The homogeneous nucleation of NAT is kinetically suppressed (Koop et al., 1995). Laboratory experiments have revealed that the homogeneous nucleation rate of NAT on liquid STS is extremely low under stratospheric conditions (Hanson and Ravishankara, 1991, 1992). Thus, NAT forms through heterogeneous nucleation processes such as (i) ice-assisted NAT nucleation and (ii) ice-free NAT nucleation. During the ice-assisted NAT nucleation process, the ice particles serve as nuclei for NAT formation upon warming of the air parcels, and this occurs at a high saturation ratio over NAT of > 500 (Luo et al., 2003). This mechanism has been supported by several field observations (Carslaw et al., 1999) and by bulk-phase lab experiments (Koop et al., 1995, 1997), which indicated that the deposition nucleation of NAT on exposed ice surfaces is also a possible pathway for NAT formation. During the ice-free NAT nucleation, instead of ice, STS with solid foreign nuclei inclusion, such as meteoritic dust, volcanic ash, soot, or H₂SO₄ hydrates, serves as a NAT nucleus (Iraci et al., 1995; Koop et al., 1997; Peter and Grooß, 2012), and this even occurs at a low saturation ratio over NAT (SNAT) of <10 (Voigt et al., 2005). Hoyle et al. (2013) made a new parameterization scheme for NAT nucleation on foreign nuclei in the immersion mode and reproduced the CALIPSO PSC observations for the entire Arctic winter of 2009–2010.

Ice PSCs can form either homogeneously under high-supersaturation conditions, with a saturation ratio over ice (Sice) > 1.5, at a temperature of 3 K less than the frost point (Tice) (Koop et al., 1998) or heterogeneously on STS with foreign nuclei inclusion at a temperature of 1.5 K less than Tice (Koop et al., 1998, 2000; Engel et al., 2013). Apart from these foreign nuclei, NAT is also known to act as potential nucleus for ice formation (Gao et al., 2016; Hanson and Mauersberger, 1988; Iannarelli and Rossi, 2015; Khosrawi et al., 2011; Weiss et al., 2016). Based on the WALES (Water Vapor Lidar Experiment in Space – airborne demonstrator) lidar aboard HALO (High Altitude and Long Range Research Aircraft), Voigt et al. (2018) showed that the depolarization ratio of ice PSCs exhibited a bimodal distribution: the low depolarization mode corresponds to the ice nucleated on STS with foreign nuclei inclusion, whereas the high depolarization mode corresponds to the ice nucleated on NAT particle. Engel et al. (2013) made a new parameterization scheme for ice nucleation on foreign nuclei in immersion mode and reproduced the CALIPSO PSC observations for the entire Arctic winter of 2009–2010.

In the present study, along with investigating the impact of the Black Summer event on lower-stratospheric chemistry, in order to explain the observed PSC anomalies, we also aim to quantify the percentage of the liquid–NAT mixture formed via the ice-assisted and ice-free nucleation processes. Furthermore, we quantify the percentage of ice PSC that nucleates on NAT (NAT-assisted nucleation pathway) and STS (NAT-free nucleation pathway).

As liquid–NAT mixture and ice are the first (with 48 %) and third (with 21.4 %) most abundant types of PSCs in terms of areal coverage (Pitts et al., 2018), performing backward trajectories for all of these detected PSC types is computationally expensive. Furthermore, the ERA5 operational analysis data are available at a temporal resolution of 1 h, whereas CALIPSO's temporal resolution is one profile per second. The accuracy of the trajectory relies on the accuracy of the wind field and diabatic heating rate. Hence, assigning an hourly wind field and diabatic heating rate to all CALIPSO profiles measured during the specific hour may contribute to uncertainty in the calculated trajectory. Although the degree of uncertainty is unclear, as a conservative measure, we impose another criterion to minimize any uncertainty that may contribute. Thus, the two criteria utilized are as follows:

Criteria 1. Firstly, CALIPSO profiles that are measured within 5 min of (before and after, i.e., 10 min time windows) any hour of the ERA5 operational analysis data are chosen. For instance, if ERA5 operational analysis data are available for the time 01:00 UTC for a specific day, CALIPSO profiles measured between 12:55 and 01:05 UTC on the same day are considered.

The backward trajectories for these selected grid points are calculated for a 48 h period using the CLaMS trajectory module. The rationale behind choosing 48 h is that, once the air parcel's temperature drops below TNAT and following the nucleation of NAT particles with a number density of 5×10-4 to 5×10-5 cm⁻³, the NAT particles exceed the CALIPSO perpendicular backscatter threshold and become detectable within ∼ 19 h (0.8 d) (Lambert et al., 2016). Similarly, Voigt et al. (2005) provided observational evidence from aircraft campaigns showing NAT formation within approximately 20 h after the temperature drops below TNAT. In the case of ice formation, a 48 h period should be sufficient, considering the average cooling rate in the stratosphere. The model runs are driven by meteorological data from the ERA5 operational analysis, with a temporal resolution of 1 h and a spatial resolution of 1° × 1°. The time step is set to 60 s in the CLaMS trajectory module. The vertical coordinate is potential temperature. The vertical motion of the air parcels is modulated by total diabatic heating rates derived from ERA5, following Ploeger et al. (2021).

To determine the PSC composition along each trajectory, we first identify the intersection points where the trajectory intersects with the CALIPSO scan track, with a temporal coincidence of ±30 min. Intersections with time differences exceeding this threshold are disregarded, thereby ensuring that the trajectory and CALIPSO scan track are both spatially and temporally co-located within a 30 min time window. The PSC composition at each trajectory intersection point is then determined from the CALIPSO measurement profile with a potential temperature closest to that of the trajectory's intersecting point. If the PSC composition determined in this manner differs from the initial composition observed at the start of the trajectory (i.e., at time t=0 h), we assign this PSC composition to the intersection point of the trajectory. Conversely, trajectories with no change in PSC composition are excluded from further consideration. Moreover, we fill the MLS-observed HNO₃ and H₂O mixing ratios along the selected trajectories to study the uptake of these gases during PSC formation.

