Atmospheric Environment 171 (2017) 317–329
Contents lists available at ScienceDirect
Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv
Strong influence of deposition and vertical mixing on secondary organic aerosol concentrations in CMAQ and CAMx
MARK
Qian Shua, Bonyoung Koob, Greg Yarwoodb, Barron H. Hendersona,∗ a b
University of Florida, Engineering School of Sustainable Infrastructure and Environment, United States Ramboll Environ, Novato, CA, United States
A R T I C L E I N F O
A B S T R A C T
Keywords: SOA Deposition Vertical mixing Model bias Volatility basis set CMAQ CAMx
Differences between two air quality modeling systems reveal important uncertainties in model representations of secondary organic aerosol (SOA) fate. Two commonly applied models (CMAQ: Community Multiscale Air Quality; CAMx: Comprehensive Air Quality Model with extensions) predict very different OA concentrations over the eastern U.S., even when using the same source data for emissions and meteorology and the same SOA modeling approach. Both models include an option to output a detailed accounting of how each model process (e.g., chemistry, deposition, etc.) alters the mass of each modeled species, referred to as process analysis. We therefore perform a detailed diagnostic evaluation to quantify simulated tendencies (Gg/hr) of each modeled process affecting both the total model burden (Gg) of semi-volatile organic compounds (SVOC) in the gas (g) and aerosol (a) phases and the vertical structures to identify causes of concentration differences between the two models. Large differences in deposition (CMAQ: 69.2 Gg/d; CAMx: 46.5 Gg/d) contribute to significant OA bias in CMAQ relative to daily averaged ambient concentration measurements. CMAQ's larger deposition results from faster daily average deposition velocities (VD) for both SVOC (g) (VD,cmaq = 2.15 × VD,camx) and aerosols (VD,cmaq = 4.43 × Vd,camx). Higher aerosol deposition velocity would be expected to cause similar biases for inert compounds like elemental carbon (EC), but this was not seen. Daytime low-biases in EC were also simulated in CMAQ as expected but were offset by nighttime high-biases. Nighttime high-biases were a result of overly shallow mixing in CMAQ leading to a higher fraction of EC total atmospheric mass in the first layer (CAMx: 5.1–6.4%; CMAQ: 5.6–6.9%). Because of the opposing daytime and nighttime biases, the apparent daily average bias for EC is reduced. For OA, there are two effects of reduced vertical mixing: SOA and SVOC are concentrated near the surface, but SOA yields are reduced near the surface by nighttime enhancement of NOx. These results help to characterize model processes in the context of SOA and provide guidance for model improvement.
1. Introduction Atmospheric fine particulate matter impacts human health (Pope et al., 2004) and global climate due to its high radiative activity (Novakov and Penner, 1993). A significant fraction of global atmospheric aerosols consists of secondary organic aerosol (SOA) (Zhang et al., 2007), formed during oxidation of both biogenic and anthropogenic volatile organic compounds (VOCs). Current air quality models offer new techniques to research SOA, but these models’ capability of accurately predicting SOA is a prerequisite. Recent SOA studies state that it is necessary to improve SOA formation modeling methods due to the complexity of emitted VOC mixture and degradation chemistry (Hallquist et al., 2009). There is a strong body of literature characterizing SOA formation and applying it to field scales. However, SOA formation modeling methods are insufficient to explain SOA
∗
underprediction because external uncertainties (eg., emissions, removal mechanisms, etc.) also limit model prediction. Recent literature has improved understanding of SOA partitioning/ formation processes through a combination of laboratory and modeling work. Early SOA modeling studies at a laboratory scale focused on the absorptive partitioning theory of Pankow (2007), and the two-product model extended by Odum et al. (1996). There are also many semi-explicit studies on mechanisms that predict SOA formation. For example, a semi-explicit mechanism of D-limonene was developed and tested against experimental results obtained from large outdoor Teflon film chambers by Leungsakul et al. (2005). However, with increasing complexity of SOA formation, two-product model and semi-explicit treatment show the limitations of representing additional laboratory data. Donahue et al. (2006) and Robinson et al. (2007) developed the Volatility Basis Set (VBS) approach. This has the potential to provide a
Corresponding author. 420 Black Hall, Gainesville, FL 32611, United States. E-mail address:
[email protected] (B.H. Henderson).
http://dx.doi.org/10.1016/j.atmosenv.2017.10.035 Received 13 July 2017; Received in revised form 12 October 2017; Accepted 14 October 2017 Available online 20 October 2017 1352-2310/ © 2017 Elsevier Ltd. All rights reserved.
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
distribution analyses, mass balances and detailed deposition parameters. Our results show that differences between models and their biases can be largely attributed to vertical mixing and deposition.
unified framework for gas-aerosol partitioning and chemical aging of SOA, and it could be used under a larger range of atmospheric conditions (Donahue et al., 2006; Robinson et al., 2007). More recently the Unified Partitioning-Aerosol phase Reaction (UNIPAR) model has been developed to predict SOA formation from both partitioning and aerosol phase reactions (Beardsley and Jang, 2015; Im et al., 2014). Regional chemistry transport models (RCTMs) have also been utilized to estimate SOA concentrations, although their results are historically underestimated (Tuccella et al., 2012). reported that the Weather Research and Forecasting model with Chemistry (WRF-Chem) underestimated PM2.5 mass concentrations by 4.0–14.0 μg/m3 (10–50%) over Europe (Tesche et al., 2006). did comparative performance evaluation of the Comprehensive Air-quality Model with Extensions (CAMx) ozone/PM model (ENVIRON, 2004) and the Community Multiscale Air Quality (CMAQ) modeling system (Byun and Ching, 1999) and reported that organic carbon (OC) was typically underestimated by nearly 80% in the summer and 60% in the winter. However, they did not find evidence that one model was superior to the other. Results of a model performance study on estimating OC of CAMx and CMAQ by Morris et al. (2006) also showed that the two models underestimate OC concentrations with CAMx exhibiting better model performance than CMAQ in Atlanta, the region of study with a lower fractional bias (−18% versus −57%). Even with the latest SOA formation mechanisms, RCTMs may have poor model performance. Chen et al. (2006) incorporated the Caltech Atmospheric Chemistry Mechanism (CACM) and the Model to Predict the Multiphase Partitioning of Organics (MPMPO) into CMAQ, but CMAQ still yielded substantial underestimates for EC and OM. The study noted that underestimates were typically attributed to EC and OM formation. In a recent study on understanding sources of OC, CMAQVBS under-predicted total OC compared to observations from the Chemical Speciation Network (CSN) (−25.5% Normalized Median Bias (NMdnB)) and Interagency Monitoring of Protected Visual Environments (IMPROVE) (−63.9% NMdnB) locations (Woody et al., 2016). The authors indicated that providing additional certainty in intermediate volatility compounds (IVOCs) and SOA formed from IVOCs could close the gap between modeled and measured SOA. Most previous studies attribute model underestimates of SOA to chemical and phase partitioning processes. However, Koo et al. (2014) recently reported that CMAQ tended to underestimate SOA by 42–46% more than CAMx even when using identical emission and meteorological inputs and the same 1.5-dimensional volatility basis set approach (1.5-D VBS). In other words, there were significant differences in performance even though the models used identical inputs. This suggests that the organic aerosol sub-model probably does not dominate this bias. Instead, the fidelity of the host RCTM significantly influences the outcome of applying laboratory-derived models. It is critical to understand which RCTM process results in this discrepancy between CMAQ and CAMx since the two models are widely used as SOA host models and their uncertainties may influence SOA prediction. There are a few studies that point out that deposition is a large source of uncertainty but not limited in SOA prediction. Textor et al. (2006) recognized the large differences between global models with respect to dry deposition fluxes and efficiencies. Fowler et al. (2009) concluded that many regional pollutant deposition models were struggling to correctly incorporate physic-chemical properties of aerosols using realistic coupled sectional approaches, particularly with respect to SOA. Yarwood et al. (2012) reported that underestimated dry deposition of ozone over water and overestimated CAMx boundary conditions for ozone could contribute to over-predicting ozone over the Gulf of Mexico. These concerns indicate a potential direction for in-depth investigations on CAMx and CMAQ. Our present work investigates and quantifies processes of CAMx and CMAQ that produce biases. In Section 2, we describe the models used, development of case studies, subsequent evaluation, and process analysis. Section 3 shows results from concentration time series, vertical
2. Methodology 2.1. Model descriptions This study uses two commonly applied regional photochemical grid models, CMAQ (version 5.0.1) and CAMx (version 5.41), both of which are used as host models. Both models are adapted to incorporate a hybrid 1.5-dimensional Volatility Basis Set (1.5-D VBS) approach, which was developed by Koo et al. (2014) by combining the simplicity of the 1-dimensional (1-D) VBS with the ability to describe the evolution of OA in the 2-dimensional space of oxidation state and volatility. Further details on the 1.5-D VBS set up can be found in Koo et al. (2014). The simulations are set up to reflect regulatory modeling purposes. The EPA's Cross-State Air Pollution Rule (CSAPR) base year (2005) database (EPA, 2011) is used to run monthly, weekly and daily episodes in August. The modeling domain is made up of a 36 km horizontal grid over the entire continental U.S. with a 12 km nested grid covering the eastern U.S. (Fig. 1). Meteorological data is from the fifth-generation Pennsylvania State University-NCAR Mesoscale Model (MM5) (Dudhia, 2012; Grell et al., 1994). The previous application (Koo et al., 2014) incorrectly reported the meteorological model as the Weather Research and Forecasting (WRF) model version 3.4 (Dudhia, 2012). Biogenic emissions are based on the Model of Emissions of Gases and Aerosols from Nature (MEGAN) version 2.1 (Guenther et al., 2012). Detailed comparisons of model configuration, input data are provided in Table 1. 2.2. Model simulations Our model simulations are designed to isolate differences and minimize the number of variables in CAMx and CMAQ simulations. The first step is to reduce the simulation burden by minimizing the temporal and spatial extent needed to produce differences. We do this using a case study.
