Upper-air radiosonde data were obtained from the University of Wyoming dataset . The radiosonde data encompassed the balloon's precise trajectory, detailing latitude, longitude, altitude, and time. This information is crucial, as radiosondes can drift significantly from their launch sites during ascent. Balloons may drift approximately 5 km in the mid-troposphere, around 20 km in the upper troposphere . Knowing the balloon's trajectory enables matching its observations to the nearest model grid point in both space and time. For example, a balloon launched at 12:00 UTC requires several minutes to reach cruising altitude, so temporal alignment is also necessary. The drift distance was therefore calculated between the launch site and the 300 hPa pressure level (approximately jet cruising altitude) for multiple stations at 12:00 UTC on 25 November 2023 (Table ).

Assuming a balloon ascent rate of 5 ms-1, the time to reach the 300 hPa pressure level is approximately 30 min if the balloon drift is ignored. However, considering potential variations in ascent rates and balloon drift, the actual time can range between 40 and 110 min. For modeling purposes, we assume an ascent time of approximately 50 min to align the model data with the balloon's arrival at the 300 hPa level.

The Advanced Baseline Imager (ABI) aboard the GOES-R series satellites is a passive imaging radiometer featuring 16 spectral bands, including 10 in the infrared spectrum. The spatial resolution for these infrared bands is 2 km . In its operational modes, the ABI provides full-disk imagery every 10 min, images of the contiguous US every 5 min, and two mesoscale images every 60 s (or one every 30 s). Among its capabilities, the ABI utilizes a “Dust” RGB (Red-Green-Blue) composite to detect and monitor airborne dust. This product is also especially useful to detect contrails. This composite combines data from infrared channels 8.4 µm (Band 11), 10.3 µm (Band 13) and 12.3 µm (Band 15). The GOES-16 data was downloaded from and the “Dust” RBG was generated using the software package from .

Data Source: flight data was obtained from Flightradar24 .

The Ceilometer, CHM 15k Cloud Height Meter, is an advanced light detection and ranging (LIDAR-based) remote sensing instrument deployed at Toronto Pearson International Airport (CYYZ) to measure cloud height, penetration depth, and vertical visibility. Operating at a 15 s temporal resolution, it employs the Ne-YAG laser (λ=1064 nm) to emit short laser pulses into the atmosphere. These pulses scatter upon interaction with aerosols and cloud particles, with the backscattered signal being analyzed to determine cloud structure and visibility conditions.

The CHM 15k is capable of measuring up to 15 km in altitude with a range resolution of 5–15 m, depending on the measurement mode. Its full waveform analysis allows for the identification of multiple cloud layers (up to 9), 3 layers in its current configuration, and provides high-resolution backscatter profiles. It uses photon counting technology, enhancing detection sensitivity and minimizing background noise. The system records parameters such as cloud base height, penetration depth, maximum detectable range, vertical visual range, and sky condition, making it crucial for aviation meteorology and atmospheric research.

The Global Environmental Multiscale (GEM) model is a versatile atmospheric modeling system widely used for high-resolution simulations of atmospheric processes . GEM serves as the operational NWP model for all atmospheric prediction systems at Environment and Climate Change Canada. GEM has a non-hydrostatic, fully compressible dynamical core using semi-Lagrangian advection and a terrain-following hybrid vertical coordinate system . This configuration is suitable for a wide range of resolutions, from mesoscale to cloud-resolving scales. For this study, we use a horizontal grid spacings of 1 km × 1 km over a 1600 × 1000 limited-area domain centered at 45.2° N and 79.9° W (eastern Canada, Fig. a). Our simulations use lateral and surface boundary conditions from the operational 2.5 km High Resolution Deterministic Prediction System HRDPS;, updated hourly. We use 121 vertical levels to enhance the vertical resolution in the upper troposphere with grid spacings of ∼230 m at 300 hPa for a more accurate representation of ice-supersaturated regions. A 30 s model timestep was used, with the model initialized at 06:00 UTC and running for a duration of 13 h.

The integration of the Predicted Particle Properties (P3) bulk microphysics scheme into GEM represents a significant advancement in the modeling of ice-phase hydrometeors and mixed-phase cloud dynamics. This methodology outlines the configuration and application of the GEM-P3 setup as described in several studies .