Following the above three steps, we chose a total of 8841 backward trajectories. Of these, 3792 trajectories correspond to a liquid–NAT mixture, whereas the remaining 5049 trajectories correspond to ice PSCs. Each trajectory contains the temperature from the ERA5 operational analysis, HNO₃ and H₂O mixing ratios from MLS at the points where the trajectories intersect with the CALIPSO scan track, and the PSC composition from CALIPSO. In addition, derived quantities such as Tice, TNAT, ice and NAT surface area density from the CLaMS box model run, SNAT (for the case of backward trajectories of liquid–NAT mixture), and Sice (for the case of backward trajectories of ice) are added to the trajectories. Hence, this provides a comprehensive picture of the air parcels' temporal evolution leading to the formation of liquid–NAT mixture and ice PSCs.

In general, PSCs exhibit a strong interannual variation owing to the polar vortex dynamics, such as vortex temperature and area. Hence, to understand the association of the polar vortex variability with the observed PSCs anomalies, the statistics, such as the minimum, maximum, mean, and standard deviation of the PSC volume, for each PSC type observed from 2009 to 2020 (excluding the year 2015) and their relationship with polar vortex dynamics are studied in detail and shown in Fig. 7. Here, the polar vortex is defined as the region enclosed by the 32 PVU (potential vorticity unit) potential vorticity contour, following the definition from Kondragunta et al. (2005). The vortex area is quantified as the total area within this defined region, while the vortex temperature represents the mean temperature over the same area. The potential vorticity and temperature are taken from the ERA5 operational analysis dataset.

During early May and June of 2020, the mean vortex temperature was the highest compared to the background period (from 2009 to 2019, excluding the year 2015) (Fig. 7a). In contrast, the mean vortex temperatures for the remaining months up to September 2020 were neither the warmest nor the coldest. However, after September 2020, the mean vortex temperature dropped to its lowest levels (ranging from 194 to 200 K), which can be attributed to the prolonged vortex period and ozone depletion. Owing to this colder environment, STS PSCs continued to appear (even after September 2020), resulting in the highest STS PSC volume with respect to the background period. However, this colder condition was not sufficient to affect the formation of other PSC types (i.e., notable positive anomaly) after September 2020 (Figs. 6 and 7). In terms of total PSC volume (i.e., a sum of the volume of total PSC types), no anomalous behavior in the total PSC volume was observed, in keeping with the results reported by Li et al. (2024). However, upon investigation of the individual PSC volume, we found that ice, enhanced NAT, and STS PSCs exhibited anomalously high volumes for a certain period during 2020, exceeding 1 standard deviation (Fig. 6c, i, and l) as well as the recorded decadal high value (Fig. 7c, e, and f; marked with red-filled circles). The periods from 8 to 14 August 2020 and from 27 August to 9 September 2020 recorded the highest ice PSC volume among the years from 2009 to 2020, reaching up to 45×106 km³ and 22×106 km³, respectively. Similarly, the year 2020 recorded the highest enhanced NAT PSC volume of up to 12×106 km³ during the period from 21 to 30 July 2020. For the STS PSCs, the second week of July and the last week of August 2020 recorded the highest PSC volume. However, the liquid–NAT mixture PSC volume did not exhibit significant anomalies relative to other PSC types. As the magnitude of the PSC volume also depends on the polar vortex dynamics, we investigated the polar vortex area and mean temperature for the year 2020 and compared these values with the background period.

Early May and early June of 2020 were the warmest periods in the background period (Fig. S3a). However, on most of the days from late June to early August of 2020, the Antarctic polar vortex area was recorded to be the highest, ranging between 23×106 km² and 32×106 km². In contrast, this period was not the coldest, as revealed by the mean vortex temperature. We checked the role of this increased polar vortex area during 2020 with respect to the increased PSCs by investigating the vortex occurrence frequency anomaly at a potential temperature level of 450 K and the PSC volume from June to August 2020. For this purpose, a positive anomaly in the vortex occurrence frequency greater than 1 standard deviation with respect to the background period (the single-hatched region in Fig. 8) and a negative anomaly greater than 1 standard deviation (the cross-hatched region in Fig. 8) are overlaid on the monthly PSC volume anomaly.

During June 2020, the vortex occurrence frequency exceeded 1 standard deviation over the eastern Antarctic and northeastern regions (marked with single hatching in Fig. 8, left column), indicating a significant expansion of the polar vortex in these areas. However, no positive anomalies in any PSC types were observed over the same region, suggesting that the increased vortex area during June 2020 did not influence PSC formation (Fig. 8, left column). In July 2020, the polar vortex shifted off the South Pole toward the Antarctic Peninsula, resulting in a significant positive anomaly in vortex occurrence frequency over the northwestern region (marked with single hatching in Fig. 8, middle column) and a negative anomaly over the southeastern region (marked with cross-hatching in Fig. 8, middle column). Over the Antarctic Peninsula, a positive anomaly in the liquid–NAT mixture PSC volume slightly overlapped with this stretched vortex area, while no other PSC types exhibited anomalous behavior (Fig. 8, middle column). By August 2020, the vortex expanded further, stretching toward the Antarctic Peninsula, whereas it contracted over the eastern Antarctic region (Fig. 8, right column). Although a significant increase in PSC volume was observed during this month, particularly for ice and liquid–NAT mixtures, these PSCs remained confined close to the South Pole. This again suggests that the increased polar vortex area did not influence PSC formation or contribute to the observed positive anomalies in PSC volume. As discussed earlier in Fig. 7, except for liquid–NAT mixture PSCs, the volume of other PSCs exhibited anomalous behavior during certain periods of 2020. To further understand the observed anomalies, we compared the mean vortex temperature and PSC volumes from 2020 with those from previous years (2009–2019). For this comparison, we specifically selected time periods during which notable anomalies were observed in each PSC type during 2020. These include the ice PSC volume from 8 to 14 August (Fig. 9a) and from 27 August to 9 September (Fig. 9b), the enhanced NAT mixture PSC volume from 21 to 30 July (Fig. 9c), and the STS PSC volume from 23 to 30 August (Fig. 9d). These periods were chosen because at least one type of PSC recorded its highest volume in 2020 compared to any year between 2009 and 2019 (Fig. 7, red-filled circles). The mean vortex temperature from 8 to 14 August 2020 was ∼ 188 K, which was considerably warmer than previous years (such as 2011, 2016, 2017, 2018, and 2019). Despite this relatively warm condition, the mean ice PSC volume was observed to be 46×106 km³ (Fig. 9a), which was significantly higher than in previous years. Similarly, from 27 August to 9 September, the mean vortex temperature was ∼ 191 K. Once again, it was not the coldest period when compared to the years from 2009 to 2019. Nonetheless, the mean ice PSC volume peaked at 22.5×106 km³, which was 2-fold higher than the previously recorded values. Similar observations were made in the case of the enhanced NAT mixture and STS PSC volumes (Fig. 9c and d), for which significant increases in their respective PSC volume were observed despite the vortex temperature being comparable to the previous years.

