Aerosol particle depolarization ratio at 1565 nm measured with a Halo Doppler lidar

Abstract. The depolarization ratio is a valuable parameter for
lidar-based aerosol categorization. Usually, the aerosol particle depolarization
ratio is determined at relatively short wavelengths of 355 nm and/or 532 nm,
but some multi-wavelength studies including longer wavelengths indicate
strong spectral dependency. Here, we investigate the capabilities of Halo
Photonics StreamLine Doppler lidars to retrieve the particle linear
depolarization ratio at the 1565 nm wavelength. We utilize collocated
measurements with another lidar system, PollyXT at Limassol, Cyprus, and at
Kuopio, Finland, to compare the depolarization ratio observed by the two
systems. For mineral-dust-dominated cases we find typically a slightly lower
depolarization ratio at 1565 nm than at 355 and 532 nm. However, for dust
mixed with other aerosol we find a higher depolarization ratio at 1565 nm. For
polluted marine aerosol we find a marginally lower depolarization ratio at
1565 nm compared to 355 and 532 nm. For mixed spruce and birch pollen we
find a slightly higher depolarization ratio at 1565 nm compared to 532 nm.
Overall, we conclude that Halo Doppler lidars can provide a particle linear
depolarization ratio at the 1565 nm wavelength at least in the lowest 2–3 km
above ground.


Now, recently developed post-processing (Vakkari et al., 2019) allows the utilization of significantly weaker signals from Halo Doppler lidars than previously. Therefore, the main aim of this paper is to assess the capabilities of Halo Doppler lidars in providing particle linear depolarization ratio measurements at 1565 nm wavelength. To do so, we utilize collocated Halo Doppler lidar and multiwavelength Raman lidar PollyXT observations during two measurement campaigns, where different polarizing aerosols were observed. Overall, the comparison indicates that Halo Doppler lidars can add another wavelength at 70 1565 nm to studies on the spectral dependency of particle linear depolarization ratio, at least in the lowest 2-3 km above ground.

Materials and Methods
Here we use data from two measurement campaigns where a Halo Photonics Doppler lidar and a PollyXT Raman lidar were collocated; at Kuopio, Finland, from 9 to 16 May 2016, and at Limassol, Cyprus, from 21 April to 22 May 2017. The 75 campaigns represent quite different environments (Fig. 1) and enable the comparison of depolarization ratio at 1565 nm by the Halo instrument to depolarization ratio at 355 and 532 nm from PollyXT for a range of aerosol types. Furthermore, the campaigns were equipped with different devices of the Halo and PollyXT designs and thus potential differences between instrument individuals can be investigated.
The Vehmasmäki site (62. 738°N, 27.543°E; 190 m a.s.l.) in Kuopio is a rural location surrounded by boreal forest 80 (Bohlmann et al., 2019). The focus of the campaign in May 2016 was to investigate the capability to characterize the optical properties of airborne pollen with the multiwavelength Raman lidar PollyXT (Bohlmann et al., 2019). Here, we utilise one week of collocated measurements to compare Halo depolarization at 1565 nm to PollyXT during a spruce and birch pollination episode.
Limassol (34.675°N, 33.043°E; 22 m a.s.l.) is located at the southern shore of Cyprus in the Eastern Mediterranean. 85 Measurements at Limassol were part of the Cyprus Clouds Aerosol and Rain Experiment (CyCARE;Ansmann et al., 2019) and were performed as a collaboration between Cyprus University of Technology (CUT), Limassol, and Leibniz Institute for Tropospheric Research (TROPOS), Leipzig. During April-May, several Saharan dust episodes were observed at Limassol in addition to the regional aerosol.

