ACE-FTS aboard the Canadian ACE–SCISAT-1 platform is a solar occultation instrument that collects up to 30 daily atmospheric absorption measurements at sunrise and sunset using the Sun as a light source (Bernath, 2005, 2017). The infrared Fourier transform spectrometer measures over a wide spectral range (750 to 4400 cm⁻¹) with a high resolution of 0.02 cm⁻¹ and signal-to-noise ratio ranging between 100 : 1 and 400 : 1 (Buijs et al., 2013). The ACE-FTS processing version 5.2 provides vertical volume mixing ratio (VMR) profiles for 46 molecules and 24 isotopologues, including characteristic biomass-burning indicator molecules such as carbon monoxide (CO), hydrogen cyanide (HCN), acetonitrile (CH₃CN), and ethane (C₂H₆) (Boone et al., 2023). A pair of filtered imagers also measure atmospheric extinction at two wavelengths: visible (VIS, 527.11 nm) and near-infrared (NIR, 1020.55 nm). The NIR imager is less likely to become saturated in cases of strong aerosol extinction and was thus used in this analysis for detection of aerosol loads (Vanhellemont et al., 2008; Boone et al., 2020). Temperature profiles are also collected with every occultation and were used to calculate tropopause heights (Sect. 2.3). Data​​​​​​​ from ACE are available at https://databace.scisat.ca/level2/ace_v5.2/ (last access: 25 August 2024) starting from 2004 (Boone et al., 2023).

Given the limited data coverage of ACE, we also analyze carbon monoxide data from the MLS aboard the NASA Aura satellite (Waters et al., 2006) given its higher spatial coverage. The MLS instrument has seven radiometers that measure microwave thermal emissions between 118 GHz and 2.5 THz from the limb of the atmosphere to determine vertical profiles of atmospheric constituents. This instrument has a much higher spatial coverage than ACE, but the use of microwave spectroscopy provides data with a lower signal-to-noise ratio on individual soundings and a more limited vertical range. For example, the CO data product is recommended for scientific use between 215–0.001 hPa, while the HCN and CH₃CN valid ranges are even smaller at 21–0.1 and 46–1.0 hPa, respectively (Livesey et al., 2022). Since we are interested in the vertical profiles of biomass-burning products, MLS is not as useful as ACE for our analysis, given the lack of reliable data at our pressure range of interest in the lowermost stratosphere and its transition to the tropopause and upper troposphere. However, carbon monoxide data from MLS in the lower stratosphere are still useful as complementary data since strong signals from large perturbations would be clearly detected (Fig. S1 in the Supplement). Additionally, comparison of ACE and MLS profiles in our latitude range of interest supports our conclusions from ACE data despite more limited coverage (Fig. S2). Data from MLS are available at https://disc.gsfc.nasa.gov/datasets/ML2CO_NRT_005/summary (last access: 25 August 2024) starting from 2004 (EOS MLS Science Team, 2022).

OMPS LP on the Suomi NPP satellite is a limb profiler that measures scattered UV, visible, and near-IR radiation. Aerosol extinction coefficients are retrieved for six wavelengths (510, 600, 675, 745, 849, and 997 nm) with the V2.1 algorithm (Taha et al., 2021). OMPS LP measures along Earth's limb with three parallel vertical slits: one central slit that views along the nadir track and two side slits viewing with a cross-track separation of 250 km at the tangent point, allowing for near-global daily coverage. The 745 nm channel was used in this analysis given its high sensitivity to aerosol loading and low bias, and aerosol extinction is analyzed as an indicator of aerosol abundance. A cloud detection algorithm detects the highest cloud altitude and flags all aerosol extinction measurements below the peak cloud level, allowing for a cloud-filtered data product (Chen et al., 2016). The retrieved aerosol-to-molecular extinction ratio analyzed in this work is analogous to an aerosol mixing ratio; see Loughman et al. (2018) and Taha et al. (2022) for a detailed account of the retrieval algorithm and previous data applications. Data from OMPS are available at https://disc.gsfc.nasa.gov/datasets/OMPS_NPP_LP_L2_AER_DAILY_2/summary (last access: 25 August 2024; Taha, 2020).

Determination of stratospheric entry depends on the definition of tropopause height. Given the high variability of tropopause heights, using data that are specific to the time and location of the VMR and aerosol extinction measurements of interest is important. Concurrent temperature profiles are retrieved alongside VMR profiles from ACE; specifically, ACE temperature profiles are determined from 18 km upwards from data retrievals, while temperatures and pressures below 18 km are fixed to data from the Canadian Meteorological Centre weather model (Sica et al., 2008). Therefore, a temperature-based tropopause definition, the WMO lapse rate tropopause, was used in this study (WMO, 1957).

In practice, the lapse rate tropopause was calculated by first determining the lapse rate of the ACE temperature profile at every vertical level and interpolating to determine slope values between the 1 km vertical intervals. Every vertical level where the lapse rate reached -2 K km⁻¹ was marked, and the lowest level at which the lapse rate passed this threshold was determined to be the tropopause altitude. For temperature profiles that exhibit a double tropopause, this method provides the lower tropopause altitude (Peevey et al., 2012; Homeyer et al., 2014).

