Extreme Ice Crystal Events Linked to Biomass and Fossil Fuel Combustion

Eight years of upper tropospheric (UT) ice crystal measurements with the Backscatter Cloud Probe (BCP), operated as part of the In-Service Aircraft for a Global Observing System (IAGOS), have been analyzed to assess the 15 frequency and characteristics of extreme ice crystal events (EIE), defined in this study as encounters with clouds that have number concentrations exceeding 5000 L-1. A total of 6490 events, in clouds of horizontal extent > 2.5 km, were identified during the period from December 2011 to March 2020 in the global band between 30°S and 30°N. Evaluation of in situ measurements of carbon monoxide in these UT clouds, combined with back-trajectories and carbon monoxide emission inventories, allow the identification of regional anthropogenic sources. An evaluation of 20 lowand upper-level kinematic variables from the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2), combined with spatial distribution of aerosol optical depth and regions of biomass burning, reveal the physical mechanism by which the particles are lofted to flight levels in regions of deep convection. The maps of lightning frequency derived from the World Wide Lightning Location Network (WWLLN) contribute additional evidence of the role of deep convection lofting ice forming aerosols to the UT. The MERRA-2 analysis 25 shows clear spatial correlations that link dust, black carbon (BC), organic carbon (OC) and sulfate particles with regions of EIE. Given the composition of the source aerosols and the nature of their transport to the UT, the ice clouds are likely of liquid origin whereby droplets formed on these particles are lofted and freeze heterogeneously as have been reported in previous studies. The results from this study have important ramifications related to climate change, satellite measurement validation, weather forecasting and aircraft operations. 30

In addition to generating statistics on the number concentration (Nice), ice water content (IWC) and mass mean radius (Rice) as a function of ambient temperature, the in situ measurements were compared to 10 years of cirrus ice number 40 concentrations derived from satellite measurements. This extensive data set serves to underscore the importance of aircraft measurements, not only to document cirrus properties, but also to provide the detailed information about those characteristics that cannot be extracted from the lower spatial resolution satellite data. For example, in the comparisons discussed by Krämer et al (2020), the Nice derived from satellites was found to be, on average, a factor of two higher than in situ measurements, a significant difference that had also been noted by Gasparini et al. (2018). Furthermore, 45 in situ measurements have provided the information needed to refine parametrizations and improve the representation of cirrus clouds and their microphysical and optical properties more accurately in climate models (Kärcher, 2018;Kärcher et al., 2010).

50
The European Research Infrastructure program IAGOS (In-service Aircraft for a Global Observing System) was developed with the primary objective of building a database of global measurements, over an extended period of time, as a source of information that would be truly representative across all temporal and spatial scales (Petzold et al., 2015). Although the instrumentation on the IAGOS aircraft fleet is not nearly as comprehensive as that found on the typical cloud research aircraft (Petzold et al., 2015), Beswick et al. (2015) showed that the cirrus cloud microphysical 55 statistics, acquired with the cloud probes installed on the IAGOS aircraft, were in good agreement with those that were summarized in published studies compiled from aircraft-based research campaigns. This leads us to conclude that data taken during the IAGOS flights should be considered as an integral and complementary component of the global data set of cirrus measurements, as summarized by Krämer et al (2016Krämer et al ( , 2020.

60
The current study expands the analysis of the 3-year dataset presented by Beswick et al. (2015) to an evaluation through March 2020 and complements the results recently reported by Lloyd et al. (2021), significantly increasing the number of clouds sampled. Beswick et al. (2015) identified frequent events in which Nice exceeded 10000 L -1 , a concentration that is more than 300 times larger than the median values and 30 times larger than the 90 th percentile reported by Krämer et al. (2020) for Nice measured in situ and derived from CALIPSO and CloudSat measurements Icing (ICI), a definition used by the community of scientists who study how ice crystals impact aircraft performance.
ICI is usually associated with ice water content and not with ice number concentration, although the two variables are obviously linked. The present study focuses on understanding the processes that result in high Nice and on the identification of the sources of the precursor aerosols that form these crystals.

