In this work, we analyze the performance of the ESA profiling products derived for ATLID, the high-spectral-resolution and polarization lidar operating at 355 nm. We focus on the geophysical quantities obtained from ATLID only, i.e., Level 2A products. Synergistic products, which also use other instruments of EarthCARE, are in the Level 2B product suite but are not in the focus of this work. To allow for synergistic products, all Level 2A products from the different instruments are calculated on a Joint Standard Grid with its respective product (X-JSG).

Notable changes in the ATLID product versions, partly based on updated calibration procedures, have taken place over the entire active mission lifetime and have influenced the product quality. Different product versions are called baselines and are named in 2 capital letters for the EarthCARE mission. It started with Baseline “AA” and when minor changes were included, the second letter was changed (e.g., to AB, AC, AD...). Major updates and software versions specifically aimed for data reprocessing of the entire database result in a change of the first letter and the reset of the second letter starting from “A” (e.g., BA, CA, DA...). Level 2A data, i.e., single-sensor data, have been publicly released in March 2025 with Baseline AE, while it was earlier available to official Cal/Val teams (Baseline AC was provided to them already in December 2024). For the first reprocessing of the entire data set with all algorithm updates since the start of the measurements, Baseline BA was rolled out in July 2025. This Baseline was used to reprocess the data set, so that one common set of data products is available from August 2024 to September 2025 to be comparable to each other. An overview of the baselines deployed for the ATLID profiling products which were available for validation teams is given in Fig. .

A detailed description of what has changed between the different baselines can be found in the “EarthCARE product data handbook” at the disclaimer of each product .

Aerosol optical profiles, i.e., particle backscatter coefficient, particle extinction coefficient, and particle linear depolarization ratio from ATLID are calculated by A-PRO ATLID-PROfile, in the ESA product chain for EarthCARE. A-PRO provides different products (described below). These products depend not only on Level 1 data of ATLID (A-NOM, i.e., the attenuated backscatter coefficient profiles for the Rayleigh signal and the co-& cross-polar Mie signals), meteorological data (X-MET product) and the joint standard grid (X-JSG) but also on the outcome of the feature mask product, A-FM , to avoid averaging over strong and weak features together. While a full description of these product algorithms is beyond the scope of this work, we still want to summarize the main features and differences of the two profiling products in the following.

In the first profiling product, A-AER (ATLID – Aerosol), the aforementioned optical particle profiles are retrieved via the classical HSRL method. Guided by A-FM and other considerations (e.g. scattering ratio), a hybrid smoothing is performed on the three signals: Strong scattering features (like clouds) are smoothed less horizontally than weak scattering features (like aerosols). Based on the output of this procedure, a classical HSRL approach is applied along-track column-by-column to the horizontally smoothed signals providing fields of particle extinction and backscatter coefficient and particle lidar and depolarization ratio for aerosol and clouds. A simple multiple scattering (MS) correction is also applied. The two-dimensional fields of these products are then stored in the A-AER product file on X-JSG resolution, but one should remember that clouds are much less smoothed than aerosols regions. The main purpose of the A-AER product is to provide a priori information for A-EBD.

The second profiling product is A-EBD (ATLID Extinction, Backscatter, and Depolarization) and uses the same input as A-AER, i.e., A-NOM, X-MET, X-JSG, A-FM, but also the output of A-AER (including an initial target categorization). A-EBD applies an advanced optimal estimation (OE) scheme to optimize the extinction coefficient and the lidar ratio (and other variables) to fit the high-resolution observations (i.e. 1-km horizontal scale). The A-AER output, i.e., the lidar ratio, extinction but also the A-AER target classification are used as a starting point for the OE scheme (if available, otherwise default values are used). The geophysical products (backscatter, extinction, lidar and depolarization ratio of aerosol and clouds) are finally delivered at 3 different resolutions (low, medium, and high). The different resolution is, however, only applied to “weak” scattering features, e.g., aerosol targets, while strong scattering features are always supplied at high resolution. This means, e.g., that the low-resolution output of A-EBD is on X-JSG grid, but aerosol features are horizontally smoothed to 100 km while cloud features remain at high resolution (1 km). A multiple scattering correction is applied within the forward model of the OE scheme and is more accurate than the one applied in A-AER. Please note that within the A-EBD product, the lidar ratio is provided layer-wise and not in the same vertical resolution as the backscatter and extinction products. A-EBD also provides the final target categorization (A-TC) at the three different resolutions, but it is foreseen to provide only one resolution of this product in future. For most applications, A-EBD products are more accurate than the corresponding A-AER products and should be prioritized. We provided here only a brief description needed for the interpretation of the validation results, a detailed explanation of these products is given in .

