollution (EPA-STAR GCAP) Daniel J. Jacob

Ozone in the Atmosphere

NO X AT CAPE VERDE (CVO) Chris Reed, Katie Read, Luis Mendes, James Lee, Lucy Carpenter

Baseline Ozone in Western North America: Measurements and Models. David Parrish

model to data comparison over Europe for year 2001

Airborne Observations for Aerosol Model Assessment. Prepared by M. Kleb and G. Chen

AERMOD Sensitivity to AERSURFACE Moisture Conditions and Temporal Resolution. Paper No Prepared By:

2. Sketch a plot of R vs. z. Comment on the shape. Explain physically why R(z) has a maximum in the atmospheric column.

Transcription:

and its Importance in Ozone Non-attainment areas in Texas AQRP Project 10-015 Prepared for: Dr. Elena C. McDonald-Buller Texas Air Quality Research Program The University of Texas at Austin 10100 Burnet Rd. MC R7100, Austin, TX 78758 Prepared by: Bonyoung Koo and Greg Yarwood ENVIRON International Corporation 773 San Marin Drive, Suite 2115, Novato, CA, 94945 and James Roberts NOAA ESRL Chemical Sciences Division 325 Broadway, Boulder, CO 80305 January 31, 2012 ENVIRON Project Number: 06-25699D1

ACKNOWLEDGMENT The preparation of this report is based on work supported by the State of Texas through the Air Quality Research Program administered by The University of Texas at Austin by means of a Grant from the Texas Commission on Environmental Quality. i

TABLE OF CONTENTS EXECUTIVE SUMMARY... ES-1 1. INTRODUCTION... 1 1.1. Background... 1 1.2. Measurement Datasets... 1 1.3. Report Organization... 3 2. PLUME AND BOX MODEL ANALYSIS... 4 2.1. Selected Measurement Data... 4 2.2. Plume Analysis... 4 2.3. Box Model Chemical Mechanism... 5 2.4. Analysis of Reactive Chlorine Sources... 6 2.5. Aerosol Equilibrium Models... 6 2.6. Model Results and Discussion... 6 3. PHOTOCHEMICAL GRID MODELING... 32 3.1. Model Description... 32 3.2. Base and Sensitivity Cases... 33 3.3. Development of Chlorine Emissions from Swimming Pools... 34 3.4. Model Results and Discussion... 38 3.5. Testing Hypotheses... 56 4. SUMMARY AND RECOMMENDATIONS... 66 4.1. Box Model Analysis... 66 4.2. CAMx Photochemical Grid Modeling... 66 4.3. Future Work... 67 5. REFERENCES... 69 ii

LIST OF TABLES Table 1-1. Measurements used in the analysis of the CalNex dataset.... 3 Table 2-1. Plumes identified in the TexAQS 2006 ship board measurements and maximum mixing ratios in the plume.... 8 Table 2-2. The initial model conditions for Plume 1 (no NO 2 source case).... 16 Table 2-3. The initial model conditions for Plume 1 (2 ppbv/hr NO 2 case).... 17 Table 2-4. The initial model conditions for Plume 2 (no NO 2 source case).... 19 Table 2-5. The initial model conditions for Plume 2 (4 ppbv/hr NO 2 case).... 20 Table 3-1. Parameter values estimated by Bertram and Thornton (2009).... 33 Table 3-2. Average daily total chlorine/chloride emissions (tons Cl per day) for the sensitivity cases.... 34 Table 3-3. Per pool reactive chlorine emissions (kg Cl/pool-day).... 36 Table 3-4. Summer season average day swimming pool emission estimates (tons/day).... 37 iii

LIST OF FIGURES Figure 2-1. The top two panels show the ambient measurements versus time (UTC). The box model results are shown in the bottom panel, where time is hours since sunset. The high N 2 O 5 values at 07:37 UTC correspond to the model at 4.3 hours.... 10 Figure 2-2. The top two panels show the ambient measurements versus time (UTC). The box model results are shown in the bottom panel, where time is hours since sunset. The high ClNO 2 values at 12:16 PM UTC correspond to the model at 8.3 hours.... 11 Figure 2-3. The results of our box model plotted on top of Figure 15, from Chang et al. (2011). The red cross represents Case 1 and the green cross Case 2.... 12 Figure 2-4. The timeline of measurements made on the night of 5/19/2009. The box model was run for the period outlined by the gray box.... 15 Figure 2-5. Mixing ratios calculated by the box model using the initial conditions listed in Table 2-2. The gray line denotes roughly the time after sunset corresponding to the period outlined in Figure 2-4.... 16 Figure 2-6. Mixing ratios calculated by the box model using the initial conditions listed in Table 2-3. The gray line denotes roughly the time after sunset corresponding to the period outlined in Figure 2-4.... 17 Figure 2-7. The timeline of measurements made on the night of 5/29/2009. The box model was run for the period outlined by the gray box.... 18 Figure 2-8. Mixing ratios calculated by the box model using the initial conditions listed in Table 2-4. The gray line denotes roughly the time after sunset corresponding to the period outlined in Figure 2-7.... 19 Figure 2-9. Mixing ratios calculated by the box model using the initial conditions listed in Table 2-5. The gray line denotes roughly the time after sunset corresponding to the period outlined in Figure 2-7.... 20 Figure 2-10. Measurements of HNO 3, gas phase soluble chloride and particle NO 3 - and Cl - for Period 1 (top), and HNO 3 and gas phase soluble chloride for Period 2 (bottom).... 22 Figure 2-11. Panel (a) shows the timeline for HCl (red line) and aerosol soluble chloride for the PM 2.5 (blue circles) and PM 1 (green crosses) measured at the Pasadena site during CalNex 2010. Panel (b) shows the timeline for HNO 3 (green line) and aerosol soluble nitrate for the PM 2.5 (black circles) and PM 1 (red crosses). The particle mass and gas phase mixing ratio axes are scaled to be equivalent.... 24 Figure 2-12. Measurements of HCl and HNO 3 from the Pasadena site during CalNex 2010, colorcoded by day.... 25 Figure 2-13. PiLS measurements at Pasadena during CalNex 2010. Data courtesy of Rodney Weber, Georgia Tech. Figure courtesy of Jennifer Murphy, University of Toronto. 27 iv