80
Ice crystals form on aerosol particles via a number of processes (Kanji et al., 2017). The two most frequent pathways are through immersion or contact freezing: 1) water droplets form on cloud condensation nuclei (CCN) and, when advected to regions of low temperatures, freeze due to the presence of ice nucleating particles (INP) embedded in the water droplets or 2) by surface contact with INP (Krämer et al., 2016).
85 Beswick et al. (2015) suggested that the high ice crystal concentrations detected in the maritime-continental, tropical region over Malaysia were possibly linked to biomass burning. Observational and theoretical studies have shown how gaseous and particulate emissions from not only biomass burning (BB), but also from urban pollution (UP), can reach aircraft cruising altitude where ice crystal clouds are found (Osman et al., 2016;Ditas et al., 2018;Hirsch and Koren, 2021). Some fraction of particles emitted from biomass and fossil fuel burning will act as CCN or INP especially as 90 they age while lofted to the UT (Brooks et al. 2014;Gorbunov et al. 2001;Grawe et al. 2016;Popovicheva et al. 2008;Umo et al. 2015;Knopf et al. 2018;Jahl et al., 2021). Ding et al. (2015) used spaceborne Measurements of Pollution in the Troposphere (MOPITT) data, in combination with aircraft measurements from the Measurement of Ozone and Water Vapor by Airbus In-Service Aircraft 95 (MOZAIC) program to document the presence of anomalously high carbon monoxide (CO) in the upper troposphere over East Asia. They used back-trajectory analysis to link anthropogenic sources over the Sichuan Basin and the North China Plain and from forest fires over Indochina. The CO was transported to such altitudes through strong frontal lifting, embedded deep convection and by orographic forcing.
The importance of the Asian Summer Monsoon (ASM) for lofting anthropogenic pollution and forming the Asian Tropopause Aerosol Layer (ATAL) is discussed by Vernier et al. (2015). Aircraft measurements identified particles composed of carbonaceous and sulfate materials and CALIPSO observations indicated that deep convection associated 115 with the ASM transports surface pollution particles to the ATAL. Figure 1 shows the locations of EIE observed worldwide between December 2011 and March 2020 along with megacities, i.e., those urban areas with populations larger than 10 million. These high ice concentration events are encountered over almost all the continents where the IAGOS aircraft fly; however, the majority of the EIE are found 120 between ±30° latitude where the larger number of megacities are located. The EIE are clustered along the flight routes from Europe to South America and to Africa, but the majority of events are found in the large cluster centered around Taiwan and radiating towards India, Japan, Indonesia and Australia/Oceania. The color-coded markers indicate the ice crystal number concentration within each EIE and, while the majority of concentrations range between 5000 and 20000 L -1 (purple markers), there are a number of regions where concentrations exceeding 50000 L -1 were detected, 125 as indicated by green, yellow and red circles.
This study will show that high ice concentrations along commercial flight tracks are frequently found along with evidence from BB and UP emissions. The case for linking surface, anthropogenic emissions with the EIE is made in the following sections. An analysis of in situ cloud and CO measurements, coupled with back-trajectory modeling of light. Some fraction of the backscattered light is collected through the same window and the collected photons are converted to an electrical signal that is processed and used to derive an equivalent optical diameter (EOD). The nominal size range is 5 -90 µm with an estimated uncertainty of ±30% when deriving the EOD of ice crystals.

150
Detected crystals are binned by size into a 10-channel frequency histogram that is transmitted to the data acquisition system every four seconds. The number concentration is calculated as the number of particles detected divided by the volume of sampled air. The BCP samples 250 cm 3 every four seconds, a spatial resolution of approximately one kilometer at normal cruise altitude airspeeds of 250 ms -1 for Airbus A330/A340 IAGOS aircraft.

155
The ice crystal concentrations reported in this study are the maximum values from individual clouds, where clouds are identified as those encountered at an altitude > 8000 m and that have a measured Nice > 50 L -1 . This latter threshold is imposed to ensure that the number of crystals sampled are statistically representative of a cloud's parent population with an uncertainty < 30% (Beswick et al., 2015;Lloyd et al., 2020).

160
Carbon monoxide (CO) concentrations are measured with a Thermo Scientific Model 48 that has been modified to improve its performance, including the use of an optimized IR detector, cell pressurization, air drying and frequent zeroing (Nédélec et al.,2003). A 30-s integration time is used to obtain a minimum detection limit of 10 ppbv and thus, the CO samples are obtained at a horizontal resolution of about 7.5 km at cruise speeds.

