Airborne and ground-based measurements of aerosol optical depth of freshly emitted anthropogenic plumes in the Athabasca Oil Sands region

. In this work we report the airborne aerosol optical depth (AOD) from measurements within freshly-emitted anthropogenic plumes arising from mining and processing operations in the Athabasca Oil Sands Region (AOSR) in the context of ground-based AERONET climatological daily averaged AODs at Fort McMurray (Alberta, Canada). During two flights on June 9 20 and June 18, 2018, the NASA airborne 4STAR (Spectrometers for Sky-Scanning, Sun-Tracking Atmospheric Research) sunphotometer registered high fine-mode (FM, < 1µm) in-plume AODs of up to 0.4 and 0.9, respectively, in the vicinity of the plume source (< 20 km). Particle composition shows that the plumes were associated with elevated concentrations of sulphates and ammonium. These high AODs significantly exceed climatological averages for June and were not captured by the nearby AERONET instrument (mean daily AODs of 0.10±0.01 and 0.07 ± 0.02, maximum AOD of 0.12) due possibly to 25 horizontal inhomogeneity of the plumes, plume dilution, and winds which in certain cases were carrying the plume away from the ground-based instrument. The average 4STAR out-of-plume (background) AODs deviated only marginally from AERONET daily-averaged values. While 4STAR AOD peaks were generally well correlated in time with peaks in the in-situ measured particle concentrations, we show that differences in particle size are the dominant factor in determining the 4STAR derived AOD. During the two flights of June 24 and July 5, 2018 when plumes likely travelled distances of 60 km or more, 30 the average 4STAR FM AOD increased by 0.01-0.02 over ~50 km of downwind particle evolution which was supported by the increases in layer AODs calculated from the in-situ extinction measurements. Based on these observations as well as the increases in organic mass, we attribute the observed AOD increase, at least in part, to secondary organic aerosol formation. The in-plume and out-of-plume AODs for this second pair of flights, in contrast to clear differences in in-situ optical and other measurements, were practically indistinguishable and compared favorably to AERONET within 0.01-0.02 AOD. This means that AERONET was generally successful in capturing the background AODs, but missed some of the spatially constrained high-AOD plumes with sources as close as 30-50 km, which is important to note since the AERONET measurements are generally thought to be representative of the regional AOD loading. The fact that industrial plumes can be associated with significantly higher AODs in the vicinity of the emission sources than previously reported from AERONET can potentially 5 have an effect on estimating the AOSR radiative impact. aerosols local of the oil plumes. focussed on either naturally produced aerosols as forest significantly aged plumes or of the OSMC flights of June 9 and 18, we discuss how the 4STAR AOD variations associated with the AOSR emitted plumes relate to the corresponding aircraft in-situ measurements. While direct comparisons between the 4STAR column-based AODs and flight-level in-situ aerosol properties are not strictly possible, evaluating the AOD measurements in the context of time-synchronous in-situ measurements provides insight into the nature of the plume and the associated AOD variations (e.g., Stone et al., 2010, Shinozuka et al., 2013, Pistone et al., 2019). In particular, understanding 25 how changes in aerosol intensive properties such as particle size and chemical composition might affect the AOD response is necessary to evaluate the radiative importance of the emitted plumes and the associated secondary organic aerosol production. particles within plume A (FM effective radius increasing from 0.10 to 0.18 µm) relative to the out-of-plume values and plume B. Particle composition analysis shows that plume A contained elevated levels (by up to a factor of 6) of sulphates and ammonium while the organic mass remained largely comparable between the two plumes.

measurements, were practically indistinguishable and compared favorably to AERONET within 0.01-0.02 AOD. This means that AERONET was generally successful in capturing the background AODs, but missed some of the spatially constrained high-AOD plumes with sources as close as 30-50 km, which is important to note since the AERONET measurements are generally thought to be representative of the regional AOD loading. The fact that industrial plumes can be associated with significantly higher AODs in the vicinity of the emission sources than previously reported from AERONET can potentially 5 have an effect on estimating the AOSR radiative impact.