This analysis indicates that the increased PSC areal coverage and volume cannot be attributed to vortex dynamics, such as variations in vortex area or temperature. Instead, we attribute these anomalously high PSC areal coverage and volume values to the intrusion of bushfire aerosols from the Black Summer event. In summary, the stratospheric intrusion of bushfire aerosols resulted in the increased production of HNO₃ via enhancement of the N₂O₅ hydrolysis process while also increasing HNO₃-containing PSCs, such as STS and enhanced NAT mixture PSCs. However, despite the relatively high availability of HNO₃ during 2020, no significant anomaly in the liquid–NAT mixture PSCs was observed. One possible explanation for this observation could be the rapid conversation of NAT particles into ice PSCs, as NAT can serve as an efficient nucleus for ice formation (Hoyle et al., 2013). To verify this hypothesis, we performed Lagrangian backward-trajectory analysis of the ice and liquid–NAT mixture PSCs to study their respective formation pathways as discussed Sect. 3.3; the results of this analysis are discussed in the next section.

In this section, we discuss case no. 1, in which NC (No Cloud) precedes the liquid–NAT mixture along the backward trajectory. In this case, there is no evidence of the formation of ice in either the CALIPSO observations or based on the presence of conditions conducive to ice formation, suggesting the possibility of the ice-free nucleation of liquid–NAT mixture PSCs.

At 18:00 UTC on 10 July 2020, CALIPSO detected a liquid–NAT mixture PSC at a latitude of -69.1° and longitude of 99.15°, with a potential temperature of 484 K. This observation is marked by a yellow diamond in Fig. 10a, and the corresponding CALIPSO scan track is shown as a solid gray line; the CALIPSO PSC orbit curtain plot is shown in Fig. S7a. The dashed black line in Fig. 10a represents the calculated 48 h backward trajectory of this PSC, with the color indicating the temperature history of the air parcel in Tice coordinates, i.e., ambient temperature (T) - the frost point (Tice). The temperature (T) is obtained from the ERA5 operational analysis, and Tice is estimated using the ERA5 pressure and the mean MLS H₂O mixing ratio found along the trajectory, following Marti and Mauersberger (1993).

The backward trajectory revealed that CALIPSO observed “No Cloud (NC)” along this trajectory 20.2 h earlier (at t=-20.2 h), on 9 July 2020, marked by a white circle in Fig. 10a. The temperature history shows that the temperature did not decrease below Tice between these two observations, indicating that the conditions were not conducive to ice formation. At the time of the NC observation, the temperature was ∼ 189 K, which is 2 K below the NAT temperature (TNAT). During this time, MLS-observed gas-phase HNO₃ and H₂O mixing ratios were 3.5 ppb and 4 ppm, respectively (Fig. 10c and d). Using these as initial conditions, a CLaMS box model run was performed from t=-20.2 to 0 h, simulating the evolution from NC to a liquid–NAT mixture PSC. After 20.2 h, the MLS HNO₃ decreased from 3.5 to 0.5 ppb, with no significant change in MLS H₂O. Figure 10c and d show that the CLaMS-modeled uptake of HNO₃ and H₂O agrees well with the MLS observations. Furthermore, the CLaMS box model run indicates that the NAT surface area density (SAD) increased to nearly 3 µm² cm⁻³ (shown in Fig. 10e), while the ice SAD remained at 0 µm² cm⁻³, supporting that no ice formation occurred before the observation of the liquid–NAT mixture PSC. During the transition from NC to the liquid–NAT mixture PSC, the saturation ratio over NAT stayed well below 30, further supporting the absence of ice involvement in the formation of the liquid–NAT mixture PSC (Fig. 10b; Luo et al., 2003; Voigt et al., 2005). As the liquid–NAT mixture PSC represents the mixture of liquid STS and solid NAT, it should be noted that a STS PSC should have formed between this transition and, specifically, before the formation of NAT. From this detailed case study of ice-free nucleation of the liquid–NAT mixture PSC discussed above, it is evident that the temperature of the trajectories did not decreased below Tice and that SNAT remained below 30, thus confirming no ice involvement during the liquid–NAT mixture PSC formation. Another example of an STS PSC preceding the observation of a liquid–NAT mixture PSC along the backward trajectory is shown in Fig. S8. Even in this case, the conditions are not conducive to ice formation between the observation of an STS PSC and a liquid–NAT mixture PSC, confirming that the NAT formed without ice. For brevity, further details of this case are not discussed here.

Here, two cases are discussed: case no. 2 and case no. 3. In the former case, the direct CALIPSO observation of ice preceding a liquid–NAT mixture PSC along the trajectory is presented. In the latter case, CALIPSO-observed NC precedes a liquid–NAT mixture PSC, but the conditions conducive to ice formation are found before the formation of the liquid–NAT mixture PSC, indicating the possibility of an ice-assisted nucleation process.

