General comments

This manuscript describes the results of a field experiment with three low-cost eddy-covariance systems over a patchy agroforestry system and a patchy monocropping system. By analyzing these data from two growing seasons, the authors attempt to answer the question that is raised in the title? The topic of agroforestry is also highly relevant. Overall, the manuscript is well written and clearly structured. The data processing is described in detail. The figures are also clear and easy to read. However, I see major deficits in the experiment design which is not really suited to address the title question, at least not in a general sense as it is formulated. Moreover, I cannot agree with some of the data-processing choices that were made and transparently communicated in the manuscript. As a consequence, data of poor quality and consequently large uncertainty are included in the analysis as the underlying assumptions of the EC-method are compromised. Moreover, I find that gap-filled fluxes should not be included in such an analysis as these modelled data are inherently much smoother than actual measurements. These choices in the data processing limit the ability to draw valid conclusions regarding the hypothesis that is posed by the authors in the introduction section. However, I believe this can still be corrected and the formulation of the objectives can be adjusted. Hence, I recommend major revisions before this manuscript can be accepted.

Specific comments

L37: Since the topic is surface heterogeneity, it would make sense to put this specific type of heterogeneity of and agroforestry system in a more general context of heterogeneity, also stressing that the effects depend on the type of heterogeneity and the scale of heterogeneity (Bou-Zeid et al. 2020)

L95:  The random uncertainty if low cost sensors is not necessarily larger than for conventional EC. This is certainly the case for a systematic error.

L96:  In my mind, the statistical robustness could only be improved through more sampling points (i.e. EC towers) if the surface can be considered homogeneous and footprints are comparable in nature. Otherwise you measure the spatial variability over a heterogeneous surface but you cannot really average those into an overall estimate that would then possibly have a lower uncertainty.

L98: The third objective is not really related to the overarching hypothesis and the title.

Figure 1: I would not call it a monocropping system if the EC tower is located at the edge of a field between two different crops, and, hence, is measuring fluxes from both crops to a certain extent (or even another crop) depending on the specific footprint.

L144: Was the flow turbulent inside the tubes for this flow rate, which depends on the Reynolds number and hence the diameter? This would be important to minimize diffusion along the tube. The given flow rates seem to be rather low. What are reasons for this choice and what are the consequences for the frequency response characteristics of these measurements?

L173: How does the RH-dependent fit look like? Could you please also give some indicators on the quality of the fit?

L175: What is the reasoning behind this threshold of quality flags <7? Normally, only data with flags <=3 are considered high quality and flags 4-6 are only suitable for calculating annual or monthly sums as they are at least better than gap filling as they have deviations of up to 100%. If data are restricted to flags 1-3, the test on well-developed turbulence can for example ensure that measurements are conducted above the RSL, and hence are not influenced by single roughness elements, i.e. single trees, and the steady state test can ensure that the footprint does not vary too much within a 30-min averaging interval due to variable wind conditions, so that the time series becomes non-stationary and a covariance calculation or any other calculation of Gaussian statistics are not meaningful anymore. Hence, I highly recommend to use only data with flags 1-3 for this study.

L210: In my mind, it does not make sense to apply gap filling for the objectives of this study. Only actual measurements should be used to analyse the spatiotemporal variability and heterogeneity effects, no modelled data, which are inherently much smoother that actual flux measurements.

Table 1: Be aware that these number represent just the error of the gap-filling and not the error of the EC measurements. These can be estimated based on other methods (e.g. Lenschow et al. 1994, Finkelstein and Sims 2001, Billesbach 2011, Richardson et al. 2012).

L241: How was the zero-plane displacement height calculated for the towers between two adjacent fields with different canopy height?

L289: In principle, it would be fine to determine the uncertainty from an intercomparison experiment. But then, it should be guaranteed that the underlying surface is homogeneous and the footprints are overlapping. This was clearly not the case in the study of Callejas-Rodelas et al. (2024) and hence this study cannot be used for this purpose. Moreover, other measures than the slope of a regression are better suited to describe the uncertainty based on an intercomparison experiment, for example comparability (RMSD) and bias.

Figure 4: Which of these data are actually measured and which are gap-filled? How do the measurements compare for 30 min flux estimates?

L738: The random error should be considered for this study as it is necessary to assess whether the spatio-temporal variability is actually larger than the measurement error.

L763ff: This statement is too simplistic and does consider the enormous complexity of this question. Homogeneous conditions within the footprint are still main prerequisite for eddy-covariance measurements. Otherwise, additional transport terms become relevant which are usually neglected and almost impossible to measure. Please also consider that this kind of thermal surface heterogeneity induces secondary circulations and local advection. As a consequence, dispersive fluxes can develop, so that the eddy-covariance system measuring only the temporal covariance with the w-component severely underestimates the actual surface flux.

