Five satellite sensor study of the rapid decline of wildfire smoke in the 1 stratosphere 2

10 11 Smoke from Western North American wildfires reached the stratosphere in large amounts in 12 August 2017. Limb-oriented satellite-based sensors are commonly used for studies of wildfire 13 aerosol injected into the stratosphere ( OMPS-LP (Ozone Mapping and Profiler Suite Limb 14 Profiler) and SAGE III/ISS (Stratospheric Aerosol and Gas Experiment III on the International 15 Space Station)). We find that these methods are inadequate for studies the first 1 – 2 months after 16 such a strong fire event due to event termination (“saturation”). The nadir-viewing lidar CALIOP 17 (Cloud-Aerosol Lidar with Orthogonal Polarization) is less affected due to shorter path in the 18 smoke, and, further, provides means that we could use to develop a method to correct for strong 19 attenuation of the signal. After the initial phase, the aerosol optical depth (AOD) from OMPS-LP 20 and CALOP show very good agreement above the 380 K isentrope, whereas the OMPS-LP tends 21 to produce higher AOD than CALIOP in the lowermost stratosphere (LMS), probably due to 22 reduced sensitivity at altitudes below 17 km. Time series from CALIOP of attenuation-corrected 23 stratospheric AOD of wildfire smoke show an exponential decline during the first month after the 24 fire, which coincides with highly significant changes in the wildfire aerosol optical properties. 25 The AOD decline is verified by the evolution of the smoke layer composition, comparing the 26 aerosol scattering ratio (CALIOP) to the water vapor concentration from MLS (Microwave Limb 27 Sounder). Initially the stratospheric wildfire smoke AOD is comparable with the most important 28 volcanic eruptions during the last 25 years. Wildfire aerosol declines much faster, 80 – 90% of 29 the AOD is removed with a half-life of approximately 10 days. We hypothesize that this dramatic 30 decline is caused by photolytic loss. This process is rarely observed in the atmosphere. However, 31 in the stratosphere this process can be studied with practically no influence from wet deposition, 32 in contrast to the troposphere where this is the main removal path of sub-micron aerosol In this study we investigate massive injections of smoke into the stratosphere from the August 603 2017 North American wildfires using five satellite sensors. Methodology was developed to 604 correct CALIOP data for attenuation of the laser signal. The CALIOP AOD and extinction 605 coefficients were compared with OMPS-LP and SAGE III/ISS. From 1 – 2 months after the fire 606 we find that OMPS-LP and CALIOP AOD agree very well at altitudes above the 380 K 607 isentrope, where the former demonstrates high sensitivity with small statistical fluctuations. The 608 methods differ dramatically during the first 1 – 2 months after the fire when the smoke layers are 609 dense, because the long optical path through the smoke of the limb-oriented instruments OMPS- 610 LP and SAGE III/ISS cause event termination (“saturation”). This is clearly demonstrated by the 611 low daily maximum extinction coefficients of the two instruments, being orders of magnitude 612 lower than the peak extinction coefficients of CALIOP. The nadir viewing CALIOP experiences 613 a much shorter optical path, because the vertical extension of smoke layers usually are orders of 614 magnitude shorter than for limb orientation. We find that CALIOP is an indispensable tool for 615 studies of dense smoke layers entering the stratosphere after intense wildfires, providing signal 616 along the laser path that can be used to correct for attenuation. Once the smoke layers are 617 sufficiently thin, the limb technique OMPS-LP provide sensitive measurements of the AOD that 618 can be used together with CALIOP.

within each swath were clustered depending on their location. Noise in the data led to some lone 140 pixels within layers of either ice or smoke. These were reclassified depending on the surrounding 141 pixels, making sure that no single pixel marked as aerosol occurred within the ice-cloud layers.

142
Layers of ice-clouds were then expanded upwards and horizontally to capture faint edges of the The evolution of the lidar, color and depolarization ratios were investigated using 32 separate 153 smoke layer measurements over the period 3 -59 days after the fire. CALIOP has a statistical 154 disadvantage compared with lidars at the ground (Baars et al., 2019), because of small solid angle 155 due to long distance to the stratosphere (~700 km) and short measurement time. Optical where is the attenuated backscattering and T 2 the two-way transmissions from both particles 185 and molecules. The two-way particle transmission is obtained by first computing the AOD: where Δ is the height of the altitude pixel, , is backscattering from air molecules, and the 190 lidar ratio of the aerosol particles. The molecular lidar ratio, for computation of the molecular 191 extinction, was set to 8.70477 sr (Prata et al., 2017). CALIOP measurements are affected by 192 multiple scattering (Wandinger et al., 2010), causing overestimation of the backscattering. The Therefore, estimations of the attenuation were undertaken also for the long wavelength. The 222 molecular backscattering is assumed to be 1/16 of that at 532 nm (1 ⁄ dependence of Rayleigh 223 scattering). Weak molecular scattering at 1064 nm prohibits lidar ratio estimation at that 224 wavelength by CALIOP. Instead, the lidar ratio was assumed to be 60 sr, inducing uncertainties 225 in the color ratio. The volume color ratio is obtained from: where we made use of the wavelength dependence of Rayleigh scattering for molecular 239 scattering, and the scattering ratio for the 532 nm wavelength was obtained from eqn. 4.