In this case, an ice PSC appeared along the backward trajectory before the appearance of a liquid–NAT mixture PSC, indicating the nucleation of NAT on the pre-existing ice PSC. At 23:00 UTC on 5 July 2020, CALIPSO detected a liquid–NAT mixture PSC (yellow diamond in Fig. 11a) at a latitude of -66.89°, longitude of 25.89°, and potential temperature of 510 K. The CALIPSO PSC orbit curtain plot for this observation is shown in Fig. S9. Before this, CALIPSO observed an ice PSC along this trajectory at t=-43 h on 4 July 2020 (blue star in Fig. 11a).

At the time of ice detection, the temperature was approximately 181.5 K, which is 1.5 K below Tice. During this time, MLS-observed gas-phase HNO₃ and H₂O mixing ratios were 0.37 ppb and 3.9 ppm, respectively. Furthermore, at this time, according to the CLaMS box model run, ice SAD was ∼ 10 µm² cm⁻³ and NAT SAD was 0 µm² cm⁻³ (Fig. 11e and f). After this, the temperature gradually increased, exceeding Tice, which decreased the ice SAD to 0 µm² cm⁻³ and increased NAT SAD to 0.14 µm² cm⁻³, possibly due to the nucleation of NAT particles on this ice. During the transition from ice to the liquid–NAT mixture PSC, the maximum SNAT was observed to be 43 at t=-43 h. The MLS-observed HNO₃ increased from 0.37 to 0.47 ppb, whereas H₂O decreased from 3.7 to 3.4 ppb. It should be noted that these changes in HNO₃ and H₂O are within MLS measurement uncertainty.

In this case, NC is observed before the observation of a liquid–NAT mixture PSC. However, the temperature along the trajectory decreased well below the ice formation temperature (i.e., below Tice-1.5 K), indicating the possible formation of ice and, thus, the nucleation of NAT on this ice PSC.

At 21:00 UTC on 10 August 2020, CALIPSO detected a liquid–NAT mixture PSC (indicated by a yellow diamond in Fig. 12a) at a latitude of -66.22°, longitude of 55.57°, and potential temperature of 399 K. The CALIPSO PSC orbit curtain plot for this observation is shown in Fig. S10. Before this observation, CALIPSO observed NC along this trajectory at t=-42.6 h on 9 August 2020. During that period, the temperature was around 197 K, which is 0.4 K above the TNAT, and MLS-measured gas-phase HNO₃ and H₂O mixing ratios were 5 ppb and 4.1 ppm, respectively. These conditions were not favorable for the formation of any type of PSC. Approximately 4.6 h later (i.e., at t=-38 h), the temperature decreased below TNAT (by 3 K), which likely promoted the formation of NAT, as supported by the CLaMS box model run, with the NAT SAD reaching up to 6 µm² cm⁻³ (Fig. 12e). No ice formation occurred at this stage, as the temperature remained above Tice. However, the temperature continued to decline, reaching Tice at t=-35 h, at which point ice began to appear in the CLaMS model, as indicated by the ice SAD (Fig. 12f).

As the temperature fell to 184 K, which is 5 K below Tice, the ice SAD peaked at 35 µm² cm⁻³, suggesting that these ice particles may have nucleated on the previously formed NAT. By this time, the CLaMS model shows that HNO₃ had been depleted from an initial value of 5 ppb to a value of 0.002 ppb, while H₂O decreased from 4.1 to 1.6 ppm. Additionally, SNAT peaked at ∼ 3000 (Fig. 12b). Unlike in case nos. 1 and 2, in which ice was not involved in NAT formation, a significant change in H₂O was observed in this instance due to the formation of ice. Subsequently, the temperature began to increase gradually and reached a value above Tice at t=-11 h, resulting in an ice SAD of 0 µm² cm⁻³. At this point, all condensed H₂O evaporated back into the gas phase, returning to the initial level of 4.1 ppm. In contrast, the gas-phase HNO₃ remained below 0.1 ppb. The temperature continued to increase, and CALIPSO detected a liquid–NAT mixture PSC at t=0 h; during this time MLS observed HNO₃ levels of 3.1 ppb, while the CLaMS-modeled HNO₃ was 3.5 ppb. Throughout this process, MLS HNO₃ decreased from 5 to 3.1 ppb, leaving 1.9 ppb in the condensed phase.

In summary, when the temperature was above TNAT, no cloud was observed by CALIPSO. As the temperature decreased below TNAT-3 K, NAT formed first, followed by ice nucleation when the temperature decreased below Tice. As the temperature increased, additional NAT particles nucleated on the existing ice. Although the CLaMS box model suggests the presence of ice between t=-35 and -11 h, there were no co-located CALIPSO observations along the trajectory to verify the presence of an ice PSC. However, we identified another backward trajectory related to the CALIPSO-detected liquid–NAT mixture PSC at a potential temperature of 388 K (which is 11 K lower than the current trajectory of 399 K), as shown in Fig. S11. The temperature fluctuations along this trajectory are similar to those in case no. 4, in which a co-located CALIPSO observation of an ice PSC was recorded at time t=-29.4 h. A similar case, in which an STS PSC precedes a liquid–NAT mixture PSC but in which the temperature along the trajectory decreased below the Tice, creating conditions favorable for ice formation, is presented in Fig. S12, along with the CLaMS box model run. Given the similarities between this case and case no. 3, which has been described above, we have not provided a detailed analysis of the case in which STS PSC precedes a liquid–NAT mixture PSC.

In these two cases of an ice-assisted nucleation process of liquid–NAT mixture PSCs, the temperature decreased below Tice and SNAT increased more than 30. Therefore, the extent to which the temperature decreased below Tice and the extent to which the SNAT increased along the trajectories from the instant of observation of an NC/STS/ice to the instant of the liquid–NAT mixture PSC observation can act as proxies to determine whether the liquid–NAT mixture PSC is formed via an ice-assisted or ice-free nucleation process.

