the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Evaluation of Aeolus L2B wind product with wind profiling radar measurements and numerical weather prediction model equivalents over Australia
- 1Wind Energy, Technical University of Denmark, Roskilde, 4000, Denmark
- 2Danish Meteorological Institute, Copenhagen, 2100, Denmark
- 3Royal Netherlands Meteorological Institute, De Bilt, 3731 GA, Netherlands
- 1Wind Energy, Technical University of Denmark, Roskilde, 4000, Denmark
- 2Danish Meteorological Institute, Copenhagen, 2100, Denmark
- 3Royal Netherlands Meteorological Institute, De Bilt, 3731 GA, Netherlands
Abstract. Carrying a laser Doppler instrument, the Aeolus satellite was launched in 2018, becoming the first mission for atmospheric wind profile measurements from space. Before utilizing the Aeolus winds for different applications, evaluating its data quality is essential. With the help of ground-based wind profiling radar measurements and the European Centre for Medium-Range Weather Forecasts (ECMWF) model equivalents, this study quantifies the error characteristics of Aeolus L2B (baseline-11) near real time horizontal line-of-sight winds across Australia by using both inter-comparison and triple collocation analysis. The results of the inter-comparison analysis indicate that both Rayleigh-clear winds and Mie-cloudy winds are in good agreement with the ground-based radar measurements with overall absolute mean biases smaller than 0.7 m s-1 and correlation coefficients larger than 0.9. Moreover, taking radar measurements as reference data set, Mie-cloudy winds are shown to be more precise than Rayleigh-clear winds with an overall random error of 5.81 m s-1 for Rayleigh-clear winds and 4.14 m s-1 for Mie-cloudy winds. Similar results were also found from triple collocation analysis, with error standard deviations of 5.61 m s-1 and 3.50 m s-1 for Rayleigh-clear winds and Mie-cloudy winds, respectively. In addition, the Mie channel is shown to be better capable of capturing the wind in the planetary boundary layer (< 1,500 m). The findings of this study demonstrate the good performance of space-borne Doppler lidar for wind profiling and provide valuable information for data assimilation in numerical weather prediction.
Haichen Zuo et al.
Status: final response (author comments only)
-
RC1: 'Comment on amt-2022-63', Anonymous Referee #1, 12 Mar 2022
General Comments
Overall well presented and useful comparison of Aeolus / wind profiles / NWP wind measurements
Specific Comments :
Bias’ is discussed in several places ( e.g. Line 59 / Sect. 3.1, Table 3 ) but no confidence limits are given for these biases. This makes it impossible to understand if they are significant or if differences between ‘bias’ in different cases are significant. Please add confidence limits for the biases
In Figs 5 and A2, Fig. 5. ‘uncertainty’ in bias for different height bins is shown by shaded areas - these look surprisingly small given the very low number of samples in the height bins in many cases. How is ‘uncertainty’ defined ? 95% confidence limits or something else ?
The conclusions ( Lines 361-364) say : “When comparing with the ground-based radar measurements, no significant biases (absolute mean bias < 0.7 m s-1) and good agreements (R > 0.9) were found for both Rayleigh-clear and Mie-cloudy winds. For the Rayleigh channel, the wind detection during ascending orbits has higher accuracy than during descending orbits, while for the Mie channel, a large bias was obtained during ascending orbit. “
This says first there are ‘no significant biases’ and then ‘a large bias was obtained’. Which is it ? Adding the confidence limits for the biases should help with getting this right.
Minor points :
There are numerous small grammar / language errors which are distracting - probably a copy editor can take care of most of these, although any co-authors who are proficient in English should also check.
For example : Lines 36-38
“Wind retrievals of ALADIN are based on light scattering by atmospheric molecules and particulates (aerosol, cloud droplets, and ice crystals) which move with the ambient wind and the Doppler effect (Ingmann and Straume, 2016). “ - this says particulates ..move with … the Doppler effect. It needs changing to
“Wind retrievals of ALADIN are based on light scattering by atmospheric molecules and particulates (aerosol, cloud droplets, and ice crystals), which move with the ambient wind, and on the Doppler effect (Ingmann and Straume, 2016). “
In Line 56 “Ray-clear “ is used - everywhere else it is not shortened so it should be “Rayleigh-clear”
- AC1: 'Reply on RC1', Haichen Zuo, 12 May 2022
-
RC2: 'Comment on amt-2022-63', Anonymous Referee #2, 15 Apr 2022
Since 22 August 2018, ESA's wind satellite Aeolus is circling the Earth at around 320 km altitude and capturing global wind profiles with its Doppler wind lidar ALADIN. To further improve the data quality and to use Aeolus observations in NWP models, the systematic and random errors must be understood. Besides global validations by means of NWP model comparison also regional/local validations with independent ground-based or in-situ reference measurements were already performed in several studies. This manuscript focuses on the validation of Aeolus wind measurements in the Australian domain with wind profiling radars as well as NWP model data and thus provides a useful contribution to the ongoing Aeouls Cal/Val activities.