Fig. 1. The 36 and 12 km simulation domain; Four sets of symbols represent U.S. Regional Planning Organization (RPO) regions.
318
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
2.3. Evaluation
Table 1 CMAQ and CAMx model configurations. Model option
CMAQ
CAMx
Model version Horizontal resolution Vertical layers Horizontal advection Vertical advection Horizontal diffusion Vertical diffusion
Version 5.0.1 36/12 km NZ = 14 PPM PPM Spatially varying Eddy diffusion with ACM2a MM5 CSAPR and MEGANv2.1 Pleim and Ran (2011) CB05
Version 5.41 36/12 km NZ = 14 PPM Implicit Spatially varying Standard K-theory approachb MM5 CSAPR and MEGANv2.1
1.5D-VBS
1.5D-VBS
Meteorology Emissions Dry deposition Oxidant chemistry mechanism SOA scheme a b
Model evaluation is used both to select the case study from candidates and to guide process analysis. The case study selection relies on performance evaluation since it could directly reflect whether each case study can reproduce previous biases shown in the Koo et al. (2014) study. It is also vital to isolating potential processes causing variances and quantifying their contributions to biases. 2.3.1. Performance evaluation Model performance is evaluated using the OC measurement dataset from the Interagency Monitoring of Protected Visual Environments (IMPROVE) and the artifact-corrected dataset from EPA's Speciation Trends Network (STN) monitoring sites (Malm et al., 2011). Fractional bias is calculated using equation (1) to evaluate model performance (Yu et al., 2006).
Zhang et al. (2003) CB05
FB =
CMAQ uses PBL directly from MM5. CAMx uses a CAMx pre-processor to calculate Kz from MM5 PBL.
2 N
M −O
∑ Mi + Oi i
i
(1)
where FB is fractional bias; N is the number of data points; Mi is modeled concentration; Oi is observed concentration. 2.2.1. Case study development Significant bias differences that have been represented in a 1-month 12 km grid simulation for the summer episode (August) in Koo et al. (2014) can be represented on shorter time scales. We develop sets of case simulations to minimize the simulation duration that is necessary to reproduce the biases found in a 1-month simulation. The initial case study candidate (C1) is a 1-month 12 km simulation reproduced from Koo et al. (2014). The second case study (C2) is a 1-week 12 km simulation to include two days of observations (Aug. 2nd and 5th). As shown in Fig. 2, MOZART provides initial conditions (IC) for the 36 and 12 km simulations and boundary conditions (BC) for 36 km simulation only. The 36 km simulation then creates BC for the 12 km simulation, which is used in the case study evaluation. The third case study candidates (C3) are two sets of 1-day 12 km grid simulation corresponding to observations (Aug. 2nd and 5th).
2.3.2. Process investigation We separately quantify the processes that could cause model differences in burden. We started with a general mass balance for the burden (Mi,t) and each process (see equation (2)) for each model cell and integrated all cells over the entire 12-km horizontal and vertical domain.
Mi, t 0 + Δ t = Mi, t 0 + Ei,Δ t + Fi,Δ t + Ni,Δ t − Di,Δ t − Wi,Δ t
(2)
where Mi is mass in atmosphere (aka regional burden); E is emissions; F is net transport flux (i.e., sum of advection (ADV), diffusion (DIF), and dilution due to a change in the PBL (DIL) from process analysis (PA) outputs); N is net chemical production (Pi, Δt) and loss (Li, Δt); D is dry deposition; W is wet deposition. For example, the burden for each species (i) is calculated as the sum of instantaneous time (t) masses in all grid cells (x) calculated from concentrations and volumes for aerosols (Mi,t = ∑x Cx,i,t0 * Vx,t) and mixing ratios and moles of air for gases (Mi,t = ∑x Xx,i,t0 * nx,t). The process masses are calculated similarly for each hourly PA output and then cumulatively summed from the simulation start to that each instant (Δt = t – t0) for all grid cells. The initial masses (Mi, t 0 ) are defined by the case study IC, and subsequent changes in mass are a result of one of the processes in equation (2) and mass continuity is evaluated. First, the emissions are known to be equal and by definition are added to the atmospheric mass. Second, because net transport fluxes are integrated over the whole domain any transport between cells is offset and can be considered the net boundary fluxes (Fi,t = Fin,i,t – Fout,i,t). We use net chemistry (Ni, t = Pi,Δ t − Li,Δ t ) to represent the change of masses caused by chemistry. For chemistry process, EC is chemically inert so net chemistry is zero. We define organic semi-volatiles and aerosols (OSA = SVOC (g) + SOA (a)) to investigate mass changes without considering partitioning. In both models, OSA is completely conserved from a mole perspective. From a mass perspective, OSA is produced by oxidation that creates higher molecular weight products from lower molecular weight reactants. OSA, by definition, cannot be removed from mass balance system by chemistry or partitioning processes. Detailed VBS mechanism is provided in Table A2 (see Appendix) to elucidate our definition. Because transport is integrated across the entire domain, the net value is expected to be very small relative to other processes. As CAMx results are less biased compared to observations, it will be used as a reference when comparing processes. Differences in processes (P∈{Ei,Δt, Fi,Δt, Ni,Δt, DiΔt, Wi,Δt}) will be shown as the difference between CMAQ and CAMx (ΔP = Pi,cmaq – Pi,cmax).
