The Government Flats Complex was a large system of three lightning-sparked fires: the Government Flats fire, the Blackburn fire, and the Wells Road fire. They burned near Mt. Hood in Wasco County, Oregon from 16 August to 26 August 2013. Cumulatively, these fires burned over 11,350 acres of land, although the Blackburn fire was responsible for 98% of this area. According to the US Forest Service, the fuel model in this area was summer hardwood litter [
63]. The plume produced by this fire complex was observed by MISR on 21 August 2013, at ~19:07 Universal Coordinated Time (UTC), ~2 h before the BBOP G-1 aircraft began sampling the same area; the aircraft continued sampling this plume for an additional hour. At both observation times, the plume displays unique aerosol optical and physical property patterns that vary with downwind distance. We divide the MISR-observed plume into four regions based on these patterns: a near-source region (designated region I), a near-downwind region (II), a mid-downwind region (III), and a far-downwind region (IV) (
Figure 1a). We calculate the approximate age of the smoke at regional boundaries using the ratio of the along-plume distance and mean wind vectors in the area derived from MINX (see
Supplementary Materials, Figure S1). Based on plume age calculated by Kleinman and Sedlacek [
18], and similar trends in particle properties, the BBOP-observed plume was divided into regions of similar age (
Figure 1b), although the aircraft did not sample the part of the plume corresponding to region IV in the MISR imagery. In this section, we compare the particle properties retrieved from space with those obtained in situ from the aircraft measurements, paying particular attention to the
differences between these regions. We then apply this information to current knowledge on how different aging processes affect particle properties in order to infer the active mechanisms at hand.
Only observations taken within the plume boundaries seen in
Figure 1 are included in our analysis, except when comparing measurements to background values. For the MISR dataset, the boundary is a user-drawn plume outline created during the MINX retrieval. Some areas of optically very thin smoke can be seen in the imagery outside the boundary, but MISR had difficulty performing retrievals beyond a certain distance downwind (where the plume becomes thin), as can be seen by the lack of MINX plume results, and as indicated by some empty pixels in the RA retrievals shown subsequently. For the BBOP dataset, the RGB MODIS-Aqua image, acquired about 2 h after the MISR overpass, was used to draw the plume outline (
Figure 1b). However, significant cloud contamination made it challenging to determine exact boundaries in the two most downwind regions; thus, carbon monoxide and total aerosol count measurements from the aircraft were used to refine the shape (
Figure 1c,d, respectively). This is also important, as plume geometry can change during the course of the flight itself; thus, the MODIS snapshot might not perfectly represent the plume as sampled by the aircraft, even without cloud contamination. Where applicable, all figures display the plume boundary as a dotted-line polygon and the dividing lines between regions as thin solid gray lines.
3.1. In Situ Observations: A First Look
We provide an overview of the aircraft observations first, with an emphasis on the differences in smoke properties between regions.
Figure 2 and
Figure 3 illustrate, as latitude/longitude plots, the chemical, optical, and physical properties, measured using the methods described in
Section 2.3. Each point represents the aggregated ~1.1-km median value of a given measurement (except for
Figure 2c, which did not need to be aggregated, as discussed in
Section 2.3.3).
Table 4 quantifies the mean, standard deviation (SD or σ), and median values of the measurements from
Figure 2 and
Figure 3 for each region, and a comprehensive discussion of these statistics can be found in the
Supplementary Materials.