In this paper, we focus on the analysis of the ATLID profiling products (i.e., A-AER, A-EBD) from Baselines B (i.e., BA and BB), but, whenever possible (e.g., in Sect. ), we also show the changes introduced due to updated calibrations and processors with respect to earlier baselines, partly pre-operational and thus not provided to the public. All operational ATLID-only L2A processors A-FM, A-PRO, A- LAY, and their respective products have benefited from early Cal/Val activities and algorithm updates e.g., . These activities will improve the synergistic L2B processors (and their products) as well, such as the synergistic products AM-COL using both, ATLID and MSI, or the ACM-CAP making use of ATLID, MSI, and CPR, until the final radiation closure processors ACMB-DF using all 4 instruments onboard EarthCARE. The algorithm updates include improved corrections (at levels 1 and 2) for radiation noise effects, a hot pixel correction, the removal of the 20-km spike feature, and improved depolarization ratio calculations, among many others . Here, we focus on the performance of ATLID L2A profiling products of Baseline BA (BB) as is, without discussing the corrections themselves.

As ground-truth we utilize the measurements of the well-established PollyXT lidar systems multiwavelength-Raman-polarization lidars, using tailored data products of the PollyNET processing chain . The lidar systems are part of the European research infrastructure for aerosol and clouds ACTRIS , and follow the quality assurance procedures of this infrastructure originally implemented in the framework of the European Research Lidar Network EARLINET . To be ready for EarthCARE validation, these lidar systems took part in the preparation activities in the framework of an ATMO-ACCESS pilot project and proved their readiness already before the EarthCARE launch . While the capabilities of the different systems slightly differ , and are continuously improved over time , all of the used PollyXT lidar systems operate at 3 wavelengths (355, 532, 1064 nm), have polarization and Raman capabilities, and a dedicated near-range telescope to also cover the lowermost atmosphere. As all systems are part of ACTRIS/EARLINET, they follow the standard operation and quality assurance procedures, e.g., telecover test and routine Δ90° calibrations for polarization observations . Thus, the systems provide independent reference profiles of the particle backscatter coefficient, particle extinction coefficient, and particle depolarization ratio at (but not only) 355 nm – ATLID's operating wavelength.

For the validation work, we utilize observations from PollyNET lidars at five different locations listed in Table  and shown in Fig.  to obtain a broad geographical coverage.

TROPOS operates permanent PollyNET lidar sites west of continental Africa at Mindelo (Cabo Verde), in Europe at Leipzig (Germany), and in Central Asia at Dushanbe (Tajikistan). Furthermore, we utilize measurements from two ACTRIS exploratory platforms hosting a PollyNET lidar, namely OCEANET and LACROS. OCEANET was operated on board the German research vessel Meteor in the Tropical Atlantic during voyage M207 in January and February 2025. LACROS started operating in Invergargill (Aotearoa New Zealand) in August 2025 as part of the goSouth-2 campaign . A brief description of these sites is given in the following sections.

The first case study to be discussed targets the eastern Tropical Atlantic on 1 June 2025. Figure  shows an overview of the EarthCARE A-EBD products in this region. The “frame” (the name for the time-height section covered by EarthCARE) used is 05729A (i.e., the first segment of orbit 5729) and actually extends from 22.5° S, 17.2° W to 22.5° N, 25.9° W, covering 5064 km. However, only the region between 14.1 and 19.7° N is shown in Fig.  to focus on the atmospheric conditions around the ground-reference site Mindelo, which is indicated on the maps as well as in the time-height plots of the ATLID products by a red vertical line.