241
We also investigated the depolarization of the scattered laser beam at 532 nm by first forming the 242 volume depolarization ratio: we will make use of two of wavelengths: 745 nm because of the reduced problem with 271 sensitivity, and 510 nm because it is the wavelength closest to that of CALIOP (532 nm).

273
The OMPS-LP aerosol extinction coefficients are provided on a grid with a vertical resolution of  Water vapor concentrations (mixing ratio) in individual smoke layers was obtained from the MLS 309 instrument aboard the Aura satellite (Waters et al., 2006) in 12 vertical steps per decade of 310 pressure (version 5.0-1.0a, level 2). In nighttime measurements from days 6 -59 after the fire,

344
Here we use an approach based on five satellite sensors to study the influence on the stratosphere American wildfire aerosol is compared with volcanic influence on the stratospheric AOD.

353
Finally, the fifth data set, water vapor from the MLS, is introduced in the discussion section, 354 where the evolution of the wildfire aerosol in the stratosphere is analyzed.

358
The smoke layers usually were 1 -3 km thick and could extend several degrees in longitude and 359 latitude. Measurements with the CALIOP lidar provide, in addition to short, nadir-viewing 360 measurement path in dense layers, the advantage that the signal is retrieved as a function of 361 position along the laser path with high resolution, which can be used to correct for attenuation of 362 the signal. Figure 3a shows the attenuated scattering ratio (R'; the measured backscattering 363 divided by the calculated molecular backscattering) from an example-smoke-layer measured on 364 August 16, 2017. The scattering ratio should be close to 1 in air layers with low aerosol 365 concentration, whereas values below 1 is caused by attenuation from particles. As can be seen in 366 Figure 3a, the attenuated scattering ratio first increases (starting from above the layer). Then the 367 signal decreases and reaches well below unity from 11 km altitude and downwards, i.e., well 368 below the scattering ratio of particle-free air. By techniques described in the Methods section we The evolution of wildfire aerosol from day 3 to 59 after the North American PyroCbs on August 373 12, 2017, is first investigated by comparing 32 smoke layers from individual CALIOP swaths.

374
The influence from attenuation is shown in Figure 3b. Clear deviation from the 1:1 line appears 375 already at layer attenuated (uncorrected) AODs (AODatt) of 0.12, and 50% reduction of the signal 376 appears at layer AODatt of approximately 0.25. Reduction by more than 50% appears until day 10 377 after the fire, whereas those measurements close to the 1:1 line were taken after day 30. The 378 AOD, i.e., the AOD corrected for attenuation, exceeds the AODatt by more than a factor of 5 in 379 the densest layers of this study (Figure 3b).

383
To study the evolution of the stratospheric AOD, we form a 3-dimensional box in the

402
The stratosphere above the LMS (above the 380 K isentrope) shows good agreement between the 403 two methods, except for the first 1 -2 months after the fire ( Figure 5d). contrast, the wildfire aerosol displays a rapid decline during the first few weeks, before the AOD 497 stabilizes (Figure 9). This is followed by a period of rather stable AOD of more than 6 months, 498 before the AOD evolution turns to a slower decline towards background conditions, with similar 499 seasonality as the aerosol from the volcanic eruptions discussed (Figure 9). This latter decline is 500 mainly caused by springtime transport out from the stratosphere at mid and high latitudes

505
The smoke aerosol is distributed both in the LMS and in the lower BD branch like aerosol from  Moreover, the change in the particle depolarization ratio (Figure 2c) indicates change of the 512 aerosol particle properties, and the particle color ratio decrease after the fire (Figure 2b) is the 513 expected outcome for reduced particle sizes. Based on these arguments we turn the attention to on the strong and rapid decline of the stratospheric AOD during the first month after the fire, we 574 find that photolytic loss of organic aerosol is a highly likely explanation. The rate of photolytic 575 loss is likely better described by the evolution of R/CH2O than by the AOD, because the latter 576 could to some degree be affected by transport across the tropopause. Our strong experimental 577 evidence leads us to the hypothesis that the rapid decline of the wildfire aerosol in the 578 stratosphere with a half-life of 10 days is caused by photochemical loss of organic material. This 579 should be further investigated by modeling, but that is outside the scope of the present study.

581
To further put the strong early decline of wildfire aerosol into context, we compare the AOD Finally, we investigate the stratospheric aerosol load from the wildfire by comparing with the 593 more studied volcanic impact (Table 1)   during one year after the fire and after the two volcanic eruptions in Figure 9.   color ratio (1064 nm divided by 532 nm wavelength backscattering) with exponential fit (R 2 = 985 0.48, P < 10 -10 ), and c) particle depolarization ratio with exponential fit (R 2 = 0.76, P < 10 -10 ).

986
The color and depolarization ratios were divided in two equal groups by number of observations 987 to illustrate the highly significant changes with time of the optical properties, where the long and 988 short error bars are the standard error and the double-sided 95% probability range of the 989 geometric means.