Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, Richland, Washington, USA
Corresponding author: E. Kassianov, Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, P.O. Box 999, MSIN K9-24, Richland, WA 99352 USA. (firstname.lastname@example.org)
 The majority of previous studies dealing with effect of coarse mode aerosols (supermicron) on the radiation budget have focused primarily on regions where total aerosol loadings are substantial. We reexamine this effect for a relatively clean area using a unique 1-month dataset collected during the recent Carbonaceous Aerosol and Radiative Effects Study (CARES, June 2010) in the central California region near Sacramento. Here we define “clean” as aerosol optical depths less than 0.1 at 0.5μm. We demonstrate that coarse mode particles contributed substantially (more than 50%) and frequently (up to 85% of time) to the total aerosol volume during this study. In contrast to conventional expectations that the radiative impact of coarse mode aerosols should be small for clean regions, we find that neglecting large particles may lead to significant overestimation, up to 45%, of direct aerosol radiative forcing despite very small aerosol optical depths. Our findings highlight the potential for substantial impacts of coarse mode aerosols on radiative properties over clean areas and the need for more explicit inclusion of coarse mode aerosols in climate-related observational studies.
 Over the past two decades, considerable progress has been achieved in sampling coarse mode aerosols (supermicron) and in estimating their contribution to extinction and aerosol optical depth in rural areas [e.g., Malm et al., 1994, 2007; Hand et al., 2004]. In contrast, however, the majority of previous studies dealing with the effect of coarse mode aerosols on regional and global radiation budgets and, consequently, on climate change, have focused primary on polluted regions [e.g., Eck et al., 2010; McFarlane et al., 2009; Yu et al., 2012]. Comparatively little is known about the relative importance of the coarse mode aerosols in shaping the radiation budget under clean conditions. Conventional wisdom suggests that coarse particles have diminished radiative impact in clean areas because the relative contribution of the coarse mode to the total aerosol mass is reduced with decreased aerosol loading.
 It is still unsettled whether these expectations are reasonable and have sufficient observational support. The required data are nuanced by well-known observational challenges. For example, the coarse mode fraction of atmospheric aerosols has been largely neglected by many routinein situ measurements due to limitations, such as line loss and instrument inlet issues, which prevent the sampling of large particles [e.g., Atkinson et al., 2010; Barnard et al., 2010]. Also, estimating aerosol intensive optical properties using passive remote sensing typically has been difficult under clean conditions [e.g., Dubovik et al., 2002].
 To quantify the relative impact of coarse mode aerosols on the radiation balance in a clean region, arbitrarily defined here as an aerosol optical depth smaller than 0.1 at 0.5 μm, we take advantage of an integrated dataset collected during the recent Carbonaceous Aerosol and Radiative Effects Study (CARES, http://campaign.arm.gov/cares/) held in June 2010, in central California near Sacramento. This area is routinely influenced by emissions from the urban San Francisco area, and seems to exhibit significant local summertime secondary organic aerosol production from both anthropogenic and biogenic sources [e.g., Fast et al., 2012; Zaveri et al., 2012]. The month-long clean period considered in this study is attributed mostly to the northwesterly flow and rain events just before, and during the early part, of the CARES campaign. The two major objectives of our study are to: (1) present observational evidence of the surprisingly large and frequent contribution of the coarse mode aerosol to the total volume usingin situ CARES data, and (2) demonstrate the unexpectedly important role of the coarse mode aerosol in influencing direct aerosol radiative forcing through application of radiative transfer calculations using CARES data.
2. Ground- and Aircraft-Based Observations
 The CARES campaign combined ground- and aircraft-basedin situobservations and aerosol remote sensing to examine, among other important subjects, the spatio-temporal changes of aerosol microphysical, optical and radiative properties [e.g.,Fast et al., 2012; Zaveri et al., 2012]. The CARES dataset also provides a great opportunity to understand the role of the coarse mode aerosols in influencing direct aerosol radiative forcing (DARF) [e.g., Ramaswamy et al., 2001] and aerosol radiative forcing efficiency [e.g., García et al., 2008], which is the ratio of the DARF over aerosol optical depth (τa) at 0.5 μm. Here, the DARF is defined as the difference between upwelling and downwelling clear-sky fluxes obtained with and without aerosols at the top-of-atmosphere (TOA).