As seen on the zoomed-in map in Fig. , left, EarthCARE was very close to the Mindelo site, with a minimum distance of around 20 km. The particle backscatter coefficient, particle extinction coefficient, particle lidar ratio, and particle depolarization ratio (from top to bottom, respectively) measured by ATLID are shown in the right-side panels of Fig. . A pronounced dust layer (strong backscatter and extinction values indicated by yellow and reddish colors) at an altitude range from 1 to 6 km is visible in these products for the entire geographical region. The dust is characterized by a lidar ratio between 40 and 60 sr and an elevated particle depolarization ratio >0.2, in agreement with many previous studies on the intensive properties of Saharan mineral dust e.g.,. Below the Saharan dust layer (or Saharan Air Layer – SAL), the marine boundary layer (BL) is visible, which was partly cloud-topped (white colors in the backscatter product). The marine boundary layer is usually characterized by lidar ratios below 30 sr and low depolarization ratio , which is in qualitative agreement with the lower lidar ratio values (bluish colors) measured by ATLID in this geometrically thin layer. Ice clouds are visible above 12 km, extending to 17.5° N. Due to their high variability, these clouds are not considered for comparison with the ground-site observations.

In agreement with the EarthCARE observation, stable atmospheric conditions were observed at Mindelo below the cirrus level throughout the day, as indicated by the combined target classification a combination of elastic only and elastic + Raman channels,, shown in Fig. . The marine boundary layer, consisting of large, spherical particles, was present the whole day up to a height of about 1.2 km above ground (sea). On top of the marine BL, liquid clouds occasionally formed (bluish colors, e.g., at around 01:00 UTC). The dust layer, which was observed by EarthCARE (around 03:30 UTC in Fig. ), was omnipresent the whole day over Mindelo, but faded out towards the end of the day, starting at 15:00 UTC. Broken high-altitude clouds (cirrus – indicated by greenish colors) were also observed throughout the day as partly seen by EarthCARE. Considering the temporal evolution of the atmospheric conditions over Mindelo as seen by ground-based lidar, together with the observed horizontal distribution seen by EarthCARE, we are extremely confident that the atmospheric conditions below cirrus level at Mindelo during the time of overpass have been representative for the EarthCARE aerosol profiles, taken with a minimum distance of 20 km away, even when considering the 100 km horizontal smoothing of the A-EBD low-resolution product. Such conditions make this case study ideal for evaluating the performance of ATLID L2A products, as representativeness issues, which usually exist in the validation of space-borne profiles , can be neglected for this case (and most of the following ones).

Figure  shows an overview of the ground-based profiles considered for the validation of EarthCARE. As PollyXT is a multiwavelength-Raman-polarization lidar, optical aerosol profiles can be acquired at several wavelengths. Even though in the following we will concentrate on the ATLID wavelength of 355 nm only, we always evaluate the full spectral range available from the ground-based measurements to quality check the lidar performance itself, perform extended aerosol classification using the spectral dependence as additional information , and gather the intensive properties at all possible wavelengths to extend the collection of intensive properties for several aerosol types DeLiAn,. PollyXT has near-range capabilities using a second telescope. For the analysis of close-to-ground profiles, like the marine boundary layer in this case, we use the near-range profiles, while for the lofted aerosol layers or clouds, the far-range profiles are considered.

The observations shown in Fig.  for Mindelo on 1 June 2025 show the typical signature of a SAL above the marine BL. Wavelength-neutral backscatter and extinction (at 355 and 532 nm) values were observed for both aerosol layers of interest, while the 1064 nm backscatter is slightly lower (only about 80 % of the 355 and 532 nm backscatter) in the dust layer. Marine aerosol and dust are characterized by large particles, which introduce little to no wavelength dependence. The marine and dust particles can be instead clearly separated by the depolarization ratio, which is close to 0 for marine particles and above 0.2 for dust particles (see the second panel to the right in Fig. ). The lidar ratio values of about 60 sr at 355 nm in the SAL reveal slightly higher values compared to what is expected for pure dust, indicating slight pollution in this layer. In the marine boundary layer, lidar ratio values of around 20 sr were observed, thus indicating pristine conditions at Mindelo. All in all, the derived optical properties provide a consistent picture of the atmospheric conditions on this day, with a clean marine boundary layer below 1 km and a partly polluted dust layer above, extending up to 5 km.