Halo Doppler lidar 90
Halo Photonics Stream Line lidars are commercially available 1565 nm pulsed Doppler lidars equipped with a heterodyne detector (Pearson et al., 2009). Halo Stream Line lidars emit linearly polarized light and the optical path is constructed with fibre-optic components, which can be equipped with a cross-polar receiver channel. The cross-polar channel is implemented through a fibre-optic switch between the normal receiver path and path with a fibre-optic polarizer. Thus, the measurement of the co-and cross-polar signals is not simultaneous, but consecutive in vertically-pointing mode. For instance, if the integration time per ray is set to 7 s then co-polar signal is collected for 7 s and then cross-polar signal is collected during the next 7 s.
For research purposes, the most commonly used variants of Stream Line lidars are Stream Line, Stream Line Pro and Stream Line XR. The Stream Line and the more powerful Stream Line XR lidars enable full hemispheric scanning. The Streamline Pro is designed without moving parts on the outside, which limits the scanning to a cone of 20° from vertical. All Stream 100 Line variants can be used for depolarization ratio measurements, but an important difference between XR and other Stream Line versions is that the XR background noise level cannot be determined as accurately in the near range as for the non-XR versions (Vakkari et al., 2019). This difference is attributed to the more sensitive amplifier used in the Stream Line XR (Vakkari et al., 2019).
In this study we utilise vertically pointing measurements only from two Stream Line Pro systems. The operating 105 specifications of these systems are given in Table 1. Stream Line lidars send and receive pulses through a single lens, which avoids issues with overlap and leads to a minimum range of 90 m due to impact of the outgoing pulse. At Vehmasmäki, we focused on boundary layer aerosol and set integration time per ray to 7 s and telescope focus to 2000 m. At Limassol, we expected to encounter elevated aerosol layers frequently and set integration time per ray to 11.5 s and telescope focus to infinity. The integration time is set to balance between signal strength and good enough time resolution for retrievals of 110 turbulent properties.
Halo Stream Line lidars provide three parameters along the beam direction: radial Doppler velocity, signal-to-noise ratio (SNR), and attenuated backscatter ( ), which is calculated from SNR taking into account the telescope focus. A background check to determine background noise level is performed automatically once per hour. We post-processed SNR according to Vakkari et al. (2019), which is essential to be able to further reduce the instrumental noise floor by averaging the SNR. 115