Figure 2-14. Comparison of measured and model HCl and aerosol chloride for the CalNex Pasadena data. The data are from our laboratory (HCl) and R. Weber at Georgia Tech. The figure is courtesy of Trevor Vandenboer and Jennifer Murphy, University of Toronto.... 29 Figure 2-15. E-AIM results for aerosol run for the PiLS data and measured Total Chloride (HCl g + Cl - aq), with and without Na +, and run for just the AMS measurements. Figure courtesy of Jennifer Murphy, Univ. of Toronto.... 30 Figure 2-16. Partitioning of total chloride predicted by ISORROPIA for the CalNex 2010 ground site data that included Na+ measurements. Figure courtesy of Trevor VandenBoer and Jennifer Murphy.... 31 Figure 3-1. Maps of the Southern California Ozone Study (SCOS) and Houston-Galveston- Brazoria (HGB) sub-regions... 34 Figure 3-2. Residential swimming pool frequency by census tract.... 35 Figure 3-3. US swimming pool emissions of (a) Cl 2 and (b) HOCl.... 37 Figure 3-4. Harris County swimming pool emissions of (a) Cl 2 and (b) HOCl.... 38 Figure 3-5. Contributions and impact of additional chlorine/chloride emissions of each sensitivity case at the Pasadena station.... 39 Figure 3-6. Contributions and impact of additional chlorine/chloride emissions of each sensitivity case at the Moody Tower station.... 40 Figure 3-7. Box and whisker plots for modeled and observed concentrations of ClNO 2 at the Pasadena site during the simulation or observation periods.... 42 Figure 3-8. Box and whisker plots for modeled and observed concentrations of HCl at the Pasadena site during the simulation or observation periods.... 43 Figure 3-9. Box and whisker plots for modeled and observed concentrations of PCl at the Pasadena site during the simulation or observation periods.... 44 Figure 3-10. Box and whisker plots for modeled and observed concentrations of HNO 3 at the Pasadena site during the simulation or observation periods.... 45 Figure 3-11. Box and whisker plots for modeled and observed concentrations of ClNO 2 at the Moody Tower site during the simulation or observation periods.... 47 Figure 3-12. Box and whisker plots for modeled and observed concentrations of N 2 O 5 at the Moody Tower site during the simulation or observation periods.... 48 Figure 3-13. Box and whisker plots for modeled and observed concentrations of HCl at the Moody Tower site during the simulation or observation periods.... 49 Figure 3-14. Box and whisker plots for modeled and observed concentrations of HNO 3 at the Moody Tower site during the simulation or observation periods.... 50 Figure 3-15. ClNO 2 and N 2 O 5 concentrations measured near Barbour s Cut on (a) 9/2/2006 and (b) 9/8/2006 during TexAQS 2006.... 52 Figure 3-16. N 2 O 5 uptake coefficients and ClNO 2 yields estimated by CAMx (blue circle and red solid line) and Box model (black dotted line) at the Pasadena site.... 54 v

Figure 3-17. N 2 O 5 uptake coefficients and ClNO 2 yields estimated by CAMx (blue circle and red solid line) and Box model (black dotted line) at the Moody Tower site.... 55 Figure 3-18. Episode mean hourly concentrations from the 1 st hypothesis test run (10x chlorine emissions) along with those from the base case and observations at the Pasadena site.... 58 Figure 3-19. Episode mean hourly concentrations from the 2 nd hypothesis test run (acid displacement of deposited sea salt by HNO 3 deposition) along with those from the base case and observations at the Pasadena site.... 60 Figure 3-20. Episode mean hourly concentrations from the 3 rd hypothesis test run (HCl dry deposition without zero surface resistance assumption) along with those from the base case and observations at the Pasadena site.... 62 Figure 3-21. Episode mean hourly concentrations from the 4 th hypothesis test run (reactive coarse sea salt particles) along with those from the base case and observations at the Pasadena site.... 64 vi

EXECUTIVE SUMMARY Background and Project Objectives Results from the 2006 TexAQS/Gulf of Mexico Atmospheric Composition and Climate Study (GoMACCS) in Houston showed that reactions at night between ozone (O 3 ), nitrogen oxides (NOx), hydrogen chloride (HCl) and particulate matter (PM) gave rise to nitryl chloride (ClNO 2 ). This finding was confirmed by other studies and is significant because ClNO 2 undergoes rapid photolysis in the morning and can influence photochemistry and O 3 formation at the start of the day. Sea salt PM is an important source of chloride in coastal regions but ClNO 2 also has been observed far from the ocean (in Boulder, Colorado) indicating that other sources of chloride can give rise to ClNO 2 and that its influence on photochemistry may not be limited to coastal regions. The primary goals of this project are to analyze nitryl chloride formation in urban areas utilizing the existing field campaign data sets and to implement this chemistry in photochemical grid models to aid the Texas SIP development. Assessment of Nitryl Chloride Formation in Urban Areas Observations during the 2006 TexAQS study brought up a number of questions about whether or not the ClNO 2 chemistry was self-consistent, how the chemistry depended on N 2 O 5 uptake, and aerosol chloride concentration, if there was enough soluble chloride to produce the observed ClNO 2, and how these aspects of the chemistry could be incorporated into regional air quality models that describe ozone production in non-attainment areas. To answer these questions, the project team further examined the ambient data set acquired during the 2006 TexAQS study and assessed the data sets from the Study of Houston Atmospheric Radical Precursors (SHARP) in 2009, and the CalNex study conducted in the Los Angeles area in May and June of 2010. Several approaches were used to estimate N 2 O 5 uptake rate and ClNO 2 conversion efficiencies from ambient measurements: the odd-nitrogen budget of isolated nighttime plumes, and box modeling of the few reactions that govern N 2 O 5 formation and N 2 O 5 to ClNO 2 conversion. In addition, gas phase HCl, HNO 3 and particle nitrate and chloride measurements for these three studies were analyzed to assess the importance of sea salt acidification as a source of soluble aerosol chloride, and the adequacy of aerosol measurements in providing the information necessary to model ClNO 2 formation. ES-1

Three-Dimensional Photochemical Grid Modeling The Comprehensive Air-quality Model with Extensions (CAMx) photochemical grid model was updated to include a parameterization of the ClNO 2 chemistry and applied to a summer 2006 ozone and PM modeling episode using the EPA s nationwide 12-km grid modeling database. The reactive chlorine and particle chloride emission inventories in the EPA s modeling database were extended to include additional chlorine/chloride emission sources, e.g., swimming pools, sea salt and wildfires. The simulation results were compared with two ground site measurement datasets, the CalNex 2010 LA site at Pasadena and the SHARP 2009 Moody Tower site near downtown Houston. At the Pasadena site, CAMx significantly underestimates gaseous HCl and PM 2.5 chloride (PCl) while overestimating HNO 3 which may indicate a shortfall in the amount of chloride in the emission inventory. The missing chloride could be sea salt or additional chlorine and/or chloride emission categories that are missing in the current emission inventory. The missing chloride may also be attributed to the CAMx CF aerosol scheme used in this study where coarse mode sea salt is treated as an inert species (we have conducted a sensitivity simulation to test this hypothesis). At this site, almost all of the total chloride resides in the gas phase, which could also result in less chloride available to form ClNO 2 because HCl is efficiently removed from atmosphere by deposition process. Another possible explanation for the discrepancy between the model and observation is that there exist abundant sea salts deposited on the surface which release HCl by acid displacement following HNO 3 deposition. CAMx underpredicts HCl and N 2 O 5 at the Moody Tower site, but predicts similar to or higher ClNO 2 concentrations than the measurements. The observed ClNO 2 concentrations at this site are quite low compared to measurements made on board the NOAA R/V Ron Brown during the 2006 TexAQS campaign. On average, the Pasadena site observed lower HCl + PCl concentrations but higher ClNO 2 than the Houston site. Conclusions The results of ambient data analyses have illustrated several key features of the ClNO 2 chemistry. The highest ClNO 2 concentrations were observed when N 2 O 5 uptake coefficient was high but N 2 O 5 to ClNO 2 conversion efficiencies were fairly modest. Episodes when high N 2 O 5 was observed, but ClNO 2 was very low corresponded to low N 2 O 5 uptake and there was very low conversion. Relative humidity appears to be one of the more important parameters controlling N 2 O 5 uptake, but high aerosol organic fraction may also suppress uptake. The main source of chloride in the Los Angeles basin appears to be sea salt, and volatilization by other acids is the main source of HCl. For complete quantification of the chloride source, the large aerosol mass fraction must be measured. ES-2