In general, the plume contains fresh, highly absorbing smoke near the source that becomes both increasingly bright (SSA
522) and oxidized (−log[NO
x/NO
y]) with downwind distance (
Figure 2a,b, respectively). Both absolute and mean CO-normalized BC mass concentration decrease along the length of the plume (
Figure 2c,
Table 4); however, rates of BC dilution vary between the plume center (no increasing dilution with downwind distance) and the plume edges (increasing dilution with downwind distance). The disparities between the median and mean values in
Table 4, as well as the differences in sampling seen in
Figure 2, illustrate this pattern. CO-normalized CCN and aerosol concentration both increase along the length of the plume, particularly in the southern flank of region III, where smoke is the most oxidized (
Figure 2e–f). CAS and PCASP data indicate the plume is comprised almost entirely of particles with r
e of 0.46 µm or less (i.e., “very small,” “small,” and “medium” particles, similar to the aerosols of the same name identified within the MISR retrieval sensitivity limitations, discussed in
Section 2.2 above). “Large” aerosols (0.46 µm < r
e < 1.7 µm) are essentially absent, never representing more than 0.35% of all particles, which is substantially below MISR retrieval sensitivity. Aerosol size decreases along the central plume transect, with the most dramatic decrease seen in the southern flank of region III (
Figure 3). PTI (532 nm) and PAS (355 nm) absorption measurements used to distinguish spectrally flat from spectrally steep aerosols show that BC-only absorption (which is approximately equal at both wavelengths) declines more dramatically downwind than BrC-only absorption (the residual of 355-nm absorption minus 532-nm absorption), and that CO-normalized BrC absorption and CO-normalized 355-nm scattering actually increase dramatically in region II (
Table 4). In their final campaign report,
Kleinman and Sedlacek [
18] also found that CO-normalized BrC absorption and scattering increased after ~30 min of plume aging. Although particle size is known to influence the spectral slope of SSA, the observed jump in 355-nm absorption and scattering does not correlate with
significant changes in particle size, and thus it is likely that particle composition is the dominant factor driving the observed changes.
The in situ data alone do not provide enough information to fully explain these changes in aerosol properties. However, from available observations, we suggest several potential processes that could be affecting the plume particle properties: (1) gravitational settling and/or dilution that leads to the preferential decrease in medium-size particle concentration and BC particle concentration in regions II and III compared to region I; (2) oxidation and/or SOA production, leading to increased hygroscopicity and decreased light absorption contributions in regions II and III; (3) changes in burning conditions at the source, which alter the emitted smoke properties, reflected in particle-type differences between regions. Gravitational settling cannot be confirmed by the in situ data acquired at a single elevation. Particle oxidation state can be measured in situ, but can only be inferred from remote sensing, e.g., based on decreasing particle light absorption, although particle hydration is also possible, especially if particle size increases, hydrophilic particles (e.g., BrC) are present, and RH is high. If in situ measurements of particle size but not composition are available, the formation of secondary particles would have to be confirmed by total column measurements, as it is possible that decreases in particle size, as measured in situ, might instead be due to preferential gravitational settling of larger particles. However, the CO-normalized, absolute increase in small-particle concentration strongly supports the idea of particle formation in this case, as it indicates that dilution is minimal. Other particle evolution processes can occur, such as aggregation; however, based on the available in situ data, this does not seem likely here, unless it occurred very near source, where aerosol loading was extremely high. Changes in the burning conditions at the source cannot be confirmed or ruled out by remote sensing, and in situ measurements only allow us to infer the degree to which the fire may be changing. Several overlapping transects made by the BBOP aircraft at different times suggest measurements at a given distance from the source (i.e., similarly aged smoke at different times) are not dramatically different between these observations, indicating that changes in burning conditions may not be important here.
3.2. Satellite Observations: Validation and Providing a Broader Context
As demonstrated above, aircraft measurements provide highly detailed observations of the plume that allow us to identify general particle-type differences and to narrow down the list of potential aging mechanisms. However, the limited area sampled in situ across two hours, during which changes could have occurred, leaves us with an incomplete understanding of plume properties, and these data alone cannot resolve key uncertainties regarding the atmospheric particle processes that operated. Here, we provide an overview of the satellite observations, with the same emphasis on
differences between regions as in
Section 3.1, and we compare them with the aircraft measurements to assess the fidelity of the RA. We also demonstrate how the RA places in situ observations into a broader context, providing information critical for understanding the processes responsible for plume evolution.