Halo depolarization ratio
We estimate the instrumental uncertainty in Halo Stream Line SNR from the standard deviation of SNR in the cloud-and aerosol-free part of the profile. Using mean values for the atmospheric number density taken from the U.S. Standard Atmosphere, 1976(COESA 1976, the molecular backscatter coefficient at 1565 nm is about 1.9 x 10 -8 m -1 sr -1 at mean sea level. Given the long wavelength and low pulse energy, no contribution from molecular scattering is observed in the signal, 120 and the two-way atmospheric transmittance at 1565 nm is still 0.9994 at 2 km altitude above a lidar situated at mean sea level. Hence, the measured depolarization ratio can be safely assumed to represent the particle linear depolarization ratio. In Fig. 2a, consecutive co-and cross-polar SNR profiles are presented, where aerosol signal is observed up to 800 m above ground level (a.g.l.) and a liquid cloud base is observed at 840 m a.g.l.. In liquid cloud the signal attenuates quickly and above 1 km the profiles represent instrumental noise only. We use the measurements above 1 km to calculate standard 125 deviations of co-polar SNR ( co) and cross-polar SNR ( cross). In Fig. 2b, raw depolarization ratio ( ) is calculated simply as the ratio of cross-polar SNR to co-polar SNR and uncertainty is estimated from co and cross by Gaussian error propagation.
The construction of Halo Stream Line lidars does not enable user calibration of the depolarization ratio, unlike PollyXT lidars for instance (Engelmann et al., 2016). However, we can evaluate the Halo depolarization ratio at liquid cloud base.
Spherical cloud droplets do not polarize the directly back-scattered radiation and thus non-zero depolarization signal at liquid 130 cloud base is an indication of incomplete extinction (or bleed-through) in the lidar internal polarizer. It should be noted, though, that multiple scattering results in increasing depolarization signal inside a liquid cloud (e.g. Liou and Schotland, 1971). This increase in in-cloud is clearly seen also in Fig. 2b. The magnitude of this effect depends on both cloud and lidar properties (e.g. Donovan et al., 2015); for Halo Stream Line lidars this effect is moderate as seen in Fig. 2b.
For the purpose of determining the polarizer bleed-through we minimize the effect of multiple scattering by considering 135 only at the cloud base and determine average bleed-through from measurements in several clouds. Furthermore, we note that cloud-base should be determined from relatively high time resolution data to ensure that both co-and cross-polar measurements represent the same part of the cloud. Finally, it should be noted that, especially in higher latitudes, it is not always trivial to find purely liquid phase clouds. Typically, mixed-phase clouds can be distinguished in the histogram of cloud-base as a secondary peak with higher than liquid clouds, but this requires the collection of data from a larger set of 140 clouds.
To characterize the Halo polarizer bleed-through, we determined the depolarization ratio at liquid cloud base during both campaigns (Fig. 3). During the campaign at Limassol, we determined at cloud base on 25 April and on 2 May 2017. From the distribution in Fig. 3a, the bleed through is 0.011 ± 0.007 (mean ± standard deviation). At Vehmasmäki, we utilized clouds on 13, 14 and 16 May 2016 as shown in Fig. 3b. At Vehmasmäki, the estimated bleed-through is 0.016 ± 0.009 (mean 145 ± standard deviation).
We attribute the spread in the distributions in Fig. 3 mostly to variability of the clouds at the measurement sites and to the fact that co-and cross-polar measurements are consecutive and not simultaneous. Given that the cross-polar measurement channel is constructed with fibre-optic technology, we do not expect changes in the performance of the polarizer.
Considering the large natural variability of depolarization ratio (e.g. Illingworth et al., 2015;Baars et al., 2016) we find the 150 spread of observations in Fig. 3 tolerable. The standard deviation in Fig. 3 is included in the uncertainty calculation of Halo depolarization ratio.
We account for the observed bleed-through (B) in calculating Halo depolarization ratio ( 1565) as where SNRco and SNRcross are observed co-and cross-polar SNR, respectively. Uncertainty in SNRcross corrected for bleed-155 through (i.e. numerator in Eq. 1) is estimated as where B is standard deviation of the distribution in Fig. 3. Finally, uncertainty in 1565 taking into account instrumental noise and uncertainty in bleed-through is estimated as (3) 160

PollyXT
PollyXT is a multiwavelength Raman lidar capable of depolarization ratio measurement at one or two wavelengths depending on instrument configuration (Baars et al., 2016;Engelmann et al., 2016). PollyXT emits simultaneously 355, 532 and 1064 nm wavelength pulses at a repetition frequency of 20 Hz. All PollyXT lidars are built with the same design, but there are small differences in the number of receiver channels equipped in each individual system. A detailed description of 165 PollyXT design is given by Baars et al. (2016) and Engelmann et al. (2016).
At Vehmasmäki, PollyXT was configured with elastic backscatter channels (355, 532 and 1064 nm), Raman-shifted channels at 387, 407 and 607 nm and a cross-polar channel at 532 nm (Bohlmann et al., 2019). Due to the biaxial construction of emission and detection units, complete overlap is reached at 800-900 m a.g.l. (Engelmann, et al., 2016) and thus, only measurements > 800 m a.g.l. are utilized for this study (Bohlmann et al., 2019). The original spatial resolution is 170 30 m and temporal resolution 30 s for the Vehmasmäki system (Bohlmann et al., 2019).
At Limassol, PollyXT operated the same receiver channels as the Vehmasmäki system had and additionally a cross-polar channel at 355 nm, together with a near-range telescope with 355 and 532 nm receiver channels. The near-range channels enable retrieval of optical properties down to 150 m a.g.l. (Engelmann et al., 2016). Raw spatial resolution is 7.5 m and temporal resolution, 30 s. 175 During night-time, the Raman-method (Ansmann et al., 1992) is used to retrieve aerosol optical properties from the raw signals. For daytime measurements, the method of Klett (1981) can be utilised. Here, we present only measurements when the Raman-method was applied. The calibration of depolarization ratio was performed at both Vehmasmäki and Limassol using the so-called 90°-method (Freudenthaler, 2016) and the relative uncertainty in particle linear depolarization ratio was estimated to be 10 %. 180

