| 25 Aug 2025
Signal processing to denoise and retrieve water vapor from multi-pulse-length lidar data
Abstract. Recent hardware developments for the MicroPulse DIAL enable the transmitter to switch between output pulses that are "longer" (higher pulse energy) and "shorter" (low pulse energy) on a shot-to-shot basis. While the longer laser pulses broadly result in higher signal-to-noise ratio, they have the shortcoming of blanking the detector in the lowest ranges and smearing out the scene in range. Conversely, shorter pulses enable observations closer to the instrument, smear the scene relatively little, but have low signal-to-noise ratio. In this work, we show that leveraging Poisson Total Variation with forward modeling enables merged estimates of backscatter and water vapor. This signal processing technique leverages the advantages of each pulse length configuration, providing better data availability and higher resolution over a broader altitude range than data processed using only one of the pulse lengths. An intercomparison with radiosondes demonstrates that this new hardware configuration and processing approach enable retrievals of absolute humidity starting at 100 m extending up to 6 km, capturing complex water vapor structure throughout this range. The retrievals are also contrasted with ERA5 reanalysis which suggests that there are instances where the model and reanalysis products are unlikely to produce accurate representation of water vapor fields in the atmosphere, thus emphasizing the value of continuous, high-vertical-resolution active thermodynamic profiling observations.
I find that the author make sufficiently clear that their method of signal processing improves the results from multi-pulse-length LIDAR measurements, in that it enables to merge data obtained from longer and shorter laser pulses and extract a better picture of water vapor distribution in the atmosphere. Therefore I recommend this paper for publication. I have only minor issues that I hope the authors might clarify.Â
1) Abstract: It is not clear to me why longer pulses should blank the detector. If the energy is distributed over a longer time, the dynamic range of the detector should not be the limit. Please explain it better. Is this only due to the fact that no data can be recorded while the pulse is still on its way out of the laser? Or there are other processes involved (see following point)?
2) Page 7: "In addition to masking due to potential errors in the noise model, the long pulse channels tend to experience a bias in the lower altitudes associated with the pulse length and recovery time of the detector (the exact causes of this effect are still not fully understood and may be related to stray light, detector recovery time, afterpulsing or a combination of all three)."
How is the dependency of this effect on the pulse length. Can you give a quantitative answer?
3) Figure 5, bottom right panel. The authors discuss the noisy data (blue line), but the average value of the blue line seems to be much shifted toward higher humidity as well. Can they explain it?
4) What happens if instead of short and long pulses one used only short pulses with more or less energy? I understand that this might not be possible with the present setup, but what if?