An analysis of seasonal surface dust aerosol concentrations in the western US (2001–2004): Observations and model predictions

Atmospheric Environment 41 (2007) 6585–6597 An analysis of seasonal surface dust aerosol concentrations in the western US (2001–2004): Observations and model predictions Kelley C. Wells a,/C3, Marcin Witek b,c, Piotr Flatau d, Sonia M. Kreidenweis a, Douglas L. Westphal e aDepartment of Atmospheric Science, 1371 Campus Delivery, Colorado State University, Fort Collins, CO 80523, USA bInterdisciplinary Centre of Mathematical and Computational Modeling, University of Warsaw, Pawinskiego 5a, 02-106 Warsaw, Poland cInstitute of Geophysics, University of Warsaw, Pasteura 7, 02-093 Warsaw, Poland dScripps Institution of Oceanography, University of California, San Diego, CA 92093, USA eNaval Research Laboratory, Marine Meteorology Division, 7 Grace Hopper Ave, Stop 2, Monterey, CA 93943, USA Received 8 November 2006; received in revised form 12 April 2007; accepted 13 April 2007 Abstract Long-term surface observations indicate that soil dust represents over 30% of the annual fine (particle diameter less than 2.5 mm) particulate mass in many areas of the western US; in spring and summer, it represents an even larger fraction. There are numerous dust-producing playas in the western US, but surface dust aerosol concentrations in this region are also influenced by dust of Asian origin. This study examines the seasonality of surface soil dust concentrations at 15 western US sites using observations from the Interagency Monitoring of PROtected Visual Environments (IMPROVE) network from 2001 to 2004. Average soil concentrations in particulate matter less than 10 mm in diameter (PM 10) were lowest in winter and peaked during the summer months at these sites; however, episodic higher-concentration events (410 mgm /C0 3) occurred in the spring, the time of maximum Asian dust transport to the western US. Simulated surface dust concentrations from the Navy Aerosol Analysis and Prediction System (NAAPS) suggested that long-range transport from Asia dominates surface dust concentrations in the western US in the spring, and that, although some long-range transport does occur throughout the year (1–2 mgm /C0 3), locally generated dust plays a larger role in the region in summer and fall. However, NAAPS simulated some anomalously high concentrations ( 450 mgm /C0 3) of local dust in the fall and winter months over portions of the western US. Differences between modeled and observed dust concentrations were attributed to overestimation of total observed soil dust concentrations by the assumptions used to convert IMPROVE measurements into PM 10 soil concentrations, lack of inhibition of model dust production in snow-covered regions, and lack of seasonal agricultural sources in the model. r 2007 Elsevier Ltd. All rights reserved. Keywords: Dust; Long-range transport; Particulate monitoring; Rural aerosol ARTICLE IN PRESS www.elsevier.com/locate/atmosenv 1352-2310/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2007.04.034 /C3Corresponding author. Tel.: +1 970 491 7484; fax: +1 970 491 8483. E-mail address: kcjohnso@atmos.colostate.edu (K.C. Wells).

1. Introduction Atmospheric aerosols can have a marked impact on health, air quality, and visibility on both global and regional scales. Furthermore, the role of aerosols in the global energy budget has been examined in many observational and modeling studies and their estimated radiative forcing can be similar in magnitude to that of greenhouse gases (IPCC, 2001 ). While many studies have focused on particulate sulfate and black carbon, natural atmo- spheric aerosols such as mineral dust represent a large fraction of the total atmospheric aerosol mass loading, and their impact is not as well understood. Between 1000 and 3000 Tg of dust are emitted into the atmosphere each year ( Duce, 1995), exerting the largest annual mean direct radiative effect of all aerosol types over ocean at solar wavelengths (Myhre et al., 2004 ). Additionally, soil dust particulate matter typically lies in a larger size mode than anthropogenic species; their larger size enables them to exert a significant longwave radiative effect ( Haywood et al., 2005 ). Dust is often transported far from its region of origin, providing a source of nutrients to ocean biota (Duce, 1995 ) and possibly enhancing the formation of ice clouds at warmer temperatures (e.g. Sassen, 2002). At the Earth’s surface, dust storms can lead to the exceedance of air quality standards in the US (Reid et al., 1994 ; Prospero et al., 2002 ). Analyses of long-term data sets have shown that soil dust is a particularly important component of the fine aerosol mass in the western US, comprising over 30% of the annual fine aerosol mass in the desert Southwest and up to 40% of the fine aerosol mass over most of the region in spring ( Malm et al., 2004). Although there are significant dust-produ- cing regions in the western US (e.g. Prospero et al., 2002), the region is impacted by Asian dust in the spring months (e.g. VanCuren and Cahill, 2002 ; Thulasiraman et al., 2002 ), leading to surface aerosol mass concentrations that can exceed 100 mgm /C0 3 (Husar et al., 2001 ). The potential for contributions to surface dust concentrations from both local and long-range dust transport compli- cates attempts at source apportionment. Indeed, previous studies have found difficulties with using chemical aerosol signatures to determine dust origin (VanCuren and Cahill, 2002 ; Liu et al., 2003 ); thus, models may currently provide the best means for estimating source contributions to surface soil dust concentrations. This study investigates seasonal variations in surface dust concentrations in the western continental US from 2001 to 2004 in in situ, filter-based aerosol observations from the Inter- agency Monitoring of PROtected Visual Environ- ments (IMPROVE) network. To examine the contributions of local production and long-range transport to the observed dust concentrations, we compare the observations to simulations of surface soil dust concentration from an operational global transport model, the Navy Aerosol Analysis and Prediction System (NAAPS). 