Auxiliary data
Air mass history was estimated with the Hybrid Single-Particle Lagrangian Integrated Trajectory model HYSPLIT (Stein et al., 2015). HYSPLIT was run through the READY website (Rolph et al., 2017) using the NCEP Global Data Assimilation System (GDAS) meteorology at 0.5° horizontal resolution. 96 h back-trajectories were calculated arriving at the elevation of aerosol layers of interest. 185 were observed during the Limassol campaign in Eastern Mediterranean and pollen was observed during the Vehmasmäki campaign in a boreal forest region in Finland. We conclude this section with an overall comparison of depolarization ratio 190 measurements with the two instruments.

Limassol 21 April 2017
Right at the beginning of Halo measurements at Limassol on 21 April 2017, several elevated layers were observed as seen in Fig. 4. Although Halo can observe elevated layers up to 6 km a.g.l. on this day, the signal is too weak to retrieve their 195 depolarization ratio. This is clearly visible in the uncertainty in the Halo depolarization ratio in Fig. 4c. At 300 s integration time (i.e. 10 minutes of alternating co-and cross-polar measurement), the depolarization ratio can be determined up 1-1.5 km a.g.l. with < 0.05 on this day (Fig. 4d). The depolarization ratio can be retrieved also for the relatively strong elevated layer at 3 km a.g.l. during the morning hours (Fig. 4d).
Increasing both temporal and spatial averaging enables the utilization of some of the weaker signals. vertically with a 300 m running mean. In the lowest layer < 1 km a.g.l., practically no difference is observed in the depolarization ratio at the different wavelengths. Back-trajectory calculations (Fig. 6) indicate this layer to be mostly regional air from Eastern Mediterranean and the relatively large lidar ratio is in the range of observations of smoke or smoke mixed with dust (e.g. Gross et al., 2011;Baars et al., 2016). On the other hand, for the layer from 1.5 km to 2 km a.g.l. a 205 clear increase in with increasing wavelength is observed. For this layer air mass history indicates origins over Northern Africa (Fig. 6) and the lidar ratio (42±4 at 355 nm, 47±5 at 532 nm) is in the range of dust (Ansmann et al., 2011). For this layer the mean (± standard deviation) at 355 nm, 532 nm and 1565 nm are 0.19±0.008, 0.23±0.008 and 0.29±0.008, respectively. Above 2 km a.g.l., the uncertainty in at 1565 nm increases rapidly and is not used for quantitative analysis here. 210

Limassol 27 April 2017
Stronger elevated aerosol layers were observed at Limassol on 27 April 2017. On this day, depolarization ratio can be retrieved by Halo up to 3 km a.g.l. (Fig. 7). For an averaging period of 01:25-02:30 UTC, depolarization ratio is retrieved for the elevated layer at 1600-2200 m a.g.l.. For this layer, the depolarization ratio at 1565 nm is 0.30±0.005, which is a little lower than for the shorter wavelengths: 0.36±0.01 at 355 nm and 0.34±0.002 at 532 nm, respectively. For this layer, the air 215 mass history indicates southerly origins.
On the same day (27 April 2017) at 19:00-20:00 UTC, the depolarization ratio can be retrieved from the surface up to 2.6 km a.g.l. (Fig. 8). in the lowest 500 m, depolarization ratio at 1565 nm is clearly higher than at the shorter wavelengths, suggesting a mixture of larger mineral dust particles with smaller particles of lower depolarization ratio. For the layer at https://doi.org/10.5194/acp-2020-906 Preprint. Discussion started: 23 October 2020 c Author(s) 2020. CC BY 4.0 License. 1500-2500 m a.g.l., practically no wavelength-dependency is observed for depolarization ratio, indicating that backscatter at 220 all wavelengths is dominated by the same aerosol. The layer-averaged depolarization ratios are 0.31±0.006, 0.33±0.005 and 0.32±0.008 at 355 nm, 532 nm and 1565 nm, respectively. This high depolarization ratio and lidar ratio of 47±5 at 355 nm (38±3 at 532 nm) indicate almost pure dust (Ansmann et al., 2011;Baars et al., 2016). Air mass history, on the other hand, indicates northerly or north-westerly origins at both 2 km a.g.l. and at the surface (Fig. 9).