2.2.2. Isolating influence Even when evaluating a single day or domain, larger-scale influences persist. Although we evaluate only the 12-km domain results, the 36 km domain results influence our evaluation as lateral boundary conditions (BC) for the 12-km domain. Similarly, the initial conditions (IC) for each day are derived from the previous day's simulation and propagate biases from previous days. These variations in IC and BC could lead to different results for 1-day simulations between the two models. In this case, we only consider Aug. 2nd in C3 since results of Aug. 5th could be influenced by that of Aug. 2nd. Differences could be accumulated in the long-term simulation and cause large biases. To truly isolate the process for evaluation, we must further modify the case study. To remove the accumulating differences, we created a test case where IC and BC are identical. We converted CAMx IC and BC for the 12 km domain for use by CMAQ's 1-day simulation (Aug. 2nd) of which we created five combinations of IC/BC (X/X, Q/Q, XQ/Q, Q/XQ, XQ/ XQ) where X denotes CAMx, Q denotes CMAQ, and XQ denotes CMAQ with CAMx IC or BC. Results indicate that for 1-day simulation, BC does not cause significant differences because there is not enough time to spin up (see Table A1 in Appendix). In Section 3, we will only show the case study candidate (C4) that uses a single day (Aug. 2nd) with both CAMx and CMAQ using BC/IC from CAMx. Case study candidates (C1, C2, C3, and C4) have varying duration and IC/BC options. In the results section (3.1), all candidate evaluations will be compared and the case study will be selected. The evaluation technique is described below.
319
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
Fig. 2. Scheme of CAMx and CMAQ simulation. Both models run 36 km and 12 km simulations, where 36 km BC are from MOZART and 12 km BC are from the 36 km simulation. Solid arrows represent data exchange from the 36 km–12 km. Dashed arrows represent data exchange from 12 km to 36 km in CAMx only. First, the initial conditions IC are from MOZART, and each 1-day simulation creates IC for next-day simulation. The case study is developed based on the 12 km simulation and the influence of IC/BC is tested using Aug. 2nd simulation.
selection (Section 3.1) shows our evaluation and identifies the shortest possible duration and minimum variables necessary to reproduce the bias differences in Koo et al. (2014). Characterization (Section 3.2) provides detailed time and vertical differences between CAMx and CMAQ case studies to inform process investigation (mass balances and deposition). Section 3.3.1 shows mass balances results that identify the important processes using the fate of conservative model tracers and the dry deposition process differences are further detailed in Section 3.3.2.
As a consistency check, we compared burdens at each hourly instant (Mi,t) to the process integral (Mi,t+Δt, equation (2)) for both CAMx and CMAQ (see Table A3 in Appendix). CMAQ and CAMx PA results show perfect mass conservation for aerosols. For semivolatile gases, CAMx results show perfect conservation, while CMAQ mass changes and summed process show artifacts that can be as large as 6% of the largest hourly process. This is likely due to small net changes and artifacts of the process unit. CAMx outputs processes in mole concentrations while CMAQ outputs in mixing ratio. Since MM5 volumes are constant while moles vary, aggregating time integrals can produce artifacts for mixing ratios. The artifacts are very small compared to differences between CMAQ and CAMx in ErrorSVOC ,Δ24 ErrorOSA,Δ24 = 1.28%, Δ W = 1.23%). deposition ( Δ W SVOC ,Δ24 + Δ DSVOC ,Δ24 OSA,Δ24 + Δ DOSA,Δ24 Studies on mass continuity would need to alter output units to ensure conservation. Given our results, special attention is paid to the dry deposition process. Dry deposition is calculated as the time integral of concentration (Ci) and deposition velocity (VD,i) of species based on equation (3).
DΔ t =
∫ Ct t
3.1. Case study selection and demonstration The first step is to choose a case study for further evaluation. The four possible case studies include a 1-month simulation (C1), a 1-week simulation (C2), a 1-day simulation (C3), and a 1-day simulation with CAMx-based IC/BC (C4). Fig. 3 shows that all case study candidates reproduce SOA bias differences between the two models. When each model uses its own IC/BC, the mean fractional bias (MFB) for CAMx (−0.14,-0.04) is consistently lower than CMAQ (−0.59,-0.47). Based on the results, the 1-day simulation can be chosen as the minimum simulation duration since it reproduces bias differences that appeared in the 1-month simulation. If CAMx IC/BC is used in the CMAQ 1-day simulation (Aug. 2nd), CMAQ MFB (−0.18) is greatly reduced but still more biased than that of CAMx (−0.04) (as shown in Fig. 3). Conversion of IC/BC from CAMx to CMAQ reduces parts of bias differences between the two models. These results show differences between CAMx
× VD, t × Δt (3)
3. Results The results are organized into case study selection and characterization, mass balances, and deposition investigation. The case study 320
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
Fig. 3. Fractional biases of OC modeled by CAMx and CMAQ at all monitoring sites including IMPROVE and STN for the summer episodes. 1-Month represents results of the 1-month simulation (August). 1-Week represents results of the 1-week simulation (Aug. 1st to 7th) including two days of observations (Aug. 2nd and 5th). 1-Day represents results of the 1-day simulation (Aug. 2nd). CMAQ (XIC, XBC) represents the modified CMAQ simulation using conversions of IC and BC from CAMx.
(1.68 ± 0.91 μg/m3) for Northeast, Southeast, Central, West, and West Pacific regions over the U.S. based on IMPROVE and the Southeastern Aerosol Research and Characterization project (SEARCH) networks provided by Yu et al. (2004). EC and OA share similar diurnal patterns after initialization, but model differences are species-specific. Both CAMx and CMAQ exhibit expected diurnal patterns for EC and OA. EC and OA concentrations both increase during the nighttime (0–12 UTC, Coordinated Universal Time) while decreasing sharply during the daytime (12–24 UTC). This is caused by sudden increases in the planetary boundary layer (PBL) height during the daytime. Differences between modeled and predicted EC (ΔCEC = CMAQ – CAMx) also have a clear diurnal pattern. During the nighttime, Fig. 4(a) shows that CMAQ EC is greater than CAMx by 0.04 ± 0.02 μg/m3, which suggests less nighttime ventilation. During the daytime, CMAQ EC is marginally smaller than CAMx by 0.01 ± 0.03 μg/m3, which could either indicate a higher boundary layer or faster deposition. This assumption will be proved in the following section. Fig. 4(b) shows that CMAQ OA is smaller than CAMx at most times
and CMAQ accumulate even within a day and can extend to longer periods. 3.2. Case study characterization In Section 3.1, we demonstrated that the minimum simulation duration could be 1 day and we selected Aug. 2nd as the projected case study. In this section, we characterize model differences on Aug. 2nd using concentration time series and vertical profiles of model differences. In the following sections, CMAQ is used to represent the modified CMAQ (XIC, XBC) for convenience. 3.2.1. Concentration time series We spatially averaged surface concentration time series of EC and OA. This provided their detailed lifetime trends, shown in Fig. 4(a–b) at the 12 km simulation domain. CAMx and CMAQ use the same initial concentrations (EC: 0.23 μg/m3; OA: 1.98 μg/m3) which is within the range of average measured EC (0.33 ± 0.21 μg/m3) and OA
Fig. 4. Characterization of the case study (Aug. 2nd) shown with time series of (a) EC and (b) OA concentrations at all domains on Aug. 2nd modeled by CAMx and CMAQ.
321
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
Fig. 5. Vertical distributions of (a) EC and (b) OA concentrations in CAMx and vertical differences of (c) EC and (d) OA concentrations between CAMx and CMAQ at all domains and atmospheric layers on Aug. 2nd. Each grid represents hourly concentration at each vertical layer.
3.3. Process diagnostic
of a day. During the daytime, CMAQ OA is much smaller than CAMx by 0.37 ± 0.07 μg/m3. Over the whole day, the CMAQ model prediction is 25% lower than CAMx. These results support conclusions by Koo et al. (2014) that the discrepancy between the two models is caused by differences outside the OA modules because inert species such as elemental carbon also show differences. However, differences between modeled and predicted OA (ΔCOA = CMAQ – CAMx), unlike EC, do not change sign between night and day. EC, as a chemically inert species, is concentrated near the surface by reduced vertical mixing. For OA, although SOA and SVOC are concentrated near the surface, SOA yields are also reduced by nighttime enhancement of NOx.