Figure 4 contains standard MISR RA latitude/longitude plots of AOD, SSA, and ANG for the plume, with the MISR–MINX stereo heights included for reference. In addition to these, we map, in terms of both retrieved absolute AOD and AOD fractional component, the spectral dependence of aerosol light absorption via the separate sums of (1) retrieved spectrally flat (BC analog) and (2) spectrally steep (BrC analog) light-absorbing components (
Figure 5a–d), (3) the retrieved aerosol shape, via the non-spherical component LaNsphWab (
Figure 5e–f), and (4) further information on absorption via the sum of all non-absorbing components (
Figure 5g–h). These particle properties are, for the most part, neatly distinctive between the four plume regions, as discussed subsequently. It is important to reiterate that, when considering specific particle types (e.g., black carbon, brown carbon, dust, non-absorbing aerosols), the retrieved aerosol properties represent the optical equivalent of the aerosol present, i.e., the true aerosol is not necessarily a mixture of the specific components retrieved from among those in the RA climatology (e.g., Reference [
30]).
Table 5 quantifies the above-mentioned RA observations, giving the mean, standard deviation (SD or σ), and median for each region; a more in-depth analysis of these statistics can be found in the
Supplementary Materials.
The MISR–MINX stereo heights indicate the plume is concentrated mostly within 2 km above sea level (ASL), with a median height of 1.78 km and a maximum height of 2.13 km (
Figure 4a–b). The RA-retrieved particle properties indicate peak AOD of up to 3.5 near the source, decreasing with downwind distance to a minimum of ~0.5 in some pixels within the area we defined as the observable plume in the satellite imagery (
Figure 4d). The plume REPA also decreases significantly with downwind distance, and SSA increases from < 0.84 near the source to nearly 1.0 in region IV. Although most particles can be considered fine mode, having ANG > 1, REPS increases slightly in region II, dramatically decreases in region III, and then dramatically increases again in region IV (some pixels having ANG < 1), indicating significant changes in particle size throughout the length of the plume. Component analysis suggests that smoke near the source is largely a BC-like mixture, but that subsequent regions transition to mixtures interpreted as mostly non-absorbing and BC particles in region II, then non-absorbing, BC, and BrC particles in region III, and non-absorbing and dust- or soil-like optical analogues in region IV (
Figure 5). Bear in mind that these retrieved component AOD values and fractions represent an interpretation of the MISR column-integrated reflectances in terms of the components included in the RA (
Table 2), whereas the REPS and REPA are less dependent on algorithm assumptions. This is one reason why validation with the in situ data is so important.
The satellite and in situ measurements are substantially similar in their characterization of particle property evolution as a function of smoke age. In both cases, particle absorption systematically decreases downwind, accompanied by a transition from BC to BrC particle-type analogues as the dominant light-absorbing aerosol. Significantly smaller particles appear in region III and partway into region IV, between 48 and 90 min of aging. However, the MISR-retrieved particle sizes for smoke between 24 and 48 min in age (region II) are larger than those near the source (although still fine-mode), whereas the aircraft observes the opposite trend. Moreover, unlike the aircraft measurements, MISR indicates that particles in region III are “small” in size, whereas the aircraft measures “very small” particles (
Table 3). Lastly, MISR coverage continues past 90 min in age, observing region IV, where the retrieval results indicate particles eventually appear somewhat coarser in size and significantly non-spherical. Despite these differences, both observation sets indicate that, overall, from the source until ~78 min of aging, most of the plume comprises a majority of small, very small, and medium particles, with monotonically decreasing light absorption downwind.
Due to the comprehensive coverage, the MISR observations provide greater context for assessing the aerosol aging mechanisms that are suggested by the in situ data analysis. In region II, the increased REPS is in contrast with the decrease in particle size observed in situ. This may be explained by the difference in vertical sampling, as the aircraft measures at a single altitude, whereas the satellite observes the entire column. Column aerosol loading is dominated by the smoke plume in this case, but the plume is at least 0.75 km thick (
Figure 4b); thus, measurements at a single elevation might not be representative. This would bolster the idea that gravitational settling is playing a role in this region. It is also possible that particle hydration, condensation of gaseous components onto existing particles, and/or particle aggregation are acting to increase particle size. The observed increase in relative humidity in this region (see
Supplementary Materials, Figure S2) points toward hydration in particular, and the smaller sizes observed by the aircraft can be explained by differences in horizontal and possibly also vertical sampling. Aerosol oxidation is also acting in this region, as indicated by the increased oxidation values seen in
Figure 2b and supported by the increased hygroscopicity seen in
Figure 2e, as well as the decreased REPA in the MISR retrievals (
Figure 4f).