The manuscript is well structured and written, presenting the obtained results with adequate figures. The paper deserves publication after some minor revisions.
General comments
- For quality control, error estimate thresholds of 8 m/s for Rayleigh-clear and 4 m/s for Mie-cloudy winds are applied. Did you try other values and check how this affects the number of data points and the determined random and systematic error of Aeolus wind measurements?
- A horizontal collocation radius of 75 km around the WPR sites was chosen. This is rather strict compared to other validation studies and recommendations which applied at least 100 km. The mentioned paper (Zhang et al.) focuses on aerosol comparison in the PBL were it can be quite variable. For wind this must not be the case. Did you try to increase the radius to a larger value (100 km or even higher) to check if this could improve the statistics by using more data points?
- Have the authors included the random errors of the WPR measurements as well as an estimate of the representativeness error in the determination of the Aeolus wind observation errors? Otherwise the determined random error of Aeolus would be a combination of different errors. Or can this be assessed from the triple collocation?
- Range bin thickness has an influence on the random error especially for Rayleigh wind measurements. Although the altitude dependence of the random and systematic errors was investigated, this was not mentioned or analyzed. The applied regrouping as well as the different range bin setting for tropics and extratropics covered by the domain hide this fact. It would be interesting to show this in the analysis, for example by separating the two different range bin settings areas.
Specific comments
- L.14: Please include the data set time period in the abstract in addition to the baseline.
- L.37: You could cite the ADM-Aeolus Science Report here (https://www.esa.int/About_Us/ESA_Publications/ESA_SP-1311_i_ADM-Aeolus_i)
- L.56: Change to Rayleigh-clear to be consistent
- L.104: Please specify the latitude regions for both settings (30 deg S) to see which sites are affected by which range bin settings
- L.115: Change to WPR
- L.160: How was the temporal collocation performed for WPR-comparisons? These have 30 min resolution. Did you average consecutive WPR-profiles?
- L.180: On what is the spacing of these new groups based?
- L.195 Table 2: Where do the 90 km come from? You mentioned 3 seconds temporal resolution corresponding to about 21 km above.
- L.238: Change to WPR
- L.239: large -> larger
- L.246 Fig.4: plot axes could be made symmetrical; change desending to descending (also in Appendix)
- L.262: As pointed out above, range bin thickness has an influence on the random error especially for Rayleigh-clear observations. This should be mentioned here.
- L.270 Fig.5c: Do you have an idea why there is a larger bias between 6 and 7.5 km?
- L.290 Table 6: second Var(u) --> Var(v)
L.290 Table 6: Does the wind variability has influence on the representativeness (random error) of Aeolus observations? For example, did you try to exclude times where the variability is high? Is the variability changing for ascending and descending orbits? (more convection for ascending orbits) - L.306: Why only at 5 km over such a long time period? What is the reason for this peak?
- L.311: Smaller range bin thickness in the PBL region could also contribute to higher random errors
- L.349: Please shortly summarize the improvements of these processor updates (non-linearities are already mentioned in the Appendix...). Are only Mie-cloudy observations affected or also Rayleigh-clear?
- AC2: 'Reply on RC2', Haichen Zuo, 12 May 2022
Haichen Zuo et al.
Data sets
Aeolus Online Dissemination System European Space Agency https://aeolus-ds.eo.esa.int/oads/access/
Wind Profiler Observations, Part of the Met Office MetDB System Met Office https://catalogue.ceda.ac.uk/uuid/9e22544a66ba7aa902ae431b1ed609d6
Haichen Zuo et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 279 | 88 | 15 | 382 | 4 | 6 |
- HTML: 279
- PDF: 88
- XML: 15
- Total: 382
- BibTeX: 4
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1