A parameterized mechanism for ClNO 2 chemistry was implemented in the CAMx photochemical grid model. The base case simulation with updated chlorine/chloride emissions showed that the parameterized mechanism implemented in CAMx was able to simulate formation of ClNO 2 albeit underpredicting it. However, the model significantly underpredicted HCl and PCl. Several hypotheses were proposed and tested to explain the discrepancies between model and observations. The test results suggest that reacting a fraction of coarse mode sea salts (in addition to fine mode sea salts) in combination with a reduced dry deposition velocity of HCl would improve model performance. An alternate aerosol modeling scheme in CAMx (the CMU scheme) provides multiple approaches to modeling coarse mode aerosol components, and can be used to simulate reactions of coarse mode sea salt chloride. ES-3

1. INTRODUCTION 1.1. Background Results from the TexAQS 2006 field study in Houston showed that reactions at night between ozone (O 3 ), nitrogen oxides (NOx), hydrogen chloride (HCl) and particulate matter (PM) gave rise to nitryl chloride (ClNO 2 ). This finding was confirmed by other studies and is significant because ClNO 2 undergoes rapid photolysis in the morning and can influence photochemistry and O 3 formation at the start of the day. Sea salt PM is an important source of chloride in coastal regions but ClNO 2 also has been observed far from the ocean (in Boulder, Colorado) indicating that other sources of chloride can give rise to ClNO 2 and that its influence on photochemistry may not be limited to coastal regions. This study analyzed the ambient measurements made during TexAQS 2006, along with the other ambient measurement and laboratory chemistry studies pertinent to the Texas nonattainment areas, to provide the sound technical basis for the inclusion of this important chemistry in air quality models. This new chemistry was included in the Comprehensive Airquality Model with Extensions (CAMx) photochemical grid model that is used by the TCEQ for SIP modeling. The CAMx model was applied using a national modeling database that includes all of the field study locations. The emission inventories for the national database were reviewed and expanded to include as many sources of chloride as possible, including sea salt, HCl, molecular Cl 2 and PM chloride. Performance of the national CAMx model was assessed to evaluate the chemistry included for ClNO 2 and the completeness of the chloride emission inventory. 1.2. Measurement Datasets The datasets analyzed in this project were acquired during the TexAQS 2006/GoMACCS, Study of Houston Atmospheric Radical Precursors (SHARP) 2009, and CalNex 2010 field projects. The measurements of nitryl chloride and associated chloride/chlorine-containing species in these projects allow some comparison with box models and 3-D photochemical grid models. However, these datasets also had limitations with respect to chemical species not measured, or limited temporal coverage. This section will summarize the datasets, and point out limitations where applicable, as a guide to interpretation of model results and formulation of further research. 1

TexAQS II/GoMACCS The TexAQS II Gulf of Mexico Atmospheric Composition and Climate Study (hereafter shortened to TexAQS II) was conducted from July 27 to September 11, 2006 in the near shore Gulf of Mexico and in and around the Houston/Galveston region. The details of the experiment and key important findings are discussed by Parrish et al. (2009). This study was the setting for the first ever ambient measurements of ClNO 2 as reported by Osthoff et al. (2008), which were made on board the NOAA R/V Ron Brown. The measurements used in analysis have been described previously by Osthoff et al. (2008). This project involved the further analyses of that dataset, focusing on the observations in or near the Houston/Galveston area. SHARP 2009 Study Study of Houston Atmospheric Radical Precursors was conducted from April 15 to May 31, 2009 at the top of Moody Tower near downtown Houston. A description of the project can be found in the material located on the project website: http://sharp.hnet.uh.edu/sharp09/. Nitryl chloride measurements were reported for only the period between May 19 and May 31, 2009. Along with the standard urban pollution measurements, NOx, NOy, O 3, CO, there were also measurements of gas-phase soluble chloride (HCl) and nitrate (HNO 3 ) along with measurements of NO 3, N 2 O 5, and aerosol mass spectrometer (AMS) measurements. CalNex 2010 Study The part of the CalNex Study that will be analyzed in this report was conducted in and around the Los Angeles air basin from April 30 to June 22, 2010. Details of the study can be found on the project website: http://www.arb.ca.gov/research/calnex2010/calnex2010.htm. The two measurement platforms that had pertinent measurements were the NOAA WP-3 aircraft and the CalNex LA ground site that was located in Pasadena on the campus of Caltech. Measurements used in the analysis are listed in Table 1-1. Data Acknowledgements The TexAQS 2006 data were produced by the co-authors of Osthoff et al. (2008). The SHARP 2009 data were produced by Barry Lefer, Jack Dibb, Greg Huey, Renyi Zhang, Jose Jimenez and co-workers. The CalNex 2010 data were produced by Patrick Veres, Steve Brown, Barry Lefer, Hans Osthoff, Rodney Weber, Jose Jimenez, Jennifer Murphy, Reluca Ellis, Trevor VandenBoer, and co-workers. 2

Table 1-1. Measurements used in the analysis of the CalNex dataset. Platform Measurement Method Time Res. Det. Limit WP-3 NO, NO 2, NOy, O 3 Chemiluminescence 1 sec 10-40 pptv NO 3, N 2 O 5 Cavity ring-down 1 sec 1 pptv HNO 3 CIMS 1 sec 12 pptv PANs, ClNO 2 CIMS 2 sec 5, 50 pptv NH 3 CIMS 1 sec 200 pptv Aerosol Composition Aerosol mass 10 sec spectrometer Particle Surface Area Optical particle 1 sec 5 cm -3 counter Ground Site NO, NO 2, NOy, O 3 Chemiluminescence 1 min 20 pptv HCl, HNO 3 CIMS 1 min 25 pptv Aerosol Particle into liquid 6 min 40sec Composition>2.5mm sampler ClNO 2 CIMS 1 min 20 pptv Chemical ionization mass spectrometer. 1.3. Report Organization Section 2 describes the plume and box model analyses using selected datasets. Threedimensional photochemical grid model simulations with more comprehensive chlorine emission inventory are discussed in Section 3. Section 4 presents the summary and our recommendations. 3