The P3 microphysics scheme uses a property-based treatment of the ice phase i.e., in contrast with the traditional approach of predefined ice-phase categories (i.e. ice, snow, graupel and hail), whereby all ice-phase or mixed-phase particles are represented by one or more generic or “free” categories whose bulk physical properties, such as mass, number, density, and rime fraction, evolve freely and continuously. This enables each ice category to represent a wide range of dominant types of ice and removes the need for artificial “conversion” between categories. With the use of multiple free ice-phase categories, multiple modes, i.e., populations of ice with different bulk properties, can exist in the same point in time and space. A complete description of the original P3 scheme can be found in and ; descriptions of subsequent major developments can be found in , , and . In this study, the two-moment ice configuration, with prognostic liquid fraction off, and with 3 ice categories is used.

Hydrometeor size distributions, including those of the liquid categories (“cloud” and “rain”), are represented by complete 3-parameter gamma functions, with shape and slope parameters evolving dynamically based on the prognostic variables. The configuration uses the two-moment treatment of ice particles instead of the triple-moment approach , the latter being more important for deep convective systems to resolve size sorting better and improve the simulation of hail and heavily rimed particles. With triple-moment ice, the spectral dispersion changes due to deposition/sublimation; this could be a topic of future work for contrails.

To examine the impact of the deposition coefficient (αD, from Eq. ) on the vapor depositional mass growth of ice crystals (Eq. ) and its subsequent effect on contrail persistence, we conducted sensitivity simulations. These simulations introduce the depositional mass growth ratio (dgr), which was implemented into P3. 1dmdtαD=dmdtαD=1×dgr(T,p,RHi,r)

The dgr is computed offline as the ratio of depositional growth rates for αD=1 and for αD varying with T, p, humidity, and ice particle size (r). This dgr, being dependent on ambient conditions and ice particle size, serves as a multiplicative factor in the P3 scheme (Eq. ). The deposition adjustment was implemented only in the upper troposphere for T<-38 °C consistent with laboratory and modeling studies . The use of αD used here in P3 is not included in the CoAT framework; hence, ice crystal formation and loss during the vortex phase are independent of αD. Further information on this approach can be found in Appendix A.

The formation of contrails, governed by the SAC, requires specific thermodynamic conditions that depend on both atmospheric and aircraft parameters. A critical parameter in this framework is the slope of the mixing line, G, which characterizes the relationship between the parameters of the exhaust plume and the ambient atmosphere. According to , G is expressed as: 2G=cppEIH2O(MH2O/Mair)Q(1-η) where cp is the specific heat capacity of air at constant pressure (1004 Jkg-1K-1), p is the ambient pressure (Pa), EI_H2O is the water emission index (1.25 kg of H2O per kg of fuel burned), MH2O/Mair is the molar mass ratio of water to air (approximately 0.622), Q is the specific combustion heat of the fuel (43.2 MJkg-1 for kerosene), and η is the propulsion efficiency (0.29).

For contrails to form, the ambient temperature T must be below the maximum threshold temperature, defined as the temperature at which the mixing line intersects the liquid water saturation curve . This ensures that the relative humidity over water exceeds the critical threshold required for condensation of exhaust water vapor. Under these conditions, water vapor condenses onto soot and ambient aerosol particles to form liquid droplets, which then undergo near-instantaneous freezing into ice crystals that subsequently grow by vapor deposition. If ice supersaturation persists (i.e., RHi≥100 %), contrails can remain and evolve into long-lived cirrus. Although contrails form in ice-supersaturated conditions, demonstrated that contrails may persist in ice-subsaturated conditions. However, to remain consistent with the parameterization, which assumes both formation and growth occur within ice-supersaturated environments, we restrict simulated persistence to regions where ice supersaturation is maintained.