165
The sampled air mass histories are evaluated using the FLEXPART Lagrangian particle dispersion model (Stohl et al., 2005) that simulates the long-range and mesoscale transport of tracers released from point, line, area or volume sources. The CO concentrations detected at cruising altitude are linked to emissions of fossil fuel combustion from urban regions and from biomass burning using the SOFT-IO tool, as described in detail by Sauvage et al. (2017a). We are particularly interested in identifying anomalous CO mixing ratios within the database that would indicate plumes originating from surface UP or BB emissions. Sauvage et al. (2017a) identify these anomalies through a methodology implemented to detect pollution layers in the IAGOS database (Petetin et al., 2018;Cussac et al., 2020).

180
One of the products that is available through the IAGOS central database is flight-by-flight netCDF files that provide the CO mixing ratio that has been transported from either biomass burning or urban regions and that exceeds the background (Sauvage et al., 2017a, b), i.e., the CO anomaly. For the UT the background CO mixing ratio is determined by calculating seasonal median values (over the entire IAGOS database) for 10 different regions worldwide. These

195
The modeled CO anomalies were compared with anomalous values derived from the in situ measured CO, using a similar approach for establishing background values. For each flight, the median CO values were calculated at altitudes between 8 and 12 km and used as the background. A positive difference between the measured value and the median is designated as an anomaly. These anomalies were compared with those from the SOFT-IO simulated anomalies and 200 the differences were generally less than 20%, serving as a quality check of the model results that are used to identify the sources of these anomalies.  The new generation reanalysis from the European Centre for Medium-Range Weather Forecasts (ECMWF), ERA5 (Hersbach et al., 2018) at 1° horizontal resolution is used to estimate climatological values of several meteorological variables such as upper-level horizontal wind divergence and outgoing longwave radiation (OLR).
In addition, selected variables from the reanalysis MERRA-2 (The Modern-Era Retrospective Analysis for Research 225 and Applications, Version 2) are analyzed over the same time period as the IAGOS measurements. MERRA-2, produced by NASA's Global Modeling and Assimilation Office (GMAO), is a multi-decadal global dataset that combines hyperspectral radiance, microwave data, GPS-Radio occultation data, ozone profile observations, and several other ground and satellite-based datasets (Randles et al., 2017;Gelaro et al., 2017). The horizontal resolution is 0.625° (long.) x 0.5° (lat.). The M2I3NVAER product in 72 vertical levels is generated from the output of the 230 Goddard Chemistry Aerosol Radiation and Transport model (GOCART), which simulates an external mixture of 15 particle types. The importance of MERRA-2 in this study is that it provides a gridded output of speciated aerosol mixing ratios (Bocquet et al., 2015;Gelaro et al., 2017), i.e., dust, organic carbon and black carbon (OC, BC) and sulfate (SO4). The mixing ratios of the different particle types in the layers between 8 and 12 km were averaged (weighted) and converted to standard temperature and pressure to determine concentrations.

235
Lighting flash counts and locations over the period 2011 to 2018 were obtained from the World Wide Lightning Location Network (WWLLN) (Dowden et al., 2002) and were gridded to coincide with the MERRA-2 horizontal grid.

Results
The measurements were made with BCPs and CO analyzers installed on eight Airbus A330/A340 IAGOS aircraft.

240
The data analyzed cover the period from December 2011 to March 2020, a total of 22857 flights, during which 67172 ice crystal clouds were sampled, accumulating 1483 hours in clouds.

Cloud statistics
The cumulative frequencies are calculated for ice crystal number concentrations grouped into each of the five regions mentioned above (Fig. 2a) and as a function of season (Fig. 2b). The distribution representing all clouds (black curve) 245 reaches 5000 L -1 at the 85 th percentile, the concentration that we will designate as the EIE threshold (indicated by the vertical dashed lines in Fig 2). The northern and southern NHAA and SHAA regions (orange and red curves) show similar frequency distributions above the EIE threshold, with 15% of the clouds in the EIE category, those in the AO region (teal curve) have 25% of the clouds with concentrations > 5000 L -1 while the PIA regions (blue curve) had 20% of the clouds with concentrations greater than the EIE. Differences in the frequency distributions as a function of dry 250 vs rainy seasons, independent of the regional areas, are slight (Fig. 2b), 13% vs 18%, respectively.
All of the ice number concentrations in the region of interest were mapped as a function of temperature similar to how Krämer et al (2020) presented their cloud climatology. Figure 3 shows   Table 1. The data are binned into 0.25° x 0.25° bins (28 km x 28 km) and the maximum BB and UP anomalies were calculated at each grid point. All of the gridded maxima are averaged to produce 265 Fig. 4a. The average maximum CO anomalies due to BB are larger in the AO and PIA regions while those due to UP and BB are comparable in the NHAA and SHAA (Fig. 4a). These differences in averages, however, are not statistically significant because of the large standard deviations (Table 1) Similarly, although the average CO anomalies appear to be twice as large during the rainy season for the AO, PIA and SHAA, the standard deviations are also quite large (Table 1).
270 Figure 5 highlights the frequency with which the CO anomalies measured over each region are coming from source regions different than the ones in which they were measured. The frequency of CO anomalies due to CO from biomass burning sources is shown in Fig. 5a and from urban pollution in Fig. 5b. The emissions from BB and UP in the PIA region dominate the anomalies in CO over the AO and PIA (Figs. 5a,b). Over the NHAA and SHAA regions it is the 275 BB emissions from the NHAA that have the largest influence on CO anomalies encountered over the Africa and Atlantic regions (Fig 5a). The CO anomalies from UP sources, over the NHAA and SHAA, are primarily from surface sources in these same regions (Fig. 5b).