2. Methodology and data sets IMPROVE samplers have modules that collect 24-h integrated samples on Teflon filters, every third day, of both PM 10 (aerodynamic diameter less than 10 mm) and PM 2.5 (aerodynamic diameter less than 2.5 mm, also called the fine mode) aerosol fractions (Malm et al., 2004 ). The total aerosol mass in both fractions is measured gravimetrically. Two addi- tional filter samples (on Nylasorb and quartz fiber filters) are also collected from the PM 2.5 inlet stream. In routine operations, only the IMPROVE PM 2.5 filters are analyzed for chemical composition, including the concentrations of most elements. Since ‘‘soil dust’’ is estimated indirectly from elemental concentrations, IMPROVE sites only report soil dust aerosol amounts that appear in the fine mode. The fine soil concentration ( mgm /C0 3) is reconstructed from the mass concentrations of soil-derived ele- ments (Al, Si, Ca, Fe, K, and Ti) and their normal oxides (Al 2O3, SiO2, CaO, K 2O, FeO, Fe 2O3, TiO2), along with a correction factor to account for other species such as water and carbonate, using ( Malm et al., 1994 ) ½Soil/C138¼ 2:20½Al/C138þ 2:49½Si/C138þ 1:63½Ca/C138 þ 2:42½Fe/C138þ 1:94½Ti/C138 . ð1Þ We used IMPROVE data from 15 sites in the western US in operation for the entire time period studied, 2001–2004 ( Table 1, with locations denoted as points in Figs. 1 and 2 ). The time period was chosen to include a year with a known significant Asian dust event in the western US (2001), as well as subsequent years in which the IMPROVE stations of interest were in operation. The 15 sites studied were chosen based on their location (to capture a wide spatial range in the western US, and to investigate specific model behaviors) and data availability during the study period. We removed ARTICLE IN PRESS K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–65976586

all data that were flagged in the database as not meeting quality control standards, with the excep- tion of those flagged for moderate ( o20%) changes in flow rate, and zeroed all fine soil concentrations that were below 3 times the minimum detection limit, which was determined to be 0.03 mgm /C0 3 (the sum of the uncertainties in the components in Eq. (1)). The mass concentration of the coarse aerosol fraction was then estimated for the valid IMPROVE samples at each site by subtracting the fine mass from the PM 10 aerosol mass, [PM 10]– [PM2.5], where [PM 10] and [PM 2.5] are gravimetric mass concentrations. The coarse fraction was assumed to contain only insoluble soil particles ARTICLE IN PRESS Table 1 Site name, location, and elevation information for the 15 western IMPROVE sites investigated in this study Site code Site name Latitude (N) Longitude (W) Elevation (m) BADL Badlands National Park, South Dakota 43.74 /C0 101.94 736 BAND Bandelier National Monument, New Mexico 35.78 /C0 106.27 1987 KALM Kalmiopsis, Oregon 42.55 /C0 124.06 90 LABE Lava Beds National Monument, California 41.71 /C0 121.51 1469 LOST Lostwood, North Dakota 48.64 /C0 102.40 692 MELA Medicine Lake, Montana 48.49 /C0 104.48 605 MEVE Mesa Verde National Park, Colorado 37.20 /C0 108.49 2177 MONT Monture, Montana 47.12 /C0 113.15 1293 MORA Mount Rainier National Park, Washington 46.76 /C0 122.12 427 ROMO Rocky Mountain National Park, Colorado 40.28 /C0 105.55 2755 SAWT Sawtooth National Forest, Idaho 44.17 /C0 114.93 1980 SEQU Sequoia National Park, California 36.49 /C0 118.83 535 SYCA Sycamore Canyon, Arizona 35.14 /C0 111.97 2040 THRO Theodore Roosevelt, North Dakota 46.89 /C0 103.38 853 YELL Yellowstone National Park, Wyoming 44.57 /C0 110.40 2425 Fig. 1. The 2001–2004 seasonally averaged NAAPS simulated non-North American surface dust concentration ( mgm /C0 3). K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–6597 6587

(Malm et al., 2007 ). Thus, we define the total IMPROVE PM 10 soil dust aerosol mass concentra- tion as the sum of the total mass concentration in the coarse fraction and the computed soil mass concentration in the fine mode: PM 10 Soil ¼½ PM10/C138/C0½ PM2:5/C138þ½ Soil/C138 . (2) In addition, a special study was conducted at nine selected IMPROVE sites from 19 March 2003 through the end of 2004 ( Malm et al., 2007 ). During the special study, Nylasorb and quartz fiber filter samples were also collected from the PM 10 inlet, and both PM 10 and PM 2.5 fractions were analyzed for chemical composition. Differencing the mass con- centrations of aerosol species found in the PM 10 and PM2.5 fractions provides an estimate of the compo- sition of the coarse mass. Results showed that although soil is the single largest contributor to coarse mass concentrations, it is certainly not the only contributor. The average annual fractional contribution of soil to coarse mass at the five western US sites studied ranged from 32% to 79%. The soil fraction at individual sites also varied seasonally. Since there is no simple correction that can be applied to Eq. (2) for all the network data, in this work we report PM 10 soil mass concentrations estimated for the limiting assumption that 100% of the coarse mass is due to soil, recognizing that this should represent a maximum contribution. We consider the impacts of this assumption on the comparisons between observations and simulations. Global simulations of the generation and trans- port of soil dust aerosol were performed using the NAAPS ( Witek et al., 2007 ), a transport model driven by wind fields from the Navy Operational Global Atmospheric Prediction