The tropopause acts as a strong barrier to troposphere-to-stratosphere transport, and the high vertical resolution of ACE is useful in determining whether smoke entered the stratosphere. Individual occultation measurements from ACE provide simultaneous constituent VMR and temperature profiles where biomass-burning product VMRs can be compared against self-consistent tropopause heights as measured on the same occultation and on broader averages. HCN is a robust wildfire tracer for the Northern Hemisphere because its primary source is biomass burning (Li et al., 2000; Roberts et al., 2020). In contrast, other tracers like CO and ethane are emitted in large amounts by both wildfires and anthropogenic sources such as transportation and industry (Xiao et al., 2004), complicating source attribution for those compounds. Additionally, wildfires are not the only source of particles in the stratosphere, especially in the recent context of the Hunga Tonga–Hunga Ha'apai and Shiveluch eruptions in the last couple of years. This is why we focus on chemical data from ACE-FTS in addition to aerosol data to be able to confidently attribute wildfires as the source.

We first investigate average monthly HCN VMR profiles. Some individual occultations do not extend into the troposphere, so monthly averages in absolute altitude capture the maximum amount of data to compare 2023 with the rest of the ACE-FTS measurement period (2004 to 2022). A tropopause-relative framework is used in Sect. 3.2 to detect individual measurements of stratospheric smoke. Chemical anomalies detected in the region from 40 to 70° N that are likely associated with wildfires also reflect other active burning regions such as Siberia (MODIS Land Rapid Response Team, 2023, https://modis.gsfc.nasa.gov/gallery/individual.php?db_date=2023-07-05, last access: 27 February 2024). Figure 1 suggests an enhancement of HCN in the upper troposphere and lowermost stratosphere (UTLS) between June and September 2023 relative to preceding years, with the monthly average tropopause heights calculated from ACE-FTS temperature profiles representing a rough delineation between the troposphere and stratosphere. On average, enhancements do not extend more than a couple kilometers above the tropopause. Similar results can be seen for single profiles (Fig. S3).

ACE–SCISAT-1 takes a very limited number of measurements over our latitude range of interest during August, as seen in Table S1; only 20 measurements were taken in August 2023 and all on the first couple of days of the month between 40 and 45° N. This may explain the unusually high tropopause altitude for this month as tropopause altitudes are higher closer to the Equator, and previous studies with ACE in August also report a higher tropopause height (Doeringer et al., 2012). Additionally, a low number of measurements are likely more susceptible to transient meteorological influences such as the Asian monsoon (Basha et al., 2020). ERA5 reanalysis data collocated at ACE measurement points and times for August yield a similarly high average tropopause height of 14.6 km (Hersbach et al., 2023).

The range given by the standard deviation of ACE data from 2004 to 2022 indicates that there were likely other wildfire years that also produced high HCN mixing ratios from June to September, but 2023 is on the upper end of this spread from about 8 to 11 km. July 2023 in particular has noticeably higher abundances compared to the range of previous years. Although this HCN enhancement is indicative of wildfire smoke reaching the upper troposphere, the occurrence of stratospheric injection is less clear given the variability of tropopause heights and the limited number of observations, particularly in August (see Table S1). This limited coverage and amount of August ACE data are supplemented by the higher coverage of OMPS as shown in the following section and MLS in the Supplement.

The average July 2023 profiles of other biomass-burning markers such as CO, CH₃OH, HCOOH, and C₂H₆ also exhibit elevated VMRs in the upper troposphere but not significantly so in the lowermost stratosphere (Fig. 2).

These data provide strong evidence of the vertical transport of wildfire smoke to the upper troposphere due to high pyroCb activity during the summer wildfire season, which can influence upper tropospheric composition and chemistry (e.g., ozone production). However, significant stratospheric penetration is not observed in the averages.

Inspection of every individual occultation measured by ACE-FTS over the 40 to 70° N latitude band in the 2023 burning season revealed a few occultations indicative of stratospheric smoke, exhibited by enhanced chemical signatures and aerosol extinction above the tropopause. These measurements offer evidence of a small amount of smoke in the lowermost stratosphere. The July occultations that exhibit multiple biomass-burning products and aerosol extinction measured in the stratosphere are shown in Fig. 3, and similar profiles for other months are shown in Fig. S5.

Elevated VMRs of multiple biomass-burning products measured above the tropopause in these data offer evidence of a nominal amount of wildfire smoke in the stratosphere. To further validate this conclusion, the ACE infrared absorption spectra were also analyzed for features characteristic of wildfire smoke, as in Boone et al. (2020). Carbonaceous aerosols typically exhibit C–H, O–H, and C–C features associated with alkanes and oxygenated organics, and these are indeed present for these occultations, for example as shown in the IR spectrum in Fig. 4 (Zhong and Jang, 2014; Boone et al., 2020; Bernath, 2020). The O–H stretch is commonly seen in smoke particles and indicates that they are oxygenated (Boone et al., 2020). The C=O stretch, which is also associated with oxygenated smoke particles and identified using a feature around 1750 cm⁻¹, cannot be seen because that region is saturated at lower altitudes by strong absorption of water vapor in the same spectral region.