Combined carbon monoxide and cloud statistics
The cumulative frequency distributions of CO anomalies, differentiated by the source of the anomalies based on the 280 model and grouped by the EIE and non-EIE, are shown in Fig. 6. In addition, CO anomalies were estimated from the measured CO along the flight track, where the anomalies were calculated by subtracting the average background CO.
In Fig. 6 the horizontal black line at 50% intersects all the curves at their median values. The median, modeled CO anomalies for EIE and non EIE are 2.5 and 3.5 ppb for CO from BB, respectively (green curves). The median value for CO from fossil fuel emissions is 10 ppb (black curves), with no difference between anomalies encountered in EIE 285 and non-EIE.
Compared with the modeled anomalies, the measurement-derived CO anomalies (red curves) have median values of 15 and 18 ppb, respectively for the EIE and non-EIE. Although these results suggest that the anomalies are larger in the non-EIE, the differences are not large enough to conclude that surface emissions from combustion could have a larger impact on the non-EIE. The frequency distributions do suggest that emissions from UP sources are potentially a larger source of nucleating particles in the ice clouds, in general. Figure 7 shows the locations of 1) CO anomalies that are not associated with ice clouds (black dots), 2) ice clouds with no CO anomalies (blue dots) and 3) clouds with anomalies (yellow dots, Fig. 7a and magenta dots in Fig. 7b).
295 Figure 7a includes all ice clouds and Fig. 7b is with EIE only. With the exception of several flights between South America and Southern Africa, a large fraction of the flight routes within this region of the world have ice clouds not only accompanied by CO anomalies, but also have EIE with the CO that is associated with air lofted from surface sources. The black star markers are cities with populations larger than ten million, i.e., megacities (these are listed along with their locations and populations in the Supplementary Material, in Table SM1).
300 Figure 8 further clarifies the regional distribution of the average CO anomalies, for all ice clouds, separated by the type of emission source, similar to how the CO data are presented in Fig. 4

Discussion
Based on the results that are presented in the previous section, we can conclude that the sources of a large fraction of the CO that is encountered within the sampled ice clouds are biomass burning and urban emissions at the surface. We will now link these emissions to the primary source regions and discuss how the CO produced in these areas is able to reach aircraft altitudes. This linkage is an important key to supporting our hypothesis that a significant number of 320 ice clouds in the tropics and sub-tropics are formed on CCN or INP produced by BB and fossil fuel combustion. Asia, in particular southeast Asia, the primary burning season is from July through October.

330
We highlight here only the spatial distribution for July and December in Figure 10, in which the magenta and blue correspond to the more intense fires. Also drawn on these maps are the EIE (black markers) from IAGOS flights during the same time periods.The EIE over Africa are north of the BB areas in July; however, as discussed below in section 5, and as was illustrated in Fig. 5, the CO anomalies encountered in the NH Africa ice clouds have some of 335 their sources located in SH Africa. The transport of BB emissions from the surface in one region to the ice clouds in another region is further highlighted in the next section, 4.2, where we discuss vertical motions. In December there are EIE along the airline route between Northern Africa and South America, these would appear to be related to enhanced emissions BB in Northern Africa and westward transport, as is discussed below. The co-location of EIE and BB in Southeast Asia, particularly in the region of Vietnam and Indonesia, is clearly observed.