System (NOGAPS) (Hogan and Rosmond, 1991 ). Two NAAPS simula- tions for the time period 2001–2004 were run for this study: one in which all of the model’s dust sources were kept active, and one in which North American dust sources were suppressed to isolate long-range transport to North America. The model was run on a 1 1 /C2 11 horizontal grid with 24 vertical levels, using archived NOGAPS data fields as meteorological input every 6 h during the simula- tion. The integration time step during the simula- tions was set to 30 min for vertical diffusion, and 1 h for all other model processes. Global sources of atmospheric mineral dust are defined in NAAPS by dust-producing regions as they are identified in the USGS Land Cover Characteristics Database with modifications based ARTICLE IN PRESS Fig. 2. The 2001–2004 seasonally averaged NAAPS simulated surface dust concentration ( mgm /C0 3). Note that scale differs from that in Fig. 1 . K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–65976588

on an analysis of TOMS AI data ( Prospero et al., 2002). NOGAPS friction velocity, a parameter accounting for both surface wind speed and surface roughness, is utilized to predict dust mobilization from these source regions. Dust is emitted whenever the friction velocity exceeds 60 cm s /C0 1 and the predicted surface moisture fraction is below 0.3. These thresholds are based upon a study of mobilization and transport of Saharan dust ( West- phal et al., 1988 ), and thus may not be entirely appropriate for all dust-producing areas of the world. In NAAPS, the dust flux is scaled to represent the fraction of dust particles smaller than 10 mm in diameter, assumed to have a mass mean diameter of 2 mm. Studies such as that by Maring et al. (2003) show that particles larger than 7.3 mmi n diameter are preferentially scavenged within the first two days of transport, so the assumed size distribu- tion of dust in NAAPS is likely reasonable for long- range transport events. Seasonal average maps of NAAPS simulated surface dust concentration from the run with suppressed North American sources are displayed in Fig. 1 . Very little non-North American dust aerosol is simulated on average in the fall and winter months over the western US, whereas NAAPS simulates a widespread influence of non-North American dust sources in the spring months. The simulated annual cycle is consistent with studies that have argued the presence of Asian dust aerosol in the western US throughout the year, with larger events occurring in the springtime (e.g. VanCuren and Cahill, 2002 ). Higher surface concentrations of non-North American dust are simulated over Florida and the Caribbean in the summer, the season in which Saharan dust events are most often observed in the eastern US ( Perry et al., 1997 ). Fig. 2 contains seasonal average maps from the NAAPS simulation in which all dust sources remained active. NAAPS simulates strong, localized dust emissions from source regions in the Colorado Plateau and the upper Great Plains, reaching a maximum in the fall months. Simulated dust mass concentrations at the loca- tions of IMPROVE sites were extracted from the model output to compare with observations. To account for potential small discrepancies in the spatial distribution of NAAPS versus real-world soil dust that might be due to the model resolution and topographical smoothing, the computed con- centrations for the lowest NAAPS layer were averaged over 2 1 /C2 21 boxes with the midpoint chosen to correspond most closely to the IM- PROVE site location. Measured surface aerosol mass concentrations are reported as 24-h averages (midnight-to-midnight local time) every three days. NAAPS predictions were thus also averaged over 24-h time periods using the simulations valid at 12, 18, 00 and 06 UTC. 3. Results 3.1. Observations from selected IMPROVE sites This section contains results from select time periods at four of the 15 sites investigated in this study. We chose to focus our discussion on these four sites because they each highlight different aspects of the model-measurement comparison. In the following timelines, the gray line corresponds to the NAAPS simulation in which all dust sources were active, the black line represents the NAAPS simulation with North American sources sup- pressed, and the black circles correspond to the IMPROVE measurements. Whenever a gray line peaks above the black line in the following time- lines, the simulation indicates the presence of North American dust. At times when only a black line is visible, the model simulates dust of only non-North American origin. 3.1.1. Kalmiopsis, Oregon Kalmiopsis is an IMPROVE site located in western Oregon, about 28 miles inland from the coast. The sampler is at an elevation of 90 m. The site’s location is ideal for examining the influence of long-range transport of dust aerosol from Asia on aerosols in the boundary layer, the importance of which has been shown in previous studies (e.g. Husar et al., 2001 ; McKendry et al., 2001 ). The site receives mostly marine air masses from the west, so the potential influence of US and North African dust sources is minimal. Timelines of NAAPS simulated soil dust and IMPROVE measurements at Kalmiopsis for 2001 are shown in Fig. 3 . The IMPROVE data at Kalmiopsis indicate very low concentrations in the winter ( o2 mgm /C0 3), concentrations up to 5 mgm /C0 3 in other seasons, and episodic larger peaks in the spring months. Conditions at the site are relatively clean overall; IMPROVE soil concentrations never exceed 12 mgm /C0 3 during the entire study period at Kalmiopsis. The NAAPS simulation indicates no influence of North American dust sources at this ARTICLE IN PRESS K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–6597 6589