Polluted marine aerosol 225
On 20 May 2017 at Limassol, very low aerosol depolarization ratio is observed throughout the day as seen in Fig. 10. During the morning and afternoon liquid clouds are observed but during the evening Raman retrievals with PollyXT were possible. respectively. The layer-averaged lidar ratio at 355 nm is 39±4 sr, whereas the lidar ratio at 532 nm is very noisy at 47±35 sr.
The low depolarization ratio is typical of marine aerosol, smoke and pollution (Gross et al., 2011;Illingworth et al., 2015).
The 355 nm lidar ratio lies between the values reported for marine aerosol and smoke (Illingworth et al., 2015).
Above 1 km a.g.l., an optically thin aerosol layer is observed (Fig. 11). Halo indicates a higher depolarization ratio for this layer than at the surface, but the signal is so weak that the uncertainty in depolarization ratio at 1565 nm becomes very large 235 (Fig. 11b). Back-trajectories arriving over Limassol at 21 UTC indicate different, but mostly northerly origins for the air mass at 500 m a.g.l. and at 2 km a.g.l. (Fig. 12).

Pollen in boreal forest
On 15 May 2016, substantial amounts of spruce and birch pollen were observed at Vehmasmäki with both an in-situ sampler and the PollyXT lidar (Bohlmann et al., 2019). The presence of more polarizing spruce pollen (Bohlmann et al., 2019) in the 240 boundary layer is observed also with Halo lidar as seen in Fig. 13d. However, the backscatter (Fig. 14a) is low compared to the case studies presented for Limassol and the low signal results in significant noise in the lidar ratio (Fig. 14c).
Comparing the depolarization ratios measured with Halo and PollyXT (Fig. 14b) shows a nearly constant depolarization ratio at 1565 nm, while the depolarization ratio at 532 nm decreases with height. At 1565 nm, the Halo signal is probably dominated by pollen grains, which are tens of micrometres in diameter. At 355 nm and 532 nm wavelengths, the backscatter 245 is increasing with height (Fig. 14a) and thus the decreasing depolarization ratio at 532 nm may reflect an increasing fraction of signal from non-pollen aerosol with increasing height. For the layer from 800 m to 1 km a.g.l. in Fig. 14, the mean depolarization ratios are 0.236±0.009 and 0.269±0.005 at 532 nm and 1565 nm, respectively.