2. PLUME AND BOX MODEL ANALYSIS A simple box model was used to explore the chemistry that forms ClNO 2 from N 2 O 5. The model involved only the chemical reactions that effect the formation of N 2 O 5 and its conversion to ClNO 2. Ambient data from select periods of the above-described datasets were used to initialize the model and were compared to the model results. The two main parameters of interest that came out of this analysis were N 2 O 5 uptake coefficients and the efficiency of conversion to ClNO 2 relative to the N 2 O 5 taken up. Ultimately these parameters will be related to the properties of the aerosol: surface area, deliquescence, and chloride content. These are all parameters that a 3-D air quality model will need to be able to describe to make an accurate simulation. 2.1. Selected Measurement Data Measurement periods were selected for box model analysis based on several criteria; the presence of significant ClNO 2, availability of supporting measurements, and in the case of TexAQS II, proximity to the Houston/Galveston area. Not all relevant parameters were available for each dataset, as will be pointed out in the respective results sections. 2.2. Plume Analysis The conversion of NO or NO 2 all the way to nitryl chloride can be described by a few chemical equations, which can then be used to interpret the chemical measurements. This analysis only yields meaningful results for well-defined plumes where, in the ideal case, it is known that that plume is initially emitted as NO. The equations simply describe the steps in the conversion of NO and O 3 to final products; ClNO 2 and HNO 3 : NO + O 3 NO 2 + O 2 NO 2 + O 3 NO 3 + O 2 NO 3 + NO 2 N 2 O 5 N 2 O 5 + aerosol φ ClNO 2 + (2-φ) HNO 3 (R1) (R2) (R3) (R4) where φ is the efficiency of ClNO 2 formation. Plots of the individual species versus total NOy for the entire plume give an estimate of φ. For a discrete, well-correlated plume, this gives 4

reasonable results, but can involve some uncertainty if there is a significant primary emission of NO 2 for example. 2.3. Box Model Chemical Mechanism The box model describes the NOx reactions that form N 2 O 5 and then has 2 reactions of N 2 O 5 that form either two nitric acid molecules or nitryl chloride and nitric acid molecule (Brown et al., 2004; Osthoff et al., 2008). NO + O 3 NO 2 + O 2 NO 2 + O 3 NO 3 + O 2 NO 3 + NO 2 N 2 O 5 N 2 O 5 + aerosol 2 HNO 3 N 2 O 5 + aerosol ClNO 2 + HNO 3 NO 3 + reactant HNO 3 NO + NO 3 NO 2 Source NO 2 (optional) (R1) (R2) (R3) (R4a) (R4b) (R5) (R6) (R7) The rates of the gas phase reactions are well known (NIST, IUPAC references), and were adjusted for known temperature dependences, using measurements. The key reactions that are most sensitive to the parameters we would like to know are Reactions 4a and 4b. These reactions are written as second order above, but in fact are put in the actual model as first order reactions: N 2 O 5 2HNO 3 N 2 O 5 ClNO 2 + HNO 3 (R4a ) (R4b ) The first order rate constants k 4a and k 4b are then relatable to the uptake rate and ClNO 2 production efficiency as the following: Uptake rate = k 4a + k 4b ClNO 2 formation efficiency = k 4b / (k 4a + k 4b ) (Eq1) (Eq2) The net uptake rate of N 2 O 5 can be related to fundamental quantities that are needed for an accurate 3-D model by the following expression: 5

Uptake rate = γωs a / 4 (Eq3) where S a is the total aerosol surface area, γ is the N 2 O 5 uptake coefficient, which is essentially the fraction of collisions of N 2 O 5 with the aerosol surface that leads to reaction, and ω is the mean molecular speed of N 2 O 5. 2.4. Analysis of Reactive Chlorine Sources The main source of chloride/chlorine that drives ClNO 2 production may be gas-phase HCl rather than particle chloride. This HCl can originate from a variety of sources: acidification of sea salt or dust (crustal material, road salt) and direct emission from combustion sources (power plants, cement kilns). For the purposes of this report, the box model and ambient HCl results will be discussed in the context of particle chloride sources, and the 3-D model will include a more complete source inventory. 2.5. Aerosol Equilibrium Models Several open-source aerosol solution models were used in this work to explore the equilibrium between gas-phase HCl and particle-phase chloride. This equilibrium is essential to understanding the efficiency of ClNO 2 formation chemistry since laboratory studies have measured the chloride-concentration dependence of this process. The two aerosol chemistry models are the extended Aerosol Inorganics Model (Clegg et al., 1998; Friese and Ebel, 2010) (http://www.aim.env.uea.ac.uk/aim/aim.php), and ISORROPIA (Fountoukis and Nenes, 2007) (http://nenes.eas.gatech.edu/isorropia/). These models were applied to the CalNex 2010 data set since that constituted the most complete set of measurements of the relevant gas and particle phase species. 2.6. Model Results and Discussion Plume Model Results Discrete plumes were observed in the TexAQS 2006 data that are amenable to analysis by the odd nitrogen (NO y ) and odd oxygen (O x ) budget methods described by Brown et al. (2004; 2006). This analysis can estimate the net yields of the reaction N 2 O 5 + H 2 O (aer) ClNO 2 + HNO 3 if we assume that this is the fastest loss for NO 3 and N 2 O 5 in this environment and the coproduced HNO 3 stays in the aerosol. These are both reasonable starting points for this environment (high humidity, high surface area). The analysis relies on the correlation of the successive sums of individual NOy species with total NOy, hence requires reasonably well- 6

defined plumes. Also, this part of analysis does not yield an uptake coefficient. Table 2-1 lists the plumes, their characteristics and the resulting reaction efficiency that were identified in the TexAQS 2006 data. Efficiencies spanned the range between 6 and 88%. 7