location during 2001, and generally low total concentrations of soil dust. The bulk of the dust mass simulated annually at Kalmiopsis appears in spring and early summer. A period of episodic non- North American dust incursions over Kalmiopsis from March to June is simulated each year in the four-year NAAPS simulation (not shown). Given that this is the period of peak Asian dust transport to the US, we conclude that Kalmiopsis is impacted by Asian dust transport each spring, resulting in surface dust concentrations 2–6 times background conditions (1–2 mgm /C0 3). The large peak on 14 April 2001 is consistent with the presence of an optically thick Asian dust plume, which was observed over the US via Sun photometry ( Thulasiraman et al., 2002); however, NAAPS indicates that the back- ground-level concentrations in winter are also likely of Asian origin. This result is consistent with that of VanCuren and Cahill (2002) , who argue that Asian dust reaches the US throughout the year. The secondary peak of non-North American dust simulated by NAAPS on 24 October 2001 occurs during the secondary period of Asian dust influence in the western US found in the 2002 study. 3.1.2. Sawtooth National Forest, Idaho The potential effects of surface elevation on dust concentration due to long-range transport are demonstrated in a comparison of measured and simulated dust mass at Kalmiopsis and Sawtooth National Forest, Idaho ( Fig. 4), a station located at an elevation of 1980 m. Measured and simulated PM 10 soil amounts at Sawtooth National Forest are nearly double those at Kalmiopsis in the spring, indicating the elevated transport of Asian dust to North America. VanCuren and Cahill (2002) found a more pronounced Asian dust influence at elevated western US sites as well. The large spikes in North American dust concentration simulated by NAAPS in fall at Sawtooth National Forest will be addressed in Section 3.1.4, and in a sensitivity study discussed in Section 4. Most of the spring episodic peaks in dust aerosol simulated by the model coincide with peaks in the IMPROVE data, though NAAPS tends to simulate higher peak concentrations than are reported by IMPROVE. This may be due to an underestimation of removal processes in NAAPS, resulting in a surplus of dust being transported in elevated plumes from Asia. For the four-year simulation, NAAPS simulated a dry dust lifetime of /C24 24 days, a value that is over twice as long as estimates from five other models ( Miller et al., 2004 ). An overestima- tion of total dust emission could also be a contributing factor (see Section 3.1.4). 3.1.3. Sequoia National Park, California The Sequoia National Park IMPROVE station is located in the southern region of the Sierra Nevada Mountains, at an elevation of 535 m. The site was one of the few included in the special PM 10 speciation study ( Malm et al., 2007 ). A timeline of IMPROVE PM 10 soil (computed by applying ARTICLE IN PRESS 01/01/01 04/01/01 07/01/01 10/01/01 01/01/02 0 5 10 15 20 25 30 35 40 Date Concentration (μg/m3) Kalmiopsis, OR 2001 NAAPS NAAPS No N.A. Dust IMPROVE [PM10 −PM2.5]+[Soil] Fig. 3. The 2001 Kalmiopsis, Oregon timelines of NAAPS simulated dust concentration ( mgm /C0 3) with and without North American sources and IMPROVE [PM 10/C0 PM2.5]+[Soil] con- centration ( mgm /C0 3). 01/01/01 04/01/01 07/01/01 10/01/01 01/01/02 0 10 20 30 40 50 60 70 80 90 100 Date Concentration (μg/m3) Sawtooth NF, ID 2001 NAAPS NAAPS No N.A. Dust IMPROVE [PM10−PM2.5]+[Soil] Fig. 4. The 2001 Sawtooth National Forest, Idaho timelines of NAAPS simulated dust concentration ( mgm /C0 3) with and without North American sources and IMPROVE [PM 10/C0 PM2.5]+[Soil] concentration ( mgm /C0 3). Note that scale differs from that in Fig. 3 . K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–65976590

Eq. (1) to the speciated PM 10 data) and NAAPS dust simulations for the special study period is presented in Fig. 5 . Like the other sites examined, Sequoia is impacted by dust of almost exclusively Asian origin in the spring, resulting in surface concentrations greater than 10 mgm /C0 3. Although NAAPS accurately simulates the spring Asian dust concentrations at the site, it misses the large (420 mgm /C0 3), persistent presence of PM 10 soil at Sequoia during summer and fall. Very little local dust production is simulated near Sequoia, despite the fact that the model surface is highly erodible in the vicinity of the site. Anthropogenic activities have been shown to play a large role in the seasonal cycle of soil dust aerosol at the measurement site in Sequoia National Park. A study of PM 10 source apportionment in the San Joaquin Valley (SJV) from 1988 to 1989 revealed that 54% of the summer and fall PM 10 mass in the valley was attributable to fugitive dust from tilling of fields, travel on unpaved roadways, and con- struction ( Chow et al., 1992 ). The measurement site in Sequoia National Park is at an elevation (535 m) that would be impacted by the SJV boundary layer during the day in summer and fall, and thus the observed high PM 10 soil levels in these seasons may reflect such a source. NAAPS has no anthropogenic or seasonal component in its dust source function, and thus cannot reproduce such a trend. A low bias in NOGAPS wind speed (see Section 4) could also be to blame for this. The influence at Sequoia of additional dust sources not represented in NAAPS was confirmed because of the availability of the special study speciated PM 10 data, which allowed us to rule out an overestimation larger than /C24 15% by Eq. (2) of dust contributions during summer. The PM 10 soil computed using Eq. (2) ( Fig. 5) tracks the speciated PM10 soil concentration fairly closely throughout the year, with some periodic overestimations of 2– 10 mgm /C0 3 (which, during the clean winter months, can represent 50–100% of the total measured soil mass). The coarse mode was composed largely of soil at Sequoia during the