Overview of depolarization ratio wavelength dependency
An overall comparison of the depolarization ratio at different wavelengths for the Limassol and Vehmasmäki campaigns is 250 presented in Fig. 15, where the Halo vertical resolution of 30 m has been smoothed with a 300 m running mean. The original time resolution observations by Halo have been averaged to match the temporal resolution of PollyXT Raman retrievals (ranging from 45 min to 2 h).
In Fig. 15a, three regions can be observed in the scatterplot. For 532 < 0.05, 1565 matches very closely with the shorter wavelength. For 532 ranging from 0.05 to 0.25, 1565 is systematically larger than 532. For 532 > 0.3, 1565 is lower than the 255 depolarization ratio at the shorter wavelength. A very similar pattern is present in Fig. 15b: for 355 < 0.05, 1565 matches 355 closely; for 355 ranging from 0.05 to 0.25, 1565 is larger than 355 and for 355 > 0.3, 1565 is lower than 355. Even comparing the two shorter wavelengths (Fig. 15c), similar regions appear: for 355 < 0.05, 532 is lower than 355; for 355 ranging from 0.1 to 0.3 depolarization ratio is on average equal on both wavelengths and for 355 > 0.3, 532 is lower than 355.
Figs. 15a-c show also similar correlations between the depolarization ratios at different wavelengths. Therefore, bearing in 260 mind the similar patterns in all three scatterplots in Figs. 15a-c, we consider the scatter to originate mainly from the atmospheric aerosol properties rather than in instrumental effects. For instance, any bias in the estimated bleed-through in the Halo polarizer would show up as bias in Fig. 15a and 15b. However, such bias is not present in the cases when 355 and/or 532 are low.
Considering the sources at Limassol during the campaign, the higher 1565 for intermediate depolarization ratios ranging from 265 0.1 to 0.25 likely represents mixtures of dust with other aerosol types. A mixture of coarse, polarizing dust with less polarizing and smaller aerosol would result in the observed spectral dependency of depolarization ratio. For aged dustdominated cases, lower depolarization ratios at longer wavelength could be due to the faster removal of coarse particles compared to submicron aerosol (e.g. Burton et al., 2015). In any case, the observed wavelength dependency in Figs. 15a-c for large suggests that, for dust-dominated cases, smaller particle sizes have, on average, higher depolarization ratio at 270 Limassol.
Another type of polarizing aerosol, i.e. pollen, was observed with a collocated Halo and PollyXT at Vehmasmäki (Bohlmann et al., 2019). Comparatively low signal levels, together with 800 m minimum range for the PollyXT system at Vehmasmäki (Bohlmann et al., 2019), reduce the amount of data available for comparison of Halo and PollyXT depolarization ratio during the campaign (Fig. 15d). During this campaign, the depolarization ratio at 1565 nm is a little larger than at 532 nm, 275 but the difference is small compared to the scatter observed at Limassol.
A further look into the distribution and spectral dependency of the depolarization ratio at Limassol is presented in Fig. 16. In Figs. 16a and 16b, the 2D-histograms of depolarization ratio show that both 532 nm and 1565 nm wavelengths present a bimodal distribution below 1 km a.g.l.. In other words, there are also less polarizing aerosols frequently present in the lowest 1 km in addition to dust and dusty mixtures with depolarization ratio > 0.2. However, above about 1.5-2 km a.g.l., almost all retrievals indicate dust or dusty mixtures. Note that the vertical extent of the data is limited by the sensitivity of the Halo instrument, as Figs. 16a and 16b are limited to cases when both wavelengths are available.
In Figs. 16c and 16d, the ratio of depolarization ratios at 1565 nm and 532 nm exhibits clear height-dependency. Above about 1.5 km a.g.l., the majority of the observations present a lower depolarization ratio at 1565 nm than at 532 nm, while below 1.5 km a.g.l., the depolarization ratio is higher at the longer wavelength. In previous studies (Freudenthaler et al., 285 2009;Gross et al., 2011;Burton et al., 2015;Haarig et al., 2017), a lower depolarization ratio at longer wavelengths has been attributed to faster removal of coarse mode dust. However, our observations indicate the presence of a small coarse mode, probably mineral dust, for sub-1.5 km aerosols most of the time at Limassol.

