Evaluation of the quality of a UAV-based eddy covariance system for measurements of wind and turbulent flux
Abstract. Instrumentation packages of eddy covariance (EC) have been developed for a small unmanned aerial vehicle (UAV) to measure the turbulent fluxes of latent heat (LE), sensible heat (H), and CO2 (Fc) in the atmospheric boundary layer. This study evaluates the measurement performance of this UAV-based EC system. First, the precision (1σ) of the measurements was estimated at 0.04 m s-1 for wind velocity, 0.08 µmol m-2 s for Fc, 1.61 W m-2 for H, 0.15 W m-2 for LE, and 0.02 m s-1 for friction velocity (u*). Second, the effect of calibration parameter and aerodynamic characteristics of the UAV on the quality of the measured wind was examined by conducting a set of calibration flights. The results shown that the calibration improved the quality of measured wind field, and the influence of upwash and leverage effect can be ignored in the wind measurement. Third, data from the standard operational flights are used to assess the influence of resonance on the measurements and to test the sensitivity of the system by adding an error of ±30 % to their calibrated value. Results shown that the effect of resonance mainly affect the measurement of CO2 (~5 %). The pitch offset angle (εθ) significantly affected the measured vertical wind (~30 %) and H (~25 %). The heading offset angle (εψ) only affected the horizontal wind (~15 %), and other calibration parameters had no significant effect on the measurements. The results lend confidence to use the UAV-based EC system, and suggest future directions for optimization and development of the next generation system.
Yibo Sun et al.
Status: final response (author comments only)
- RC1: 'Comment on amt-2022-321', Andrew Kowalski, 23 Feb 2023
- RC2: 'Comment on amt-2022-321', Anonymous Referee #2, 11 Mar 2023
Yibo Sun et al.
Yibo Sun et al.
Viewed (geographical distribution)
The authors are to be commended for confronting a major challenge. I believe in much of their introduction regarding the need to increase the spatial density of flux observations, and the potential for UAV platforms to fulfill this need. The paper contains much methodological detail regarding UAV wind measurements and assessment, aspects of the paper that I do not feel qualified to evaluate. However, I can assess eddy covariance methodologies for determining surface fluxes, and I am afraid that the authors as yet fall short regarding both the implementation and error assessment.
From reading the text, it is unclear to me whether the authors have (1) incorrectly determined the fluxes of CO2 and H2O, or simply (2) incorrectly described the methodology that they applied. I suspect the former, based on the comment below regarding the text at line 479. In any event, the paper requires major revision to clarify these points, and possibly to modify both the eddy covariance methodology and also the assessment of its errors and sensitivity to environmental parameters. I also believe that the presentation of the data could be improved significantly as described below.
82 - As the authors note "The EC method is a well-developed technology for directly measuring vertical turbulent flux...". Decades of experience that has shown that the covariances between the vertical wind and the densities of CO2 (ρc) and H2O (ρv) - measured directly by the EC150 - do not define the turbulent fluxes of these gases, as the authors seem to believe (lines 250-251). This is because fluctuations in these variables are predominantly caused by temperature fluctuations (due to heat exchange), and ρc also fluctuates because of varying humidity (due to evaporation). See comment regarding line 269 below.
162 - The choice of the Bohai Sea as the place for the in-flight calibration campaign is quite unfortunate. Its waters are cool, particularly relative to continental temperatures in September, and therefore the magnitude and spectra of the temperature fluctuations that tend to dominate fluctuations in ρc and ρv are not representative of what might be encountered in many other environments. Indeed, the authors note that stable atmospheric conditions prevailed during the campaign (lines 174-175), implying that turbulence is supressed during this assessment of the ability to measure turbulent fluxes. If this limitation cannot be removed from the analysis, it should at least be noted.
261-262 - "In this study, the objective is not to quantify the actual flux exchange between the surface and the atmosphere, but rather to assess the sensitivity of the calculated turbulent flux to external parameters." The quality of a UAV-based eddy covariance system for measurements turbulent flux (reflecting the title of the paper) cannot be assessed without determining whether it quantifies the actual surface exchange. As an example, if the system reports an unbelievable uptake of 50 μmol m-2 s-1 of CO2 uptake, its quality is likely low whatever its sensitivity to external parameters. For this reason, I believe that the authors should indeed provide magnitudes of the fluxes that they are characterizing. This is particularly so given methodological uncertainties regarding how the fluxes are determined (see comment regarding line 82 above, and 269 below).
269 - The authors cite Metzger et al. (2012) regarding the calculation of turbulent fluxes. Since Metzger et al. (2012) did not measure CO2 fluxes, the citation at line 269 is insufficient to document the data processing and corrections needed to determine the turbulent CO2 flux. Users of an open-path IRGA must address the issue of "density corrections", and cite an appropriate reference (Webb et al., 1980, Correction of flux measurements for density effects due to heat and water vapor transfer. Quart J Roy Meteorol Soc 106:85–100; or perhaps Kowalski et al.,2021, Disentangling Turbulent Gas Diffusion from Non-diffusive Transport in the Boundary Layer. Boundary-Layer Meteorol 179, 347–367. https://doi.org/10.1007/s10546-021-00605-5). Otherwise, the results are inconsistent with what we know about biological activity.
273-278 - There is a problem with randomly generated errors for input variables, because some of these variables tend to be correlated. For example, over the sea in stable atmospheric conditions, the heat flux is downward and the vapor flux is presumably upward. Therefore, temperature and humidity are negatively correlated. If, for some reason, errors in the measurement of temperature and humidity are correlated, then this can badly bias the eddy covariance. Correlated errors could arise for many reasons including faulty instrumentation, sampling errors, and flow distortion. A Monte Carlo simulation that presumes independence of such variables will miss this sort of problem, and therefore is not an appropriate tool for error assessment.
396 - As noted above (see comment regarding line 273), the method used to determine the least resolvable flux magnitude is not believable.
399 - Regarding sensor drift, the authors should take care to examine the effects of any lens contamination (Serrano-Ortiz et al., 2008, Consequences of uncertainties in CO2 density for estimating net ecosystem exchange by open-path eddy covariance, Boundary-Layer Meteorology, 126, 209-218.), which would seem to be a problem in an environment rich in sea salt. To be clear, such errors arise as a consequence of the necessary density corrections when measuring gas densities with an open-path IRGA.
479 - "a sensitivity test was conducted by adding an error of ±30 % to the calibrated value of each calibration parameter." Serrano-Ortiz et al. (2008) showed that just a 5% in the CO2 density can cause CO2 flux errors in excess of 13%, due to the influence of density corrections. The fact that the authors of this study found such small errors in the CO2 flux (Table 4) strongly hints that they are not correcting for density effects, and therefore that their error analysis is inadequate. It also causes me to strongly doubt the claim of a 0.4 μmol m-2 s-1 least resolvable magnitude for the CO2 flux.
583 - Change "Forth" to "Fourth".