| 25 Aug 2026
Online zero-Doppler reference tracking and wind speed correction for an iodine-cell Rayleigh Doppler lidar: method and validation
Abstract. In an iodine-cell-based Rayleigh Doppler lidar (RDLD), slow drift in the relative spectral position between the transmitted laser frequency and the discriminator response can shift the zero-Doppler reference and introduce systematic wind errors. This study proposes and validates an online zero-Doppler reference tracking and wind correction method based on time-division multiplexed measurements of seed reference light. The seed reference light and atmospheric backscatter share a single iodine-cell frequency discriminator, and the seed transmittance is used to determine the spectral position of the emitted laser relative to the current discriminator response and update the zero-Doppler reference online. In a dynamic frequency-tracking experiment, the difference between Doppler frequency-discrimination receiver (DFDR) and wavemeter measurements had a standard deviation of 1.25 MHz. During the 8 h continuous measurement, hour-scale variations of several megahertz were observed in the relative spectral position, indicating that the zero-Doppler reference may not remain strictly constant during long-term operation. With integration times of 100–1000 s, the Allan deviation of the frequency measurements was reduced to approximately 0.29 MHz. The frequency difference between the seed reference light and pulsed laser had a standard deviation of 1.14 MHz, supporting the use of the seed reference light to track pulsed-laser frequency variations. In atmospheric observations, the mean vertical wind speed shifted from approximately 1.2 to −0.34 m s⁻¹ after zero-Doppler reference correction. In a controlled frequency-offset experiment, RDLD retrieved equivalent wind speeds agreed well with values derived from independent wavemeter measurements. Measured zonal and meridional winds also agreed well with ERA5 reanalysis in their major vertical structures and temporal evolution. These results demonstrate that the method enables online monitoring and correction of the RDLD zero-Doppler reference without continuous reliance on an external wavemeter, improving the stability of long-term wind measurements.
Wind lidars require measuring the Doppler shift of the backscattered light that is in the range of 10-8 or 10-9 of the laser wavelength. This sets very high demands on the wavelength stability of the used laser system and/or on the monitoring of the true laser wavelength. The paper by Tan et al. describes a wind lidar setup where the wavelength-selective part of the detector (i.e., an iodine cell) is used to monitor the wavelength of the emitted light. By this, they achieve a simpler setup with potentially less wavelength drift compared to alternative methods. Therefore, the method is worth publication. The manuscript, however, needs a careful revision because of some inconsistencies, lengthy sections, and missing descriptions.
General comments:
The manuscript is lacking a discussion of the advantages of the proposed method against other methods, as, e.g., described by Souprayen et al., Applied Optics, 1999a, Hildebrand, 2014, and Yan et al., 2017. So far, there is only a short reference to other methods in the introduction, but a clear motivation and justification are lacking.
Furthermore, the authors may consider shortening the description of the method. So far, it appears a bit wordy. From my understanding, the method is combining an idea of feeding the seed light into the detection system (described in Hildebrand, 2014) with the fact that the pulsed laser nearly perfectly matches the frequency of the seed laser (published by Yang et al., 2025). This is a good idea, and the description may be limited to this method without the basics of iodine-absorption based lidar wind measurements.
Specific comments:
- l. 134 – l. 145, Figure 3: The description and the figure strongly disagree. Please double-check and correct accordingly. Besides inconsistencies with the text, the figure is partly inconclusive (e.g., when the seed light stops before the chopper is closed).
- l. 148 – l. 153: The numbers are only roughly in agreement with Fig. 4. I recommend being more precise here, e.g., by calculating the respective altitudes from the trigger timing.
- l. 174: The symbol Tobs for a transmission is confusing if in the same paper T is already used for temperature. I recommend using τ or t instead.
- l. 179: “horizontal LOS” is contradictory. It is the horizontal wind that is derived from the LOS Doppler frequency shift.
- l. 193 – l. 197: It would be helpful to mark the linear part in Fig. 5. This would describe the selected part of the spectrum much easier.