340
The CO anomaly analysis identifies the potential sources of the air that is mixed with ice clouds, but cannot provide any information about the particles co-emitted with CO. The AOD derived from the MODIS on the Aqua satellite measures the particle light extinction in a vertical column of air, which is a very good indicator of the particle mass in that column. The AOD, shown in Fig. 11, compared with the fire maps and the CO anomalies and ice cloud 345 distributions, provides a correlative link between the particles that are the source of CCN/INP for the ice crystals and the source of emissions. Note that AOD is derived after filtering for clouds so that only aerosol particles are contributing to the light extinction, and values shown in Fig. 11 have been normalized to the maximum value measured in the domain shown.

350
Looking first at Africa and comparing AOD and fire radiative power, while the widespread regions of biomass burning in western Africa in December (Fig 10b) are not directly associated with EIE, the AOD spatial distribution (Fig. 11b) indicates a westward transport of BB emissions over the Atlantic by the dominant upper-level winds. The ice clouds sampled by aircraft near deep convection in the Atlantic Intertropical Convergence Zone (ITCZ) show several EIE which are associated with high AOD originated over the BB in Africa, supported by the modelling to determine CO 355 source collocated with EIE. In contrast, AOD in July (Fig. 11a) is high over northern Africa but not collocated with BB, which is virtually absent there. Nevertheless, this region adjoins the area of most frequent EIE indicating high aerosol particle concentrations associated with the ice clouds.
Some of the higher AOD (Fig. 11b) in Asia may correspond to region with BB in December across India (Fig. 10b) 360 but in July, high AOD observed in the same region, and particularly over the Indian Ocean (Fig. 11a), are clearly unrelated to BB emissions when compared with the FRP mapped in Fig. 10a. The proximity of the EIE to regions with large magnitude of AOD suggests that these clouds have likely formed on aerosol particles from relatively nearby sources. However, the maps suggest that these BB emissions are the source 365 of some, but not all of the particles that lead to EIE. Recall, that the back trajectory analysis indicated that UP was another significant source of CO, but additional information is needed to better support the argument that UP is a source of aerosol particles as well as of CO.
The concentration of different particle composition (considered as external mixtures) is derived from the the MERRA-2 reanalysis, as detailed in the methodology. Figure 12 show the spatial distribution of the mass concentrations of 370 particle mass by composition, averaged over the atmospheric layer from 8-12 km and converted to standard temperature and pressure in dimensions of grams per cubic meters. The composition considered relevant for this study are dust ( Fig. 12a and b), the sum of organic and black carbon ( Fig. 12c and d) and sulfate ( Fig. 12e and f). Only the months of July and December are shown here to relate directly with sources of BB (Fig. 10) and AOD (Fig. 11); all months are shown in the supplementary material (Fig. SMx). The seasonality of the particle composition in the upper 375 troposphere (at altitudes between 8 and 12 km) is quite evident, especially with the dust and sulfate distributions that are much larger in concentration during July than during December.
A comparison of the maps of particle composition (Fig. 12) and AOD (Fig. 11) reveals that the high AOD over northern Africa (Fig. 11a) in July is collocated with the high dust concentrations (> 10 µg m -3 ) in the upper troposphere and also with EIE over the same area. This strongly suggests that the EIE in this region is likely related to dust, in 380 addition to the BB that is also adjoining this region during July and whose presence is confirmed by the CO analysis.
Not only is northern Africa inundated with dust at flight level, but the region over the Arabian Peninsula, the Indian Ocean, India and China also have dust concentrations in excess of 10 µgm -3 , coinciding with significant AOD. The The high OC/BC concentrations in the upper troposphere ( > 10 µgm -3 ) near the ITCZ in central Africa (Fig. 12c) in July and in December over Africa and South America (Fig. 12d), reflect the similarly large AOD in these regions (Fig.   11b) and they overlap with the area of high FRP during the same time periods in July (Fig. 10a) and December (Fig.   390 10b). Hence, the particles associated with the BB emissions are clearly linked with the fires, CO, OC/BC and EIE.
In a similar analysis, the high concentrations of sulfate particles in the upper troposphere in July (Fig. 12e), which stretch from the western Africa to the Pacific Ocean east of Japan, are mixed with dust over part of the region (Africa, Arabia and partially India) but dominate the particle composition over southern China, Thailand, Cambodia, Vietnam 395 and Japan. These high concentrations are partially reflected in the larger AOD, but the particularly striking feature are the many EIE in the region over eastern Asia. Recall that the CO anomaly analysis also identified UP as the primary source of CO in the ice clouds over this area.