Introduction
Atmospheric aerosols play an important role in the Earth's climate. Depending on their properties (such as size, composition, and concentration) and climatological factors (such as ambient radiative energy budget and surface albedo), aerosols can either cool the surface by reflecting sunlight back to space or have a net warming effect by directly absorbing this shortwave radiation 10 (e.g. Kaufman et al., 2002, Charlson et al., 2005. Aerosols can also affect cloud properties by modifying droplet size, cloud lifetime, and cloud burn-off rates (Ackerman et al., 2000, Wilcox, 2010, McComiskey and Feingold, 2012. Despite recent progress, aerosols, together with clouds, remain the greatest source of uncertainty in climate models (Boucher et al., 2013). This is in part due to the complexity of aerosol-generating processes (natural and anthropogenic), and aerosol transport mechanisms resulting in particles of various size, concentration, chemical composition, and mixing state that 15 ultimately determine aerosol radiative impact. A recent work by Matus et al., 2019, based on the multi-sensor A-train observations, estimates that the global mean aerosol direct radiative effect is -2.40 W/m 2 with anthropogenic direct radiative forcing accounting for 21%.
The Athabasca Oil Sands Region (AOSR) in Alberta, Canada, is home to the third-largest oil deposit in the world (behind 20 Venezuela and Saudi Arabia) with an estimated 164.1 billion barrels of recoverable crude oil (as of December 2018, see NRCAN, 2020). In addition to the emissions of anthropogenic pollutant and greenhouse gases (e.g., Charpentier and Bergerson, 2009, Bytnerowicz et al., 2010, Baray et al., 2018, Liggio et al., 2019 the AOSR generates a substantial amount of aerosols, either directly as sulfate, black carbon (BC), primary organic aerosol (POA) and dust (e.g., Landis et al. 2017, Cheng et al., 2018, or from secondary atmospheric processes. In fact, Liggio et al., 2016 showed that the AOSR is a large 25 source of secondary organic aerosol (SOA, formed through gas-to-particle conversion) with formation rates comparable to major metropolitan areas. Understanding the radiative impact of anthropogenically produced aerosols in the AOSR through better characterization of their radiative properties is therefore of significant importance.
The estimation of aerosol radiative impact requires knowledge of key aerosol properties. One of the most important of these 30 properties is the aerosol optical depth (AOD) (e.g., Kahn, 2012), defined as the total vertical column of aerosol extinction (the combined capacity of aerosol particles to scatter and absorb light). The mid-visible AOD (typically reported at wavelengths https://doi.org /10.5194/acp-2020-1218 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License. between 440 and 550nm) can vary significantly from less than 0.1 for rural sites with background aerosols to more than 3 for aerosol plumes from intense forest fires, dust storms or volcanic eruptions. Model analysis suggests that 95% of global AOD at 440 nm, however, are below 0.4 . AOD measured from or near ground level is commonly retrieved using passive sunphotometry. A sunphotometer measures AOD in several spectral channels and can yield an estimation of particle abundance as well as aerosol size indicators (effective radius (reff) of submicron ("fine mode", FM) and supermicron 5 ("coarse mode", CM) modes for example) from the spectral information (O'Neill et al., 2003 The analysis of local AOD climatologies including that of AOSR, serves as a basis for a global aerosol climatology and is 10 essential for validating satellite retrievals (Toledano et al, 2007, Sioris et al, 2017b. Sioris et al, 2017a analyzed AOD data records for several Canadian AERONET stations, including Fort McMurray, in the context of ground-based particulate matter (PM2.5) concentrations and presented annual climatologies for both the FM and CM AODs. They found that Fort McMurray FM AODs (at 500 nm) are generally < 0.1 for Sep.-Apr. and < 0.2 for May-Aug. The summer-time AOD increase was attributed to forest fires plumes and no significant trends in the long-term AOD record were observed. In addition to ground-15 based measurements, Shinozuka et al., 2011 reported on AOD measurements acquired on board the NASA P-3 aircraft using the AATS-14 sunphotometer (AOD from 14 wavelength bands in the 354-2139 nm range) in the vicinity of Fort McMurray in June-July 2008. While the AATS-14 measured AOD499 values that frequently exceeded 1 and at times reached 4, the study was entirely focused on biomass burning aerosols from local forest fires and did not include measurements of the oil sands industrial plumes. Indeed, most airborne studies have predominantly focussed on either naturally produced aerosols (such as 20 forest fires or dust events) and/or significantly aged plumes that have been transported for hundreds or thousands of kilometers (Heald et al., 2006, Chin et al., 2007, Stone et al., 2010. While multiple studies have described cases of extreme pollution in China (e.g., Gu et al., 2018, Qin et al., 2018, Sun et al., 2018 significantly less attention has been given to AOD values from isolated industrialized sources outside of Asia. Given that the anthropogenic plumes in the AOSR, far from large urban regions, are nearly entirely attributed to a singular industry, the obtained airborne AOD measurements can be used to evaluate the 25 representativeness of AERONET measurements in the context of industrial emissions. This is especially relevant to the AOSR, where secondary aerosol formation is likely to occur away from source areas and associated AERONET sites. In the spring-summer of 2018, the NASA 4STAR sunphotometer (Spectrometers for Sky-Scanning, Sun-Tracking Atmospheric Research, Dunagan et al, 2013, Pistone et al, 2019, LeBlanc et al, 2020 was integrated 30 on the National Research Council (NRC) Convair-580 aircraft, as part of state-of-the-art suite of in-situ and remote sensing instruments. The NASA 4STAR acquired airborne hyperspectral AOD data during the Oil Sands Measurement Campaign of 2018 (OSMC), and represented the first airborne sunphotometer deployment to study industrial pollutant plumes from the Alberta oil sands mining and upgrading operations.
In the current work, based on four OSMC flights, we evaluate whether the existing ground based AERONET measurements capture the full extent of AOD variations in the AOSR by comparing the 4STAR AODs of anthropogenic plumes from mining and processing operations with AERONET climatological and diurnal AODs. In addition, we investigate how the AOD evolves as a function of distance downwind of the plume source, in order to better understand the potential effect of secondary 5 aerosol production, i.e. whether the AOD increases as new particles are being formed -a phenomenon that would be missed by a fixed ground-based instrument located near the source. Finally, we consider how the 4STAR FM and CM AODs relate to other in-situ measurements such as particle size, optical absorption and scattering coefficients, and particle chemical composition. The latter analysis provides additional insight into the observed AOD variations and ensures that the 4STAR measurements are physically consistent with other instruments. 10 2 Instrumentation and data processing