spring, summer and fall of the PM 10 speciation study, ranging from /C24 70 to 80% of the total coarse mass. In winter, however, soil made up just 6% of the average aerosol coarse fraction at Sequoia. Other sites revealed a different seasonal pattern for 2004. The same analysis for Mount Rainier National Park in Washington (Fig. 6 ) reveals that Eq. (2). consistently over- estimates the actual soil mass in the fall months by 5–8 mgm /C0 3, which is 2–4 times the average total PM10 measured at the site in fall. An analysis of the special study data at this site revealed that organic species (OC) actually comprise most of the coarse mass at this site in fall. This could be the case at other sites examined in this study as well. OC ARTICLE IN PRESS 09/01/03 12/01/03 03/01/04 06/01/04 09/01/04 12/01/04 0 10 20 30 40 50 60 70 80 90 100 Date Concentration (μg/m3) Sequoia NP, CA 2004 Special Study NAAPS NAAPS No N.A. Dust IMPROVE [PM10−PM2.5]+[Soil] IMPROVE PM10 Soil Fig. 5. Special study Sequoia National Park, California timelines of NAAPS simulated dust concentration ( mgm /C0 3) with and without North American sources, IMPROVE PM 10 soil con- centration ( mgm /C0 3), and IMPROVE [PM 10/C0 PM2.5]+[Soil] con- centration ( mgm /C0 3). The IMPROVE PM 10 soil concentration values were computed by applying Eq. (1) to the speciated PM 10 data. Note that scale differs from that in Fig. 3 . 11/01/03 02/01/04 05/01/04 08/01/04 11/01/04 0 10 20 30 40 50 60 Date Concentration ( μg/m3) Mount Rainier National Park, WA 2004 Special Study NAAPS NAAPS No N.A. Dust IMPROVE [PM10−PM2.5]+[Soil] IMPROVE PM10 Soil Fig. 6. Special study Mount Rainier National Park, Washington timelines of NAAPS simulated dust concentration ( mgm /C0 3) with and without North American sources, IMPROVE PM 10 soil concentration ( mgm /C0 3), and IMPROVE [PM 10/C0 PM2.5]+[Soil] concentration ( mgm /C0 3). Note that scale differs from that in Fig. 3 . K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–6597 6591

constitutes the second-largest fraction of coarse mass in the western United States ( Malm et al., 2007), due to the extensive biogenic sources and fire activity in the region. We conclude that the assumption that the coarse mode contains only dust aerosol (Eq. (2)) leads to an overestimation of total PM 10 dust aerosol concentrations for most western US IMPROVE sites, which is most apparent in the fall months. 3.1.4. Theodore Roosevelt, North Dakota Frequency distributions of IMPROVE [PM 10/C0 PM2.5]+[Soil] for Theodore Roosevelt, North Dakota are shown in Fig. 7 . The median observed dust concentration is lowest at the site during the winter months. The median concentra- tion is highest in the summer and fall; the summer distribution exhibits the smallest geometric standard deviation, indicating a less frequent occurrence of episodic very high or low concentrations. In spring, the median concentration is between that of winter and summer, but the highest frequency of concentrations above 31.5 mgm /C0 3 occurs during this season as well. This behavior points to the influence of sporadic, higher concentration events, consistent with the seasonality and temporal pattern of Asian dust transport. The site is elevated with respect to Kalmiopsis (elev. 853 m), and exhibits higher concentrations of dust in spring transport events than the latter site, which is again consistent ARTICLE IN PRESS 0 0.1 0.2 0.3 0.4 Relative Frequency Theodore Roosevelt, ND DJF 2001 −2004 <0.20.30.50.81.252 3.25.17.912.519.931.550 79 125.2198.4314.4>498 Concentration (μg/m3) 0 0.1 0.2 0.3 0.4 Relative Frequency Theodore Roosevelt, ND MAM 2001 − 2004 <0.20.30.50.81.252 3.25.17.912.519.931.550 79 125.2198.4314.4>498 Concentration (μg/m3) 0 0.1 0.2 0.3 0.4 Relative Frequency Theodore Roosevelt, ND JJA 2001 − 2004 <0.20.30.50.81.252 3.25.17.912.519.931.550 79 125.2198.4314.4>498 Concentration (μg/m3) 0 0.1 0.2 0.3 0.4 Relative Frequency Theodore Roosevelt, ND SON 2001 − 2004 <0.20.30.50.81.252 3.25.17.912.519.931.550 79 125.2198.4314.4>498 Concentration (μg/m3) NAAPS NAAPS No N.A. IMPROVE Median = 2.55 σg = 2.29 Median = 5.22 σg = 2.57 Median = 7.48 σg = 1.86 Median = 5.55 σg = 2.40 Fig. 7. Theodore Roosevelt, North Dakota 2001–2004 seasonal IMPROVE [PM 10/C0 PM2.5]+[Soil] and NAAPS simulated surface concentration distributions ( mgm /C0 3). Numbers in the upper left corner are median and geometric standard deviation concentrations for the IMPROVE data. K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–65976592

with the transport of Asian dust in elevated plumes. Frequency distributions of dust concentrations from NAAPS simulations with and without North American dust sources for the Theodore Roosevelt site are also shown in Fig. 7 . The distributions are generally similar to those of the measurements, especially in the spring months, when the model simulates dust of mostly non-North American origin. The NAAPS and IMPROVE distributions peak in the same concentration bin in summer, but the IMPROVE measurements contain fewer low concentration days than are simulated by NAAPS. In the fall and winter, the mode of non-North American simulated dust concentrations is shifted to much smaller ( o2 mgm /C0 3) concentrations than the mode of those simulated with all sources intact, indicating the strong production of dust from local sources in the model. All seasons have simulated mass concentrations greater than 50 mgm /C0 3, whereas few 24-h average concentrations of this magnitude appear in the IMPROVE data record for 2001–2004. Thus, whereas NAAPS underestimated the contributions of local dust events to summer and fall dust concentrations near Sequoia National Park, it overestimates local production throughout the upper Great Plains, as is evident at Theodore Roosevelt, North Dakota. 