The wake vortex phase of contrail evolution involves distinct processes in the primary and secondary wakes, critical for understanding contrail dynamics and their transition into cirrus clouds. The primary wake, associated with the counter-rotating vortex pair generated by the aircraft's lift, experiences strong downward motion, leading to adiabatic heating and partial sublimation of ice particles. These dynamics, which trap a significant portion of the exhaust within the vortex pair, are strongly influenced by environmental factors such as temperature and relative humidity . In contrast, the secondary wake forms a “curtain” of detrained exhaust between the original emission altitude and the descending vortex. Ice particles in the secondary wake grow through deposition in an ice-supersaturated environment and retain the majority of the contrail ice mass by the end of the vortex phase. This secondary wake, less affected by adiabatic heating, plays a crucial role in the persistence and spreading of contrails .

When the conditions for SAC and for contrail persistence (RHi≥100 %) are satisfied, GEM uses the wake vortex model from , which focuses on the interaction between ice microphysics and wake vortex dynamics. The parameterization provides a framework for quantifying key characteristics of young contrails during their early development phases. It calculates the maximum vertical displacement of wake vortices and the vertical extent of the contrail, which corresponds to the maximum vortex displacement if ice particles can survive the warming effects associated with the adiabatic descent of the vortices. Additionally, it estimates the survival fraction of contrail ice particles, accounting for their loss due to changes in relative humidity resulting from adiabatic warming and incorporating the influence of ice supersaturation, T, and the Kelvin effect on crystal growth . The wake vortex model uses the aircraft information from Table  and the atmospheric conditions (i.e. T, p, RHi and the static stability of air) from GEM to determine ice particle survival, the CINC and the vertical contrail extent of young contrails during the first 5 min. The wake vortex model enables parameterizations to incorporate additional aircraft characteristics, such as weight, speed, and fuel flow rate. However, for our purposes, we utilized only the parameterization based on wingspan. Given the aircraft wingspan bws and the ice crystal emission index EIiceno (equal to the soot emission index), the water vapor emission rate I0 is estimated by 3I0=0.02kgm-1bws80m2 from which the initial number of emitted ice crystals N0 can be derived as 4N0=I0EIH2OEIiceno.

After the vortex phase the contrail ice crystals which survived the vortex phase (CI_surv), is defined by 5CIsurv=N0×fNs where fNs is the fraction of ice crystals surviving the vortex phase. The I0 and CI_surv per flight meter are distributed over the flight segment within each GEM grid cell and converted to the volumetric prognostic variables during one time step as 6CINC=CIsurvfdVg,  and7CQI=I0fdVg, where CINC (m-3) and the contrail ice mass concentration (CQI, kgm-3) is of young contrails, fd is the flight distance within a grid box and Vg is the volume of the grid box. When assessing contrail persistence for individual the flights (Table ), not contrail-forming regions and their associated properties, the CINC and CQI remaining after the vortex phase are integrated into the cloud ice number and mass concentrations within the P3 microphysics scheme. Consequently, these contrails evolve through the same microphysical growth and loss processes as naturally occurring clouds. These combined quantities, representing the total ice number and mass concentrations, are subsequently advected by GEM’s dynamical core.

If the contrail's vertical extent exceeds that of the model grid, the contrail's ice content is distributed uniformly into the model grid box below the flight level (Appendix A2). This adjustment is necessary because the scheme was developed for a single, constant-RHi layer. In contrast, our model resolves contrail properties within discrete grid cells that may be shallower than the diagnosed contrail depth. When the simulated contrail depth exceeds a grid layer, the CINC can be confined within that grid box – leading to an overestimation of the CINC and an underestimation of the wind-shear effects – or redistribute it downward into the grid box below, where RHi may be lower than the value used to diagnose CINC. Both choices have limitations, but we also adopt the latter approach to include impacts from vertical wind shear, promoting vertical spreading.

found that, on average, CI_surv shows the strongest sensitivity to EI_iceno and RHi. Soot number emission indices, represented here by EI_iceno are often explored over several orders of magnitude to represent engine or fuel regimes. In contrail formation studies, soot emission indices (EI_soot) delineate microphysical regimes controlling the origin and number of ice crystals. The Very-High-Soot (VHS) and High-Soot (HS) regimes (EIsoot≥1015 kg-1) represent the soot-rich regime described by , where nearly all emitted soot particles act as freezing nuclei and ice forms predominantly by homogeneous freezing around soot cores. The Normal-Soot (NS) regime (EIsoot∼1014 kg-1) corresponds to the intermediate regime, in which ice formation transitions from soot-driven to mixed liquid–soot contributions, leading to a tenfold reduction in ice number concentration compared to the soot-rich case. Finally, the Low-Soot (LS) regime (EIsoot≲1013 kg-1) aligns with the soot-poor regime, where contrail ice forms primarily on entrained atmospheric particles or ultrafine plume particles rather than soot, yielding fewer but larger ice crystals and substantially lower contrail optical depth .