Oil Sands Measurement Campaign (OSMC)
The Oil Sands Measurement Campaign 2018 (OSMC2018, hereafter referred to as OSMC) was a follow-up to the OSMC2013 campaign, where the Convair-580 aircraft was deployed to the Oil Sands region for the first time (see e.g. Liggio et al., 2016, Li et al, 2017. Figure 1 shows the location of the AOSR facilities overlaid on recent satellite imagery. Two types of flights 15 performed during the OSMC focused on (i) pollutant emissions and (ii) their transformation (i.e., chemical evolution/aging) using "box" (latitude/longitude grid) and "screen" (vertical grid) flight patterns, respectively. Box flights encompassed an entire facility to characterize pollution emissions from individual oil sands processing facilities, while screen flights were conducted to study the evolution of the pollutants transported away from the source. Each box/screen started at 500 ft above ground level (AGL) and progressively increased in altitude at intervals of approximately 250 feet or more, until reaching the 20 top of the boundary layer typically resulting in 3-6 altitude levels. The transformation flights were conducted as a Lagrangian experiment designed to intercept the same plume at increasing distances from the source perpendicular to the wind direction.
The measurements presented here were obtained during the second phase of the campaign between May 30 and July 5, 2018.

AERONET and CIMEL CE-318 25
The Cimel CE-318 used at Fort McMurray and Fort McKay are standard ground-based sunphotometers which are part of the global AERONET sunphotometer network. The CE-318 takes measurements of the direct solar irradiance, at nominal sampling intervals of three minutes in eight spectral bands (340,380,440,500,675,870,1020, and 1670 nm) from which spectral AODs can be calculated. For more information about the AERONET and CIMEL CE318 the reader is referred to Giles et al., 2019 In this work we used AERONET Level 2 (cloud-screened and quality controlled) Version 3 data. 30 https://doi.org/10.5194/acp-2020-1218 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License.

4STAR (Spectrometers for Sky-Scanning Sun-Tracking Atmospheric Research)
The NASA 4STAR instrument combines airborne sun tracking and sky scanning with diffraction spectroscopy. It is comprised of a movable optical head that protrudes through the top of the aircraft fuselage, an accompanying instrument rack inside the aircraft and optical cables that connect the two ( Figure S1). The hyperspectral measurements are obtained with two Zeiss spectrometers resulting in a total of 1556 spectral channels between 210 and 1703 nm with sampling resolution of 0.2-1 nm 5 below 1000 nm and 3-6 nm at longer wavelengths. The nominal calibration accuracy of AOD measurements from 4STAR are dependent on wavelength, time of day, tracking stability, and various corrections (such as removal of light absorption by trace gases). This accuracy is typically near 1% in transmittance, for wavelengths outside gas absorption bands between 355 nm and 1650 nm, resulting in an AOD uncertainty of 0.01 when the sun is at zenith. A more detailed technical description of the 4STAR can be found in Dunagan et al, 2013 andShinozuka et al, 2013. 10