 The CARES campaign combines data from two surface sites referred to as “T0” and “T1” located within and northeast of the Sacramento urban area, respectively. These sites, separated by about 40 km, were equipped with a wide range of in situ, passive and active instruments for measuring aerosol properties and surface spectral and broadband fluxes (http://campaign.arm.gov/cares/). In particular, an Aerodynamic Particle Sizer (APS) at the T0 and T1 sites measured the aerosol size distribution in the aerodynamic diameter (Da) range 0.5–20.0 μm, which represents mostly the coarse mode. Another instrument, the Scanning Mobility Particle Sizer (SMPS), measured fine particles in the mobility diameter range 0.012–0.737 μm and 0.008–0.858 μm at the T0 and T1 sites, respectively. It should be noted that the SMPS and APS instruments use different measurement principles and, thus, the SMPS and APS data are not directly comparable (SMPS mobility diameter versus APS aerodynamic diameter). To obtain combined SMPS-APS size distributions (Figures 1a and 1b), we convert the SMPS-measured aerosol mobility size distributions into their aerodynamic size counterparts by assuming the particles are spherical and using well-known relationships and an assumed density of the aerosol (1.8 g/cm3), similar to Barnard et al.  .
 The DOE G-1 research aircraft was equipped with a suite of meteorological, aerosol and cloud instruments. The Cloud and Aerosol Spectrometer (CAS), as part of the Cloud, Aerosol and Precipitation Spectrometer probe, measured the aerosol optical size distribution in the optical diameter range 0.51–50μm. Similar to the APS, the CAS data represent mostly coarse mode aerosols. To compare coarse mode aerosols on the ground and aloft, we consider the CAS data when the G-1 was within a 5-km radius of a given ground site, either T0 or T1. Also, we disregard the CAS data during steep climbs and descents of the aircraft associated with changing sampling altitudes, takeoff and landing.
 In addition to the in situobservations, each site includes data from radiometric, active and passive remote-sensing measurements. The passive remote sensing instruments include the Multifilter Rotating Shadowband Radiometer (MFRSR), the National Aeronautics and Space Administration (NASA) Aerosol Robotic Network (AERONET) Sun photometer [e.g.,Holben et al., 1998], and an Eppley PSP broadband radiometer. The MFRSR measured the total all-sky surface downwelling irradiance and its diffuse component at six wavelengths (0.415, 0.5, 0.615, 0.673, 0.87, and 0.94μm) at both sites. The spectral values of columnar τa were derived from the direct (total minus diffuse) component [Harrison and Michalsky, 1994] (uncertainties of the derived τaare 0.01–0.02), while information on the columnar fine and coarse modes and intensive aerosol optical properties, such as single-scattering albedo (ω0) and asymmetry factor (g), were estimated from the direct and diffuse irradiances using a MFRSR-based aerosol retrieval [Kassianov et al., 2007, 2011a]. AERONET data (level 2) were collected at the T0 site only during 4 days and contained size distributions and spectrally resolved τa. The AERONET- and MFRSR-derivedτa and size distributions (radius range of 0.05–15 μm) were comparable for these 4 days [Kassianov et al., 2011b]; both the AERONET- and MFRSR-derived columnar size distributions show the frequent occurrence of a large coarse mode fraction [Kassianov et al., 2011b; Zaveri et al., 2012] (see auxiliary material).
 We previously evaluated the MFRSR-derived columnar aerosol optical properties (τa, ω0, and g) using a radiative closure experiment [Kassianov et al., 2011b], where these optical properties served as input to the Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART) model [Ricchiazzi et al., 1998]. Results from the radiative closure experiment indicate that calculated broadband irradiances match measured irradiances quite well (bias is about 3 Wm−2; see auxiliary material). The favorable agreement between model output and observations suggests that, on the average, the MFRSR-derived aerosol optical properties are reasonable. We also note that our retrieval technique has been applied to MFRSR data from a very dusty site, Niamey, Niger [Slingo et al., 2006], and these retrievals successfully captured the large coarse mode often found in this region.
3. Aerosol Microphysical Properties
 We use the SMPS-APS data (Figures 1a and 1b) for calculating the coarse mode fraction of the total volume concentration, defined as a ratio of the coarse mode volume (Cv,coarse) over the total (Cv,total) volume: ηv = Cv,coarse/Cv,total [e.g., Hand et al., 2002]. We further define the coarse mode fraction as that portion of the observed size distribution that is larger than a specified threshold. Accordingly, we adopt two popular thresholds for defining fine mode particles: particles with an aerodynamic diameter, Da, less than 1.0 μm (called PM1.0 aerosols) and 2.5 μm (called PM2.5aerosols), respectively. Time series (1-h averages) of the volume coarse mode fraction are used for creating the corresponding cumulative percentages for the PM1.0 and PM2.5 aerosols (Figure 1c). The cumulative percentage indicates the percentage of time when ηv ≥ ηv*, where ηv* is a specified level.