415
There is much less seasonal variation over Asia where we observe appreciable divergence that corresponds to the locations of the EIE during July and December. Although the amount of sulfate is much more pronounced in July than in December, there is an increase in the OC/BC during December over this region as well as persistent dust and sulfate particles. Hence, with the widespread vertical motion, there is sufficient uplift of material from the surface to transport

430
The regions of high lightning counts are in close correspondence with the low OLR and large upper-level divergence, further reinforcing the correlative relationship between these independent indicators of vertical motion and the presence of clouds. Note the large clusters of EIE located in the same regions of lightning high frequency, low OLR and elevated divergence.

Relationship of EIE to ice crystal icing hazards
Ice clouds, in addition to having a major influence on climate through their interactions with incoming and outgoing radiation, are also a potential hazard to commercial aviation as a result of the clogging of air data sensors or by being ingested into engines, a condition that can lead to engine flameout or rollback (e.g., Mason et al., 2006;Grzych and Mason, 2009;Mason and Grzych, 2011;Bravin, et al., 2015;Haggerty et al., 2019;Bravin and Strapp, 2019). Figure   14 superimposes The juxtaposition of nearly all of the ICI events with the EIE underscores the importance of the link between hazardous 445 flying conditions and the presence of ice clouds with extremely high crystal concentrations, clouds that in this current study have been associated with ground based emissions of anthropogenic CO and aerosol particles.

Relationship with previous studies
The results that we have presented provide a framework for linking ice clouds in general, and EIE in particular, to 450 surface sources of dust, BB and UP in tropical latitudes. However, this study lacks in situ measurements of the actual ice crystal residuals that could identify the composition of the ice nuclei, we are unable to unequivocally conclude that extreme ice events are due to emissions from BB and UP. Instead, we will place our results in the context of studies  The potential influence of anthropogenic emissions was evaluated using SOFT-IO, a model that incorporates back trajectory analysis to identify anomalous levels of carbon monoxide (CO) in the air masses sampled by the aircraft.

475
The CO anomalies, values that exceeded a background concentration at flight level, were averaged into 0.25° by 0.25° grid cells and compared with ice crystal concentrations averaged into the same cells. Of all the cells that had positive CO anomalies or ice concentrations > 50 L -1 along the aircraft flight tracks, 52% of the cells had CO anomalies with no clouds, 10% had clouds with no CO anomalies, and 48% had concurrent observations of clouds and CO anomalies.
When only EIE clouds were evaluated, the numbers were 75%, 4% and 21%, respectively. In other words, when 480 considering all clouds, 79% were accompanied by positive CO anomalies while only including EIE clouds, 84% of these had anomalous concentrations of CO. Hence, there is nothing that would suggest that EIE clouds are being influenced significantly more often than non-EIE clouds.
The modal frequency of the temperature EIE clouds was five degrees colder than non-EIE clouds; however, this 485 difference is insignificant when taking the temperature variance into account. Likewise, neither the differences in EIE over the study region or the differences in the magnitude of CO anomalies between EIE and non-EIE clouds were significant.
We conclude that the two, primary factors that are associated with the EIE encounters are the proximity to sources of   especially over Africa and Asia, which is highlighted in the BCP measurements compared with pilot results (Fig. 14).
More recently, Rugg et al. (2021) analyzed 11 years of Cloudsat and CALIOP measurements and showed that, on average, more than 2% of the UT clouds in this region contain ice water content > 1.0 gm -3 . Hence, a better knowledge 520 of the sources of aerosol particles, their composition and their concentrations, will provide valuable information that can improve flight planning and the forecast of potential areas of EIE.
The current study has established a pathway that links extreme ice concentrations in ice clouds and the surface emissions of particles from biomass burning and fossil fuel combustion; however, there remain a number of critical, 525 unresolved questions: • What is the relationship between ice cloud CO and the properties of the CCN or INP from which the crystals formed?
• What is the relationship between age of the CO and the ice crystal concentrations, in particular, what leads to the evolution of EIE clouds?

530
• What additional information is needed to accurately forecast ice clouds in general and EIE in particular, not only for improved flight safety, but for a better understanding of the climate impact of these clouds?
In final summary, the results presented in this study emphasize the importance of programs like IAGOS for expanding the data base of cloud measurements and the need to broaden such programs to include more aircraft to provide a more 535 complete spatial and temporal climatology. 545 at no cost for users who supply data to this network; otherwise, there is a charge for the data.