Figure  shows the direct comparison of particle backscatter and extinction coefficients, lidar ratio, and particle depolarization ratio between the ground-based reference and the respective EarthCARE A-EBD low-resolution profiles. Uncertainties are shown as shaded areas. An excellent agreement is found for the particle backscatter and extinction coefficient for the height region of 1.3 to 6 km, i.e., within the SAL. Within the marine BL (below 1 km height), these extensive quantities deviate slightly, most probably due to local differences in the lowermost atmosphere (e.g., cloud contamination).

Focusing on the aerosol intensive quantities, namely the particle lidar and depolarization ratio, one sees an excellent agreement of the lidar ratio in the SAL. Within the marine BL however, ATLID overestimates the lidar ratio with values of about 38 sr, which is clearly too high for pure marine particles . As shown in Fig. , marine conditions are expected at both measurement locations: at Mindelo, but also at the EarthCARE ground-track location, being 20 km east, directly above the ocean.

Regarding the particle depolarization ratio, excellent agreement is found in the height range of maximum dust load at 3.5 km altitude with values of 0.24. However, at regions with less dust load (lower backscatter) above and below 3.5 km, the retrieved values deviate from the PollyXT profiles. Generally, A-EBD provides higher values (maximum of 0.4 at 2 km) than measured with the ground-based reference. Within the marine BL, A-EBD reports a depolarization value of 0.06, while PollyNET values are about 0.02.

Besides the A-EBD low-resolution products, there are three more products providing vertical profiles of aerosol properties from ATLID only, namely A-AER and A-EBD medium & high resolution. All of these products are shown together for illustration in Fig. . Here it becomes obvious that for these stable atmospheric conditions in the lofted aerosol layer, all products provide reasonable results besides the issues already discussed.

However, as expected, the A-EBD high-resolution product is the noisiest product, having a large uncertainty and deviating partly from the reference values, but all still within the provided uncertainties. The A-EBD medium-resolution product becomes noisier when getting closer to the ground due to a decreasing signal-to-noise ratio while penetrating the dust layer. The lidar ratio values above 2.5 km, however, are interestingly accurate for all products (even for high resolution). Likely, this is an effect caused by the usage of the A-AER (lidar ratio) product as a priori for the optimal estimation in A-EBD. Below 2.5 km, the lidar ratios of the high and medium-resolution products start deviating and show values already close to marine conditions (about 25 sr) around 1 km a.s.l. However, in the marine BL itself, the values are too high for all products as discussed previously.

Concerning the particle depolarization ratio in the SAL, it becomes obvious that the high-resolution product is not reliable for this parameter, with depolarization ratio values much lower than the reference and being noisier at the same time. The tendency of the medium-resolution product to get worse close to the ground is also evident in the particle depolarization ratio profile, while above 2.5 km, the values of medium- and low-resolution A-EBD and A-AER are similar. Nevertheless, the general and most obvious issues with respect to the particle depolarization ratio for weakly scattering targets as discussed above (significant deviation from the reference) remain for all four products and need to be tackled in new baseline versions in the future.

In summary, we can conclude that the A-AER and A-EBD low-resolution products perform well within the Saharan dust layer, except for overestimating the depolarization ratio. Efforts to improve the polarization ratio have been made since the beginning of the mission and are discussed further below (Sect. ) and in a dedicated work , but still leave room for improvement. But generally, the excellent high-spectral-resolution lidar products, including the lidar ratio in the lofted aerosol layer, can be seen as a proof of concept for ATLID HSR profiling capabilities.

A second case study to be discussed is covering Frame 01508B on 2 September 2024. The ground-reference measurements were taken in Dushanbe, Tajikistan, in the eastern part of the dust belt (see Fig. ), and thus in a completely different geographical region than the first case study.

Figure  shows the corresponding products of ATLID (A-EBD low resolution) together with the local topography from 35.4 to 41.6° N, centered around the overpass over Dushanbe at 21:04:35 UTC (nighttime). Minimum distance to the ground site was as low as 17 km. A remarkable dust layer trapped between the mountain ridges is seen in all four ATLID products. The dust layer is characterized by extinction values of about 50 Mm⁻¹ (greenish/yellowish colors) to 200 Mm⁻¹ (orange colors), with lidar ratios around 40 to 50 sr (green colors), and depolarization ratios of 0.15 to 0.3 (bluish colors). Occasionally, ice clouds are indicated by high depolarization ratios (red) at the top of the dust layer, e.g., around 37.8° N, but could also be an artifact as result of edge effects (discussed below). Interestingly, the top of the highest mountain ridges (above 4 km: Hindu Kush in Afghanistan at around 35.4° N and Zarafshan ridge at 39.1° N) seem to coincide with the top of the dust layer trapped between these ridges.