 Let us take, for example, ηv* = 50%. The PM2.5 cumulative percentage curves indicate that about 16% and 26% of the time ηv ≥ ηv* at the T0 and T1 sites, respectively. In other words, for about 26% of the time at the T1 site, the coarse mode contribution to the total aerosol volume exceeds 50%, for a coarse mode defined as having Da exceeding 2.5 μm. For the PM1.0 aerosols, the corresponding percentage of time is about 85% for both the T0 and T1 sites (Figure 1c). These figures imply that comparable (∼50%) contributions of the coarse mode and fine mode particles to the total volume occur frequently -- a finding that is roughly consistent withMalm et al. . Thus, ignoring coarse mode particles can frequently lead to underestimating the 1-h total volume averages by a factor of 2. The observed quantitative difference for the PM2.5 aerosols (T0 versus T1, Figure 1c) is attributed mostly to local sources and meteorology described in previous studies [Fast et al., 2012; Zaveri et al., 2012].
 Occurrence of the coarse mode particles above the T0 and T1 sites is illustrated by the aircraft-basedin situ data (Figure 2). When aircraft sampling altitudes are low (≤ 0.6 km), there is a very good agreement between the coincident and co-located ground-based APS and the aircraft-based CAS total concentrations (Figure 2a). The relation between them can be approximated by a linear expression with high correlation coefficient (0.97) and large regression slope (0.88). Such periods with low sampling altitudes occurred only over the T0 site (Figure 2a). For the T1 site, the sampling altitudes were higher (Figure 2b) and the regression slope becomes smaller (0.34) while the correlation coefficient remains high (0.90). In the absence of elevated aerosol layers, total particle concentrations typically decrease with altitude, which likely explains why the APS-CAS regression slope is reduced substantially from 0.88 (flight altitude ≤ 0.6 km; T0 site) to 0.34 (flight altitude > 0.6 km; T1 site).
 Day-to-day variations ofin situnear-surface (Figures 1a and 1b) and MFRSR-derived columnar (seeauxiliary material) coarse mode fractions generally have similar temporal patterns. Since the MFRSR-derived columnar microphysical properties (size distribution and size independent refractive index) provide “radiatively consistent” columnar optical properties of the aerosol load [Kassianov et al., 2011b], we apply these microphysical properties for assessing the influence of coarse mode particles on TOA aerosol radiative forcing.
4. Aerosol Optical Properties and Radiative Forcing
 The assessment is performed using these three steps. First, we convert the MFRSR-derived size distribution (called “original”) into a corresponding PM1.0 size distribution by removing coarse particles (aerodynamic diameter Da ≥ 1.0 μm). As a result, the PM1.0 size distribution is a truncated version of the original distribution and includes only particles with Da less than 1.0 μm. Second, we calculate the corresponding PM1.0 intensive aerosol optical properties (ω0 and g) by assuming that particles can be approximated by homogeneous spheres. Finally, we compute PM1.0 DARF and forcing efficiency at the TOA using the calculated PM1.0 intensive properties (ω0 and g) and the MFRSR-derived, unmodifiedτa. This scheme mimics the well-established technique of calculating radiative fluxes when a columnarτa is measured, but only surface measurements of ω0 and g are available [e.g., Michalsky et al., 2006; McComiskey et al., 2008]. The daily-averaged (averaged over 24 hours) values of DARF are obtained through application of the SBDART model and previously described approaches [e.g.,Barnard and Powell, 2002; Michalsky et al., 2006]. We repeat these three steps for obtaining the radiative forcing associated with the PM2.5 aerosols.
