Interactive comment on “ Performance of a mobile car platform for mean wind and turbulence measurements ”

Abstract. The lack of adequate near-surface observations of the stable atmospheric boundary layer spatial structure motivated the development of an instrumented car for mobile turbulence measurements. The calibration and validation of the car measurements are performed using controlled field experiments and a comparison with an instrumented tower. The corrections required to remove the effects of the car motion are shown to be smaller and simpler than the corrections for research aircraft measurements. A car can therefore satisfactorily measure near-surface turbulence using relatively low-cost equipment. Other natural advantages of a car, such as the ability to drive on any road at any time of day or night and follow the terrain slope, as well as its low cost of operation, make it applicable to observations of a variety of flow regimes that cannot be achieved with the usual platforms, such as research aircraft or networks of flux towers.

P. 957, L. 7-17: This experiment can be seen as a separate test, which is not only important for the correction of the speed bump test as it was used here, but is of general use for evaluation of wind speed and turbulence data from moving cars.If so, I suggest putting it in an extra section or changing the title of this section accordingly.The results described here would then be presented in Sect.4.4 or 4.5.R: It actually was a separate section in an early version of this manuscript, but since there were no sufficient data for further analysis, we choose to report on it within this section.We have, therefore, opted for changing the title of this section.P. 957, L. 19: Were the corrections in 4.1 applied to the data both in test 1 and 2? Please specify.If different, also specify which data the found corrections (p.958) are based on (only data from test 1 or also from test 2)?

R:
The corrections are applied to all instrumented car data, which is now specified in the manuscript.The sources of the found corrections are also given now.
P. 958, L. 12: "This" may be misleading in this context.I suggest using "The method applied here", or similar.

R: Done.
P. 958, L. 15-16: What do you conclude here?I.e., were non-stationarities and hori-C688 zontal heterogeneities large, so that it had an effect on the described corrections?How large are the corrections and could uncertainties in the correction explain some of the discrepancies between the car and tower measurements?R: We have included additional discussion of our conclusions.During the first test, the non-stationarity was large and has most likely influenced the results.However, it is difficult to quantify the error of the correction factors, because the correction is based only on the car measurements.Using the tower measurements to correct for non-stationarity is not feasible due to different averaging times and unknown effects of horizontal heterogeneity.P. 959, L. 16: There seems to be a systematic deviation between car and tower measurements, e.g. for the horizontal wind speed in u-direction (Fig. 3).u measured by the car seems to have a negative deviation from the tower measurements.This may hint to an influence of the car speed, which is not corrected for, since u is orientated closely to the direction of the track.If so, are there suggestions how the corrections or the alignment of wind vector could be improved?Or is the result satisfactory, i.e. within the expected range of uncertainty?R: The bias in u would seem to be a reflection of the limitations in using a fixed-point wind measurement to compare with the car measurement along a horizontal traverse; that is the differences would likely be due to nonstationarity and horizontal heterogeneity in the wind field.The car speed shouldn't directly enter into this (although perhaps not completely if there are systematic wind changes), as this should mostly be taken care of by our removing any bias detected when the car traverses the track in the opposite direction.This is now discussed in the text.P. 959, L. 24-25: Is there a reason why the tower measurements were most likely subject to non-stationarities in the temperature measurements and the car measurements were not?Could the non-stationarities be verified using a stationarity criteria?R: The main reason is that a single tower flux data point is an average over 10 min, C689 while a car data point is obtained during less than 1 min.In that respect, the car is an almost-instantaneous measurement influenced by horizontal heterogeneity, while the tower is influenced only by non-stationarity.Furthermore, in the cases where the Taylor's hypothesis cannot be applied (e.g., when the mean wind is perpendicular to the car tracks), there is no reason to expect the tower and car fluxes to result from the same flow structures.
As for testing the non-stationarity, there is a myriad of tests and assumptions that could be used, one of which would certainly find the non-stationarity since it is always present to a certain extent in ABL time series.But in our opinion, this would be of limited use for understanding the differences between the car and tower fluxes.P. 959, L. 25: There is factor of up to 2 between the different heat fluxes, which might be also due to varying fetch conditions of car and tower measurements.Also, there are significant differences due to the measurements with sonic and TC as it is described below.The expression "agree well" is hence misleading.I would suggest writing: "are in the same range between XX and YY K m s −1 , or similar.R: Done.P. 959, L. 25: A general question on the flux calculations: Did you apply other correction methods which are not mentioned (e.g.coordinate rotation (for tower fluxes), crosscorrelation between w and TC).If so, please state it in the methods section.If not, could this account for some of the systematic differences?R: We did not apply other corrections to the measurements of the variables on the tower and car.This may account for some small systematic differences in the fluxes, but we do not think that it would change any of our conclusions.Additionally, the correction procedures might not work properly in the current conditions.For example, the very weak winds and variable wind direction over 40 min (total record length) could result in spurious corrections from double coordinate rotation, while the record length is too short for the planar-fit method.Finally, the tower sonic was carefully levelled, and C690 the terrain is flat and rather homogeneous, so the corrections are not expected to be significant.
P. 959, L. 27: I suggest writing "This heat flux difference: : :" or specify it more clearly to distinguish between the differences between sonic + TC and car + tower.R: Done, together with modifications addressing other reviewers' comments.P. 960, L. 6: Is it possible to give a measure of the goodness of the agreement?From first visual inspection the difference in the u v flux does not seem to be significantly better than, e.g., for u w .As the comparison of the two approaches is challenging, it might be therefore better to give a statistical measure of the flux agreement or the fluxes itself.R: Bias and normalized RMSE have been calculated and are included in the Supplement.We have also clarified what we mean by better agreement for u v .P. 960, L. 15-16: Did you also try the opposite test, i.e. extending the sampling period of the car measurements by merging the measurements from the each track at one car speed?If this is possible, the comparison with the tower fluxes would be interesting.

R:
We agree -this would smooth the car fluxes and remove some of the noticed issues.However, one of the main issues with the system was that the sign of u w changed when driving in opposite directions along the same track.Likewise, some other errors of the car measurements can be quantified only when comparing measurements from opposite directions.Therefore, merging the measurements for opposite directions would solve some problems, but at the expense of disguising possible errors.Since the main goal of this paper is to detect and analyse all potential errors of the system, we choose to keep the discussion in its current limits.P. 961, L. 6-14: To me the effect on w T is even more significant than as for the other scalars (when comparing the tower fluxes with the TC-fluxes from the car), hence, it might be worth mentioning here.