| 17 Aug 2026
Multi-year ground-based multi-wavelength Raman lidar aerosol dataset over the interior Tibetan Plateau (Yangbajing, 4284 m a.s.l.; 2021–2024)
Abstract. The interior of the Tibetan Plateau exerts a first-order influence on the Asian monsoon and regional climate, yet vertically resolved aerosol observations there are scarce, and no openly archived, multi-year, ground-based profile record has existed for this exceptionally clean high-altitude environment. We present such a dataset from a six-channel elastic/Raman/depolarisation lidar (355, 387, 408, 532, 607, 1064 nm) at Yangbajing (30.10° N, 90.52° E, 4284 m a.s.l.), covering September 2021 to December 2024: 437 quality-controlled nighttime profiles on a 75 m grid (157 levels, 0.30 to 12 km a.g.l.) and 279 daytime column-AOD retrievals. The independent products are the nitrogen-Raman extinction at 355 and 532 nm and the Raman-derived lidar ratio. The aerosol backscatter at 355 and 532 nm is retrieved by the elastic (Fernald) inversion with an assumed lidar ratio of 50 sr, and the 1064 nm backscatter is a non-independent colour-ratio transfer. Volume and particle depolarisation ratios (532 nm), water vapour mixing ratio (408/387 nm; radiosonde-calibrated C = 90.4 g kg-1, ~17 % uncertainty), Ångström exponents and column aerosol optical depth (AOD) complete the product set, and a scattering-ratio cloud screen supplemented by a particle-depolarisation ice test separates mineral dust from ice. The site is among the cleanest continental settings sampled by lidar: the nighttime detection-night median column AOD at 532 nm is 0.053 (n = 269) and the daytime detection median is 0.078 (n = 208). The Raman-derived site-mean lidar ratio is 51.6 sr (IQR 41 to 80 sr, n = 85), consistent in the median with the assumed 50 sr and with the CALIPSO clean-continental value, but it varies by a factor of two from night to night. This assumption, not photon noise (median relative σ ~1.7 % for the 532 nm extinction, rising to a 90th percentile of ~24 % in the weak-signal upper troposphere), dominates the column-AOD uncertainty (+60 %/-18 %), for which per-profile S = 41 and 80 sr bounds are provided. Daytime AOD tracks AERONET Nam Co (~85 km) in rank order (Pearson r = 0.61, Spearman ρ = 0.72, both p < 0.001, n = 41). To our knowledge this is the first openly archived, multi-year, ground-based multi-wavelength aerosol profile dataset from the clean interior Tibetan Plateau, a region pivotal for the Asian monsoon and the Third Pole cryosphere where high surface albedo and complex terrain defeat passive retrievals and leave spaceborne lidar the de facto reference. It supplies the vertically resolved ground truth needed to evaluate and correct spaceborne-lidar aerosol detection and typing, to benchmark next-generation missions such as EarthCARE over bright, high terrain, and to constrain reanalysis, model, and aerosol radiative studies of dust transport and monsoon aerosol over the Third Pole.
Review of manuscript entitled “Multi-year ground-based multi-wavelength Raman lidar aerosol dataset over the interior Tibetan Plateau (Yangbajing, 4284 m a.s.l.; 2021-2024)” by Yinan Wang et al.
The manuscript presents and documents an openly archived, multi-year, ground-based multi-wavelength lidar dataset from Yangbajing (30.10°N, 90.52°E, 4284 m a.s.l.) on the interior Tibetan Plateau, covering September 2021–December 2024. The archive contains 437 quality-controlled nighttime profiles (157 levels, 75 m grid, 0.30–12 km a.g.l.) and 279 daytime column-AOD retrievals, with products including elastic backscatter at 355/532/1064 nm, nitrogen-Raman extinction at 355/532 nm, volume and particle depolarisation at 532 nm, water-vapor mixing ratio, Ångström exponents, column AOD, and per-bin uncertainties and quality flags. The paper is framed as a data-description paper and includes instrument description, retrieval methodology, a detailed uncertainty budget, quality control, seasonal characterisation, and independent corroboration against AERONET Nam Co, CALIPSO, and Lhasa radiosondes.