Using mass balance analysis based on the Aug. 2nd case study, we investigate and quantify potential processes that cause bias differences between CAMx and CMAQ. Our results demonstrate that deposition is the primary source of differences during the day. Based on this result, we characterize CMAQ and CAMx deposition velocities to understand the source of the differences. Detailed results of mass balances analysis and deposition velocities are discussed in the following sections. 3.3.1. Mass balances Fig. 6(a–b) show that atmospheric masses in CAMx and CMAQ start with identical values that diverge over the day. CAMx and CMAQ have the same initial (t0 = 0 UTC) atmospheric masses (MEC,t0: CMAQ = 3.47 Gg, CAMx = 3.46 Gg; MOSA,t0: CMAQ = 186.06 Gg, CAMx = 185.72). For 1-day simulation, chemistry process and emissions produce masses into the atmosphere but removed by deposition and transport over the day. Chemistry process causes a small OSA mass difference between CAMx and CMAQ (ΔNOSA ) compared to that of deΔN position (ΔTOSA = ΔDOSA + ΔWOSA, Δ TOSA ,Δ24 = 6.12% ). This confirms OSA,Δ24 that 1.5D-VBS scheme shows consistency in both models. CMAQ deposits more EC (ΔTEC = 0.2 Gg) and OSA (ΔTOSA = 22.71 Gg) masses than CAMx by wet and dry deposition mechanisms. As expected, OSA mass differences caused by transport are small compared to that of ΔF deposition ( OSA,Δ24 = 11.2%) and differences in transport are likely
3.2.2. Vertical profiles We also did vertical profiles of EC and OA as averages over all rows and columns for the 12 km simulation. Fig. 5(a–b) show CAMx vertical profiles and Fig. 5(c–d) show vertical distribution of concentration differences between CAMx and CMAQ (ΔCi = CMAQ - CAMx). From Fig. 5(a–b), we found EC has a stronger vertical gradient while OA has a weaker gradient. It indicates that nighttime mixing is less influential on secondary organic aerosols than primary pollutants. Fig. 5(c–d) confirm discrepancies in nighttime mixing between CAMx and CMAQ. CMAQ has lower concentrations in layer three at night for both EC and OA. For EC, this is roughly balanced by the high biases in layer 1. For OA, however, the low biases aloft are not offset. In the afternoon, both OA and EC differences show CMAQ's consistently lower concentrations. Outside the PBL, the differences are more modest.
Δ TOSA,Δ24
symptoms. Based on above discussion, deposition process accounts for a 322
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
Fig. 6. CAMx (hatched) and CMAQ (solid) distributions of (a) EC and (b) OSA fates (atmosphere: white; dry deposited: green; wet deposited: blue; transport: yellow). (c) Detailed deposition proportion of gases and aerosols in OSA deposition. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
within the reported range of OA sensitivity to deposition. Bessagnet et al. (2010) reported that omitting dry deposition can lead to an overestimation of SOA concentrations by as much as 50%. Knote et al. (2015) also stated that dry deposition of gas phase SVOCs is found to be more effective than wet deposition in reducing SOA concentrations (−40 vs. −8% for anthropogenic, and −52 vs. −11% for biogenic). They also point out that deposition has similar removal effects on SOA over the PBL and at the surface. Results of the above studies confirm that dry deposition is able to cause ∼25% differences in SOA predictions.
majority part of mass differences between CAMx and CMAQ and needs follow-up investigation. Understanding the differences for OSA require distinguishing between wet and dry mechanisms as well as between gas and aerosol phases. Fig. 6(c) shows the proportion of gases and aerosols in OSA deposition and highlights three important findings. (1) CMAQ deposits more OSA mass by both dry and wet deposition, but CMAQ dry deposition is 94.6% higher than CAMx (ΔDOSA = 15.35 Gg) compared to just 24% higher for wet deposition (ΔWOSA = 7.36 Gg). (2) CMAQ's dry deposition of aerosols is 336% higher than CAMx and 68.7% higher for gases (SVOCs). Despite lower relative increases, the magnitude of model differences for SVOC dry deposition (ΔDSVOC = 10.08 Gg) is larger than that for SOA (ΔDSOA = 5.27 Gg). Because of its higher dry deposition, CMAQ has less SOA mass for wet deposition (ΔWOSA = −4.43 Gg). Even though the relative difference for gas is smaller, the gases contribute substantially more to the overall difference. (3) Simulated differences for aerosols and gases have their own distinct diurnal pattern. For aerosols, CMAQ has less dry deposition than CAMx (Dcmaq = 0.92 × Dcamx) at nighttime (0–12 UTC), but deposits much more than CAMx (Dcmaq = 5.99 × Dcamx) during the daytime (12–24 UTC). For gases, on the other hand, CMAQ deposits more than CAMx (Dcmaq = 1.66 × Dcamx) the whole day (0–24 UTC). Based on our results, enhanced dry deposition (SVOC and OA) is causing CMAQ OA predictions to be 25% lower than CAMx, and resulting in the low bias compared to observations. The 25% difference is
3.3.2. Deposition investigation Results of the mass balances analysis reveal that CMAQ's larger deposition, especially that for dry deposition, causes underestimates of OA concentration. Therefore, we investigate discrepancies in dry deposition velocity between CAMx and CMAQ. Atmospheric turbulence, the chemical properties and the nature of surface govern the dry deposition of gas and particles. In the two models, deposition velocity was calculated based on several species properties (see Table A4 and A5 in Appendix). It is important to note that gas and particle phases use different algorithms to calculate dry deposition velocity. In general, dry deposition velocity (VD) of gas is defined as the inverse of the total resistance (rt), which is the sum of the aerodynamic resistance (ra), the quasi-laminar layer resistance (rb) and the surface or canopy resistance (rc), expressed as equation (4). 323
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
VD =
1 1 = rt ra + rb + rc
SVOC deposition velocities are more sensitive to species properties and have meaningful differences in day and night. Fig. 7(b) shows the CMAQ's SVOC deposition velocities are 2 times higher than in CAMx at night and 2.2 times higher during the day. The SVOC properties in CAMx produce only small differences in deposition velocities ( ± 0.2 mm/s). Acetic acid dry deposition velocity calculated in CAMx is consistently slower than in CMAQ. If we also use acetic acid as a surrogate species for SVOCs in CAMx, SVOCs' dry deposition velocity slows, causing CMAQ to be 3.5 times higher than CAMx at night and 3 times higher during the day. During the daytime, CAMx deposition velocities are less sensitive to the acetic acid properties. Our results suggest that species properties and algorithm both contribute to differences in gas dry deposition velocities between the two models. The similarity between CAMx SVOC dry deposition velocities demonstrates that using species-specific deposition velocities contributes very little to the differences between the models. However, it is theoretically preferable to calculate SVOC dry deposition velocity with its own property. The difference between the acetic acid and CAMx's species velocities does, however, suggest that acetic acid should be replaced with a more appropriate surrogate. For the gas algorithm, CAMx and CMAQ both use equation (4) to calculate dry deposition velocity. Further, two models use a similar way to calculate ra and rb but a different way for rc. Since Zhang et al. (2003) stated that the uncertainties in ra and rb from the different models are small, we presume that rc could be the major factor to cause deposition differences. For example, CMAQ uses a simple in-canopy resistance (rac) like equation (7) suggested by Erisman et al. (1994):
(4)
For particles, the particle settling velocity (Vs) was considered because of sedimentation. It is also usually assumed that particles adhere to the surface on contact so that the surface or canopy resistance rc = 0. Thus, the particle dry deposition velocity can be expressed as equation (5) (Seinfeld and Pandis, 1998).