Atmospheric conditions over Dushanbe, as observed with the ground-based PollyNET lidar, were characterized by intense dust layers persistent for several hours to days. The derived optical profiles (geophysical parameters) from the PollyXT around the time of the EarthCARE overpass (2 September, 21:00 UTC, Fig. ) reveal particle extinction coefficient values between 50 and 200 Mm⁻¹ and thus indicate a medium aerosol load for this case study. The intensive optical properties, i.e., lidar ratio and depolarization ratio at all wavelengths (not shown), clearly indicate Central Asian dust as the main aerosol type in the lofted layer. Below 400 m above ground level (a.g.l. – 864 m offset to above sea level), a polluted planetary boundary layer was found, which will not be compared to EarthCARE products due to its shallowness and local occurrence.

The direct comparison between the ground-based lidar observation and the closest EarthCARE A-EBD product at low resolution is shown in Fig. . Focusing on the main dust layer above 1.2 km altitude above sea level (a.s.l.) to 3.5 km a.s.l., an excellent agreement is seen for the backscatter coefficient between ATLID A-EBD and the Raman retrieval of PollyXT. From 3.4 to 4.5 km, slight deviations become evident, which might be related to local heterogeneity in these uppermost layers. For the extinction, a good agreement is also found, but with some more fluctuations. Instead, the lidar ratio of A-EBD low resolution matches perfectly within the uncertainties of the two measurements. Both instruments measure lidar ratio values between 45 and 50 sr at 355 nm in this dust layer. Please remember that the lidar ratio of ATLID is provided layer-wise and not within the same resolution as the backscatter and extinction products, thus, larger steps are visible compared to the backscatter and extinction profiles of ATLID. For the depolarization ratio, the agreement is fair, but a systematic discrepancy characterized by a slight overestimation by ATLID is obvious. While the ground-based lidar measured depolarization ratios of 0.17 to 0.21, indicating slightly polluted dust, ATLID observed values of around 0.18–0.23, but always higher than the reference.

The early date of the case study in EarthCARE's lifetime also allows us to study the improvements/changes that have been implemented in the processing chain since the start of the mission. Figure  shows the comparison for the same atmospheric scene, but as processed with the operational Baseline AC (at the very beginning of the mission) and compared to the reprocessed Baseline BA, which is the focus of this paper. A-AER (top panels) and A-EBD low-resolution products (bottom panels) are shown. From Baseline AC to BA, several improvements/modifications have been implemented in the processing chain to account for, e.g., hot/cold pixels, background correction issues, biases due to improper vertical smoothing (in A-AER), and many more. The explanation of all these developments is beyond the scope of this work, which will focus on the results of the improvements only. Further details have been presented in dedicated publications, for example, improvements in the Level 1 data quality in general and in the depolarization ratio . With Fig. , we want to illustrate the overall effect of all these changes and to highlight the ongoing efforts for better data quality.

The most prominent change in EarthCARE data from Baseline AC to BA is observed with respect to the particle depolarization ratio for both products, A-AER and A-EBD. Here, too low values were reported in the beginning of the active mission . By introducing an improved background correction of the cross-polarized channel with Baseline AD and inserting the transmission coefficients measured in the pre-launch on-ground characterization with Baseline AE, the depolarization ratio has significantly improved. Now, the particle depolarization ratios are close to the ground truth, mostly overlapping within the measurement uncertainties of both systems. Other changes can be seen in the backscatter and extinction profiles, which result from several bug fixes (e.g., issues related to the screening of clouds when producing the medium and low-resolution optical property outputs). The values of backscatter and extinction have significantly increased (compared to Baseline AC) for the dust layer in A-AER and below 3 km altitude in A-EBD low resolution and are now (Baseline BA) much closer to the ground-truth. The lidar ratio values instead, have only changed slightly but mostly increased. The A-AER and A-EBD lidar ratio values have been already in the valid range (compared to the ground-truth) in Baseline AC. Now, the increase of the lidar ratio on top of the dust plume as identified with the ground-based PollyXT reference observations can be also reproduced with the A-EBD BA low-resolution product.