Given the extent of Canada's wildfire season, using fire databases to identify potential source fires is non-trivial given the large number of events in 2023 with extensive burned areas (Canadian Interagency Forest Fire Centre, 2024, https://www.ciffc.ca/publications/canada-reports, last access: 17 October 2024). Thus, more targeted methods for identifying pyroCb-specific fires are a more promising approach for this analysis. Cross-referencing with an online crowdsourced pyroCb database (https://groups.io/g/pyrocb, last access: 25 August 2024) suggests a few pyroCb events with brightness temperatures below -55 °C. A cold cloud top of this magnitude is indicative of deep overshooting convection that may penetrate the tropopause and enter the stratosphere (Romps and Kuang, 2009). It is therefore plausible that these pyroCbs injected smoke into the stratosphere.

Back trajectories initialized from ACE stratospheric smoke measurements on NOAA HYSPLIT with GDAS 1° meteorological data do not directly intercept any of these reported pyroCbs, but this is not surprising given the limited spatial coverage of ACE (Stein et al., 2015). Trajectories initialized from two occultations, ss107346 and ss107570, pass within tens of kilometers and a few hours of reported pyroCbs (Fig. S6). Despite the lack of direct detection, chemical signatures of smoke after dispersion in the lower stratosphere are clearly measured and show the limited vertical range of wildfire influence in the stratosphere: within 2 km above the tropopause.

Aerosol extinction data from OMPS LP also indicate a minor increase in aerosol burden in the lower stratosphere averaged between 11.5 and 16.5 km, in late July 2023 (Fig. 5). However, this increase above background levels is both small in magnitude and short-lived, which indicates that not enough smoke was injected into the stratosphere to significantly impact extinction measurements. Throughout the entire wildfire season, there are no significant aerosol extinction signals at this altitude range that would suggest deep convective wildfire events. This validates our general finding that very little smoke managed to make it well above the tropopause despite high pyroCb activity, suggesting that the many convective events that occurred were mostly limited to the upper troposphere. There is a constant background level of aerosols above 60° N, but this feature is present from the beginning of 2023. Therefore, it is not linked to the 2023 Canadian wildfire events but rather to other aerosol sources such as volcanic eruptions.

The altitude range between 10.5 and 11.5 km contains either tropospheric or stratospheric air depending on the latitude and time of year, and an analysis of OMPS data at this altitude reveals substantial average aerosol loading in the region around the tropopause in 2023 (Fig. 8, middle panel). This suggests that the many pyroCbs of the 2023 wildfire season alongside other aerosol sources managed to inject particles right around the tropopause. However, the significantly lower aerosol extinction signal past 11.5 km as seen in Figs. 5 and S7 constrains these inputs to the region at or just above the tropopause as opposed to higher in the stratosphere, where more impacts would be realized.

To contextualize the 2023 Canadian events, we compare its ACE-FTS profiles with measurements of the 2017 PNE, which also occurred in western Canada during the summer. Analysis of this event was previously reported in Boone et al. (2020), with the occultation sr75758 probing the plume a few weeks after initial stratospheric injection. Figure 6 shows a substantial difference in smoke altitude, with the 2017 PNE leading to chemical signatures of biomass-burning products measured above 20 km compared to the much lower vertical extent of the 2023 fires.

The top panel of Fig. 7 also shows with OMPS data that 2017 featured a much larger injection of aerosols between 11.5 and 16.5 km relative to 2023, where there is no evident increase in average stratospheric aerosol extinction throughout the entire summer. Similarly, the bottom panel of Fig. 7, compared to Fig. 5, shows that the magnitude and duration of stratospheric perturbation from the PNE were much more substantial than those of any pyroCb activity from 2023. Since ACE-FTS measured only a handful of occultations with stratospheric signatures of smoke and the OMPS LP aerosol data do not feature any significant perturbations in 2023 past 12 km, we conclude that the 2023 Canadian wildfire season did not significantly impact stratospheric composition relative to previous years despite burning a far larger area and exhibiting frequent pyroCb activity.

However, the aerosol loading near the tropopause is similar between the 2 years, as seen in Fig. 8. It is clear that the active pyroCb activity in the 2023 wildfire season did influence upper tropospheric composition given the increase in aerosol extinction starting in May, which coincides with the start of the burning season. This is in stark contrast to the stratospheric aerosol loading from Fig. 5, which is the largest at the beginning of the year. Thus, the stratospheric impact of the 2017 PNE is visibly more significant since smoke following the event is measured well above the tropopause, whereas smoke from the 2023 fires remains around the tropopause throughout the entire wildfire season. This lower altitude of perturbation corresponds to a shorter smoke lifetime in the stratosphere, further suggesting the lower impact of the 2023 fires compared to the PNE (D'Angelo et al., 2022).