Table 2-1. Plumes identified in the TexAQS 2006 ship board measurements and maximum mixing ratios in the plume. Index Start Stop likely origin Age (hrs.) [NO 2 ] (ppb) [NO 3 ] (ppt) [N 2 O 5 ] (ppt) [ClNO 2 ] (ppb) 1 July 29 05:07-05:16 ship 1.8 28 37 2.0 n/a 30 2 July 30 03:54-04:13 ship 8.4 2.1 6 6 0.65 n/a 18 3.3 22 61 ±17 3 Aug 01 08:58-09:21 uncertain 7.9 1.1 9 4 0.50 n/a 23 1.6 32 59 ±15 4 July 30 02:41-02:46 ship 2.8 5 7 0.27 n/a 19 3.3 18 5 Aug 04 07:33-07:42 ship 3.1 3.2 26 56 0.55 n/a 34 0.8 32 57 ±20 6 Aug 04 11:21-11:27 urban 6.2 13 48 0.56 n/a 30 0.77 14 7 Aug 08 02:10-02:14 ship 1.8 37 34 0.46 2.1 24 14 8 Aug 09 05:55-05:55 uncertain 3.1 1.5 38 38 0.16 2.2 28 13 26 ±12 9 Aug 16 03:20-03:30 uncertain 13 1.2 15 18 0.46 1.3 18 0.8 19 10 Aug 16 11:29-11:32 urban 13 31 0.46 5.0 22 0.45 8.6 11 Aug 18 07:37-07:48 ship 4.9 64 47 0.53 2.2 61 0.064 8.6 17 ±4 12 Aug 18 10:53-11:00 ship 2.1 57 102 0.69 5.2 51 0.08 1.4 57 ±19 13 Aug 18 10:31-10:39 ship 35 39 0.61 7.6 51 0.09 1.4 14 Aug 23 02:54-03:14 urban 8 30 77 0.95 3.0 45 0.12 1.4 6 ±1 15 Aug 25 02:48-03:00 ship 2.4 2.0 41 73 0.65 2.3 25 0.62 5.3 60 ±36 16 Aug 25 10:25-10:34 ship 3.4 29 28 0.44 4.8 24 1.6 12 17 Aug 30 06:27-06:35 urban 4.2 1.7 10 18 0.72 2.2 24 0.97 12 88 ±33 18 Sept 01 01:30-01:51 ship 2.4 2.2 150 161 0.59 2.8 71 0.041 8.4 19 ±5 19 Sept 01 08:38-09:10 uncertain 3.1 1.0 162 117 0.58 1.3 70 0.044 0.4 18 ±4 20 Sept 01 11:25-11:42 urban 3.9 0.9 143 174 0.61 1.1 65 0.049 0.4 16 ±4 23 Sept 02 01:59-02:09 urban 21 26 168 0.75 25 67 0.049 4.8 24 Sept 03 01:46-01:50 oil platform 19.5 25 147 0.53 24 83 0.006 4.8 25 Sept 03 05:43-05:45 ship 2.2 2.2 132 241 0.53 2.9 72 0.015 1.1 25 ±7 27 Sept 04 04:09-04:22 ship 3.5 2.2 220 211 0.51 3.4 67 0.22 1.1 13 ±3 28 Sept 04 01:39-01:42 uncertain 9.9 125 467 0.50 12.7 70 0.11 2.8 30 Sept 04 09:40-09:53 ship 3.3 0.8 85 95 0.61 1.6 57 0.48 2.8 17 ±4 31 Sept 08 03:00-03:29 uncertain 8.3 168 934 2.3 10.3 67 0.03 2.8 32 Sept 08 02:14-02:33 ship 7.4 177 1190 2.1 9.0 70 0.028 2.0 33 Sept 08 03:32-03:36 urban 1.7 8.0 63 301 0.89 9.3 66 0.031 2.0 26 ±9 35 Sept 09 07:36-07:40 ship 1.3 7.5 20 105 0.80 8.6 56 0.21 5.0 66 ±31 36 Sept 09 07:44-07:47 ship 2 2.6 17 44 0.44 3.0 55 0.27 5.0 30 ±10 37 Sept 09 08:36-08:39 ship 1.3 7.7 24 139 0.67 8.7 55 5.0 56 ±28 38 Sept 09 11:33-11:46 ship 0.85 7.5 18 101 0.38 8.5 55 0.021 5.0 31 ±22 39 Sept 11 08:42-08:57 ship 3.6 5.0 54 127 0.90 5.4 23 62 ±28 40 Sept 11 05:35-05:49 ship 2.4 5.9 31 165 0.99 6.2 18 68 ±66 41 Sept 11 07:36-07:45 urban 5.8 1.4 67 94 0.57 1.7 26 32 ±8 1 The O 3 mixing ratio just outside the plume. 2 The fraction of sea salt surface area to total aerosol surface area. [NO y ] (ppb) [O 3 ] 1 (ppb) [Cl - ] (M) F 2 (%) φ 8

Box Model Results The analysis of the CalNex 2010 WP3 aircraft data using the box model considered two contrasting cases. In this analysis, uptake rates were adjusted to obtain the best fit to the observations. The results showed a range of N 2 O 5 uptake rates and N 2 O 5 ClNO 2 conversion efficiencies. The low humidity, high organic aerosol case (Case 1) is shown in Figure 2-1. The high humidity, high inorganic aerosol case is shown in Figure 2-2. The model results for these cases are consistent with moderate N 2 O 5 uptake coefficient (4.4 x10-3 ) and low conversion efficiency (6%) in Case 1 and a relatively high uptake coefficient (2.8 x10-2 ) and moderate conversion efficiency (44%) in Case 2. Although we have not done a formal sensitivity study, these results are fairly robust in the sense that only parameters close to the ones quoted above yield N 2 O 5 and ClNO 2 mixing ratios anywhere close to the ambient measurements. There have been a number of studies aimed at developing parameterizations for N 2 O 5 uptake, as compiled in recent review by Chang et al. (2011). The results of our model cases are plotted along with other models as a function of RH in Figure 2-3. 9

Figure 2-1. The top two panels show the ambient measurements versus time (UTC). The box model results are shown in the bottom panel, where time is hours since sunset. The high N 2 O 5 values at 07:37 UTC correspond to the model at 4.3 hours. 10

Figure 2-2. The top two panels show the ambient measurements versus time (UTC). The box model results are shown in the bottom panel, where time is hours since sunset. The high ClNO 2 values at 12:16 PM UTC correspond to the model at 8.3 hours. 11

Figure 2-3. The results of our box model plotted on top of Figure 15, from Chang et al. (2011). The red cross represents Case 1 and the green cross Case 2. 12