To quantify contrail persistence and its influence on cirrus cloud properties, we identify and track the regions associated with the persistent contrails over Toronto on 25 November 2023 (Fig. ). This region spans 310 to 290 hPa, where RHi≥100 %, and corresponds to the layer in which two flights (DL384 and TK6061) generated contrails. The DA-VHS simulation, which produce the longest-lived contrail, is used to define this region. We calculate the 99.5th percentile of the total ice column density (TICD) from the DA-VHS simulation (described in the next section), which provides the strongest and most coherent contrail signal, and use it to construct a binary mask M(x,y,t) defined as 8M(x,y,t)=1,ifTICDVHS(x,y,t)≥P99.5VHS,0,otherwise. where TICD is calculated as the vertically integrated total ice number concentration between 310 and 290 hPa within ice-supersaturated regions, and P is a percentile of TICD. This mask isolates the spatial extent of the simulated contrail and is applied uniformly to all sensitivity simulations to ensure spatial consistency. For each simulation i, the difference relative to the deposition-adjusted control simulation (CNTL-DA, described in the next section), in which no contrails are initialized, is computed over the masked region (〈⋅〉M) as 9ΔI‾i(t)=〈Ii(x,y,t)-ICNTL DA(x,y,t)‾〉M,I∈{TICD,IWP} where IWP is the ice water path.

This approach allows for a consistent assessment of contrail evolution across different emission regimes. Even for short-lived contrails, this approach enables analysis of how the initially contrail-affected region evolves, whether it transitions into contrail cirrus or dissipates completely.

Two control simulations were performed: CNTL and CNTL-DA. The setups are identical except that CNTL assumes a constant deposition coefficient (αD=1), whereas CNTL-DA applies a variable αD. Both use a part of the CoAT configuration, diagnosing contrail-forming regions and properties – such as contrail depth, ice particle survival (IPS), CINC, CQI – at each timestep as if an A321 or B747 aircraft were uniformly distributed across the model domain. FlightRadar24 tracks are excluded, so no explicit contrail ice is added to the P3 microphysics or advected by GEM (CoAT excluding flights). Both simulations use an emission index of 1×1015 kg-1, representing the HS regime. Sections 3.1.1 and 3.1.2 compare the ice-supersaturated regions between CNTL-DA and CNTL, while Sect. 3.1.3 examines the sensitivity of persistent contrail formation and related properties to αD.

Subsequent sensitivity experiments employ the full CoAT configuration, which includes FlightRadar24 tracks (Table ; Fig. a), to analyze persistent contrail occurrence under varying soot-emission regimes for flights overpassing Toronto near 14:00 UTC (Sects. 3.2.1–3.2.3). When persistent contrails form, the resulting contrail ice is integrated into the P3 microphysics scheme for evolution and advected by GEM's dynamical core. All sensitivity simulations use the DA configuration. The complete set of simulations and their configurations is summarized in Table .

The approach examines the relationship between ice particle growth rates and RHi by analyzing soundings and comparing them to the GEM simulations (CNTL and CNTL-DA). The simulations are designed to illustrate how variations in ice particle growth rates, through the variation in αD, influence RHi. By contrasting the sounding observations with GEM under different growth rate scenarios, we aim to elucidate the impact of ice particle depositional growth on atmospheric humidity profiles. To account for uncertainty in balloon drift, a 5 km × 5 km area surrounding balloon location while ascending was used to compile the vertical profile for each station in the model output, which is then compared to the soundings.

