The manuscript by Walther et al. presents the technical details of a new customized and gap-filled remote-sensing product generated from MODIS and Landsat instruments across Fluxnet sites. They proposed a procedure to extract, quality-filter, correct, and gap-fill MODIS and Landsat data and develop standardized data products of surface reflectance, vegetation indices, and land surface temperature.

Overall, I think this is a great initiative, and I agree with the authors that “the data sets can widely facilitate the integration of activities in the fields of eddy-covariance, remote sensing, and modeling”. I also appreciate the authors’ efforts in presenting the details and being frank on the merits and limitations. Therefore, I would recommend the manuscript to be considered published in Biogeosciences after addressing a few general and specific comments.

[1] I agree with the general comment made by Michael Dietze on the need to differentiate this proposed data product from the one distributed under ORNL DAAC or any other ones. Consider highlighting the uniqueness of this product or main differences with others in the manuscript. It’ll help the potential users to choose a suitable product for specific use.

[2] A standardized and operationally feasible procedure for quality control and gap-filling of MODIS and Landsat data is the main focus of this manuscript. I think the authors should consider adding more analyses to validate or at least demonstrate the uncertainties/limitations of the proposed data product. The examples presented in section 4.2 are great and illustrative, but I think it needs more generalized information on the performance across sites. For example, consider comparing the data product with other available gap-filled products (e.g., MCD43GF or others like Robison et al., 2017). Also, why isn’t Landsat land surface temperature included, and why do Landsat data only cover till 2017?

[3] It’s challenging and potentially problematic to gain generalized ideas of the spatial contexts based on those few examples (section 4.3). I understand it may not be feasible to calculate flux footprints for all sites included in this study. Still, please consider leveraging the findings from previous efforts (e.g., Göckede et al., 2008 on European sites, Chen et al., 2011, 2012 on Canadian sites, Ran et al., 2016 on Chinese sites, Chu et al., 2021 on AmeriFlux sites, and Griebel et al. 2020 on Fluxnet sites heterogeneity). Those studies analyzed many sites included in this study, providing information about the flux footprints (e.g., extents, areas) that can help justify the selection of cut-out extents and area-weighted methods.

A universally 2-km cut-out may be a bit small for specific tall tower sites. I’d suggest expanding the extents for at least the tall tower sites (e.g., forests, known tall tower (e.g., US-PFa). In our recent study (Chu et al. 2021), we found a few AmeriFlux sites (e.g., US-ChR, US-Wrc) have footprints (i.e., monthly climatology, 80% contour, based on Kljun et al., 2015 model) extending beyond 2 km from the tower. And, more sites are extending beyond 2 km if using half-hourly or daily footprints or using a different footprint model (e.g., Kormann & Meixner 2001). I think it’s practically safer to start with a larger extent and then crop the images as needed.

[4] Last, I’d suggest the authors and the team consider adding other sites at this or future release, especially those with compatible processed flux datasets. For example, AmeriFlux begins rolling out processed flux data products compatible with FLUXNET2015 (see links below). Also, with other new Fluxnet initiatives (e.g., Fluxnet Co-op), it’s optimistic to anticipate similar Fluxnet products will become available at more sites in the near future. As pointed out earlier, one of the major differences between this and DAAC subset products is the number of sites that are included. It will benefit many users if this data product could be generated at more Fluxnet sites.

https://ameriflux.lbl.gov/data/download-data-oneflux-beta/

https://ameriflux.lbl.gov/data/data-availability/#/FLUXNET

Specific comment

[5] Line 3: Please consider adding AmeriFlux to the list as other regional networks.

[6] Line 6-7: This sentence “…support the training and validation of ecosystem models” is vague. Consider rewriting it.

[7] Line 95: Is the sensor difference (e.g., among Landsat 4, 5, 7, 8) or sensor drifting corrected? Also, Landsat 7 is known for Scan Line Corrector (SLC) failure and causes problematic data in certain themes. How does it be addressed?

[8] Line 135-137: This sentence is unclear. Could you explain it a bit more in detail?

[9] Line 139-161: Does the gap-filling procedure apply to the raw bands only (i.e., calculate vegetation indices based on filled bands), or separately for both the raw bands and vegetation indices? Any justification?

[10] Line 170-172: Consider adding more granular details of the flags. Does it indicate which method is being used, or is it simply a binary flag (filled/original)?

[11] 4.1 & Figure 1: Please add some discussions on the Landsat availabilities. Also, would it be more suitable to group sites by regions or biomes given geo-patterns of cloudiness? Looking at the low availability at some sites, I wonder whether it is more appropriate to leave out those sites entirely.

[12] 4.2, Figure 2-3: Please add some discussions on the Landsat time series.

[13] Table 2: Please add some details about the two cut-outs to the Method section in the main text. Consider briefly justifying the weighted approach.

[14] Line 274-290: I suggest moving this part of the literature review to an earlier section.