Spectral Deconvolution Algorithm
We applied the Spectral Deconvolution Algorithm (SDA; O'Neill et al, 2003) in order to separate the total measured AOD into its fine (submicron) and coarse (supermicron) mode components ( and or FM and CM). While AERONET uses 380, 440, 500, 675, and 870 nm channels as standard inputs to SDA, using the 380 nm channel in the 4STAR analysis was not possible because of its unrealistically high values (thought to be due to stray light at that wavelength) aggravated by high 15 relative humidity within the spectrometer enclosure. Reducing the number of channels to four is thought to produce results commensurate with the standard AERONET wavelength set with RMS differences in retrieved of less than 0.01 (O'Neill et al., 2008). We expect that, in most cases, the anthropogenically produced aerosols observed in the AOSR are submicron, since coarse-mode aerosols are usually associated with natural sources (e.g., winds lifting up dust particles). Consequently, with the exception of dust aerosols resulting from surface mining, the coarse-mode dominant AODs in the AOSR will usually 20 be associated with clouds. As will be shown below, the CM AODs associated with dust in the AOSR are small (< 0.05) and can easily be distinguished from cloud-dominated events that are associated with significantly higher AODs and larger variations over short periods of time.

4STAR data filtering and cloud screening
In this work we have eliminated all 4STAR measurements with low (raw count) signal to noise ratio and ignored measurements 25 when the aircraft was not in a straight-level track (i.e. when turning or engaged in a spiral). Moreover, in addition to eliminating improbable data (for AOSR) with AOD500 > 3, we have used the SDA as a spectral cloud-screening where the FM AOD is thought to be free of coarse-mode contamination by either clouds or dust (e.g., O'Neill et al, 2016). https://doi.org/10.5194/acp-2020-1218 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License.

In-situ measurements
The fine-mode particle size (diameter ranging from 60 to 1000 nm) was measured using a DMT Ultra-High-Sensitivity Aerosol Spectrometer (UHSAS), which employs light scattering techniques to derive particle concentrations and size distributions (Cai et al, 2008, Kupc et al, 2018. The UHSAS effective radius (reff), defined as an area weighted mean radius of the aerosol particles, was estimated using 5 (1) where r is the particle radius and n(r) is the particle size distribution (i.e., number of particles per cm 2 having a radius in the range r and r+dr microns). The particle in-situ scattering and absorption coefficients were measured using the TSI 3563 nephelometer and Continuous Light Absorption Photometer (CLAP), respectively. The TSI 3563 measures the light scattered by aerosols ultimately yielding the aerosol total scattering coefficient at 450, 550, 700 nm (Anderson et al., 1996;Bodhaine et 10 al., 1991)  Aerosol single scattering albedo (SSA) values were estimated using the ratio of the light scattering coefficient to the sum of the light absorption and scattering coefficients. Black carbon (BC or soot) mass was measured using the DMT Single Particle Soot Photometer (SP2) based on the laser-induced incandescence between approximately 100 and 600 nm (e.g., Schwarz et 15 al., 2006). Finally, non-refractory particle composition (ammonium, nitrate, sulfate and organics) was measured with a High-Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-ToF-AMS; Aerodyne Research Inc., DeCarlo et al., 2006).