Discussion
The majority of aerosol depolarization ratio measurements have been carried out at relatively short wavelengths (355 nm and 290 532 nm) with only a few previous studies investigating the spectral dependency including 710 nm (Freudenthaler et al., 2009;Gross et al., 2011) and/or 1064nm (Freudenthaler et al., 2009Burton et al., 2012Burton et al., , 2015Haarig et al., 2017Haarig et al., , 2018Hu et al., 2019). In this study we have for the first time determined aerosol particle depolarization ratios at a wavelength of 1565 nm.
From an instrumental point of view, the Halo Doppler lidar depolarization ratio seems to be of comparable quality to 295 PollyXT depolarization ratio when the aerosol signal is strong. However, Halo has a much less powerful laser than PollyXT, which limits significantly the range of usable signal. On the other hand, Halo Doppler lidars are capable of independent operation for months and are therefore suitable for operational use in meteorological measurement networks.
The integration time and range gate length are adjustable in Halo firmware and prolonging these parameters would increase the sensitivity of the system. However, high spatial and temporal resolution are preferable for utilizing the Doppler 300 capabilities of Halo lidars. Inspecting the internal polarizer performance at liquid cloud base also requires a higher resolution. Overall, the configuration of a Halo Doppler lidar needs to be considered individually for the aims of each measurement campaign.
The spectral dependency that we observed for 355 nm, 532 nm and 1565 nm particle linear depolarization ratio agrees reasonably well with previous spectral analyses for similar aerosol types as seen in Table 2. For mineral dust depolarization 305 ratio, both decreasing and increasing trends with increasing wavelength have been observed previously (Table 2). This is the case for our observations at Limassol as well, though on average 1565 tends to be a little lower than 532 (Fig. 16). Probably, the spectral dependency of mineral dust depolarization ratio depends on both the age of the dust and the origin of the dust.
Wavelength-dependent changes in mineral dust depolarization ratio are small compared to elevated smoke layers, which can help to distinguish between these two aerosol types (Burton et al., 2012). For elevated smoke, a strong decrease in 310 depolarization ratio has been reported from > 0.20 at short wavelengths to < 0.05 at 1064 nm (Burton et al., 2015;Haarig et al., 2018;Hu et al., 2019). Thus, adding a depolarization ratio measurement at 1565 nm can provide added value to the commonly-used measurements at 355 nm and 532 nm wavelengths.
For marine aerosols, the depolarization ratio is small and has practically no spectral dependency (Gross et al., 2011), which is what we observed at Limassol. For the mixture of spruce and birch pollen at Vehmasmäki, the differences in 315 depolarization ratio at 532 nm and 1565 nm are small.

Conclusions
In this paper we report for the first time remote sensing measurements of atmospheric aerosol particle linear depolarization ratio at a wavelength of 1565 nm. Using observations at liquid cloud base we have been able to characterize the Halo Doppler lidar polarizer bleed-through with sufficient accuracy to obtain useful depolarization ratio measurements; 320 uncertainty in the bleed-through is propagated to the depolarization ratio measurement. A comparison of two different Halo Doppler lidar systems with two PollyXT systems during collocated measurements at Limassol, Cyprus, and Kuopio -Vehmasmäki, Finland, show good agreement between the lidar systems. The agreement between the instruments is remarkably good considering the large wavelength difference: the PollyXT depolarization ratio is retrieved at 355 nm and/or 532 nm. However, given the much lower laser energy in Halo Doppler lidars, it is not surprising that the vertical extent of 325 usable depolarization ratio is much lower than for PollyXT.
For relatively fresh mineral dust, we find particle linear depolarization ratios at 1565 nm ranging from 0.29 to 0.32, which is in good agreement with previous observations, including measurements at 710 nm and 1064 nm wavelengths (Freudenthaler et al., 2009;Gross et al., 2011;Burton et al., 2015;Haarig et al., 2017). For polluted marine aerosol we observed very low depolarization ratio of 0.009 at 1565 nm with a small decrease with increasing wavelength. Spruce and birch pollen 330 depolarization ratio has been characterized only recently at 532 nm (Bohlmann et al., 2019). Our measurements indicate a slightly higher depolarization ratio of 0.27 at 1565 nm compared to 0.24 at 532 nm. Overall, our results indicate that Halo Doppler lidars can add another wavelength at 1565 nm to studies on the spectral dependency of particle linear depolarization ratio, at least in the lowest 2-3 km above ground.