VD =
1 1 = + Vs rt ra + rb + ra rb Vs
(5)
SVOC (g) is calculated in the gas phase while SOA (a) and EC (a) are in the particle phase. In both models, SOA and EC have practically same deposition velocities. In CAMx, each SVOC's uses its own properties to calculate dry deposition velocities. In CMAQ, SVOC's dry deposition is calculated using acetic acid as a surrogate species. CMAQ and CAMx calculate dry deposition using different physical/chemical properties, even for the same aerosol and acetic acid (CMAQ's SVOC surrogate). To understand the importance of model species properties, we calculate deposition velocities in CAMx using properties from CMAQ for aerosols and acetic acid. For acetic acid, SVOC and SOA, we compare daytime and nighttime values from CMAQ (Q), CAMx (X), and CAMx with CMAQ properties (XwQ). Differences for aerosol dry deposition velocities are stronger during the day than at night and insensitive to the difference in model species properties. Fig. 7(a) shows that the aerosol nighttime mean dry deposition velocities are similar (0.53 ≤ VD,0 − 12Z ≤ 0.63 mm/s) for all models while daytime mean dry deposition velocities vary nearly an order of magnitude (VD,12 − 24Z : X = 1.26; XwQ = 1.27; Q = 7.89 mm/ s). Using CMAQ species property parameters for SOA in CAMx only causes negligible differences in their velocities ( ± 0.1 mm/s). This indicates that algorithm results in differences of particle dry deposition velocities between CAMx and CMAQ. In CAMx, the dry deposition algorithm of particle is using Zhang et al. (2001) model that a simple parameterization of particle dry deposition as a function of aerosol size and land use that predicts higher deposition velocities for sub-micron aerosols, especially over rough vegetated surfaces. In CMAQ, equation (6) was derived by Venkatram and Pleim (1999) as dry deposition algorithm of particle because they found electrical analogy of equation (5) is not consistent with the mass conservation equation. However, they pointed that there might be little difference between the magnitudes of dry deposition velocities estimated with equation (5) and equation (6). This may explain magnitude difference of SOA daytime mean dry deposition velocities between CAMx and CMAQ.
VD =
Vs 1 − e−Vs (ra+ rb)
rac = b
hc LAI u∗
(7)
where LAI is the one-sided leaf area index that directly from CMAQ meteorological inputs, hc the canopy height, b an empirical constant taken as 14 m−1 and u* the friction velocity (m/s). In CAMx, rac is developed by Zhang et al. (2002) using equation (8) and parameterized related to LAI and u* including variations by land use category and day of the year to explain canopy structure changes (Pleim and Ran, 2011): 1
rac =
rac0 (LAI ) 4 u∗ 2
(8)
where rac0 is the reference value for in-canopy aerodynamic resistance. Overall, a further investigation is needed since different dry deposition appeared in the CAMx and CMAQ due to different deposition algorithms.
(6)
Fig. 7. Mean dry deposition velocities for (a) aerosols (SOA) and (b) SVOC and acetic acid (AACD) gases at nighttime (0–12 UTC) and daytime (12–24 UTC) for CMAQ (Q), CAMx (X) and CAMx with CMAQ species property parameters (XwQ).
324
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
4. Conclusions
difference between the two models. CAMx has neutral fractional biases of SOA concentration compared to low fractional biases of that in CMAQ. It suggests that SOA is reasonably predicted in CAMx, while there is a problem in CMAQ. Continued mass balance analysis leads to the hypothesis that CMAQ's particle and gas dry deposition velocities for SOA related species are both over-predicted. Our characterization of CMAQ-VBS deposition shows an overly simplistic representation of SVOC properties for dry deposition velocity. Even so, we would need to compare these deposition fluxes with field data to confirm the hypothesis that the total deposition is over-estimated. In addition, it is recommended to implement the deposition methodology used for these CAMx simulations in CMAQ as an option to test whether it improves the CMAQ deposition process. The results of this study are potentially limited by the use of Mesoscale Model version 5 (MM5). Discussions with the CMAQ development team suggest that the Meteorological Chemical Interface Processor (MCIP) may produce more detailed descriptions of surface characteristics from the WRF model. From one perspective, this indicates the strong sensitivity of CMAQ deposition velocities to the characterization of surface characteristics and suggests a need for caution when using meteorological data other than from WRF. Additional studies on deposition processes are needed to further constrain current aerosol predictions. Overall, our study illustrates the magnitude to which the host model representation of deposition limits accurate predictions for SOA. The importance of deposition is often overlooked in favor of a missing source. This work shows the importance of parallel efforts to develop alternate aerosol deposition algorithms.
This study identifies processes that strongly influence the simulation of organic aerosols and quantifies sensitivity to process treatments. We compared two modeling systems (CMAQ and CAMx) that shared meteorological and emission inputs but produced drastically different concentrations. Our results showed that deposition, particularly dry deposition, is uncertain and ill-constrained in regional models due to lack of deposition flux measurements for model evaluation. Despite immediate differences between the models, bias was not previously attributed to the deposition because of compensating effects of vertical mixing. CMAQ retains more mass in the surface layer during the nighttime reflecting differences in the methods by which CAMx and CMAQ calculate vertical diffusivities. Differences in deposition produced distinct biases within 24 h, but the magnitude depends strongly on the aerosol type. Primary pollutants have stronger vertical gradients. Stronger vertical mixing could cause lower surface concentrations (e.g., EC) even with slower deposition velocities. Secondary organic aerosols, however, have weaker vertical gradients, and nighttime mixing is less influential on model performance. Deposition velocity differences, therefore, create distinct signatures in primary and secondary species. This highlights both the continued need for diagnostic evaluation as well as the need for deposition evaluation studies using extended episodes. Results from process analysis show that the deposition process is the primary cause of model differences. CMAQ had faster deposition velocities of SVOC and SOA than CAMx, leading to more SVOC and SOA deposited. For SOA, the deposition difference between the model's deposition velocities is the algorithm. For the gas phase, species properties and algorithm both contribute to differences. CMAQ's larger mass deposition results in lower SOA concentrations in the atmosphere, accounting for a significant part of the fractional bias and SOA underestimation. Evaluation of CAMx and CMAQ presents SOA concentration
Acknowledgements Funding for this work included contributions from the University of Florida.
Appendix A In Appendix A, we provide (1) Detailed results of IC/BC influence; (2) Case study performance at each RPO region; (3) VBS mechanism; (4) Process quantifications in CAMx and CMAQ; (4) Deposition velocity species property. Table A1 Influence of IC/BC FB REGION
IC/BC
X/X
Q/Q
Q/XQ
XQ/Q
XQ/XQ
CENRAP
IMPROVE STN
−0.4647 −0.1795
−0.8918 −0.6355
−0.8913 −0.6355
−0.5278 −0.3665
−0.5274 −0.3665
MRPO
IMPROVE STN
0.2313 0.2692
−0.2837 −0.0085
−0.2837 −0.0085
0.0758 0.2196
0.0758 0.2196
MANE-VU
IMPROVE STN
0.0881 0.0589
−0.6055 −0.3492
−0.5929 −0.3492
−0.1610 −0.0493
−0.1541 −0.0493
VISTAS
IMPROVE STN
−0.2456 −0.1144
−0.5018 −0.5397
−0.5018 −0.5397
−0.2962 −0.3176
−0.2962 −0.3176
(CENRAP: Central Regional Air Planning Association; MRPO: Midwest Regional Planning Organization; MANE-VU: Mid-Atlantic/Northeast Visibility Union; VISTAS: Visibility Improvement State and Tribal Association of the Southeast).
325
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
Fig. A1. Fractional biases of OC modeled by CAMx and CMAQ at the IMPROVE and STN monitoring sites in four RPO regions (CENRAP, MRPO, MANE-UV, VISTAS) for the summer episodes. 1a and 1b represent results of the 1-month simulation (August). 2a and 2b represent results of the 1-week simulation (Aug. 1st to 7th), including two days of observations (Aug. 2nd and 5th). 3a and 3b represent results of the 1-day simulation (Aug. 2nd). 4a and 4b represent results of the 1-day simulation of chosen case study for characterization and process diagnoses (CAMx2CMAQ represents CMAQ simulation using converted IC/BC (XQ/XQ) from CAMx).