The SHARP 2009 effort involved approximately 10 days of intensive measurements, of which there are 2 days of ClNO 2 measurements that were over 100 pptv and have concurrent N 2 O 5 measurements. We applied the N 2 O 5 ClNO 2 box model to determine the net N 2 O 5 uptake coefficients and effective ClNO 2 yields that are implied by the observations. As described previously, we obtained the net uptake coefficient of N 2 O 5 from the first order loss rates that best fit the data (as described by Osthoff et al., 2008), and the ClNO 2 yield from the ratio of the rate going to ClNO 2 and HNO 3, to the total rate, the other channel being the reaction of N 2 O 5 to form 2HNO 3. The two nights that were analyzed were 5/19/2009 and 5/29/2009. (a) 5/19/2009 The timeline of measurements from this night, averaged to 60 seconds, is shown in Figure 2-4. The portion that was compared to the box model is highlighted by a gray rectangle. This period was selected because it had the highest N 2 O 5 on the night, and therefore would give the best signal with which to test the chemistry. In addition, the highest ClNO 2 value was observed and other related parameters (NO 2 and O 3 ) were relatively stable during that time period, which provided a basis with which to estimate a starting point for the box model. The NO, NO 2, O 3 measurements just prior to this period were used as initial conditions for the model (Table 2-2). The results of the model for this case are shown in Figure 2-5. The first order rate constants for the N 2 O 5 uptake reactions give a ClNO 2 efficiency of roughly 3.3%, and a net uptake coefficient of 0.052, assuming an aerosol surface area of 200 µm 2 /cm 3. One result of the box model is that the calculated NO 2 was about a factor of 3 lower than the observations at that time. One reason for this could be that the site, even though above ground, is impacted by urban NOx sources that continue throughout the night (e.g., vehicle exhaust emissions from surface roads). While they emit mostly NO, our simple box model does not handle a continuous NO source because it rapidly destroys NO 3 (NO + NO 3 2 NO 2 ). Instead, we use a continuous source of NO 2, assuming that by the time these emissions reach the altitude for the Moody Tower NO has been converted to NO 2 by ambient O 3. The box model loses NO 2 through eventual conversion to HNO 3 (and ClNO 2 ), which results in unrealistically low NO 2 concentrations at the times selected for analysis. To compensate this loss, a source of NO 2 was added to the box model by adding a slow first order reaction that forms NO 2 from a large reservoir, termed SO in the reaction table. This effectively yields a constant source of NO 2 that can be adjusted by slight adjustment of the rate constant. In practice, this source is in the range of a few ppbv/hr, which seems like a reasonable number, and effectively maintains the ambient NO 2 mixing ratio close to the measured values. The model result that gives the best simulation of the May 19 period is shown in Figure 2-6. These model conditions imply a slightly lower ClNO 2 formation efficiency (2.5%) and a higher N 2 O 5 uptake coefficient, 0.1. The NO 2 source brings the model NO 2 closer to that measured. (b) 5/29/2009 The second period of interest in the SHARP 2009 dataset occurred on May 29, and is shown in Figure 2-7. In this case the highest ClNO 2 occurred later in the night, almost 8 hours after sunset. The model results for this period are shown in Figure 2-8 and Figure 2-9 for 13

the straight model and NO 2 source case, respectively. The ClNO 2 efficiencies implied by these model results were low (0.5-1%) and the N 2 O 5 uptake coefficients were quite high (0.04 to 0.1) for the two cases. As above, an NO 2 source (4 ppbv/hr in this case) gave a better fit to the NO 2 ambient measurements. 14

Figure 2-4. The timeline of measurements made on the night of 5/19/2009. The box model was run for the period outlined by the gray box. 15

Table 2-2. The initial model conditions for Plume 1 (no NO 2 source case). Figure 2-5. Mixing ratios calculated by the box model using the initial conditions listed in Table 2-2. The gray line denotes roughly the time after sunset corresponding to the period outlined in Figure 2-4. 16

Table 2-3. The initial model conditions for Plume 1 (2 ppbv/hr NO 2 case). Figure 2-6. Mixing ratios calculated by the box model using the initial conditions listed in Table 2-3. The gray line denotes roughly the time after sunset corresponding to the period outlined in Figure 2-4. 17

Figure 2-7. The timeline of measurements made on the night of 5/29/2009. The box model was run for the period outlined by the gray box. 18

Table 2-4. The initial model conditions for Plume 2 (no NO 2 source case). Figure 2-8. Mixing ratios calculated by the box model using the initial conditions listed in Table 2-4. The gray line denotes roughly the time after sunset corresponding to the period outlined in Figure 2-7. 19

Table 2-5. The initial model conditions for Plume 2 (4 ppbv/hr NO 2 case). Figure 2-9. Mixing ratios calculated by the box model using the initial conditions listed in Table 2-5. The gray line denotes roughly the time after sunset corresponding to the period outlined in Figure 2-7. 20

Reactive Chlorine Sources (a) TexAQS 2006 The HCl and aerosol chloride measurements from TexAQS 2006 that pertained to ClNO 2 formation were discussed by Osthoff et al. (2008). The chloride measurements were long duration impactor measurements hence are of limited utility in assessing the HCl chloride particle chemistry. It is worth noting that the laboratory study of Roberts et al. (2009) showed that median chloride concentrations found in those samples corresponded to relatively efficient ClNO 2 formation, 40%. (b) SHARP 2009 The SHARP 2009 project had measurements of soluble gas-phase chloride (thought to be HCl) and sub-micron volatile aerosol chloride, and those are shown in Figure 2-10 for the two periods that were analyzed by the box model, although AMS measurements were only available for the first period (5/19/2009). Those measurements show that there were only modest amount of aerosol chloride present, consistent with a low N 2 O 5 to ClNO 2 conversion efficiency. 21

Figure 2-10. Measurements of HNO 3, gas phase soluble chloride and particle NO 3 - and Cl - for Period 1 (top), and HNO 3 and gas phase soluble chloride for Period 2 (bottom). 22

(c) CalNex 2010 The HCl and aerosol soluble chloride measured at the Pasadena ground site during CalNex 2010 tells the story of why ClNO 2 is so abundant in the LA Air Basin. The timelines for gas phase HCl and HNO 3 and PM 2.5 chloride and nitrate are shown in Figure 2-11. Also shown is a brief period at the end of the project during which an inlet with a 1 micron cutoff was alternately used on the system. The aerosol chloride is often abundant enough to provide efficient ClNO 2 formation. Moreover, mid-day mixings ratios are broadly indicative of the gas phase HCl concentrations in the daytime boundary layer and therefore the next night s residual layer, so there is often a substantial pool of soluble chloride available for nighttime reaction. There are several clues as to the source of the chloride and HCl. The fact that what data we do have on PM 1 chloride is substantially lower than PM 2.5 indicates a sea salt or soil source for chloride. The gas phase HNO 3 and particle nitrate are much closer to each other in magnitude and temporal profile, indicating that they are readily exchanging in this environment, a feature of LA aerosol chemistry that has been known for some time. In addition, there are tight correlations between gas phase HCl and HNO 3 the slopes of which vary a great deal from day-to-day (Figure 2-12). This is consistent with aerosol acidification by HNO 3 leading to HCl volatilization from sea salt particles. 23

Figure 2-11. Panel (a) shows the timeline for HCl (red line) and aerosol soluble chloride for the PM 2.5 (blue circles) and PM 1 (green crosses) measured at the Pasadena site during CalNex 2010. Panel (b) shows the timeline for HNO 3 (green line) and aerosol soluble nitrate for the PM 2.5 (black circles) and PM 1 (red crosses). The particle mass and gas phase mixing ratio axes are scaled to be equivalent. 24

Figure 2-12. Measurements of HCl and HNO 3 from the Pasadena site during CalNex 2010, colorcoded by day. 25

The partitioning of chloride between gas phase HCl and particle chloride is a critical aspect of ClNO 2 chemistry since the efficiency of the N 2 O 5 chemistry is dependent on the effective aqueous concentration of chloride. This partitioning can be predicted by models of aerosol particle thermodynamics and compared to ambient measurements as a test of model effectiveness. Thermodynamic models rely in part on a correct charge balance in solution. Figure 2-13 shows the major anions and cations for the PM 2.5 measurements at CalNex Pasadena. Although cations were only measured during the last 10 days or so, it can be seen that Na + was a large contributor to the cations. This confirms that even the smaller aerosol size fraction had a substantial sea salt contribution, and also implies that any chloride associated with Na + would be soluble, but probably not volatile enough to be measured by the standard Aerosol Mass Spectrometry (AMS) technique. This is an important observational result that should be kept in mind when examining other data sets or designing future experiments. 26