AERONET AOD climatology
Understanding seasonal AOD variations from the ground-based Fort McMurray instrument is useful for the interpretation of 20 the airborne measurements. While this analysis is required to evaluate whether the AERONET measurements capture the full extent of AOD variations in the AOSR, understanding climatological AODs also provides context for the 4STAR OSMC AODs. The summer-time FM AOD increase is thought to be predominantly associated with forest fires that frequently occur in the region (Sioris et al., 2017a). For example, the second highest monthly AOD on record of 0.45 occurred during May 30 2016 when intense forest fires had a major impact on air quality and resulted in the evacuation of more than 80,000 people https://doi.org/10.5194/acp-2020-1218 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License. (Landis et al, 2018). The highest hourly-and daily-averaged AODs during that month were 2.10 and 1.80, respectively. Figure   S2 shows monthly-averaged AOD time series for 2005-2018 at Fort McMurray with daily-averaged values shown in pink shading. Also, included for reference is a fine-mode AOD time series for nearby Fort McKay, however, the present discussion focuses on Fort McMurray because of its significantly longer measurement record and data availability during OSMC. In many cases, on days with persistently high AOD values (e.g., FM AOD > 0.5) forest fire smoke plumes were clearly 5 discernable in the vicinity of Fort McMurray from satellite imagery (not shown). In comparison, the highest FM monthly AOD average during the non-summer months (Sep.-Apr.) -considered to be primarily impacted by industrial pollution plumes in the absence of other major aerosol sources -was 0.13 with 96% of all hourly measurements being lower than that value.
With respect to the coarse-mode, the peak monthly-averaged AOD of 0.04 is associated with the month of April and is thought 10 to be due to dust particles. Landis et al. (2017) demonstrated through systematic chemical analysis of summertime CM aerosol from ground-based measurements that a majority of dust is due to local surface mining activities where heavy machinery (trucks and excavators) lift significant amounts of dust into the air, although long-range transport of Asian dust may be more prominent in the spring (e.g., McKendry et al., 2007, AboEl-Fetouh et al., 2020. AboEl-Fetouh et al., 2020 argued, for example, that a springtime coarse mode event was continental in scale (including the North American and European Arctic) 15 and that it was associated with (relatively small) coarse mode particles around 1.3 µm in radius.
Figure 2 also indicates that the mean total AOD during June 2018 was 0.13 ± 0.05 suggesting that the aerosol loading during the OSMC was generally representative of the climatological AOD values for June (0.15 AOD).

4STAR case studies during OSMC 20
In this section, based on the OSMC flights of June 9 and 18, we discuss how the 4STAR AOD variations associated with the AOSR emitted plumes relate to the corresponding aircraft in-situ measurements. While direct comparisons between the 4STAR column-based AODs and flight-level in-situ aerosol properties are not strictly possible, evaluating the AOD measurements in the context of time-synchronous in-situ measurements provides insight into the nature of the plume and the associated AOD variations (e.g., Stone et al., 2010, Pistone et al., 2019. In particular, understanding 25 how changes in aerosol intensive properties such as particle size and chemical composition might affect the AOD response is necessary to evaluate the radiative importance of the emitted plumes and the associated secondary organic aerosol production. On June 9, the Convair-580 conducted a transformation flight to the west of the Oil Sands facilities. The aircraft flew 4 screens perpendicular to the wind direction intercepting the emitted plumes from the majority of the largest surface mining facilities 30 ( Figure 2). Figure 3 shows the AOD time series and corresponding in-situ measurements for the relatively cloud-free screen 1, with five nominal altitude levels and "in plume" and "out of