Table A2 VBS mechanism. < SA01 > TOLRO2 + NO = NO +0.006*SV_AVB1 +0.145*SV_AVB2 +0.281*SV_AVB3 +0.432*SV_AVB4 < SA02 > TOLRO2 + HO2 = HO2 +0.006*SV_AVB1 +0.145*SV_AVB2 +0.437*SV_AVB3 +0.281*SV_AVB4 < SA03 > XYLRO2 + NO = NO +0.001*SV_AVB1 +0.127*SV_AVB2 +0.201*SV_AVB3 +0.301*SV_AVB4 < SA04 > XYLRO2 + HO2 = HO2 +0.048*SV_AVB1 +0.195*SV_AVB2 +0.252*SV_AVB3 +0.364*SV_AVB4 < SA05 > BENZENE + OH = OH + 1.0*BENZRO2 < SA06 > BENZRO2 + NO = NO +0.001*SV_AVB1 +0.079*SV_AVB2 +0.148*SV_AVB3 +0.222*SV_AVB4 < SA07 > BENZRO2 + HO2 = HO2 +0.035*SV_AVB1 +0.108*SV_AVB2 +0.185*SV_AVB3 +0.268*SV_AVB4 < SA08 > ISOPRO2 + NO = NO
< SA13 > SESQ + OH = OH +0.092*SV_BVB1 +0.188*SV_BVB2 +0.968*SV_BVB3 +0.679*SV_BVB4 < SA14 > SESQ + NO3 = NO3 +0.092*SV_BVB1 +0.188*SV_BVB2 +0.968*SV_BVB3 +0.679*SV_BVB4 < SA15 > SV_AVB1 + OH = OH + SV_AVB0 < SA16 > SV_AVB2 + OH = OH + SV_AVB1 < SA17 > SV_AVB3 + OH = OH + SV_AVB2 < SA18 > SV_AVB4 + OH = OH + SV_AVB3 < SA19 > SV_BVB1 + OH = OH + SV_BVB0 < SA20 > SV_BVB2 + OH = OH + SV_BVB1 < SA21 > SV_BVB3 + OH = OH + SV_BVB2 < SA22 > SV_BVB4 + OH = OH + SV_BVB3 < SA23 > SV_PVB1 + OH = OH +0.864*SV_PVB0 +0.142*SV_AVB0 < SA24 > SV_PVB2 + OH = OH +0.877*SV_PVB1 +0.129*SV_AVB1 < SA25 > SV_PVB3 + OH = OH +0.889*SV_PVB2 +0.116*SV_AVB2 < SA26 > SV_PVB4 + OH = OH +0.869*SV_PVB3 +0.137*SV_AVB3 < SA27 > SV_FVB1 + OH = OH +0.538*SV_FVB0 326
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
+0.000*SV_BVB1 +0.009*SV_BVB2 +0.006*SV_BVB3 +0.000*SV_BVB4 < SA09 > ISOPRO2 + HO2 = HO2 +0.004*SV_BVB1 +0.013*SV_BVB2 +0.006*SV_BVB3 +0.000*SV_BVB4 < SA10 > TRPRO2 + NO = NO +0.010*SV_BVB1 +0.101*SV_BVB2 +0.173*SV_BVB3 +0.451*SV_BVB4 < SA11 > TRPRO2 + HO2 = HO2 +0.087*SV_BVB1 +0.077*SV_BVB2 +0.309*SV_BVB3 +0.540*SV_BVB4 < SA12 > SESQ + O3 = O3 +0.092*SV_BVB1 +0.188*SV_BVB2 +0.968*SV_BVB3 +0.679*SV_BVB4
+0.464*SV_BVB0 < SA28 > SV_FVB2 + OH = OH +0.689*SV_FVB1 +0.313*SV_BVB1 < SA29 > SV_FVB3 + OH = OH +0.783*SV_FVB2 +0.220*SV_BVB2 < SA30 > SV_FVB4 + OH = OH +0.846*SV_FVB3 +0.156*SV_BVB3 < SA31 > IVOC_P + OH = OH +0.033*SV_AVB1 +0.216*SV_AVB2 +0.304*SV_AVB3 +0.447*SV_AVB4 < SA32 > IVOC_F + OH = OH +0.033*SV_BVB1 +0.216*SV_BVB2 +0.304*SV_BVB3 +0.447*SV_BVB4
CMAQ and CAMx both use the same VBS mechanism but different species names. Below table is shown as CMAQ species name (CAMx: SV_AVB = VAS, SV_BVB = VAB, SV_PVB = VAP, SV_FVB = VFP).
Table A3 Cumulative Process mass changes in CAMx and CMAQ and CMAQ mass errors. CMAQ
MEC, t 0 + Δ t EEC,Δ t FEC,Δ t NEC,Δ t DEC,Δ t WEC,Δ t
∑ PEC,Δ t MEC,Δ t ErrorEC ,Δ t MOSA, t 0 + Δ t EOSA,Δ t FOSA,Δ t NOSA,Δ t DOSA,Δ t WOSA,Δ t
∑ POSA,Δ t MOSA,Δ t ErrorOSA,Δ t MOA, t 0 + Δ t EOA,Δ t FOA,Δ t NOA,Δ t DOA,Δ t WOA,Δ t
∑ POA,Δ t MOA,Δ t ErrorOA,Δ t
CAMX
0h
6h
12 h
18 h
24 h
0h
6h
12 h
18 h
24 h
3.47 0.00 0.00 0.00 0.00 0.00 0.00
3.51 0.19 0.00 0.00 −0.02 −0.12 0.04
3.48 0.36 −0.02 0.00 −0.06 −0.27 0.01
3.40 0.74 −0.05 0.00 −0.36 −0.41 −0.07
3.30 1.15 −0.07 0.00 −0.61 −0.65 −0.17
3.46 0.00 0.00 0.00 0.00 0.00 0.00
3.45 0.21 −0.02 0.00 −0.03 −0.16 −0.01
3.45 0.40 −0.05 0.00 −0.06 −0.29 −0.01
3.56 0.80 −0.09 0.00 −0.12 −0.49 0.10
3.51 1.23 −0.13 0.00 −0.17 −0.89 0.05
0.00 0.00
0.04 0.00
0.01 0.00
−0.07 0.00
−0.17 0.00
0.00 0.00
−0.01 0.00
−0.01 0.00
0.10 0.00
0.05 0.00
186.06 0.00 0.00 0.00 0.00 0.00 0.00
183.89 0.38 −0.95 9.31 −3.69 −7.38 −2.33
180.24 0.71 0.63 16.88 −8.71 −15.74 −6.23
179.71 1.51 −0.24 38.41 −22.12 −24.01 −6.45
178.30 2.41 −2.64 61.97 −31.58 −37.63 −7.47
185.72 0.00 0.00 0.00 0.00 0.00 0.00
187.18 0.49 −1.02 10.41 −2.42 −6.00 1.46
187.86 0.93 0.48 17.41 −5.42 −11.26 2.15
195.05 1.85 −1.23 37.74 −11.36 −17.67 9.33
197.47 2.87 −5.20 60.58 −16.23 −30.27 11.76
0.00 0.00
−2.16 0.17
−5.82 0.40
−6.35 0.10
−7.76 −0.28
0.00 0.00
1.46 0.00
2.15 0.00
9.33 0.00
11.76 0.00
46.82 0.00 0.00 0.00 0.00 0.00 0.00
45.00 0.38 −0.65 −0.19 −0.23 −1.13 −1.82
42.78 0.71 −1.30 −0.19 −0.48 −2.78 −4.04
43.80 1.51 −2.43 5.67 −3.71 −4.06 −3.03
42.97 2.41 −3.64 10.89 −6.84 −6.67 −3.85
46.63 0.00 0.00 0.00 0.00 0.00 0.00
44.64 0.49 −0.73 0.41 −0.26 −1.90 −2.00
43.00 0.93 −1.45 0.93 −0.51 −3.53 −3.63
46.76 1.85 −2.55 7.67 −1.02 −5.83 0.13
47.08 2.87 −3.70 13.95 −1.57 −11.10 0.44
0.00 0.00
−1.82 0.00
−4.04 0.00
−3.03 0.00
−3.85 0.00
0.00 0.00
−2.00 0.00
−3.63 0.00
0.13 0.00
0.44 0.00
327
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
MSVOC, t 0 + Δ t ESVOC,Δ t FSVOC,Δ t NSVOC,Δ t DSVOC,Δ t WSVOC,Δ t ∑ PSVOC,Δ t MSVOC,Δ t ErrorSOVC,Δ t
139.24 0.00 0.00 0.00 0.00 0.00 0.00
138.89 0.00 −0.30 9.50 −3.46 −6.25 −0.51
137.45 0.00 1.93 17.07 −8.23 −12.96 −2.19
135.91 0.00 2.19 32.75 −18.41 −19.95 −3.42
135.33 0.00 1.01 51.08 −24.74 −30.97 −3.62
139.09 0.00 0.00 0.00 0.00 0.00 0.00
142.54 0.00 −0.29 10.00 −2.16 −4.09 3.46
144.86 0.00 1.93 16.48 −4.91 −7.72 5.77
148.29 0.00 1.32 30.07 −10.34 −11.84 9.20
150.40 0.00 −1.50 46.64 −14.66 −19.17 11.31
0.00 0.00
−0.34 0.17
−1.78 0.40
−3.33 0.10
−3.90 −0.28
0.00 0.00
3.46 0.00
5.77 0.00
9.20 0.00
11.31 0.00
Table A4 Aerosol species properties for deposition velocity Density (g/cm3)
Dry extinction efficiency (m2/μm)
EC
CMAQ CAMx
2.2 2.0
10.0 18.0
SOA
CMAQ CAMx
1.5 1.0
4.0 7.0
Table A5 Gas species properties for deposition velocity Henry's law constant (M/atm)
Henry's law temperature dependence (K)
Molecular diffusivity ratio
Wesley's reactivity Surface resistance scaling factor parameter for strong acid (0–1)
AACD CMAQ
4.10E+03
−6.30E+03
1.83
0.0
0.4
CAMx
5.00E+03
−4.00E+03
2.03
1.0
1.0
SVOC CMAQ
4.10E+03
−6.30E+03
1.83
0.0
0.4
1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05 1.00E+05
−4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03 −4.00E+03
3.05 3.01 2.96 2.92 3.05 3.01 2.96 2.92 3.91 3.89 3.86 3.84 3.40 3.42 3.44 3.44
1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 1.0 1.0 1.0 1.0
1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
CAMx VAS1 VAS2 VAS3 VAS4 VBS1 VBS2 VBS3 VBS4 VAP1 VAP2 VAP3 VAP4 VFP1 VFP2 VFP3 VFP4
(VAS[1–4], VBS[1–4], VAP[1–4], VFP[1–4] represent 16 specific SVOC's in CAMx).