Concluding on this case of dust in Central Asia, we generally see an excellent agreement and an overall improvement of the ATLID products from Baseline AC to BA. The profiles in BA are now close to the ground truth, overlapping within the uncertainties, and thus give trust that the retrievals work well, at least for this specific case.

The two cases discussed previously were measured in regions of the dust belt, where mineral dust plays the dominant role in the atmosphere. Now, we would like to discuss a case from Central Europe, namely Leipzig, Germany, to investigate the ATLID performance for a mix of dust with other aerosols, which occurred on 26 June 2025. The respective EarthCARE curtains are shown in Fig. . The atmospheric scene closest to Leipzig is again indicated by a red vertical line. On this day, an aerosol layer above the local PBL up to 6 km a.s.l. was observed by EarthCARE, coming from the Ore mountains in the south and flying over Leipzig towards northern Germany. The aerosol-layer-top height slightly decreased towards the north and the lofted layer itself obviously vanished around 52° N, thus 80 km north of Leipzig when only the local PBL remained. Backward trajectories and particle tracing simulations not shown indicate a clear pathway of air masses from the Saharan desert via the Iberian Peninsula towards the Atlantic and the North Sea before entering the continent towards Leipzig. Therefore, an uptake of marine aerosol and continental pollution, in addition to the mineral dust, is very likely. In the respective A-EBD products, some oscillations are seen (most prominent in the extinction plot, second panel in Fig. ), which are not gravity waves nor other atmospheric features, but caused by a bug related to the joint standard grid of the EarthCARE processing chain, which shifts the height array from time to time by one bin. Nevertheless, the optical properties are not affected by this issue and thus do not hinder their analysis.

Figure  shows the comparison of the closest ATLID profiles (7.5 km away, A-EBD low resolution in green and A-AER in blue) and the ground-based optical profiles retrieved from Polly. A nearly perfect agreement for this case in terms of the particle backscatter coefficient can be found between 1 and 5 km for the A-EBD low-resolution product. The extinction coefficient, however, appears to be overestimated by A-EBD except for the height region close to the ground (<1 km) and between 2.5 to 3 km. This is also reflected in the lidar ratio profiles, which are generally higher than the ground-truth (around 50 to 75 sr for ATLID compared to 30 to 50 sr for the ground reference) in the pronounced aerosol regions. The higher lidar ratio values are caused by the A-AER results (shown in Fig.  as blue line) which serve as input to the optimal estimation in A-EBD. Dedicated investigations performed for this study in the framework of DISC (EarthCARE Data Innovation & Science Cluster) have shown that one erroneous A-NOM profile in the smoothing procedure of A-AER has led to unrealistic error estimates within the lidar ratio retrieval allowing for too high lidar ratio values. In addition, a very complex atmospheric layering with a strong spatial inhomogeneity occurred within the 100 km around Leipzig. As nicely seen in Fig. , a dust layer (characterized by high depolarization ratio values) was situated around Leipzig and was partly on top of the existing near-surface layer of continental/marine aerosol (low depolarization ratio and moderate lidar ratio). During the time of the overpass, the northernmost edge of this dust layer was just 50 km away from Leipzig. But one has to remember that the horizontal averaging of the A-AER products is adaptive and can be as large as 100 km and, therefore, also includes regions more far away from the ground-based reference site than its closest distance. For these reasons, it is very likely that the complex layering together with erroneous profile reported previously, might have disturbed the A-AER retrieval and finally caused the overestimated lidar ratio values. Currently, studies on the root cause of the one erroneous profile within the 100 km around Leipzig are going on as well as investigations on how to flag such invalid profiles properly in future. The main take-home message with respect to the extinction and lidar ratio profiles is that, even though EarthCARE data is public, the data products of Baseline BA are not yet 100 % quality assured and occasional errors in the retrieval can occur which are not yet properly flagged.