plume" periods separated based on elevated values relative to the upwind (background) measurements. While the average fine-mode AODs were ~0.10 outside of the plume, the values https://doi.org/10.5194/acp-2020-1218 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License. were greater than 0.4 within the plume. Moreover, the two consistent peaks at each altitude level in the UHSAS total particle number concentration (pane 2) suggest that the Convair-580 was intercepting two spatially separated plumes where only one is associated with a significant AOD response. Particle composition data (pane 4 of Figure 3 and Figure 4) shows that the first plume (plume "A", detected by the 4STAR and shaded in pink) contained significantly more SO4 and NH4 than the second plume (plume "B", shaded in grey) indicating that the two plumes were likely originated from different facilities. The organic 5 aerosol mass, however, was comparable between the two plumes and similar to the background level. Figure 3 pane 3 and Figure 5 show a clear distinction in particle effective radius and volume size distribution between plumes A and B in that the latter lacks optically significant contribution from larger FM particles. The difference is especially pronounced, both in comparison to plume B and out of plume (background) values, for particle radii (rp) larger than ~0.12 with the largest differences observed at radii 0.3 and 0.42 . The relative differences between plumes A and B continue to persist for screen 10 two, approximately 20 km downwind (not shown) but the concentration of large FM particles larger than ~ 0.12 for plume A was significantly lower (by a factor of 2 or more relative to screen one) as the plume disperses and some of the larger particles settle out of the atmosphere.
During the June 18 emissions flight around the Syncrude Mildred Lake (SML) facility, there was a significant and clearly 15 visible industrial plume emanating from the SML stacks ( Figure S3). Figure 6 shows the corresponding data for the first three altitude levels: 500, 750, and 1000 ft with the 500 ft track repeated twice. The 4STAR measurements show that for the duration of the Syncrude box the total AOD at 500 nm was relatively low and stable around 0.05-0.07, with AOD spikes of up to 0.86 associated with the plume on the eastern side of the box. As in the case of June 9, the measurements show that the plume was heterogeneous with two distinct regions identified in Figure 6 in pink (plume "A") and grey ("plume "B"). Plume A is 20 characterized by a significant response in fine-mode AOD and increased (up to a factor of 2) UHSAS effective radius. Plume A is also associated with high values of SSA440 (up to 0.96, pane 4), indicating that the particles were mostly of a scattering nature. In contrast, the SSA values dip below 0.75 in plume B suggesting a much stronger presence of absorbing particles. This hypothesis is confirmed by the enhanced concentrations of highly absorptive BC particles (pane 5), with BC mass increasing during certain times by more than a factor of 10 relative to plume A. Similarly to June 9, plume A contains 25 significantly higher concentrations of SO4 and NH4 than plume B (not shown). It should be noted that the AOD peaks precede most of the plume A in-situ peaks, which we attribute to the 4STAR viewing geometry in relation to the vertical extent of the plume (i.e. the relative position of the sun, the plume and the aircraft carrying the 4STAR). Plume B contains fewer of the larger FM particles and goes mostly undetected by 4STAR. At the same time, plume B contains higher concentrations of CM particles from surface mining (not shown) but the resulting CM AOD (for example, around 15:22) is < 0.03 which is 30 significantly smaller than the FM AODs.
For both flights, the analysis indicates that differences in particle size within plumes are the driving factor responsible for the AOD response. As a consequence, while certain sections of the plume (e.g., plume B) are clearly associated with elevated insitu particle and gas concentrations, they have little effect on AOD when particulates are small (in this case rp < 120 nm).