 Comparison of the MFRSR-derived intensive properties (ω0 and g) with their PM1.0 and PM2.5 counterparts suggests that the removal of the large particles leads to an increase of ω0 and a decrease of g (Table 1). The consequence of these joint changes of ω0 and g is that the amount of radiation reflected back to the space becomes larger [e.g., Hansen and Travis, 1974] and, thus, the absolute magnitude of the DARF (Figures 3a and 3b) and the forcing efficiency (Figures 3c and 3d) increases. For the majority of observational conditions and surface types, the DARF is negative -- aerosol cooling -- and it becomes more negative with increasedτa and ω0 and decreased g [e.g., McComiskey et al., 2008]. In comparison with the T1 site, the T0 site has larger, on average, τa values (Table 1). Thus, the DARF magnitude has larger negative values at the T0 site (Figures 3a and 3b and Table 1), i.e., a greater cooling effect exists relative to T1. Since the forcing efficiency represents the amount of forcing per unit of τa, the efficiency is less sensitive (in comparison with the DARF) to the aerosol loading and its values are comparable at the T0 and T1 sites (Figures 3a and 3b versus Figures 3c and 3d). Overall, the results embodied in Figure 3 and Table 1 imply that neglecting large particles may substantially increase the calculated cooling effect of aerosols, up to 45% and 30% for the PM1.0 and PM2.5 cases, respectively.
Table 1. Averaged Values of Aerosol Optical Properties (τa, ω0, and g) at 0.5 μm and Direct Aerosol Radiative Forcing (DARF) at the Top of Atmosphere for the Three Cases (Original, PM2.5 and PM1.0) at the Two Sites (T0 and T1)a
The corresponding relative differences between original values and their PM2.5 and PM1.0 counterparts are included as well. The averaged values represent all twenty days considered in Figure 3. DARF units are Wm−2.
 We have assessed the role of coarse mode aerosols in direct aerosol radiative forcing (DARF) under relatively clean conditions (τa ∼ 0.1 at 0.5 μm wavelength) by using 1-month of data from the June 2010 CARES field campaign conducted near Sacramento, California, USA. We used ground-basedin situ data from the Aerodynamic Particle Sizer (APS) and Scanning Mobility Particle Sizer (SMPS) instruments at two sites to illustrate that the large coarse mode fraction, defined as ratio of the coarse mode volume to the total volume exceeding 50%, was observed frequently (up to 85% of time when coarse mode aerosols are defined as larger than PM1.0aerosols). Passive ground-based retrievals of columnar size distribution (radius range 0.05–15μm) from the Multi-filter Rotating Shadowband Radiometer (MFRSR) and AERONET observations [Kassianov et al., 2011b; Zaveri et al., 2012], and independent aircraft-basedin situdata from the Cloud and Aerosol Spectrometer (CAS), also illustrate the frequent occurrence of the large coarse mode fraction despite very low aerosol loadings. We estimated the relative contribution of the coarse mode particles to the DARF at the top-of-atmosphere using aerosol optical properties obtained from the MFRSR-derived size distributions and their truncated counterparts consisting of PM1.0 and PM2.5aerosols. The results of the radiative transfer calculations indicate that relative impact of the coarse mode particles on the DARF can be substantial (up to 30–45%) -- contrary to conventional expectations.
 The DARF is commonly estimated using measured column aerosol optical depth and surface intensive aerosol properties, such as single-scattering albedo and asymmetry factor, obtained from ground-basedin situ measurements [e.g., McComiskey et al., 2008, and references therein]. This approach should be especially relevant for clean regions where the retrievals of aerosol intensive properties from sun/sky measurements represent a great challenge and, therefore, surface-based measurements must be used. However, many routinein situ measurements were designed for sampling mostly fine mode aerosols. Results from our study suggest that estimations of the DARF based on such fine mode in situ measurements may largely overlook the important role of coarse mode aerosols in influencing radiative properties under clean conditions. Thus, reliance on such fine mode in situ measurements may hinder the ability of models to accurately predict aerosol effects on the regional and global radiation budget. Deployment of in situ instruments with increased observational capabilities for sampling both the fine mode and coarse mode aerosols would provide the needed information on the coarse mode fraction and, thus, would help with the assessment and subsequent improvement of model predictions.
 This work has been supported by the Office of Biological and Environmental Research (OBER) of the U.S. Department of Energy (DOE) as part of the Atmospheric Radiation Measurement (ARM) and Atmospheric Systems Research (ASR) Programs. The Pacific Northwest National Laboratory (PNNL) is operated by Battelle for the DOE under contract DE-A06-76RLO 1830. Our recognition is also extended to those responsible for the operation and maintenance of the instruments that produced the ARM Archive data retrieved for use in this study. Special thanks to Bertram Jobson (WSU), Chen Song (PNNL), Qi Zhang (UC Davis), Stephen Springston (BNL), and Gunnar Senum (BNL). We appreciate helpful discussions with Richard Ferrare and Chris Hostetler (NASA LRC) about the AERONET measurements. We also appreciate the comments of two reviewers that have significantly improved this paper.
 The Editor thanks two anonymous reviewers for assistance evaluating this paper.