Besides the issue of the extinction and lidar ratio, for the particle depolarization ratio, we see an amazing agreement between the two lidars. The almost linear increase from ground (more pollution than dust) to higher values at the top of the aerosol layer (highest values of around 0.2 for the Polly observations indicating a mixture of dust with other aerosols) is also shown by the EarthCARE products. However, the increasing trend continues with height in A-EBD which cannot be confirmed by the ground-truth instrument. As discussed above, the air mass transport analysis indicated a clear transport pathway from the western Sahara towards Leipzig with a high probability of mixing with other aerosols, and thus confirming the findings of the ground-based reference. But even in the case of pure dust, values of the depolarization ratio of almost 0.40, as measured between 4 and 5 km by EarthCARE, are clearly overestimated in this atmospheric region with low scattering (low aerosol load) as dust depolarization ratios at 355 nm are usually well below 0.30 . Thus, obviously, the depolarization measurements of EarthCARE are not yet fully reliable in such aerosol regimes with low backscatter signal.

Another case of interest from Leipzig for the validation is 5 July 2025. Here, the EarthCARE overpass occurred about 75 km away from TROPOS. However, as seen in the cross sections in Fig. , a spatially very homogeneous aerosol layer was measured below 3 km around the overpass, allowing us to use this case for validation as well. The multiwavelength lidar observations at Leipzig indicate the presence of smoke from fires in North America throughout the troposphere which is also confirmed by backward trajectory and model analysis e.g.,. Above the tropospheric aerosol layers, stratospheric smoke was observed by ATLID at around 11 km indicated by enhanced depolarization values and high lidar ratio values (>60 sr). The presence of smoke at this altitude was also confirmed by observations with the fluorescence lidar MARTHA at Leipzig showing clearly enhanced fluorescence capacity values as typically for smoke . Interestingly, it seems to be that ice formation is triggered at the bottom of the smoke layer in some regions indicated by the very low lidar ratio. Nevertheless, to not lose focus, we concentrate on the tropospheric features below 6 km in the following.

The comparison of the ground-based optical profiles at 355 nm and the respective EarthCARE profiles are shown in Fig. .

Despite the distance of the two observations, a very good agreement is found for the extensive quantity particle backscatter coefficient between 1 and 3 km altitude. Below, variations occur which can be explained with spatial inhomogeneity. The extinction coefficient is slightly overestimated below 2 km. The lidar ratio profile is in fair agreement (within the uncertainties) from 1.8 to 2.8 km, but overestimated from 0.8 to 1.8 km. Below 800 m, the A-EBD lidar ratio converges towards values below 30 sr, which cannot be confirmed by the ground-based PollyXT with its near-range capabilities (observing values above 40 sr in the BL below 1 km). However, considering the distance between the two observations of 75 km, not too many conclusions should be drawn for the comparison within the surface-near first 800 m, that are likely partially influenced by local planetary boundary layer processes in the summertime continental environment. Nevertheless, it still seems a bit suspicious that lidar ratio values close to clean marine conditions are provided by ATLID in this low altitude range. In contrast to the cases before, the aerosol conditions were dominated by spherical particles (continental pollution/smoke). Therefore, only low depolarization ratio values were observed by both instruments as indicated in the right panel of Fig. . Here, we see that the depolarization values of the reference measurements of PollyXT are lower compared to the values provided by ATLID. Between 1 and 2 km height, the mean particle depolarization ratio of PollyXT is 0.06–0.07, while A-EBD provides values of around 0.10. On the top of the aerosol layer at around 2.7 km, A-EBD provides significantly too high depolarization ratio values which are most probably a result of edge effects (i.e., smoothing over the sharp boundary between the aerosol layer and the cleaner atmosphere above). However, also the given error for this value is very high so that the two profiles agree within their uncertainty range. Nevertheless, ATLID's profiling product should obviously not be used when such a large error range is given, spanning basically all possible values (in the case of the depolarization ratio from 0 to 1). Potentially, data should therefore be flagged as invalid in future.

We can conclude from this case that for aerosol conditions with low particle depolarization, EarthCARE's depolarization ratio products still need to be improved even though the absolute deviations are comparably low (when the respective uncertainties are low).