CMAQdocumentation.html n.d. Chen, J., Mao, H., Talbot, R.W., Griffin, R.J., 2006. Application of the CACM and MPMPO modules using the CMAQ model for the eastern United States. J. Geophys. Res. Atmos. 111, D23S25. http://dx.doi.org/10.1029/2006JD007603. Donahue, N.M., Robinson, A.L., Stanier, C.O., Pandis, S.N., 2006. Coupled partitioning, dilution, and chemical aging of semivolatile organics. Environ. Sci. Technol. 40, 2635–2643. http://dx.doi.org/10.1021/es052297c. Dudhia, J., 2012. The WRF model: 2012 annual update. In: Presented at the 13 Annual WRF Users' Workshop, 25e29 June 2012, Boulder, CO. http://www.mmm.ucar.edu/ wrf/users/workshops/WS2012/ppts/1.1.pdf n.d. ENVIRON, 2004. User's Guide—comprehensive Air-quality Model with Extensions. Version 4.10 s ENVIRON International Corporation, Novato, California Available at: http://www.camx.com August, n.d.
References Beardsley, R.L., Jang, M., 2015. Simulating the SOA formation of isoprene from partitioning and aerosol phase reactions in the presence of inorganics. Atmos. Chem. Phys. Discuss. 15, 33121–33159. http://dx.doi.org/10.5194/acpd-15-33121-2015. Bessagnet, B., Seigneur, C., Menut, L., 2010. Impact of dry deposition of semi-volatile organic compounds on secondary organic aerosols. Atmos. Environ. 44, 1781–1787. http://dx.doi.org/10.1016/j.atmosenv.2010.01.027. Byun, D.W., Ching, J.K.S., 1999. Science Algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) Modeling System. Atmospheric Modeling Divi- sion, National Exposure Research Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, NC. http://www.epa.gov/AMD/Research/CMAQ/
328
Atmospheric Environment 171 (2017) 317–329
Q. Shu et al.
dx.doi.org/10.1016/j.atmosenv.2007.10.060. Pleim, J., Ran, L., 2011. Surface flux modeling for air quality applications. Atmosphere 2, 271–302. http://dx.doi.org/10.3390/atmos2030271. Pope, C.A., Burnett, R.T., Thurston, G.D., Thun, M.J., Calle, E.E., Krewski, D., Godleski, J.J., 2004. Cardiovascular mortality and long-term exposure to particulate air pollution epidemiological evidence of general pathophysiological pathways of disease. Circulation 109, 71–77. http://dx.doi.org/10.1161/01.CIR.0000108927.80044.7F. Robinson, A.L., Donahue, N.M., Shrivastava, M.K., Weitkamp, E.A., Sage, A.M., Grieshop, A.P., Lane, T.E., Pierce, J.R., Pandis, S.N., 2007. Rethinking organic aerosols: semivolatile emissions and photochemical aging. Science 315, 1259–1262. http://dx.doi. org/10.1126/science.1133061. Seinfeld, J.H., Pandis, S.N., 1998. Atmospheric Chemistry and Physics, From Air Pollution to Climate Change. John Wiley and Sons, Inc., NY. Tesche, T.W., Morris, R., Tonnesen, G., McNally, D., Boylan, J., Brewer, P., 2006. CMAQ/ CAMx annual 2002 performance evaluation over the eastern US. Atmos. Environ. 40, 4906–4919. http://dx.doi.org/10.1016/j.atmosenv.2005.08.046. Textor, C., Schulz, M., Guibert, S., Kinne, S., Balkanski, Y., Bauer, S., Berntsen, T., Berglen, T., Boucher, O., Chin, M., Dentener, F., Diehl, T., Easter, R., Feichter, H., Fillmore, D., Ghan, S., Ginoux, P., Gong, S., Grini, A., Hendricks, J., Horowitz, L., Huang, P., Isaksen, I., Iversen, I., Kloster, S., Koch, D., Kirkevåg, A., Kristjansson, J.E., Krol, M., Lauer, A., Lamarque, J.F., Liu, X., Montanaro, V., Myhre, G., Penner, J., Pitari, G., Reddy, S., Seland, Ø., Stier, P., Takemura, T., Tie, X., 2006. Analysis and quantification of the diversities of aerosol life cycles within AeroCom. Atmos. Chem. Phys. 6, 1777–1813. http://dx.doi.org/10.5194/acp-6-1777-2006. Tuccella, P., Curci, G., Visconti, G., Bessagnet, B., Menut, L., Park, R.J., 2012. Modeling of gas and aerosol with WRF/Chem over Europe: evaluation and sensitivity study. J. Geophys. Res. Atmos. 117, D03303. http://dx.doi.org/10.1029/2011JD016302. Venkatram, A., Pleim, J., 1999. The electrical analogy does not apply to modeling dry deposition of particles. Atmos. Environ. 33, 3075–3076. http://dx.doi.org/10.1016/ S1352-2310(99)00094-1. Woody, M.C., Baker, K.R., Hayes, P.L., Jimenez, J.L., Koo, B., Pye, H.O.T., 2016. Understanding sources of organic aerosol during CalNex-2010 using the CMAQ-VBS. Atmos. Chem. Phys. 16, 4081–4100. http://dx.doi.org/10.5194/acp-16-4081-2016. Yarwood, G., Jung, J., Nopmongcol, O., Emery, C., 2012. Final Report Improving CAMx Performance in Simulating Ozone Transport from the Gulf of Mexico. Yu, S., Dennis, R.L., Bhave, P.V., Eder, B.K., 2004. Primary and secondary organic aerosols over the United States: estimates on the basis of observed organic carbon (OC) and elemental carbon (EC), and air quality modeled primary OC/EC ratios. Atmos. Environ. 38, 5257–5268. http://dx.doi.org/10.1016/j.atmosenv.2004.02. 064. Particulate Matter: Atmospheric Sciences, Exposure and the Fourth Colloquium on PM and Human Health - Papers from the AAAR PM Meeting. Yu, S., Eder, B., Dennis, R., Chu, S.-H., Schwartz, S.E., 2006. New unbiased symmetric metrics for evaluation of air quality models. Atmos. Sci. Lett. 7, 26–34. http://dx.doi. org/10.1002/asl.125. Zhang, L., Brook, J.R., Vet, R., 2003. A revised parameterization for gaseous dry deposition in air-quality models. Atmos. Chem. Phys. 3, 2067–2082. http://dx.doi.org/ 10.5194/acp-3-2067-2003. Zhang, L., Brook, J.R., Vet, R., 2002. On ozone dry deposition—with emphasis on nonstomatal uptake and wet canopies. Atmos. Environ. 36, 4787–4799. http://dx.doi. org/10.1016/S1352-2310(02)00567-8. Zhang, L., Gong, S., Padro, J., Barrie, L., 2001. A size-segregated particle dry deposition scheme for an atmospheric aerosol module. Atmos. Environ. 35, 549–560. http://dx. doi.org/10.1016/S1352-2310(00)00326-5. Zhang, Q., Jimenez, J.L., Canagaratna, M.R., Allan, J.D., Coe, H., Ulbrich, I., Alfarra, M.R., Takami, A., Middlebrook, A.M., Sun, Y.L., Dzepina, K., Dunlea, E., Docherty, K., DeCarlo, P.F., Salcedo, D., Onasch, T., Jayne, J.T., Miyoshi, T., Shimono, A., Hatakeyama, S., Takegawa, N., Kondo, Y., Schneider, J., Drewnick, F., Borrmann, S., Weimer, S., Demerjian, K., Williams, P., Bower, K., Bahreini, R., Cottrell, L., Griffin, R.J., Rautiainen, J., Sun, J.Y., Zhang, Y.M., Worsnop, D.R., 2007. Ubiquity and dominance of oxygenated species in organic aerosols in anthropogenically-influenced Northern Hemisphere midlatitudes. Geophys. Res. Lett. 34, L13801. http://dx.doi. org/10.1029/2007GL029979.