- l. 198 – l. 213: If I understand the description correctly, it includes some unnecessary steps. Why is first a reference seed transmission assumed to calculate a response function and a Doppler frequency shift? This is not needed from my point of view. The true seed reference transmittance Tseed,real is available immediately after the WT/NT measurements are finished, and the zero-Doppler reference Δνzero,real can be calculated as well as the corresponding Doppler shift. I recommend rephrasing this section, either making the process clearer or avoiding unnecessary steps.
- l. 221: If I understand correctly, then the data acquisition is done with the normal lidar counting system in a "virtual height range" of ~200 to 240 km at a repetition rate of 200 Hz. That would result in 20 independent data points that are averaged. Please clarify.
- l. 223: More precisely, each point covers a spectral range of 5 fm if a continuous scan of the seed wavelength is done. What are the implications of this kind of smearing? Please discuss.
l. 249: The distribution around zero is a necessary result of the fit. The good thing is that the distribution is kind of “homogenous”.
- l. 252 – l. 254: a) The offset is not persistent but changes from negative to positive. b) The WS7-60 has a resolution of 2 MHz (~2 fm). The results show much smaller differences, which, I assume, are caused by some smoothing. Please explain.
- l. 257: As mentioned above (comment to l. 221), please clarify how the integration time relates to the noncontinuous data acquisition.
- l. 277 – l. 278: Why is this drift not observed in the seed wavelength if locked to an iodine cell with similar properties?? I would expect the seed wavelength as measured with the wavemeter to drift as well if the seed laser is fixed to a certain signal ratio of the seeder iodine cell. Figs. 6 and 7 of Yang et al., 2025, suggest some drift of the laser wavelength, but it is not clear if this is limited to the pulsed laser only.
- l. 309/310: Again, what is the averaging time (or number of pulses) here? The agreement between both data series is surprising, given that the accuracy of the wavemeter is 60 MHz. Is there any bias correction applied?
- l. 310/311: “followed by a small-range frequency tuning” sounds odd. Maybe "followed by 35 min of frequency tuning over +/- 2 MHz".
- l. 327: What is meant by “intermediate”?
- Section 4.3: I do not understand the relevance of this section. From my point of view, the test with the three different locking transmittances shows only that Eq. 7 and Fig. 5 are correct and that both iodine cells behave about the same. All has been shown before. The remaining differences in Fig. 11 and Table 1 are within the photon count uncertainties. Please either expand the motivation for this section or consider deleting the section.
- l. 397/398: I think the differences could also be caused by the inaccuracy of ERA5 at these altitudes, where no measurements are assimilated.
- Section 4.4, Figs. 13-16: The comparison with ERA5 wind data should be shortened. As long as the description remains general and does not include, e.g., some geophysical interpretation of the variability, etc., it does not require four different figures. This is especially true since the expected wind correction would likely remain invisible in the color plots (Figs. 13/14). It would instead be interesting to really see the effect of the wind correction. How does the horizontal wind change in case that the wind correction is not done? How does this (avoided) systematic error compare quantitatively to the uncertainty caused by photon statistics in Figs. 15/16? This would be a nice motivation for the paper.
Technical comments:
- l. 93: The sentence is duplicated.
- l. 131: Delete “that”.
- l. 165: I assume that the description of ΔνD should be moved behind Eq. 3.
- l. 205: The delta should be omitted here because the “baseline” is a frequency (if this section is kept at all, see specific comment above)
- l. 221: Please add that this is the wavelength in a vacuum.
- l. 270: Delete “relative to the DFDR frequency-discrimination response”
- l. 279: Delete “relative to the laser frequency”.
- l. 310/311: “followed by a small-range frequency tuning” sounds odd. Maybe "followed by 35 min of frequency tuning over +/- 2 MHz".
- l. 358: “tow” should read “two”.
- l. 422: “transmitted laser position on the current frequency–transmittance response curve” can be rephrased to “laser frequency”