During the formation of the liquid–NAT mixture, changes in the gas-phase concentrations of MLS HNO₃ and H₂O were observed, as previously discussed. We used the CLaMS microphysical box model to simulate the formation of the liquid–NAT mixture based on the backward trajectories and then modeled the uptake of HNO₃ and H₂O. These results were validated against the observed uptake values from MLS, as shown in Fig. 13. This result includes both the ice-assisted and ice-free nucleation of the liquid–NAT mixture.

A strong correlation is found between the MLS-observed change in HNO₃ (ΔHNO₃) and the CLaMS-modeled change in HNO₃, with a correlation coefficient (R) of 0.64 and a coefficient of determination (R2) of 0.4096 (Fig. 13a). The intercept of -2.72 suggests that the CLaMS model overestimates the HNO₃ condensation during NAT formation. This could be due to a possible cold bias in temperature from the ERA5 operational analysis that is used in this study. Furthermore, significant HNO₃ uptake of up to 12 ppb, as observed by MLS, occurred when NC preceded the observation of the liquid–NAT mixture PSC along the trajectory (indicated by the white circle in Fig. 13a). However, a relatively low HNO₃ uptake is observed for cases in which an STS PSC preceded the liquid–NAT mixture (marked by the green circle in Fig. 13a). It is important to note that ice may have formed during the transition from NC/STS to the liquid–NAT mixture, as discussed in case no. 3. Furthermore, in most cases where ice preceded the liquid–NAT mixture along the trajectory, a positive change in MLS ΔHNO₃ is observed; however, more detailed studies are required to understand these scenarios further.

In the case of H₂O uptake, a weak correlation is found between MLS ΔH₂O and CLaMS ΔH₂O, with an R of 0.23 and an R2 of 0.0529 (Fig. 13b). In addition to the cold temperature bias in the ERA5 operational analysis, this low correlation is also attributed to the fact that the initialized H₂O at the beginning of the trajectory only changes in the CLaMS model if ice forms during liquid–NAT mixture formation, resulting in most of the CLaMS ΔH₂O values remaining at zero. However, there are a few instances in which changes in H₂O are observed in both CLaMS and MLS, potentially indicating ice-assisted formation of the liquid–NAT mixture. Furthermore, for simplicity, other chemical reactions (e.g., N₂O₅ hydrolysis and HCl null cycles) that involve H₂O are not considered in the CLaMS box model run, as discussed in Sect. 3.4. Hence, even if such reactions occur and deplete/produce H₂O in a real atmosphere, the CLaMS box model run does not account for them. This simplified box model run likely contributes to the weak correlation between H₂O values from CLaMS and MLS observations. A similar reasoning could apply to HNO₃; however, as the HNO₃ depletion is primarily driven by NAT formation, the omission of other chemical reactions in the simplified box model does not significantly affect the correlation between HNO₃ values from CLaMS and MLS.

As discussed in case nos. 1 and 2, the presence of NC/STS preceding the liquid–NAT mixture along the trajectories does not exclude the involvement of ice in the formation of the liquid–NAT mixture. It is possible that ice could have formed before the observation of the liquid–NAT mixture if the temperature decreased well below Tice and reached a state of high NAT supersaturation. Therefore, the extent of the temperature decrease below Tice and the extent of the corresponding increase in SNAT along the trajectories, from the time of NC/STS observation to the formation of the liquid–NAT mixture, can serve as indicators of whether the liquid–NAT mixture formed through ice-assisted or ice-free nucleation. To investigate this, we created two sets of histograms. Figure 14a, b, and c show the histograms of minimum T–Tice and maximum SNAT along the trajectories for cases in which NC/STS preceded the liquid–NAT mixture. Figure 14d, e, and f show similar histograms for cases in which ice preceded the liquid–NAT mixture. In all cases, the minimum T–Tice and maximum SNAT were only considered from the time of NC/STS (for Fig. 14a–c) or ice (for Fig. 14d–f) observation until the liquid–NAT mixture formed.

For cases in which NC/STS preceded the liquid–NAT mixture, the histogram of minimum T–Tice (Fig. 14a) shows that most trajectories experienced T–Tice values between -0.5 and 1.5 K, with the highest peak between 0 and 0.5 K. Similarly, the histogram of maximum SNAT (Fig. 14b) shows that most trajectories had SNAT values ranging from 15 to 150, with a peak between 15 and 30. A cluster of trajectories (represented by yellow to orange grids) is visible in the joint histogram (Fig. 14c) within this range. With respect to the formation of ice along these trajectories, the temperature should have reached at least -1.5 K for the heterogeneous nucleation of ice, and a high SNAT of >500 (Luo et al., 2003) should have occurred for the nucleation of NAT on these ice particles. As neither condition is observed and, thus, conditions are not conducive to ice formation, it is highly likely that the liquid–NAT mixture formed via an ice-free nucleation process. Case no. 1, which was discussed earlier, belongs to this category. In addition, when the minimum T–Tice decreased below -1.5 K, the air parcels reached a high supersaturation, with SNAT values exceedingly reaching values as high as 500. Case no. 3, which was also discussed earlier, belongs to this category. Under these conditions, the formation of ice and the subsequent nucleation of NAT on this ice is likely, indicating that the liquid–NAT mixture may have formed via ice-assisted nucleation. This observation aligns with the previous findings by Luo et al. (2003) on NAT formation over ice under high-supersaturation conditions.

For cases in which ice preceded the liquid–NAT mixture along the trajectory, the histogram of minimum T–Tice (Fig. 14d) shows that most trajectories experienced T–Tice values between -2 and -1 K, with the highest peak between -2 and -1.5 K. Similarly, the histogram of maximum SNAT (Fig. 14e) shows that most trajectories had SNAT values ranging from 15 to 150; however, for this case, the highest peak occurs between 30–70, instead of 15–30 (as observed a previous case). A cluster of trajectories (represented by yellow to orange grids) is visible in the joint histogram (Fig. 14f) within this range. Case no. 2, which was discussed earlier, belongs to this category, as ice preceded the liquid–NAT mixture.