It's acknowledged that huge efforts were made to deploy an advanced atmospheric lidar system at the Third Pole and to maintain the system under difficult logistical conditions. The dataset addresses a genuine and well-justified gap for the research of the Asian monsoon and the Third Pole cryosphere. It should also be mentioned that the authors disclose the current measurement capabilities as well as the problems for the system with many details, making the dataset not similar to those included in the EARLINET. The manuscript bears an honest delimitation of which products are independent (Raman extinction, lidar ratio) versus conditional (elastic backscatter, column AOD). The dataset is likely to be widely used.
However, a big impression is that the manuscript is somewhat too lengthy, with several parts of similar content being present more than once in different sections. Thus, it is suggested that the authors should reorganize the structure of the manuscript, in particular Sections 4, 5, and 6, to avoid repeating similar statements. Also, it would be helpful to polish the English language, since unusual expressions are sometimes shown in the text.
In addition, it would be hard to believe that you have crosstalk in the 607 nm channel, which is rather far from the corresponding elastic channel. What’s more, it is also problematic that the authors insist on providing the particle backscatter coefficient profiles at 1064 nm although they are not independently retrieved from the 1064 Mie backscatter signal, which makes limited sense to my understanding (also see my other major comments regarding this point below).
Therefore, the current form of the manuscript also requires a major revision before it can be considered for publication in ESSD. My comments below are intended to strengthen the documentation and remove several ambiguities that could lead to misuse.
Major comments
The authors are to be commended for the explicit treatment of the S = 50 sr assumption and for providing per-profile S = 41 and S = 80 sr AOD bounds. However, several points need clarification:
Sect. 3.2 bounds the residual aerosol at the reference altitude to 0.005–0.01 in AOD. This is a useful estimate, but it should be propagated into the archived AOD uncertainty (or at least flagged in the uncertainty_note global attribute) rather than left only in the text. Is it included in the per-bin σ fields? If not, users combining the S-bounds with the photon-noise σ may underestimate the total.
The authors justify a common S = 50 sr at 355 and 532 nm using 30 paired nights with a median 532-minus-355 difference of ~3 sr (IQR −7 to +21 sr, p = 0.16). This is a reasonable argument, but the sample is small, and the IQR is wide. I suggest reporting the 355 and 532 nm lidar ratios separately in Figure 6 or the text, or at least giving the reader the per-wavelength medians so that users can judge whether a common value is appropriate for their application.
Sect. 10.1 reports that the Ansmann et al. (1992) self-calibrated backscatter agrees with the elastic product within 30% on fewer than one night in five. This is an important result and supports the decision to archive the elastic product. However, the text says "the ratio of the two has a median of 1.48 but an interquartile range of 0.79 to 2.27", i.e., a median of 1.48 implies a systematic offset, not just noise. The authors should comment on whether this offset is understood (e.g., calibration of the Raman channel, overlap residual, or the known low-signal bias of the Raman backscatter method) and whether it implies a bias in the elastic product. If the Raman backscatter is systematically higher, this could indicate that the assumed S = 50 sr is biased low for some nights, which would be worth flagging. I also suggest the authors check the SNR at the reference height for Raman retrieval. Usually, it requires a higher standard to achieve a stable aerosol backscatter profile, and even higher when it comes to the short-wavelength band. The authors can do some perturbation analysis to see the effects of SNR at the reference height.
https://www.tropos.de/fileadmin/user_upload/Institut/Abteilungen/Fernerkundung/Daten_PDF/Bachelorarbeit_Nathan_Skupin.pdf
Minor comments
Line 10: Is it really a "six-channel" lidar? Should the 532 nm cross-channel be included?
Line 13: Insert "vertical" before "grid".
Line 27-29: similar content has already been provided in the first sentence of the paragraph.
Line 30: “ground truth” is overselling
Line 35-39: please add the scientific significance of conducting height-resolved aerosol observations over the plateau interior, especially considering the low aerosol loading there.
Line 45: add previous studies if possible.
Line 47: "to our knowledge, none has been released as an openly archived, multi-year, multi-wavelength record". It should be made clear what the scientific value is of having this record.
Line 53-54: It would be better to reorganize this sentence
Line 63: mention the wavelength for the lidar ratio
Line 68: “climate” -> “climatology”
Line 81 and 82: please specify the “northern deserts” and “local dust emissions” (what kinds?)
In section 2.2, it would be better to provide the pulse energy of the laser employed
Table 1: "Backscatter" -> "Elastic".
Line 147: "~90 km distant" -> "~90 km in distance".
Line 171: "Raman lidar ratio" is an awkward expression. Use "lidar ratio" instead.