3.1.5. Summary for western US sites Mass concentration frequency distributions for the observations at all 15 western IMPROVE sites investigated are displayed in Fig. 8. As at Theodore Roosevelt, the winter distribution for all western ARTICLE IN PRESS 0 0.1 0.2 0.3 0.4 Relative Frequency All Western Sites DJF 2001−2004 <0.20.30.50.81.252 3.25.17.912.519.931.550 79 125.2198.4314.4>498 Concentration (μg/m3) 0 0.1 0.2 0.3 0.4 Relative Frequency All Western Sites MAM 2001−2004 <0.20.30.50.81.252 3.25.17.912.519.931.550 79 125.2198.4314.4>498 Concentration (μg/m3) 0 0.1 0.2 0.3 0.4 Relative Frequency All Western Sites JJA 2001−2004 <0.20.30.50.81.252 3.25.17.912.519.931.550 79 125.2198.4314.4>498 Concentration (μg/m3) 0 0.1 0.2 0.3 0.4 Relative Frequency All Western Sites SON 2001−2004 <0.20.30.50.81.252 3.25.17.912.519.931.550 79 125.2198.4314.4>498 Concentration (μg/m3) NAAPS NAAPS No N.A. IMPROVE Median = 1.25 σg = 3.16 Median = 3.06 σg = 2.82 Median = 5.24 σg = 2.29 Median = 3.17 σg = 3.09 Fig. 8. Seasonal 2001–2004 IMPROVE [PM 10/C0 PM2.5]+[Soil] and NAAPS simulated surface concentration distributions ( mgm /C0 3) for all 15 western sites. Numbers in the upper left corner are median and geometric standard deviation concentrations for the IMPROVE data. K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–6597 6593

sites has the lowest median concentration, and exhibits a relatively high frequency of extre- mely clean conditions ( o0.5 mgm /C0 3). In spring, the mode of the distribution shifts to higher concentrations. In summer, the median IMPROVE [PM 10/C0 PM2.5]+[Soil] value is highest and the geometric standard deviation is lowest, as a result of fewer low concentration days. In fall, the median concentration is just slightly higher than the spring value, and conditions become cleaner than those in summer, as can be inferred from the increased frequency of low ( o1 mgm /C0 3) dust concentrations. NAAPS simulated concentrations of dust at the 15 IMPROVE sites, with and without North American sources, are also shown in Fig. 8 . The winter mass distributions of the two data sets have very similar shapes to each other, however, NAAPS simulates a lower frequency of extremely clean conditions, that is, those with 24-h mass concentra- tions less than 0.5 mgm /C0 3, and a higher frequency of high ( 450 mgm /C0 3) concentrations from North American sources. In spring, as mentioned pre- viously, the model attributes nearly all of the dust simulated over the western US to non-North American sources. This indication of long-range transport is consistent with previous findings of an increased soil contribution to fine aerosol mass in spring, relative to other seasons in the western US (Malm et al., 2007 ). The spring dust can be considered to be from Asian sources, although a rare case of trans-Pacific transport of Saharan dust to western North America has been documented (McKendry et al., 2007 ). In other seasons, NAAPS attributes smaller, but still significant ( 450%), portions of the total dust in the western US to non-North American sources. Again, this result is consistent with the hypothesis that Asian dust transport impacts the western US in measurable amounts throughout the year, with larger events in spring ( VanCuren and Cahill, 2002 ). The model is biased toward higher-than-observed concentrations (with a median over twice that of the IMPROVE concentrations) of dust in spring, probably due to the long simulated dry dust lifetime that leads to high concentrations in dust plumes in the free troposphere. In summer, the NAAPS frequency distribution has a larger geometric standard devia- tion than the observations; that is, the model simulates too many low and very high concentration days and, on average, underestimates the total soil mass compared with IMPROVE. However, as was shown in Section 3.1.3, at some locations the IMPROVE estimates contain a significant percen- tage ( 420%) of non-soil material in the coarse fraction, and so the apparent underestimation of the IMPROVE observations may be less than suggested by this comparison. In fall, the NAAPS and IMPROVE data sets have a similar frequency of low dust concentration days, but IMPROVE does not report any soil concentrations above 125 mgm /C0 3, whereas NAAPS simulates some 24-h average surface concentrations from local sources in excess of 500 mgm /C0 3. As can be seen in Fig. 8 ,i n every season there are some very high NAAPS 24-h mass concentrations for which there is no evidence in the IMPROVE data. The behavior of this local dust production in the fall and winter seasons will be examined in the following section. 4. Sensitivity of emissions to changes in friction velocity threshold and snow cover Apparent underestimation of observed dust con- centrations by the NAAPS model can be attributed to a lack of some key western US sources, including agricultural activities, and to an overestimation of total PM 10 soil by Eq. (2). Apparent overestimation of IMPROVE dust concentrations could be the result of (1) NAAPS source regions being located too close to one or more of the IMPROVE sites due to the coarse model resolution; (2) a high bias in NOGAPS surface wind speeds near source regions; (3) excessively strong dust source functions or inappropriate emission thresholds in the model, and/or (4) underestimation of soil dust mass concentrations from the IMPROVE samples. It has already been shown that assumptions used to estimate soil mass concentrations from IMPROVE data are more likely to lead to overestimations in PM 10 soil mass. Thus, it is not likely that IMPROVE is systematically underestimating PM 10 dust concentrations in the upper Great Plains, although the every-third-day sampling schedule of IMPROVE may result in some large events being missed in the data set. A comparison of NOGAPS wind fields to NCEP reanalysis surface winds over dry and desert regions of the US suggests that NOGAPS has a bias toward lower-than-measured wind speeds, making option (2) an unlikely ex- planation for anomalously high simulated dust concentrations. The lack of snow cover information in the model and the existence of a constant friction velocity threshold throughout the model domain may ARTICLE IN PRESS K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–65976594