-3 ) mole eq uivalent s (µeq m January 31, 2012 0.6 0.5 0.4 0.3 PILS PM 2.5 Cl - NO - 3 SO 2-4 Na + NH 4 + 0.2 0.1 0.0 5/21/2010 5/26/2010 5/31/2010 6/5/2010 6/10/2010 Date (PDT) Figure 2-13. PiLS measurements at Pasadena during CalNex 2010. Data courtesy of Rodney Weber, Georgia Tech. Figure courtesy of Jennifer Murphy, University of Toronto. 27

The prediction of gas phase HCl and aerosol phase chloride with both the E-AIM and ISORROPIA models is really quite good when the measured sodium concentrations are included. This is shown in Figure 2-14 for a three day period at the end of the CalNex Pasadena campaign. A major effect of the inclusion of Alkali metal cations (and alkaline earth cations when present) is to increase the ph of the particles. This is shown in Figure 2-15 for the CalNex ground site data. The modeled ph is often 0.3 ph units higher when Na + was included. Moreover, ph predicted from PiLS data, with or without Na + included, was always higher than that predicted from AMS data, sometimes by more than 0.5 ph units. The effect of ph on chloride partitioning can be seen in Figure 2-16, which shows the fraction of total chloride predicted to be in the gas phase versus relative humidity and as a function of ph. As substantial amount of Cl -, more than 50% of the total, is predicted to be in the particle phase even at the lowest ph (ph 3.2). This result is essentially in agreement with the box model results from the CalNex aircraft measurements, in which the highest ClNO 2 formation was observed at the highest RH. Thus high RH (>85%) has the dual effect of not only increasing N 2 O 5 uptake, but also increasing the concentration of particle chloride, for a given amount of total chloride. 28

Figure 2-14. Comparison of measured and model HCl and aerosol chloride for the CalNex Pasadena data. The data are from our laboratory (HCl) and R. Weber at Georgia Tech. The figure is courtesy of Trevor Vandenboer and Jennifer Murphy, University of Toronto. 29

Figure 2-15. E-AIM results for aerosol run for the PiLS data and measured Total Chloride (HCl g + Cl - aq), with and without Na +, and run for just the AMS measurements. Figure courtesy of Jennifer Murphy, Univ. of Toronto. 30

Figure 2-16. Partitioning of total chloride predicted by ISORROPIA for the CalNex 2010 ground site data that included Na+ measurements. Figure courtesy of Trevor VandenBoer and Jennifer Murphy. 31

3. PHOTOCHEMICAL GRID MODELING 3.1. Model Description The Comprehensive Air-quality Model with Extensions (CAMx) modeling system is a publicly available, state-of-science One-Atmosphere photochemical grid model capable of addressing ozone, PM, visibility and acid deposition at regional scale (ENVIRON, 2010). TCEQ is currently using CAMx for their SIP modeling. The Carbon Bond 2005 (CB05) chemistry mechanism of the latest CAMx code (version 5.4) was updated with nitryl chloride chemistry based on the parameterization by Bertram and Thornton (2009). The heterogeneous reaction of N 2 O 5 with either liquid water or HCl on aerosol surfaces is implemented in CAMx as a pseudo gas-phase reaction: N 2 O 5 (g) + H 2 O(l) 2 HNO 3 (g) N 2 O 5 (g) + HCl(g) ClNO 2 (g) + HNO 3 (g) The pseudo first-order rate coefficient (k het ) for the heterogeneous decay of N 2 O 5 is determined as follows: k het = 1 d[n 2 O 5 ] = γωs a [N 2 O 5 ] dt 4 (Eq3) where γ is the reactive uptake coefficient for N 2 O 5, S a is the particle surface area concentration, and ω is the mean molecular speed of N 2 O 5. Based on laboratory experiments, Bertram and Thornton (2009) proposed the following parameterization: γ = Ak 2f 1 k 2f = β βe ( δ[h 2O]) Y ClNO2 = ClNO 2 N 2 O 5 1 k 3[H2O] k 2b [NO +1+ k 4[Cl ] 3 ] k 2b [NO 3 ] = 1 + k 3[H 2 O] k 4 [Cl ] 1 (Eq4) (Eq5) (Ep6) The estimated values for the parameters are presented in Table 3-1. Gas-particle equilibrium partitioning of chloride is modeled by ISORROPIA version 1.6 (Nenes et al., 1998, 1999) in CAMx. 32

Table 3-1. Parameter values estimated by Bertram and Thornton (2009). Parameter Value A [s] 3.2 10-8 β [s -1 ] 1.15 10 6 ± 3 10 5 δ [M -1 ] 1.3 10-1 ± 5 10-2 k 3 /k 2b 6.0 10-2 ± 1.0 10-2 k 4 /k 2b 29 ± 6 Median values are used in CAMx. 3.2. Base and Sensitivity Cases The updated CAMx was applied to a month-long episode (July 2006) using the current EPA 2006 US modeling database. The modeling domain, which covers entire contiguous US, consists of 459 by 299 horizontal grid cells of 12-km resolution and 26 vertical layers. The base case simulation was conducted without any chlorine/chloride emission. The first sensitivity case (S1) adds reactive chlorine (Cl 2 and HCl) emissions from the EPA 2006 modeling database which were based on EPA s hazardous air pollutant (HAP) emission inventories. The EPA modeling database was originally developed for the Community Multiscale Air Quality (CMAQ) modeling system which internally models sea salt generation from ocean, and thus does not include sea salt emissions. ENVIRON has added emissions of sea-salt particles (speciated into sulfate, sodium, and chloride) estimated from the hourly, gridded meteorological data using flux equations for open ocean (Smith and Harrison 1998; Gong, 2003) and breaking waves in the surf zone (de Leeuw et al., 2000). The second sensitivity case (S2) includes all the chlorine/chloride emissions of S1, and adds chlorine emissions from swimming pools which may provide significant contribution to total chlorine emissions in urban areas. ENVIRON developed average summer day chlorine (Cl 2 and HOCl) emissions from swimming pools for the 48 contiguous states, which will be described in the following section. The last sensitivity case (S3) retains all the chlorine/chloride emissions of S2, and adds particulate chloride (PCl) emissions that EPA developed from primary PM 2.5 emissions of the national emission inventory (NEI) using the speciation profiles by Reff et al. (2009) as well as the PCl emissions from wildfires. Table 3-2 presents average daily total emissions of reactive chloride species for the Southern California Ozone Study (SCOS) and Houston-Galveston-Brazoria (HGB) sub-regions (each subregion is defined as shown in Figure 3-1) and the contiguous US domain. 33