EPA, 2011. Air Quality Modeling Final Rule Technical Support Document. Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. http://www.epa.gov/crossstaterule/pdfs/AQModeling.pdf n.d. Erisman, J.W., van Pul, A., Wyers, P., 1994. Parameterization of dry deposition mechanisms for the quantification of atmospheric input to ecosystems. Atmos. Environ. 28, 2595–2607. Fowler, D., Pilegaard, K., Sutton, M.A., Ambus, P., Raivonen, M., Duyzer, J., Simpson, D., Fagerli, H., Fuzzi, S., Schjoerring, J.K., Granier, C., Neftel, A., Isaksen, I.S.A., Laj, P., Maione, M., Monks, P.S., Burkhardt, J., Daemmgen, U., Neirynck, J., Personne, E., Wichink-Kruit, R., Butterbach-Bahl, K., Flechard, C., Tuovinen, J.P., Coyle, M., Gerosa, G., Loubet, B., Altimir, N., Gruenhage, L., Ammann, C., Cieslik, S., Paoletti, E., Mikkelsen, T.N., Ro-Poulsen, H., Cellier, P., Cape, J.N., Horváth, L., Loreto, F., Niinemets, Ü., Palmer, P.I., Rinne, J., Misztal, P., Nemitz, E., Nilsson, D., Pryor, S., Gallagher, M.W., Vesala, T., Skiba, U., Brüggemann, N., Zechmeister-Boltenstern, S., Williams, J., O'Dowd, C., Facchini, M.C., de Leeuw, G., Flossman, A., Chaumerliac, N., Erisman, J.W., 2009. Atmospheric composition change: ecosystems–Atmosphere interactions. Atmos. Environ. 43, 5193–5267. http://dx.doi.org/10.1016/j.atmosenv. 2009.07.068. ACCENT Synthesis. Grell, G.A., Dudhia, J., Stauffer, D., 1994. A Description of the Fifth-Generation Penn State/NCAR Mesoscale Model (MM5). http://dx.doi.org/10.5065/D60Z716B. NCAR Technical Note NCAR/TN-398+STR. Guenther, A.B., Jiang, X., Heald, C.L., Sakulyanontvittaya, T., Duhl, T., Emmons, L.K., Wang, X., 2012. The Model of Emissions of Gases and Aerosols from Nature version 2.1 (MEGAN2.1): an extended and updated framework for modeling biogenic emissions. Geosci. Model Dev. 5, 1471–1492. http://dx.doi.org/10.5194/gmd-5-14712012. Hallquist, M., Wenger, J.C., Baltensperger, U., Rudich, Y., Simpson, D., Claeys, M., Dommen, J., Donahue, N.M., George, C., Goldstein, A.H., Hamilton, J.F., Herrmann, H., Hoffmann, T., Iinuma, Y., Jang, M., Jenkin, M.E., Jimenez, J.L., Kiendler-Scharr, A., Maenhaut, W., McFiggans, G., Mentel, T.F., Monod, A., Prévôt, A.S.H., Seinfeld, J.H., Surratt, J.D., Szmigielski, R., Wildt, J., 2009. The formation, properties and impact of secondary organic aerosol: current and emerging issues. Atmos. Chem. Phys. 9, 5155–5236. http://dx.doi.org/10.5194/acp-9-5155-2009. Im, Y., Jang, M., Beardsley, R.L., 2014. Simulation of aromatic SOA formation using the lumping model integrated with explicit gas-phase kinetic mechanisms and aerosolphase reactions. Atmos. Chem. Phys. 14, 4013–4027. http://dx.doi.org/10.5194/acp14-4013-2014. Knote, C., Hodzic, A., Jimenez, J.L., 2015. The effect of dry and wet deposition of condensable vapors on secondary organic aerosols concentrations over the continental US. Atmos. Chem. Phys. 15, 1–18. http://dx.doi.org/10.5194/acp-15-1-2015. Koo, B., Knipping, E., Yarwood, G., 2014. 1.5-Dimensional volatility basis set approach for modeling organic aerosol in CAMx and CMAQ. Atmos. Environ. 95, 158–164. http://dx.doi.org/10.1016/j.atmosenv.2014.06.031. Leungsakul, S., Jaoui, M., Kamens, R.M., 2005. Kinetic mechanism for predicting secondary organic aerosol formation from the reaction of d-limonene with ozone. Environ. Sci. Technol. 39, 9583–9594. Malm, W.C., Schichtel, B.A., Pitchford, M.L., 2011. Uncertainties in PM2.5 gravimetric and speciation measurements and what we can learn from them. J. Air Waste Manag. Assoc. 61, 1131–1149. http://dx.doi.org/10.1080/10473289.2011.603998. Morris, R.E., Koo, B., Guenther, A., Yarwood, G., McNally, D., Tesche, T.W., Tonnesen, G., Boylan, J., Brewer, P., 2006. Model sensitivity evaluation for organic carbon using two multi-pollutant air quality models that simulate regional haze in the southeastern United States. Atmos. Environ. 40, 4960–4972. http://dx.doi.org/10.1016/j. atmosenv.2005.09.088. Special issue on Model Evaluation: Evaluation of Urban and Regional Eulerian Air Quality Models. Novakov, T., Penner, J.E., 1993. Large contribution of organic aerosols to cloud-condensation-nuclei concentrations. Nature 365, 823–826. http://dx.doi.org/10.1038/ 365823a0. Odum, J.R., Hoffmann, T., Bowman, F., Collins, D., Flagan, R.C., Seinfeld, J.H., 1996. Gas/particle partitioning and secondary organic aerosol yields. Environ. Sci. Technol. 30, 2580–2585. http://dx.doi.org/10.1021/es950943+. Pankow, J.F., 2007. An absorption model of the gas/aerosol partitioning involved in the formation of secondary organic aerosol. Atmos. Environ. 41 (Suppl.), 75–79. http://
329