contribute to the persistent generation of dust in the region of the upper Great Plains, especially in fall, when strong katabatic winds and climatologically dry conditions result in dust generation within the model. To test this hypothesis, two test runs were performed: a simulation from January to February 2004 in which dust production was prohibited wherever NOGAPS simulated snow cover, and another from August to October 2004 with an increased friction velocity threshold of 80 cm s /C0 1. The value of 80 cm s /C0 1 was chosen based upon maximum simulated NOGAPS friction velocity values in the dust source region of eastern Montana. For the test period of January–February 2004, the addition of NOGAPS snow cover did improve the overall agreement between NAAPS and IMPROVE. The frequency of NAAPS simulated concentrations falling within a factor of two of the IMPROVE measurements increased from 27% to 43%. The simulation from August–October 2004 in which the threshold friction velocity was increased, however, resulted in a decrease in the frequency of simulated concentrations falling within a factor of two of the observations from 42% to 29%. Frequency distributions of NAAPS and IMPROVE mass concentrations at the 15 western US sites for the period of the simulation ( Fig. 9) reveal why this is so. The increased threshold friction velocity did reduce the frequency of high NAAPS mass con- centrations, but shifted the mode of the simulations to even smaller mass concentrations. Thus, an emission threshold friction velocity of 80 cm s /C0 1 is statistically worse than a value of 60 cm s /C0 1. Recent IMPROVE data do suggest, however, that coarse aerosol mass concentrations reach a maximum in the Great Plains of the US ( http://vista.cira. colostate.edu/views/). The high dust loadings in this region are likely produced locally from sources currently not included in the NAAPS dust source database. 5. Summary of findings This study investigated the seasonality of surface concentrations of PM 10 dust aerosol in the western US using data from 15 measurement sites in the Interagency Monitoring of PROtected Visual Environments (IMPROVE) network. At these sites, dust concentrations were highest in the summer; this season also saw the lowest frequency of low dust concentration days. Fall and spring had fewer high concentration days overall, however, periodic very high concentrations ( 4100 mgm /C0 3) were reported at certain sites in spring. Dust concentrations were lowest throughout the region in the wintertime. To examine the impacts of local and long-range dust transport on these data records, simulated concentrations from the NAAPS for 2001–2004 ARTICLE IN PRESS 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Relative Frequency <0.20.30.50.81.252 3.25.17.912.519.931.55079125.2198.4314.4>498 Concentration (μg/m3) 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Relative Frequency <0.20.30.50.81.252 3.25.17.912.519.931.55079125.2198.4314.4>498 Concentration (μg/m3) NAAPS IMPROVE Fig. 9. August–October 2004 IMPROVE [PM 10/C0 PM2.5]+[Soil] and NAAPS simulated surface concentration distributions ( mgm /C0 3), for threshold friction velocity set to 60 cm s /C0 1 (left) and 80 cm s /C0 1 (right), for all 15 western sites. K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–6597 6595

were also analyzed. NAAPS simulates some (1–2 mgm /C0 3) non-North American dust throughout the year in the western US, but suggests that non- local dust dominates dust concentrations in the region each spring. This indicates that Asian dust transport is the main contributor to western US IMPROVE surface dust concentrations in this season, occasionally leading to major (4100 mgm /C0 3) spikes in the aerosol record. NAAPS does not simulate the persistent high dust concen- trations ( 420 mgm /C0 3) observed in summer at locations like Sequoia National Park, although estimates of PM 10 dust concentrations from IM- PROVE data that assume all of the coarse mass is attributable to dust represent overestimates of the actual PM 10 dust mass concentration. In fall and winter, high dust concentrations ( 450 mgm /C0 3) were simulated frequently in the upper Great Plains, the Colorado Plateau, and the northern Rockies, a pattern for which there is little evidence in the IMPROVE data record. A short-term (January– February 2004) sensitivity model run in which dust production was suppressed in the presence of snow cover increased the frequency of factor-of-two agreement between NAAPS and IMPROVE dust concentrations from 27% to 43% during the test period. In a separate sensitivity run from August to October 2004, however, an increased dust produc- tion threshold friction velocity significantly de- creased the frequency of agreement between NAAPS and IMPROVE. Our study supports previous findings that Asian dust transport to the western US can occur throughout the year, with a maximum intensity in the spring months, and also suggests that agricul- tural and other anthropogenic dust sources can represent important contributions to PM 10 dust concentrations in this region in the summer and fall. Improved model characterization of these dust sources will be necessary to more accurately estimate the role of anthropogenic activities on dust concentrations in this region in the future. Acknowledgments This research was supported by the DoD Center for Geosciences/Atmospheric Research at Colorado State University under Cooperative Agreement DAAD19-02-2-0005 