4STAR-AERONET comparisons
The inherent assumption of using the AERONET data is that it is representative of the regional AOD within a certain distance 5 from the ground-based station. While this distance is often taken as 100 km, the spatial representation will ultimately depend on the local context including proximity to aerosol sources and meteorology , Holben et al, 2018.

Consequently, in this section we investigate the representativeness of Fort McMurray AERONET measurements of AOSR
AODs by studying how the in-flight AOD variations compare to climatological and diurnal ground-based AODs.  On June 9 (screen 1) and June 18 ("high-AOD flights") the plumes were sampled in the direct vicinity of the source.
Consequently, the June 9 and June 18 cases were associated with particularly high AODs and show relatively large variations in terms of the vertical distribution and AOD variability within the plume. These large AODs and corresponding variations significantly exceed climatological AODs for June and are clearly not captured by AERONET. However, the average 4STAR 25 out-of-plume (background) AODs deviate only marginally from AERONET daily-averaged values (Table 1). During the June 24 and July 5 flights ("low-AOD flights"), the plumes in general travelled longer distances than during the high-AOD flights before being measured by the Convair-580 and hence had more time to mix and disperse. The in-plume and out-of-plume AODs for these dates were indistinguishable (in contrast to clear differences in in-situ measurements, not shown) and the airborne and ground-based averages compared favorably within 0.01-0.02 AOD. In other words, the AERONET 30 measurements capture the background AODs relatively well, but miss some of the spatially constrained high-AOD plumes with sources as close as 30-50 km. https://doi.org/10.5194/acp-2020-1218 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License.
It is not clear what kind of agreement would be considered as reasonable between 4STAR and AERONET given the differences in spatial sampling between the two platforms. While several studies have reported agreements down to RMS differences of 0.01 for wavelengths between 500 nm and 1020 nm (Stone et al., 2010;Shinozuka et al., 2013;LeBlanc et al., 2020), those comparisons often involve aged plumes that are likely to be more (horizontally) homogeneous than newly-emitted AOSR plumes. Furthermore, the comparisons usually include only a few points closest in time to the aircraft overpass of an 5 AERONET station. Factors such as prevalent wind directions, local topography, and distance to emission sources can all bias comparisons between the aircraft and ground-based platforms. In particular, during the June 9 and June 18 flights, the Fort McMurray AERONET station was mostly located upwind of the Syncrude facilities and likely missed most of the plumes measured by the 4STAR. However, even when limiting the AERONET data to periods associated with northern winds (270 to 90° on a wind rose) likely carrying the industrial plumes (Figure 1), the mean and median June 2018 AODs were both 0.09 10 with only 4 individual measurements exceeding 0.2 (maximum of 0.29). We attribute these disparities to dilution as the plumes are being transported from their sources to where they are captured by the AERONET sunphotometer. Simply moving the AERONET station closer to specific facilities might yield a better agreement in certain cases, but the AOD differences will likely remain because of spatial inhomogeneity of the plumes and the inherent limitations of measurements from a single fixed ground-based station. 15  Figure S4, top pane). The bottom pane of that figure also shows that the AERONET reff,f values agree particularly well with UHSAS but the 4STAR retrievals seem to be overestimated. We suspect that the source of this latter inconsistency is due to the artificially high 4STAR irradiance signals at the shorter wavelengths. Figure 8  ) for all four flights combined. It is evident from this figure that despite the slightly shorter atmospheric column, 4STAR AODs are marginally overestimated relative to AERONET with | | mean, median, 30 and standard deviation being 0.02, 0.01 and 0.02. The interpretation of this result is not obvious. In principle, as the aircraft moves further away from sources the AOD should increasingly resemble the background and thus τdiff should decrease with distance. On the other hand, as the distance between the two instruments increases so do the differences in sampled air masses.
In fact, Sioris et al., 2017a (their Figure 4) showed that the two AERONET stations become progressively dissimilar as a function of distance between them with the correlation coefficient dropping by ~50 % in the first 500 km. This is consistent with Figure S5 which indicates that despite the significant variance in the data, the average increases as a function of distance separating the 4STAR and the Fort McMurray AERONET station (up to 160 km) based on the data from the 4 flights.
This suggests that either the 4STAR was influenced by dispersed plumes with AODs higher than the AOSR background even 5 at larger distances (i.e. not fully diluted plumes), or that the AERONET AODs were influenced by cleaner air masses.
Another potential explanation accounting for some of the 4STAR overestimation is secondary organic aerosol (SOA) production which may increases with distance/time (e.g. Liggio et al., 2016). This possibility was investigated by targeting the transformation flights of June 24 and July 5, in an attempt to observe any systematic screen-to-screen AOD changes that might 10 be due to SOA production -a phenomenon that could be missed by AERONET. As shown in Table 1 between screen 1 to 3 only a modest increase of 0.01 and 0.02 was observed. While these AOD changes are small and commensurate with the 4STAR measurement error, O'Neill et al., 2008 previously showed that changes on this scale can be linked to real underlying changes in aerosol properties. We compared the in-plume vertical profiles obtained during screens 2 and 3 of each of the two flights, excluding the data from screen 1 because on June 24 it was not properly positioned relative to other screens and on July 5 the 15 profile data only covered a short vertical range, which we judged to be inadequate for analysis. We found that for both dates the calculated in-situ extinction coefficient at 500 nm was higher at practically each level for screen 3 (Figure 9). Integrating the extinction coefficient over the common vertical range between the screens yielded (in-situ) layer AOD increases between 33 and 67%. Comparing these values to 4STAR was not always possible in part because of the missing sunphotometry data, and in part because the AOD didn't vary sufficiently (i.e. above the measurement uncertainty of 0.01) over these short vertical 20 ranges of 200-500 m to be reliably measured. However, for the two longest profiles during screen 3 on July 5 spanning 800 m, the in-situ layer AOD of 0.013 ± 0.002 compared favorably to 4STAR AOD of 0.01 (taking standard deviation within a 50 m altitude bin as a measurement error for extinction and assuming that the errors in altitude measurements are negligible). Therefore, the observed 4STAR AOD increases, while small, are consistent with layer AOD increases calculated from in-situ extinction measurements. Particle composition data shows that the plumes were dominated by the organic content (> 78%) 25 with the organic mass increasing from screen to screen despite dilution ( Figure S6). Using a Top Down Emissions Rate Retrieval Algorithm (TERRA; Gordon et al., 2015) significant SOA formation rates were determined of 1003±193 kg/hr and 443±45 kg/hr for June 9 and July 5, respectively. Consequently, while acknowledging the limited scope of the two flights and the need for more case studies, we believe that the observed AOD increase can be attributed, at least in part, to the process of SOA formation. 30 https://doi.org /10.5194/acp-2020-1218 December 2020 c Author(s) 2020. CC BY 4.0 License.