References

Bou-Zeid E, Anderson W, Katul GG, Mahrt L (2020) The Persistent Challenge of Surface Heterogeneity in Boundary-Layer Meteorology: A Review. Boundary-Layer Meteorol. https://doi.org/10.1007/s10546-020-00551-8

Billesbach DP (2011) Estimating uncertainties in individual eddy covariance flux measurements: A comparison of methods and a proposed new method. Agric For Meteorol 151:394–405

Finkelstein PL, Sims PF (2001) Sampling error in eddy correlation flux measurements. J Geophys Res 106:3503–3509. https://doi.org/10.1029/2000JD900731

Lenschow DH, Mann J, Kristensen L (1994) How Long Is Long Enough When Measuring Fluxes and Other Turbulence Statistics? J Atmos Ocean Technol 11:661–673. https://doi.org/10.1175/1520-0426(1994)011<0661:HLILEW>2.0.CO;2

Richardson AD, Aubinet M, Barr AG, et al (2012) Uncertainty quantification. In: Aubinet M, Vesala T, Papale D (eds) Eddy Covariance: A Practical Guide to Measurement and Data Analysis. Springer, Dordrecht, pp 173–210

GENERAL COMMENTS

The manuscript reports the results from the monitoring of CO2, H2O and sensible heat fluxes applying the eddy covariance method over a heterogeneous agroforestry field and a conventional cropping field. The authors deployed three low-cost eddy covariance tower in the agroforestry field to assess if the representativeness of fluxes due to the heterogeneity of the surface can be improved by increasing the number of measurement points, as stated in the title.

The application of the eddy covariance method over heterogenous surfaces, especially in terms of canopy structure (height, density, etc.) is challenging because the basic requirements for the application of the method are not fulfilled and other terms, besides the measured turbulent fluxes, should be taken into account. In my opinion, the authors do not give the right importance to this issue and only focus on the spatial representativeness, in terms of footprint area, of the vertical turbulent fluxes. I am well aware that accounting also for advection fluxes would have required a completely different and more demanding instrument setup, so I kindly ask the authors to at least acknowledge in more details the challenging aspects of making eddy covariance measurements over heterogeneous surface, as mentioned at lines 49-51, not just in terms of heterogeneity of scalar sources and sinks.

The manuscript is generally well written, but some sections are very dense and difficult to read. In particular, I think that the results section reports in too much details the patterns of the different variables observed. I suggest to include only the main and significant results so that reading might be easier.

Overall, the manuscript try to characterize carbon and water fluxes over agroforestry systems that are not yet well studied and the analysis approach and findings might be important also for studies on other heterogeneous ecosystems, so I consider that the manuscript should be considered for publication but before that some minor revisions are necessary.

SPECIFIC COMMENTS

L104: is "monocropping" the right term for this site? 3 different crops were grew in the same field, not just one. Maybe "conventional cropping system" might be more appropriate for this specific site. Please consider this comment and change the term accordingly throughout the manuscript.

L165: is there a particular reason why you decided to perform the sectorial planar fit with 8 different sectors? The sector of the planar fit should be determined based on the topography or characteristics of the surface. Why did you opt for this rotation method instead of “normal” planar fit or double rotation? Please add a sentence in the text explaining the reasons for your choice.

L194: based on which criteria did you reject the data?

L236: it is not clear to me if you used only daytime data to assess the aerodynamic parameters or if you calculated footprints only for daytime periods. If this is the case, I think that you should consider also nighttime periods because they contribute to an important part of the C flux.

L246: why did you aggregate the data in wind sectors of 30°? This is not consistent with the 45° sector of the planar fit

L322-323: I do not think you can define "large" a value of 0.5 kPa, I would delete this sentence

Figure3: I think it would be interesting to show also line of 50 or 60% contribution to fluxes so one can have an idea of the location of the area contributing more to fluxes.

L460: could you please explain better the meaning of "effect size" in terms of flux spatial variability here or in the discussion session?

L700-702: I think this is a very important point that could lead to misinterpretation of results. In such heterogenous surfaces, the development of the homogeneous surface layer is not obvious and turbulent and mean flux divergence in the horizontal and vertical directions might be important. Please add a comment on how missing information on these processes could have affected your results.

TECHNICAL CORRECTIONS

L10-11: please rephrase this sentence, I cannot find the subject of “contributed”.

L66: Markwitz and Siebicke (2019) should be in parentheses

L127: close the parentheses after “NETRAD”.

L209: 2 should be 3 instead.

L218: “developed” instead of “developped”

L258: should U be WS instead?

L274-275: should C be FC instead?

L305: please add “total” to “monthly values of P”