with the Army Research Laboratory. The support of the Office of Naval Research and the Naval Research Laboratory through program PE-0602435N is gratefully ac- knowledged as well as the ONR Global NICOP project. References Chow, J.C., Watson, J.G., Lowenthal, D.H., Solomon, P.A., Magliano, K.L., Ziman, S.D., Richards, L.W., 1992. Pm10 source apportionment in California San-Joaquin Valley. Atmospheric Environment Part a—General Topics 26, 3335–3354. Duce, R.A., 1995. Sources, distributions, and fluxes of mineral aerosols and their relationship to climate. In: Charlson, R.J., Heintzenberg, J. (Eds.), Aerosol Forcing of Climate: Report of the Dahlem Workshop on Aerosol Forcing of Climate. Wiley, New York, pp. 43–72. Haywood, J.M., Allan, R.P., Culverwell, I., Slingo, T., Milton, S., Edwards, J., Clerbaux, N., 2005. Can desert dust explain the outgoing longwave radiation anomaly over the Sahara during July 2003? Journal of Geophysical Research—Atmo- spheres 110, D05105. doi:10.1029/2004JD005232. Hogan, T.F., Rosmond, T.E., 1991. The description of the navy operational global atmospheric prediction systems spectral forecast model. Monthly Weather Review 119, 1786–1815. Husar, R.B., Tratt, D.M., Schichtel, B.A., Falke, S.R., Li, F., Jaffe, D., Gasso, S., Gill, T., Laulainen, N.S., Lu, F., Reheis, M.C., Chun, Y., Westphal, D., Holben, B.N., Gueymard, C., McKendry, I., Kuring, N., Feldman, G.C., McClain, C., Frouin, R.J., Merrill, J., DuBois, D., Vignola, F., Murayama, T., Nickovic, S., Wilson, W.E., Sassen, K., Sugimoto, N., Malm, W.C., 2001. Asian dust events of April 1998. Journal of Geophysical Research—Atmospheres 106, 18317–18330. IPCC, 2001. Climate change 2001: the scientific basis. Contribu- tion of Working Group to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Liu, W., Hopke, P.K., VanCuren, R.A., 2003. Origins of fine aerosol mass in the western United States using positive matrix factorization. Journal of Geophysical Research— Atmospheres 108, D23, 4716. doi:10.1029/2003JD003678. Malm, W.C., Sisler, J.F., Huffman, D., Eldred, R.A., Cahill, T.A., 1994. Spatial and seasonal trends in particle concentra- tion and optical extinction in the United States. Journal of Geophysical Research—Atmospheres 99, 1347–1370. Malm, W.C., Schichtel, B.A., Pitchford, M.L., Ashbaugh, L.L., Eldred, R.A., 2004. Spatial and monthly trends in speciated fine particle concentration in the United States. Journal of Geophysical Research—Atmospheres 109, D03306. doi:10.1029/2003JD003739. Malm, W.C., Pitchford, M.L., McDade, C., Ashbaugh, L.L., 2007. Coarse particle speciation at selected locations in the rural continental United States. Atmospheric Environment 41, 2225–2239. Maring, H., Savoie, D.L., Izaguirre, M.A., Custals, L., Reid, J.S., 2003. Mineral dust aerosol size distribution change during atmospheric transport. Journal of Geophysical Research— Atmospheres 108, doi:10.1029/2002JD002JD002536. McKendry, I.G., Hacker, J.P., Stull, R., Sakiyama, S., Mignacca, D., Reid, K., 2001. Long-range transport of Asian dust to the Lower Fraser Valley, British Columbia, Canada. Journal of Geophysical Research—Atmospheres 106, 18361–18370. ARTICLE IN PRESS K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–65976596

McKendry, I.G., Strawbridge, K.B., O’Neill, N.T., Macdonald, A.M., Liu, P.S.K., Leaitch, W.R., Anlauf, K.G., Jaegle, L., Fairlie, T.D., Westphal, D.L., 2007. Trans-Pacific transport of Saharan dust to western North America: a case study. Journal of Geo- physical Research—Atmospheres 112, D01103. doi:10.1029/ 2006JD007129. Miller, R.L., Tegen, I., Perlwitz, J., 2004. Surface radiative forcing by soil dust aerosols and the hydrologic cycle. Journal of Geophysical Research—Atmospheres 109, D04203. doi:10.1029/2003JD004085. Myhre, G., Stordal, F., Johnsrud, M., Ignatov, A., Mischenko, M.I., Geogdzhayev, I.V., Tanre, D., Deuze, J.L., Goloub, P., Nakajima, T., Higurashi, A., Torres, O., Holben, B., 2004. Intercomparison of satellite retrieved aerosol optical depth over the ocean. Journal of the Atmospheric Sciences 61, 499–513. Perry, K.D., Cahill, T.A., Eldred, R.A., Dutcher, D.D., Gill, T.E., 1997. Long-range transport of North African dust to the eastern United States. Journal of Geophysical Research— Atmospheres 102, 11225–11238. Prospero, J.M., Ginoux, P., Torres, O., Nicholson, S.E., Gill, T.E., 2002. Environmental characterization of global sources of atmospheric soil dust identified with the Nimbus 7 Total Ozone Mapping Spectrometer (TOMS) absorbing aerosol product. Reviews of Geophysics 40, 1002. doi:10.1029/ 2000RG000095. Reid, J.S., Flocchini, R.G., Cahill, T.A., Ruth, R.S., 1994. Local meteorological, transport, and source aerosol characteristics of Late Autumn Owens Lake (Dry) dust storms. Atmospheric Environment 28, 1699–1706. Sassen, K., 2002. Indirect climate forcing over the western US from Asian dust storms. Geophysical Research Letters 29, 1465. doi:10.1029/2001GL014051. Thulasiraman, S., O’Neill, N.T., Royer, A., Holben, B.N., Westphal, D.L., McArthur, L.J.B., 2002. Sunphotometric observations of the 2001 Asian dust storm over Canada and the US. Geophysical Research Letters 29, 1255. doi:10.1029/ 2001GL014188. VanCuren, R.A., Cahill, T.A., 2002. Asian aerosols in North America: frequency and concentration of fine dust. Journal of Geophysical Research—Atmospheres 107, 4804. doi:10.1029/ 2002JD002204. Westphal, D.L., Toon, O.B., Carlson, T.N., 1988. A case-study of mobilization and transport of Saharan dust. Journal of the Atmospheric Sciences 45, 2145–2175. Witek, M. L., Flatau, P. J., Quinn, P. K., Westphal, D. L., 2007. Global sea salt modeling: results and validation against multi- campaign shipboard measurements. Journal of Geophysical Research 112, D08215. doi:10.1029/2006JD007779. ARTICLE IN PRESS K.C. Wells et al. / Atmospheric Environment 41 (2007) 6585–6597 6597