Summary and conclusions
During the spring-summer 2018 Oil Sands Measurement Campaign, the NASA 4STAR sunphotometer acquired airborne hyperspectral AOD data in the Athabasca Oil Sands Region. In this work, we report on the airborne AOD measurements of anthropogenic plumes from mining and processing operations in AOSR in the context of AERONET climatological and diurnal AODs at Fort McMurray. The monthly-averaged AERONET AOD for June 2018, when most of the 4STAR sunphotometer 5 data was acquired, was 0.13 ± 0.05 (mean ± std) which was in line with summertime climatological averages, indicating that the observational conditions during OSMC were representative of the typical aerosol loading in the area for that month.
During the two flights of June 9 and June 18 (high-AOD flights), while flying into the visually opaque industrial plumes, 4STAR registered high fine-mode in-plume AODs of up to 0.4 and 0.9, respectively, in the vicinity of the plume source (< 20 10 km). These high AODs significantly exceed climatological averages and were clearly not captured by the nearby AERONET instrument, which reported mean daily AODs of 0.10±0.01 and 0.07 ± 0.02 (mean ± std) for those days. At the same time the average 4STAR out-of-plume (background) AODs deviated only marginally from AERONET daily-averaged values. We attribute the disparities principally to the horizontal inhomogeneity of the plumes, the distance between the plume source and the AERONET instrument as well as prevalent winds which, in certain cases, were carrying the plume away from the ground-15 based instrument.
For both flights, the 4STAR AOD peaks were generally well correlated in time with peaks in in-situ measured particle and gas concentrations for the optically active sections of the plume (plume "A", i.e. those sections that were associated with significant increases in in-situ extinction and/or AOD), while other sections, clearly associated with elevated in-situ concentrations 20 (relative to out-of-plume measurements), produced no significant 4STAR response (plume "B"). We attribute this difference to the presence of larger FM particles within plume A (FM effective radius increasing from 0.10 to 0.18 µm) relative to the out-of-plume values and plume B. Particle composition analysis shows that plume A contained elevated levels (by up to a factor of 6) of sulphates and ammonium while the organic mass remained largely comparable between the two plumes.

25
During the June 24 and July 5 flights (low-AOD flights), the plumes likely travelled longer distances before being measured by the Convair-580 onboard instruments (the Syncrude plume for example would have travelled ~60 km, or three times the distance than on June 9 before being first measured) and hence had more time to mix and disperse. The in-plume and out-ofplume 4STAR AODs for these dates were practically indistinguishable (average AODs of ~0.12 and 0.05 for June 24 and July 5, respectively) and the airborne and ground-based averages compared favorably to within 0.01-0.02 AOD. Comparing screen 30 in organic content and organic fraction, we attribute the observed AOD increase, at least in part, to the process of SOA formation.
Comparing out-of-plume 4STAR AODs to AERONET measurements shows that despite the slightly shorter atmospheric column 4STAR, AODs are marginally overestimated ( 0.02) with the difference increasing as a function of distance 5 from the AERONET station. This suggests that either the 4STAR was still influenced by plumes with AODs higher than the AOSR background (presumably measured by AERONET) even at larger distances or that the AERONET AODs were generally underestimated during those flights (due for example to being influenced by cleaner air masses).
Finally, it is noted that low AOD AERONET measurements should be interpreted with care when they are used to represent 10 the AOSR average. We showed that industrial plumes can be associated with significantly higher AODs in the vicinity of the emission sources and their full extent is not captured by the ground-based instruments even with sources as close as 30-50 km. This might potentially have an effect on estimating the regional radiative impact. From cursory radiative transfer calculations, based in part on the in-situ aerosol properties within the industrial aerosol plumes, the radiative effect of the in-plume aerosol is evaluated at up to a 25% increase over the background aerosol. 15

Acknowledgements
Funding for the NRC Convair-580 participation in the Oil Sands 2018 field project was provided by ECCC and NRC. We would like to acknowledge the NRC and ECCC project crew that supported the field project. We also acknowledge Warren Gore (retired) from NASA Ames Research Center for his efforts in bringing the 4STAR to Canada for the duration of the project. We thank Ihab Abboud and Vitali Fioletov for establishing and maintaining the Fort McMurray and Fort McKay 20 AERONET sites used in this investigation. This work was partially funded under the Oil Sands Monitoring (OSM) program.
It is independent of any position of the OSM program. The U.S. authors were supported in part by NASA Radiation Science    . From top to bottom: a) total, fine and coarse mode AODs at 500 nm derived from the SDA of the 4STAR data; b) UHSAS total particle number concentration (all sizes), # / , c) UHSAS effective radius, , d) AMS particle composition. In the top panel, the dashed vertical lines delineate each altitude level with nominal (planned) altitudes listed in feet above ground level at the top. The fine-mode AOD is reproduced on several plots to facilitate the comparisons. The pink and grey rectangles represent periods associated with plumes "A" and "B", respectively, while the green shading indicates periods when the aircraft was engaged in a spiral manoeuvre.