Line 188-189: Why not derive the 1064 nm aerosol backscatter from the 1064 nm elastic signal?
In section 3.3: please discuss why this lidar system cannot provide an independent measurement of backscatter at 1064 nm. The solution for the backscatter coefficient at 1064 nm will bring large uncertainties due to the roughly assumed Ångström exponent. It would be even a bit better to apply any possible values from the AERONET measurements of the nearest station.
Line 231-232: it's suggested to add the statistics of the lidar ratio in the supplementary material, which is very informative for dataset users.
Line 242-246: This sentence is too long and should be divided.
Line 243-246: did you consider using the simultaneous and quasi-collocated CALIPSO measurements overflying to calibrate the gain ratio for the depolarization ratio.
Line 247-248: it's suggested to add some details to explain "high-voltage-versus-gain LUT for each photomultiplier".
Line 290-296: This C is an average value for different nights. Did you make any changes to the lidar system during these several years’ measurements, which may influence the values of C. Is it in general stable over long-term operation?
Line 297: do you mean that the overlap functions at low heights are the same for both 408-nm and 387-nm channels?
Line 300: please explain “bin-level availability”
Line 305: insert "ratio" after "depolarisation".
Line 308: "By day" -> "At daytime".
Line 311: It's not clear how "a near-surface reference" is determined, considering that there are always non-negligible aerosols from the surface.
Line 323-324: "volume linear depolarisation ratio" should be used consistently in the manuscript.
Line 327: what is the value of the molecular depolarization ratio applied in the lidar system?
Line 341: "aerosol night" -> "nighttime aerosol profile".
Line 349: "molecular temperature profile" -> "US Standard Atmosphere 1976".
Line 368: "spherical layer (a liquid water cloud)" -> "liquid water cloud".
Line 380-381: remove ", rather than through telescope chromatic". This is somehow misleading, as the overlap difference was not only caused by telescope chromatic aberration.
Line 404-407 and 597-600: it is strange that you have crosstalk between 532 nm and 607 nm, which are rather far from each other in the optical spectrum. And this is not even the case for 355 nm and 387 nm. Are you sure that this problem is caused by the interference filter centered at 607 nm? Because of this issue, you lost almost 25% of the valid data as seen in Fig. 4a, which is really a pity, especially considering that you collected the data from a location difficult to reach.
Line 418-421: It should be noted that Ångström exponents for some dust layers can be negative; see Veselovskii et al. (2020)
Veselovskii, I., Hu, Q., Goloub, P., Podvin, T., Korenskiy, M., Derimian, Y., Legrand, M., and Castellanos, P.: Variability in lidar-derived particle properties over West Africa due to changes in absorption: towards an understanding, Atmos. Chem. Phys., 20, 6563–6581, https://doi.org/10.5194/acp-20-6563-2020, 2020.
Line 434-439: the contents of this paragraph have already been provided in previous sections.
Line 488-491: the contents of this paragraph have already been provided in previous sections.
Line 526-528: it would be better to have some references here
Figure 6c: use British spelling for “depolarization” for the sake of consistency
Section 6.2, distinguishing dust and ice clouds can also involve the backscatter ratio together with the particle linear depolarization ratio. In addition, clouds can only show up for a while rather than the entire night; thus, is it problematic to use night-average profiles for identifying dust nights and ice-carry nights?
Line 575-585: the contents of this paragraph have already been provided in previous sections (similarly, section 7.3).
Line 627: "near-overpass" -> "closest".
Line 688-691: The contents of this paragraph have already been provided in previous sections.
Section 7.4: This section is redundant with previous sections (sections 5.4, 6.3, and 9). I highly suggest that the authors go through the manuscript to remove the replicated content. The same applies to section 8.3.
Line 744: "Users should cite both this paper and the Zenodo dataset DOI", I think this claim should be more relevant to the dataset.
Line 784: insert "range" before gate.
Line 795: "The dataset provides the vertically resolved aerosol ground reference that is missing for the plateau interior".
Line 800: remove "forthcoming".
Table 4: it's suggested to add the China Aerosol Raman Lidar Network (CARLNET) for comparison; see Shao et al. (2025) and Bu et al. (2026). The dataset is only accessible within the China Meteorological Administration.
In the reference, there are many incorrect characters that should be checked. To name a few, "Munoz-Porcar" -> "Muñoz-Porcar"; "Muller" -> "Müller".
Line 847: “staff” -> “staffs”