Figure  presents the RHi distribution for sounding observations compared to GEM at 12:00 UTC on 25 November 2023, considering pressure levels between 100 and 600 hPa and temperatures colder than -38 °C. As the RHi approaches 100 %, the CNTL simulation underestimates the probability of higher RHi values, where it fails to capture the frequency and intensity of ice supersaturation events. The probability density of the soundings peaks at an RHi of 102 % and decreases to 122 % beyond which sparse observations, primarily from the White Lake (DTX) sounding, are present, extending up to a maximum RHi of 148 %. The CNTL simulation follows a trend commonly observed in atmospheric models, where the RHi distribution peaks around 102 % before sharply declining due to rapid humidity quenching by ice growth schemes or models using saturation adjustment schemes . Unlike models employing a saturation adjustment, P3 does not impose such a constraint. The CNTL-DA simulation exhibits a greater buildup of ice supersaturation due to reduced depositional growth, producing a peak at RHi∼108 % and a tail extending toward 113 %. The largest contributor to the underestimation is GEM’s inability to capture the DTX sounding (Fig. ).

At the 500 hPa level a deep trough was present over the Great Lakes region, indicating an area of lower pressure and cooler temperatures aloft (Fig. ). This trough was associated with enhanced jet stream activity, which led to increased upper-level divergence. Such divergence promotes rising motion, generating the thin streaks of ice-supersaturated regions that favored the cirrus cloud formation and persistent contrail development observed in the GOES-16 imagery over the Buffalo (BUF) and DTX regions (Fig. c). These satellite-observed cirrus layers are indicative of the widespread ice-supersaturated environment also captured in the sounding profiles (Fig. a and b).

The DTX sounding profile reveals that the balloon drifted through one of these supersaturated streaks, recording RHi≥140 % between 350 and 250 hPa. While the CNTL-DA simulation accurately captured the broader upper-air trough, it failed to generate the highly ice-supersaturated regions above DTX (Fig. a). When the DTX station is excluded, CNTL-DA shows markedly improved agreement with observations (Fig. ), capturing the ice-supersaturated layers between 400–200 hPa observed at most other stations (i.e., BUF, Fig. b). The improved RHi representation may be because the other soundings did not contain such a large region with many thin layers of highly ice-supersaturated air, making them easier for the model to capture. The model’s relatively coarse vertical resolution and unresolved subgrid-scale horizontal supersaturation inhomogeneity may smooth out small-scale supersaturation features, limiting its ability to resolve narrow layers of high RHi. However, since RHi in the CNTL-DA simulation remains well below the observed mean of 117 % between 350 and 250 hPa the discrepancy likely involves factors beyond vertical resolution.

In the Great Lakes region, the interaction between cold Arctic air masses and the relatively warmer lake waters frequently triggers lake-effect snow events . The synoptic conditions on 25 November 2023 and the following days facilitated such phenomena, leading to the formation of low clouds and heavy snowfall in the surrounding areas. On this day, these low clouds significantly attenuated the ceilometer signal at CYYZ, Toronto, making it nearly impossible to retrieve data from cirrus clouds between 13:00 and 17:00 UTC (Fig. ). However, a descending high cloud base was still noticeable, aligning with the descending moist layer observed in the CNTL and CNTL-DA simulations (Fig. a and b, top panels). The CNTL and CNTL-DA simulations show three panels each of time-height vertical profiles of RHi and the corresponding CINC for A321 (middle panels) and B747 aircraft (bottom panels). Between 11:00 and 12:00 UTC, the ceilometer’s backscatter intensity displayed a diffused structure, indicative of thin ice clouds like cirrus or contrails ∼10 km in altitude (Fig. ). By 12:00 to 13:00 UTC, the cirrus clouds became more structured, with a lower cloud base near 8 km, suggesting the presence of ice-supersaturated regions.

The CNTL-DA simulation captures the ice-supersaturated layer, favoring persistent contrail formation, between 8 and 10 km, while the CNTL simulation shows no ice-supersaturation in this altitude range. Here, aircraft-specific differences become evident: the A321 forms contrails at RHi≥100 %, whereas the heavier B747 requires RHi≥102 %. In these marginally ice-supersaturated conditions, the B747’s initial number of emitted ice crystals sublimates within the descending vortex. To produce contrails under the same ambient conditions (T, p, RHi, atmospheric stability) observed between 12:00 and 12:30 UTC, the B747 would require ambient temperatures ∼2 °C lower.