In addition to the observations made on land, we also had the opportunity for four direct overpasses during a research voyage with the RV Meteor on the Tropical Atlantic Ocean in Jan/Feb 2025 see Fig.  and. While for two of these overpasses, atmospheric conditions were inappropriate for a direct comparison, we can utilize the rare opportunity for validating ATLID above the Tropical Atlantic for the other 2 cases.

To contrast the previous case studies, which were characterized by significant aerosol load, we focus in the last case study on the pristine environment in the Southern Hemisphere, namely one observation from the goSouth-2 campaign in Southern Aotearoa New Zealand on 24 October 2025. As a side effect, we cover a completely different geographic region, which is usually under-sampled with respect to ground-based profiling observations and thus of high interest for the validation of EarthCARE. During the night of interest, atmospheric conditions at the measurement site in Invercargill/Waihōpai were entirely pristine. The AOD was as low as 0.03 (last measurement of AERONET photometer on the afternoon before) as shown in Fig. , top. However, the atmosphere in the lowermost 2 km was quite turbulent as the time-height plot of the PollyXT attenuated backscatter coefficient at 1064 nm reveals (Fig. , bottom).

For this reason, and because EarthCARE measurements were as close as 2.5 km to the ground-site, we do not use the A-EBD low-resolution products for comparison as before, but aim for the medium and high-resolution products. The respective geophysical quantities (backscatter and extinction coefficient, lidar and depolarization ratio) at both resolutions are shown in Fig. . The baseline provided is BB, the successor of BA.

The medium-resolution products (left panels) show a consistent but horizontally heterogeneous structure. While the high-resolution products (right panels) seem to be dominated by noise, the general atmospheric state is still identifiable. Interestingly, the lidar ratio, which is used as a priori input for the optimal estimation scheme and finally provided layer-wise, shows the least scatter and is almost identical to the medium-resolution product. The depolarization ratio product, however, is most dominated by noise (as discussed above, but here during nighttime) and obviously hardly usable in the high-resolution product for this case. Clouds around the overpass time are visible at both resolutions (remember that strong features are not smoothed in the A-EBD product) and reveal the occasional cloud formation at the top of the aerosol layer (up to 2 km), which is also visible in the PollyNET time-height cross section (see Fig. ). The direct comparison between the ground-based reference and the EarthCARE A-EBD products is shown in Fig.  for the medium and high resolutions. The low-resolution products are not shown in this Figure as well because they are not representative for near-surface atmospheric conditions due to the respective strong smoothing combined with the existing high atmospheric variability. While the high-resolution product is, of course, noisier, it shows an excellent agreement for the backscatter and extinction products, being even superior to the medium-resolution product, which shows higher values in the height region of 0.8 to 1.8 km compared to the PollyNET lidar products. Because of the strong heterogeneity of the atmospheric scene as observed by both instruments, this can be easily attributed to representativeness issues with respect to the medium-resolution products. Therefore, besides being noisy, the high-resolution products can be used and are reliable in highly dynamic regimes with low aerosol backscatter and extinction. Interestingly, the lidar ratio retrieved is as low as 10 to 13 sr for the lowermost 800 m, which is clearly a bit too low as such values have so far never been observed see e.g., and are also revealed by a slight underestimation of the extinction in this height range. The depolarization ratio products, however, suffer stronger from noise because of the instrument design of ATLID as discussed above the Mie cross-polar channel is prone to more background noise, see. The values of the high-resolution product agree within the uncertainties with PollyXT, but basically, the EarthCARE depolarization ratio is not usable at high resolution. For the medium resolution, it is even worse, because here the depolarization ratio strongly deviates from the ground-based reference and also shows a significant overestimation even when considering the uncertainties. The too high depolarization ratio values are most probably linked to inconsistencies in the spectral cross-talk correction. Also, the faint structures of the depolarization ratio in the clean marine BL as retrieved by EarthCARE (Fig. , medium resolution) might be a result of this issue, as they are not observed with the ground-based reference. Thus, it makes these ATLID depolarization ratio observations questionable.

Finally, we conclude that A-EBD backscatter and extinction products in the clean Southern Hemisphere are of excellent quality with respect to backscatter and extinction and therefore show the excellent potential of EarthCARE to study aerosol-cloud interactions in unperturbed regimes. The lidar ratios agree well except for the lowermost part of the atmosphere (below 1 km), where the values are too low and thus need some improvement.