Table 3-2. Average daily total chlorine/chloride emissions (tons Cl per day) for the sensitivity cases. Region S1 case S2 case S3 case Area Point Pool Area Point Fire SCOS HGB Cl 2 HCl HOCl PCl Cl 2 HCl HOCl PCl Contiguous US Cl 2 HCl HOCl PCl 0.85 1.5 65 0.34 1.0 15 0.025 1.4 0.97 15 2.1 2.1 0.65 0.65 0.65 0.064 4.3 1.8 1.1 0.014 4.8 84 8.6 1,168 31 31 5,359 127 21 241 Emissions shown for each sensitivity case represent the amount added for the case. (a) SCOS sub-region (b) HGB sub-region Figure 3-1. Maps of the Southern California Ozone Study (SCOS) and Houston-Galveston- Brazoria (HGB) sub-regions. 3.3. Development of Chlorine Emissions from Swimming Pools Reactive chlorine emissions from swimming pools were estimated for each county in the US based on the estimated number of public and residential swimming pools in each county and the estimated emissions per pool on a per day basis. 34

Pool Counts The number of public swimming pools in each county was estimated according to one of two methods: 1) For Los Angeles County and Harris County, current residential pool locations were obtained in GIS format from each county. These data were then multiplied by the ratio of 2005 to 2011 population in each county to generate an estimate of the total number of pools in each county. Figure 3-2 portrays pool counts by census tract. 2) For all other counties in the 48 contiguous states, an estimate of the total number of pools in the US of 8.6 million (SBI, 2007) was disaggregated to the county level based on the EPA 1999 NEI (http://www.epa.gov/ttnchie1/net/1999inventory.html) estimates of chloroform emissions from swimming pools in each county. (a) Harris County (b) Los Angeles County Figure 3-2. Residential swimming pool frequency by census tract. Emission Rate Following Sarwar and Bhave (2007) and Chang et al. (2002), three methodologies were used for reactive chlorine emission factors from swimming pools: 35

1) Mass Balance: Based on rate of chlorine addition, assuming the addition of NaOCl in the amount of 2 gallons per week per pool which yields a potential of 0.051 kg of Cl emissions per day per residential pool. 2) Mass Transfer Coefficient: Based on estimates of an overall mass transfer coefficient for chlorine of 10-100 micrograms HOCl per meter squared per minute from Rogozen (1998). 3) Free Chlorine: Based on an estimate of an average of 2 ppm chlorine mass fraction. These methods rely upon different emission factors and assumptions and may be expected to produce different emission estimates. A broad range of chlorine emission estimates are presented in Table 3-3 which represent likely bounds of average swimming pool emissions. The mass transfer method was selected, because it is most closely related to measurements, and used to estimate swimming pool emissions with the high-end mass transfer coefficient. Table 3-3. Per pool reactive chlorine emissions (kg Cl/pool-day). Public Pool Residential Pool Chlorine Addition Rate Method 0.108 0.051 Mass Transfer Method 0.0029-0.029 0.0007 0.007 Free Chlorine Stock Method 0.5 0.12 Speciation Previously, Chang et al. (2002) assumed all chlorine emissions were Cl 2 while Sarwar and Bhave (2007) assumed all emissions were HOCl; Chang et al. (2006) assumed equal amounts of chlorine released as HOCl and Cl 2. In the absence of definitive data, it was assumed that reactive chlorine emissions were split equally between HOCl and Cl 2. Spatial Allocation to Grid Cells Spatial allocation from the county level to the grid cell level was made based on EPA s population surrogate. 36

Temporal Previously Chang et al. (2006) assumed hourly emissions are evenly distributed from 12:00-20:00. Based on the assumption that emissions would be highest in the morning hours subsequent to chlorine addition, but recognizing that chlorine may be released throughout the day, 50% of emissions have been uniformly allocated between 12:00 and 20:00 and 50% of the emission were allocated uniformly to the remaining hours. Results Table 3-4 shows chlorine emission estimates for Harris and Los Angeles Counties which represent about 1.3% and 2.5% of all 48 contiguous state swimming pool emissions, respectively, along with emissions from all other counties and 48 state totals. As expected, emissions are concentrated in urban areas as shown in Figure 3-3. Figure 3-4 shows emissions by grid cell for Harris County, Texas. Table 3-4. Summer season average day swimming pool emission estimates (tons/day). County Cl 2 HOCl Los Angeles County, California 0.76 1.15 Harris County, Texas 0.40 0.59 All Other Counties 29.77 44.05 48 contiguous state total 30.94 45.79 (a) Cl 2 (b) HOCl Figure 3-3. US swimming pool emissions of (a) Cl 2 and (b) HOCl. 37

(a) Cl 2 (b) HOCl Figure 3-4. Harris County swimming pool emissions of (a) Cl 2 and (b) HOCl. 3.4. Model Results and Discussion We first focused our analysis on the results at two sites: The Pasadena ground site (CalNex 2010) and the Moody Tower site at ~70 m above ground (SHARP 2009). These sites provide continuous measurement data for a relatively longer period compared to other plume measurement data, allowing more meaningful evaluation of the grid model results. Figure 3-5 and Figure 3-6 show impact of additional chlorine/chloride emissions of each sensitivity case on the episode average concentrations of total chloride (HCl and PCl) and nitryl chloride (ClNO 2 ) along with chlorine emission contributions for each case. Anthropogenic reactive chlorine emissions and sea salt chloride (S1) dominate at both sites. At the Pasadena site, adding chlorine emissions from swimming pools (S2) resulted in a 28% increase in the episode average total chloride concentration but marginal increase (2%) in ClNO 2. Anthropogenic and wildfire PCl emissions (S3) increased total chloride and ClNO 2 concentrations by 4.4% and 18%, respectively. At this site, almost all of the total chloride resides in the gas phase, which could result in less chloride available to form ClNO 2 because HCl is efficiently removed from atmosphere by deposition process. At Moody Tower, on the other hand, a significant fraction of total chloride was predicted to be in the aerosol phase. Swimming pool chlorine emissions (S2) didn t have much impact on the episode average concentrations of total chloride (1.8%) and nitryl chloride (0.34%) at this site while increases in these concentrations between the S2 and S3 simulations are more noticeable (12% for total chloride; 66% for nitryl chloride). On average, the Pasadena site observed lower HCl + PCl concentrations but higher ClNO 2 than the Houston site. 38

(a) Average daily total Cl emissions over the SCOS sub-region (b) Modeled episode average HCl & PCl concentrations at the Pasadena site (c) Modeled episode average ClNO 2 concentrations at the Pasadena site Figure 3-5. Contributions and impact of additional chlorine/chloride emissions of each sensitivity case at the Pasadena station. 39

(a) Average daily total Cl emissions over the HGB domain (b) Modeled episode average HCl & PCl concentrations at the Moody Tower site (c) Modeled episode average ClNO2 concentrations at the Moody Tower site Figure 3-6. Contributions and impact of additional chlorine/chloride emissions of each sensitivity case at the Moody Tower station. 40