From 13:00 UTC until around 14:45 UTC, neither of the simulations show conditions suitable for persistent contrail formation. During this time, contrails were observed over Toronto (Fig. b–d), but images at 14:20 UTC and simulations at 14:00 UTC show no indication of new persistent contrail formation, only widespread older contrails that had formed upstream towards the west.

After 14:45 UTC, the CNTL simulation remains mostly ice-subsaturated to only weakly ice-supersaturated (maximum RHi≈104 %), supporting only a shallow layer with a CINC of ∼0.4 cm-3 for an A321 aircraft. In contrast, the CNTL-DA simulation develops a pronounced ice-supersaturated region (RHi up to 112 %) conducive to persistent contrail formation near 10 km. Under these conditions, contrails from the A321 appear first at RHi≥100 %, followed by those from the B747 around 15:15 UTC as RHi rises to ≈102 %. The CNTL-DA simulation produced deeper contrail forming region with enhanced CINC up to 1.4 cm-3 for both aircraft types compared to the CNTL simulation.

Overall, the CNTL-DA simulation captures deeper and more persistent contrail regions, consistent with its stronger and deeper ice-supersaturated layers.

In the following analysis, we compare the CINC produced along the A321 (medium-weighted category) flight path with those from a hypothetical B747 (heavy-weighted category) operating under identical meteorological conditions. This comparison isolates the influence of aircraft-specific wake dynamics and fuel flow rates on contrail ice production across different EI categories. Figure  illustrates the simulated cross-sectional regions of persistent contrail formation, highlighting areas where ice particles survive the wake vortex along the A321 (DL384) flight route (Figs. a and a). The aircraft, operating along the Albany-Buffalo-Gaylord (ALB-BUF-APX) corridor, maintained an altitude of 300 hPa (30 000 ft) between 74 and 84° W before ascending to 250 hPa (34 000 ft) at 84° W, where it then continued cruising.

During its cruise at 300 hPa, the A321 traversed regions favorable for persistent contrail formation, particularly over Lake Ontario as it approached the location of the APX station. Observations from GOES-16 Dust imagery confirm the presence of multiple aircraft-generated contrails between 13:00 and 15:00 UTC (Figs. b and c). Near the location of the APX station, the A321 ascends out of the contrail-forming layer into a drier atmospheric region, where contrail formation ceases.

A comparison between the A321 DA-HS simulation (B747 DA-HS) and the A321 CNTL simulation (B747 CNTL), where the CNTL simulations use an EI=1×1015 kg-1, reveals pronounced differences in CINC and the extent of persistent contrail formation when the depositional mass growth of ice is adjusted by αD (Fig. ). The CINC is substantially reduced, and regions of persistent contrail formation are smaller, particularly in the B747 CNTL simulation. The contrast between two different aircraft in the CNTL simulation indicates that the B747 produces less persistent contrails than an A321 in similar meteorological conditions where 100%≤RHi≲102 %.

In the DA simulations, with enhanced RHi, the regions of persistent contrail formation become more consistent across aircraft and EI categories, and the main differences emerge in the CINC and IPS statistics. For example, in the DA-VHS simulation, the A321 produces a mean CINC of 1.22 cm-3 with a mean IPS of 7.4 %, whereas the B747 yields a higher mean CINC of 1.90 cm-3 but a lower IPS of 6.41 % (Table ).

With identical EIs, the mass fuel flow rate (mf) differs markedly between aircraft – A321: mf≈0.68 kgs-1 vs. B747: mf≈2.54 kgs-1. The B747’s stronger wake induces a deeper descent and adiabatic heating, enhancing sublimation and lowering IPS relative to the A321; consequently, the contrail depth of the young contrail is shallower for the B747. Despite this, CINC remains higher for the B747 because the higher mass fuel flow rate of the B747. That is, more particles are injected per meter of flight, offsetting early losses and maintaining elevated CINC concentrations. This scaling also explains the observed cross-EI relationships: in the DA-LS simulation, a B747 yields ICNC similar to an A321 in the DA-NS simulation, and similarly, in the DA-NS simulation the CINC produced by a B747 approaches the CINC produced by an A321C in the DA-HS simulation. Overall, a B747 using lean-burn combustor producing lower soot (EI=6.44×1013, DA-LS) can produce as many CINC as an A321 operating with conventional JET-A fuel (EI=2.8×1014, DA-NS).