From these two sets of histograms, an apparent shift in minimum T–Tice is evident between cases in which NC/STS preceded the liquid–NAT mixture (peak between 0 and 0.5 K) and cases in which ice preceded the liquid–NAT mixture (peak between -2 and -1.5 K). In contrast, only a subtle shift is observed in maximum SNAT. Therefore, to determine whether ice was involved in the formation of the liquid–NAT mixture, we primarily rely on the minimum T–Tice reached by each trajectory, rather than the maximum SNAT.

Cases in which ice preceded liquid–NAT mixture were classified with high confidence as ice-assisted nucleation of liquid–NAT mixture. We specify that these cases were classified with “high confidence” because we have CALIPSO observations of the presence of ice along the trajectory, similar to case no. 2. Moreover, for cases in which NC/STS preceded the liquid–NAT mixture, if the temperature along the trajectories decreased to less than -1.5 K, we classified the case as ice-assisted nucleation of a liquid–NAT mixture, although with low confidence. These cases are classified with low confidence because ice formation is not directly observed by CALIPSO, but ice was possibly formed as the temperature decreased 1.5 K below Tice. For the cases in which NC/STS preceded the liquid–NAT mixture, if the temperature along the trajectories did not decrease to less than -1.5 K, we classified them with high confidence as the ice-free nucleation of liquid–NAT mixture. The final retrieved relative percentage of the liquid–NAT mixture formation pathway is shown in Fig. 14g. The figure shows that 79 % of the liquid–NAT mixture (corresponding to 2996 backward trajectories) is formed via an ice-free formation process with high confidence. Almost 12 % and 9 % of the liquid–NAT mixture (corresponding to 796 backward trajectories) formed via an ice-assisted formation process with high and low confidence, respectively.

In the case of the NAT-assisted ice formation pathway, evidence of a direct CALIPSO observation of liquid–NAT mixture along the backward-trajectory ice PSC is given in Fig. S13. For brevity, this case is not discussed in detail here. Instead, another example of NAT-assisted ice formation pathways in which “NC” precedes an ice PSC is elaborated upon in detail through case no. 4.

In this case, NC is observed before the observation of ice. At 18:00 UTC on 19 June 2020, CALIPSO detected ice (blue star symbol in Fig. 15a) at a latitude of -80.96°, longitude of 6.00°, and potential temperature of 451 K. Before this, CALIPSO had observed NC along this trajectory at t=-37.6 h on 18 June 2020. The CALIPSO PSC orbit curtain plots for these observations are shown in Fig. S14. During the observation of NC, the temperature was around 192.5 K, which is 1 K above TNAT (Fig. 15b), and MLS measured gas-phase HNO₃ and H₂O mixing ratios of 2 ppb and 3.7 ppm, respectively (Fig. 15c and d). As the temperature was above TNAT, there was no possibility of the formation of PSCs at this time. Approximately 14.6 h later (i.e., at time t=-13 h), as the temperature decreased below TNAT, a strong depletion in CLaMS HNO₃ was observed, while the CLaMS H₂O mixing ratio remained unchanged. Concurrently, NAT SAD increased to ∼ 2.7 µm² cm⁻³ (Fig. 15e). At t=-18 h, ice SAD started increasing gradually as the temperature reached Tice (Fig. 15f), and it reached a peak value of ∼ 20 µm² cm⁻³. This suggests the nucleation of ice on the NAT particles.

In this section, we discuss case no. 5, in which the NAT-free ice nucleation process leads to ice formation.

In this case, NC is observed before the observation of ice along its backward trajectory. At 18:00 UTC on 19 June 2020, CALIPSO detected ice (blue star symbol in Fig. 16a) at a latitude of -80.41°, longitude of 359.56°, and potential temperature of 497 K. Before this, CALIPSO had observed NC along this trajectory at t=-37.6 h on 18 June 2020. The CALIPSO PSC orbit curtain plots for these observations are shown in Fig. S15. There are notable similarities between case no. 4 and case no. 5. In both cases, the latitude, longitude, and temperature histories of the observed ice PSCs and the preceding NC observations along the backward trajectory are almost identical. However, the ice PSC is located at a potential temperature of 497 K in case no. 5 (Fig. 17a), whereas the PSC in case no. 4 is at a relatively lower potential temperature of 451 K (Fig. 16a). Despite having similar temperature histories, ice formed via nucleation on NAT in case no. 4, whereas ice formed without NAT in case no. 5. A key difference between the two cases is the availability of gas-phase HNO₃. For case no. 5, the MLS HNO₃ mixing ratio at the CALIPSO NC observation time is 0.6 ppb (denitrified scenario), which is 1.4 ppb lower than that of case no. 4. Hence, compared to case no. 4, relatively low MLS HNO₃ is observed for case no. 5 when the temperature was above TNAT. The CLaMS box model for case no. 5 shows no NAT formation along the trajectory, with the NAT SAD remaining at 0 µm² cm⁻³ (Fig. 16e), whereas the NAT SAD increases to 2.7 µm² cm⁻³ for case no. 4 (Fig. 16e). This is because there is not enough gas-phase HNO₃ (with just 0.6 ppb) to form NAT in case no. 5, as it might have already been removed due to the growth of NAT to a large size, leading to denitrification before the observation of NC itself. In both cases, ice forms as the temperature drops below Tice, with the ice SAD reaching ∼ 23 µm² cm⁻³ in case no. 5, which is 3 µm² cm⁻³ higher than in case no. 4 (Fig. 16f). Furthermore, during ice formation, the maximum Sice is ∼ 1.5 in case no. 4, compared to a high-supersaturation Sice of 1.8 in case no. 5 (Fig. 16b). As ice can form through a homogeneous nucleation process at Sice>1.5, the temperature needs to be 3 K below Tice. However, for this case, the temperature is just 2 K below Tice, suggesting the heterogeneous nucleation of ice without NAT. One possibility could be that STS with the inclusion of foreign nuclei could have resulted in the formation of ice in this case. Another example in which STS precedes an ice PSC but in which liquid–NAT mixture was not formed, as modeled by CLaMS, is given in Fig. S16. As the bushfire aerosols entered the lower stratosphere in the high-latitude region, as shown in Figs. 1 and 2, it is highly likely that the STS contained these aerosols along with other foreign nuclei. In short, under the denitrified scenario, in which there is the insufficient availability of gas-phase HNO₃, ice can form via a NAT-free nucleation process. Based on PSC measurement using lidar aboard HALO, Voigt et al. (2018) observed ice PSCs with a bimodal distribution of the particle depolarization ratio over the Arctic region on 22 January 2016. Lagrangian backward-trajectory analysis revealed that the ice with a high depolarization ratio is preceded by NAT, based on CALIPSO measurements. Furthermore, ice with a low depolarization ratio is preceded by STS. Hence, to investigate ice formation via NAT-free and NAT-assisted nucleation process cases, we analyzed the depolarization ratio of ice, as discussed in the next section.