In region III, there is strong in situ evidence for high levels of aerosol oxidation, leading to the formation of BrC either from SOA production (via the nucleation of new particles) or from the deposition of organic material onto BC particles, increasing hygroscopicity and decreasing light absorption. The aircraft does not observe strong changes in particle light absorption, as would be expected with these processes, although this could be due to limited sampling (
Figure 2a). MISR does appear to see the expected increase in SSA (
Figure 4f), at least along the northern part of the plume in this region. The main particle-type feature of region III is the dramatic decrease in REPS, accompanied by a decrease in total-column AOD. It is unclear from the limited in situ sampling (
Figure 3c) whether medium particles are actually lost from the column here. However, region III begins about 30 km downwind of the source, and the MISR stereo height profile indicates that the plume descends toward the surface at about this distance (
Figure 4b), suggesting selective, large-particle deposition is possible. This is supported by a decrease in the total-column AOD for medium-sized particles between regions II and III (
Figure S3, Supplementary Materials). In addition, both absolute and fractional increases in small aerosol AOD are observed, strongly suggesting the formation of new particles. The dramatic increase in the CO-normalized aerosol concentrations in the aircraft data further supports this conclusion, as does the increase in both the absolute and the fractional BrC AOD observed by MISR. The formation of weakly absorbing organic aerosols (BrC) would also explain why SSA does not change more dramatically in this region.
Although the aircraft did not observe region IV, MISR points to potential aging mechanisms occurring here as well. Most notably, there is an increase in REPS, accompanied by decreasing MINX plume heights. Both the absolute and the fractional AOD of large and medium aerosols increase, suggesting that not only do small and very small particles disappear, but medium and large particles are being added (
Figure S3, Supplementary Materials). At such a long downwind distance and with low overall AOD (
Figure 4d), it is unlikely that particle aggregation is occurring in this region. Particle hydration might be involved, although the increase in non-spherical particles is not likely explained by this. Rather, these retrieval results might be due to a lack of larger, spherical, weakly absorbing particles in the aerosol climatology included in the MISR algorithm (
Table 2), and, from a single satellite snapshot, we cannot rule out temporal changes at the smoke source.
3.3. Brief Summary of the Government Flats Smoke Plume Analysis
Overall, the satellite-retrieved particle properties are relatively well supported by the in situ observations, with minor disparities that can be explained by measurement uncertainty and sampling differences. Together, the data indicate that (1) there is a near-source region of high AOD with a mixture of dark, mostly spherical, small-medium particles, with light absorption dominated by BC, (2) there is an adjacent region downwind with reduced AOD comprising brighter, slightly larger particles that are less BC-like, (3) there is a subsequent downwind region where AOD is reduced further, with significantly smaller particles that are brighter in nature and akin to BrC, and (4) there is a region at the nominal end of the plume with relatively low AOD, high SSA, and a significant fraction of larger, possibly non-spherical particles.
Our interpretation of the RA results parallels the atmospheric–aerosol interactions suggested by the in situ data. Through column-integrated particle-type retrievals and plume stereo heights, the preferential gravitational settling of larger particles is supported in regions II and III. Furthermore, decreased REPA throughout the length of the plume bolsters the idea of particle oxidation and/or hydration in regions II and III. Lastly, a sharp decrease in REPS in region III accompanied by a dramatic increase in BrC particle AOD strongly supports the idea of new-particle SOA formation here, although the RA results alone cannot rule out condensation of gaseous compounds onto existing particles.