In this section, we incorporate the flight information into GEM-CoAT using the aircraft characteristics listed in Table . We simulate persistent contrail formation based on the variation of aircraft properties and flight routes from FlightRadar24.

From the GOES-16 observations and photographs (Fig. ), it is evident that multiple aircraft-generated contrails were present over the Toronto area at 14:20 UTC. Contrail formation began around 12:00 UTC, increasing in extent and number until 16:00 UTC. Here, we analyze the simulated contrails formed by two specific aircraft: TK6061 (B747, flying northeastward) and DL384 (A321, flying westward). At 14:20 UTC the contrail ages of the B747 and A321 are 35 and 105 min, respectively (Fig. ). The CINC produced by these aircraft is combined with the ice number concentration generated by the ice nucleation parameterizations in P3 between 310 and 290 hPa and where the RHi≥100 %, to get the TICD. The simulated contrails are then compared with GOES-16 Dust observations at 14:20 UTC.

Figure a highlights the contrasts between soot emission regimes across the different aircraft. The DA-LS simulation captures persistent contrail formation, particularly for the B747 and A321 aircraft. The contrail generated by the B747, with TICD≈240 cm-2, is clearly distinguishable from the surrounding cirrus, whereas the contrail from the A321, with TICD≈90 cm-2, is only faintly visible within the cirrus at 14:20 UTC. These flights operated at flight levels FL310 and FL300, respectively (Table ). As soot emissions increase, the TICD increases correspondingly, indicating greater persistence of the contrails (Fig. a–d). For example, for the A321, the TICD in the DA-VHS simulation is about 2.8 times higher compared to the DA-LS simulations, whereas the soot EI differs by nearly a factor of 19 for the same aircraft. Similarly, for the younger contrail from the B747, the TICD is about 3.2 times higher. This shows that although the soot EI directly influences the initial CINC, the resulting surviving ice crystals do not scale linearly with it (also true for a 5 min old contrail), because a significant fraction of particles are lost during the wake vortex descent. Consequently, the wake dynamics strongly modulate how much of the emitted soot ultimately forms surviving ice crystals. Therefore, in modeling approaches that prescribe contrail radiative properties based solely on the soot-derived ice number concentration, without accounting for wake-vortex losses, the simulated contrails may contain excessively small ice crystals, leading to biased radiative calculations.

In general, the locations of our simulated contrails from these aircraft align well with the contrails detected by GOES-16 Dust observations, a result that is only achieved in the DA simulations.

The temporal evolution of contrail properties reveals clear differences in how soot emissions influence both the persistence of simulated contrails. Following the entry of flights DL384 and TK6061 into the model domain (black box in Fig. ) at 12:30 and 13:40 UTC, respectively, all sensitivity simulations show a rapid increase in TICD (Fig. b). Although the TICD is significantly larger than in the CNTL-DA simulation, the corresponding change in IWP remains insignificant compared to that of the pre-existing cirrus before 14:00 UTC (Fig. a). Nevertheless, the enhanced TICD suggests that a greater number of contrail ice particles formed within the cirrus layer, which would amplify its radiative impact beyond that of the background cirrus. The magnitude and duration of these enhancements vary substantially across soot regimes. After 14:00 UTC, both the IWP and TICD become significantly larger for the DA-VHS and DA-HS simulations compared to the CNTL-DA simulation. The DA-VHS simulation shows the strongest and most sustained increases in both TICD and IWP, reflecting high contrail ice crystal formation and slower dissipation. In contrast, the DA-LS simulation exhibits a much weaker response: its contrail dissipates progressively after 14:00 UTC and becomes indistinguishable from the background variability of the CNTL-DA simulation by approximately 16:45 UTC. This behavior highlights the sensitivity of contrail lifetime to soot availability, with reduced emissions leading to shorter-lived, less persistent contrails.