The depolarization ratios of ice preceded by a liquid–NAT mixture and STS along the backward trajectories are analyzed, with the corresponding probability densities shown in Fig. 17. As discussed through the case studies, after the CALIPSO observation of STS, it is possible that a liquid–NAT mixture formed as the temperature decreased and that these NAT particles acted as nuclei for ice formation. Therefore, the probability density of the depolarization ratio for the STS case may reflect not only the formation of ice on STS but also ice nucleated on NAT. The probability densities of the depolarization ratio for ice preceded by a liquid–NAT mixture (solid red line in Fig. 17a) peak at 0.35. In contrast, the probability density for ice preceded by STS (solid black line in Fig. 17a) shows a bimodal distribution, with a high depolarization mode peaking at 0.26 and a low depolarization mode peaking at 0.07. This observation of a bimodal distribution of the depolarization ratio is similar to the findings reported by Voigt et al. (2018). This bimodal distribution suggests that, under certain conditions, ice nucleates on NAT/STS, contributing to the mixed depolarization signal. Using the example in Fig. S16 and case no. 5, it is shown that, under denitrified conditions, the ice can form via a NAT-free formation pathway.

Therefore, we chose ice PSCs that are preceded by STS under the denitrified condition of MLS HNO3<0.5 ppb (i.e., during the observation of STS along the backward trajectory of ice, the MLS HNO₃ should be <0.5 ppb) and under the normal condition of MLS HNO3>3.0 ppb. Figure 17b shows the depolarization ratio of ice preceded by STS, distinguishing between cases under a denitrified scenario (MLS HNO3<0.5 ppb, dashed black line) and under normal conditions (MLS HNO3>3 ppb, dotted black line). These threshold values of 0.5 and 3 ppb are chosen because they effectively separate the bimodal distribution into two clearly distinguishable modes. The lower threshold (0.5 ppb) captures conditions under which denitrification is significant and, thus, NAT is less likely to form, leading to ice nucleation directly on STS with foreign nuclei inclusion. On the other hand, the higher threshold (3 ppb) represents conditions under which sufficient HNO₃ is available for NAT formation, supporting ice nucleation via NAT-assisted processes. These thresholds help differentiate the distinct nucleation pathways reflected in the depolarization ratio distributions. The low-HNO₃ case shows a more pronounced peak at lower depolarization values (∼ 0.076), whereas the high-HNO₃ case follows a distribution more similar to the overall STS case (with a higher depolarization peak near 0.26). This suggests the potential influence of the availability of gas-phase HNO₃ on the ice nucleation processes. Under normal gas-phase HNO₃ conditions with respect to availability, ice tends to form via NAT-assisted nucleation process, whereas under the denitrified condition, in which there is not enough HNO₃ to form NAT, ice can form via NAT-free nucleation process (i.e., nucleation on STS with foreign nuclei inclusion).

A strong correlation is found between the MLS-observed change in HNO₃ (ΔHNO₃) and the CLaMS-modeled change in HNO₃, with an R of 0.94 and an R2 of 0.8836 (Fig. 18a). Similar to HNO₃ uptake during the formation of liquid–NAT mixture from NC, significant HNO₃ uptake of up to 12 ppb, as observed by MLS, occurred when NC preceded the observation of ice along the trajectory (indicated by the white circle in Fig. 18a). In contrast, a relatively low HNO₃ uptake was noted when STS (liquid–NAT mixture) preceded ice formation, as marked using the green circle (yellow diamond) in Fig. 18a. In the case of H₂O uptake, a good correlation was found between MLS ΔH₂O and CLaMS ΔH₂O, with an R of 0.76 and an R2 of 0.5776 (Fig. 18b).

Based on the Lagrangian backward-trajectory analysis and the availability of gas-phase HNO₃ as observed through MLS, we estimated the relative percentage contributions of NAT-assisted and NAT-free ice formation processes, with the color-coded bars in Fig. 19 indicating different confidence levels based on observational evidence. High confidence (blue) in NAT-assisted formation indicates that direct CALIPSO observational evidence of liquid–NAT mixture was found along the backward trajectory of the ice PSC. This provides strong evidence that the ice formed via NAT-assisted nucleation. Ice PSCs that formed via the NAT-assisted nucleation process are classified with medium confidence (orange) when there is no direct observation of NAT but there is significant availability of gas-phase MLS HNO₃ >3 ppb during the observation of NC/STS along the backward trajectory. In this case, even without the detection of NAT, the availability of HNO₃ strongly suggests that NAT could have formed, justifying a medium-confidence classification for NAT-assisted nucleation. Low confidence (yellow) in NAT-assisted formation corresponds to cases in which the MLS HNO₃ mixing ratio during NC/STS observations was between 0.5 and 3 ppb. While NAT formation cannot be ruled out, the lower HNO₃ mixing ratio suggests only a low likelihood of NAT involvement. Hence these types of cases were classified as NAT-assisted nucleation, although with low confidence. These findings on ice formation pathways are result of an analysis carried using total of 5049 backward trajectories of ice PSCs; of these, 4796 backward trajectories correspond to ice formed via the NAT-assisted formation pathway, whereas only 253 trajectories correspond to the NAT-free formation pathway.

Ice PSCs that formed under extremely denitrified conditions are classified as NAT-free ice formation with medium confidence. Here, the backward trajectories reveal very low MLS HNO₃ (<0.5 ppb), creating conditions under which NAT formation is unlikely; thus, ice must have nucleated without the presence of NAT. In this scenario, ice PSCs were likely formed on STS particles with foreign nuclei inclusion. This breakdown helps to clarify the processes of ice formation under varying atmospheric conditions and enhances our understanding of the role that NAT plays in polar stratospheric cloud nucleation. The clear separation into different confidence levels also emphasizes the influence of gas-phase HNO₃ availability on determining the nucleation pathway.

In summary, we report the strong enhancement of aerosol loading in the lower stratosphere caused by the Black Summer event during early 2020. In addition, we observed increased gas-phase MLS HNO₃ and H₂O mixing ratios at an altitude of ∼ 25 and ∼ 17 km, respectively. The increased HNO₃ is due to an enhanced heterogeneous reaction known as N₂O₅ hydrolysis, while increased H₂O is directly injected by smoke plumes from the bushfire. Subsequently, during Austral winter in 2020, an anomalously high PSC areal coverage was observed, suggesting that the modified stratospheric composition of trace gases strongly influenced PSC formation. Despite the relatively high availability of gas-phase HNO₃ during this period compared to previous normal years, the liquid–NAT mixture PSC areal coverage remained within 1 standard deviation with respect to the background mean (2009–2019), whereas the ice PSC areal coverage exceeded 3 standard deviations. To explain this behavior, we calculated the 48 h backward trajectories of ice and liquid–NAT mixture PSCs using the CLaMS trajectory module with hourly ERA5 operational analysis meteorological data. The temperature history, uptake of MLS-observed HNO₃ and H₂O mixing ratios during the formation of these ice/liquid–NAT mixture PSCs, and types of PSCs observed before the observation of these PSCs along the trajectories are studied to understand their formation pathways. Through this analysis, we report with high confidence that 79 % of the liquid–NAT mixture formed via the ice-free nucleation pathway, as an ice PSC was not observed along the trajectory and the temperature of the air parcel never decreased 1.5 K below Tice (i.e., the minimum temperature required for ice formation) from the time of the “No Cloud (NC)”/STS observation until the time of the liquid–NAT mixture observation along the backward trajectory. This confirms no formation of ice before the observation of the liquid–NAT mixture. Moreover, we report with high confidence that 12 % of the liquid–NAT mixture formed via the ice-assisted nucleation pathway, as an ice PSC was observed by CALIPSO before the formation of liquid–NAT mixture along the trajectory; we further report with low confidence that 9 % of the liquid–NAT mixture formed via the ice-assisted nucleation pathway, as there was no direct CALIPSO observation of an ice PSC before the liquid–NAT mixture but the corresponding temperature decreased 1.5 K below Tice, indicating the possible involvement of ice during this liquid–NAT mixture formation. The inability of the ERA5 operational analysis to capture the small-scale temperature fluctuations caused by gravity waves and the possible bias in ERA5 temperature may contribute uncertainty to these percentage contributions of formation pathways.

In the case of ice formation, we observed that the depolarization ratio at 532 nm of ice PSCs preceded by STS along the trajectories exhibited a bimodal distribution. The low depolarization ratio mode peaked at 0.07, and the high depolarization ratio mode peaked at 0.26. In contrast, the ice PSC preceded by liquid–NAT mixture along the trajectory exhibited a single mode of the depolarization ratio with a peak occurring at 0.35, which is closer to the high depolarization ratio mode of ice PSC preceded by STS. This hints a the fact that ice with a high depolarization ratio could be the result of the nucleation of ice on NAT particles, whereas ice with a low depolarization ratio could be the result of the nucleation of ice on STS, which is supported by a previous study from Voigt et al. (2018). To check this, we analyzed the availability of gas-phase HNO₃ at the time of the observation of STS, which was observed before ice formation along the backward trajectories, as HNO₃ availability is crucial for NAT formation. The results indicate that ice PSCs with low depolarization ratios formed under denitrified conditions, when the MLS HNO₃ mixing ratio was below 0.5 ppb during the STS observation along the backward trajectories. Under these conditions, NAT particles could not form due to low HNO₃ availability, which is also confirmed by the CLaMS box model run, and this resulted in ice nucleating directly on STS. In contrast, ice PSCs with high depolarization ratios formed under normal conditions, when the MLS HNO₃ mixing ratio exceeded 3 ppb during the STS observation. This sufficient availability of gas-phase HNO₃ led to the formation of NAT particles (again, this is also confirmed by the CLaMS box model run) and to ice nucleating on NAT. Hence, based on the availability of gas-phase MLS HNO₃ and the temperature history, we conclude the ice formation pathways and relative percentage contribution of formation pathways and degree of confidence.

Hence, we report with high confidence that 14 % of the ice is formed through NAT-assisted ice formation, as direct CALIPSO observations of liquid–NAT mixtures are present along the trajectory; we report with medium confidence that 38 % of the ice is formed through NAT-assisted ice formation, as there is significant gas-phase MLS HNO₃ (>3 ppb), suggesting possible NAT formation before the ice formation; and we report with low confidence that 43 % of the ice is formed through NAT-assisted ice formation when HNO₃ levels are between 0.5 and 3 ppb, indicating a lower likelihood of NAT formation. Furthermore, we report with medium confidence that only 5 % of ice PSCs formed via NAT-free ice formation, as they formed under extremely denitrified conditions (HNO3<0.5 ppb), suggesting less likelihood of NAT formation and, thus, of ice formed via nucleation on STS. This suggests that the availability of gas-phase HNO₃ is a crucial factor in deciding whether ice forms via a NAT-assisted or NAT-free nucleation process. To the best of our knowledge, this is the first study that has provided the relative contributions of various liquid–NAT mixtures and ice formation pathways. These results provide deeper insights into the microphysical processes through which aerosols from extreme bushfire events (like Black Summer) can influence stratospheric chemistry and PSC formation.