[15] Line 327-340 & Figure 6: The comparison is misleading. The net radiometer (for measuring long-wave radiation) has a fixed field of view depending on its mounting height and location. It is more appropriate to compare Tsurf with LST at pixels corresponding to the radiometer’s field of view or compare LSTfpa with sensible heat fluxes (or derived aerodynamic surface temperature (see Novick & Katul 2020).

[16] Figure 7: Consider adopting the same color scale for all EVI maps.   

[17] Figure D1: Consider using a similar layout (e.g., extents, x-/y-axes, color codes…) as in Figures 5-6.

Chen, B., Coops, N. C., Fu, D., Margolis, H. A., Amiro, B. D., Barr, A. G., et al. (2011). Assessing eddy-covariance flux tower location bias across the Fluxnet-Canada Research Network based on remote sensing and footprint modelling. Agricultural and Forest Meteorology, 151(1), 87-100.

Chen, B., Coops, N. C., Fu, D., Margolis, H. A., Amiro, B. D., Black, T. A., et al. (2012). Characterizing spatial representativeness of flux tower eddy-covariance measurements across the Canadian Carbon Program Network using remote sensing and footprint analysis. Remote Sensing of Environment, 124, 742-755. DOI: 10.1016/j.rse.2012.06.007

Göckede, M., Foken, T., Aubinet, M., Aurela, M., Banza, J., Bernhofer, C., et al. (2008). Quality control of CarboEurope flux data – Part 1: Coupling footprint analyses with flux data quality assessment to evaluate sites in forest ecosystems. Biogeosciences, 5(2), 433-450. doi:DOI: 10.5194/bg-5-433-2008

Griebel, A., Metzen, D., Pendall, E., Burba, G., & Metzger, S. (2020). Generating spatially robust carbon budgets from flux tower observations. Geophysical Research Letters, 47(3), e2019GL085942. doi:DOI: 10.1029/2019gl085942

Kormann, R., & Meixner, F. (2001). An analytical footprint model for non-neutral stratification. Boundary-Layer Meteorology, 99(2), 207-224. doi:DOI: 10.1023/a:1018991015119

Kljun, N., Calanca, P., Rotach, M. W., & Schmid, H. P. (2015). A simple two-dimensional parameterisation for Flux Footprint Prediction (FFP). Geosci. Model Dev., 8(11), 3695-3713. doi:DOI: 10.5194/gmd-8-3695-2015

Ran, Y., Li, X., Sun, R., Kljun, N., Zhang, L., Wang, X., et al. (2016). Spatial representativeness and uncertainty of eddy covariance carbon flux measurements for upscaling net ecosystem productivity to the grid scale. Agricultural and Forest Meteorology, 230, 114-127.

Chu, H., Luo, X., Ouyang, Z., Chan, W. S., Dengel, S., Biraud, S. C., et al. (2021). Representativeness of Eddy-Covariance flux footprints for areas surrounding AmeriFlux sites. Agricultural and Forest Meteorology, 301-302, 108350. doi:https://doi.org/10.1016/j.agrformet.2021.108350

Novick, K. A., & Katul, G. G. (2020). The Duality of Reforestation Impacts on Surface and Air Temperature. Journal of Geophysical Research: Biogeosciences, 125(4), e2019JG005543. doi:https://doi.org/10.1029/2019JG005543

Robinson, N. P., Allred, B. W., Jones, M. O., Moreno, A., Kimball, J. S., Naugle, D. E., et al. (2017). A Dynamic Landsat Derived Normalized Difference Vegetation Index (NDVI) Product for the Conterminous United States. Remote Sensing, 9(8), 863.

In the technical note “A view from space on global flux towers by MODIS and Landsat: The FluxnetEO dataset”, Walther et al. presents a standardized procedure to extract, gap-fill and quality control remote sensing observations around >300 flux sites. This contribution is critical to the reliable integration of remote sensing and eddy covariance measurements for understanding ecosystem functions and changes. I am in support of its publication, and my comments are meant to help improve the note and make it more clear to the audience.

L34: As gap-fill is a key step in producing the dataset, perhaps it would be helpful to further clarify the general assumptions under these categories of methods. Other than the realistic considerations (i.e., generalizable, no need to use ancillary data) to do gap-fill only based on the remote sensing time series themselves, are there studies that suggest this method produce comparable results to complicated ones (i.e., the one that use ancillary meteo data).

L38: “contribution” means “study”?

L51: reference for “view zenith angles”.

L136-137: I have some difficulties in understanding “The idea was to….instead of….”. I feel the authors are arguing that their method is appropriate for the study though I cannot understand the second part of the sentence. “Valid data” means ancillary data or just the good quality data of the time series.

L154: it is not easy to understand the scaling method without carefully looking into some equations in ANN C. Perhaps it is helpful to insert some equations here, such as y = ax + b, where x means MSC while y is the non-gap filled time series. Then we can get a and b from the equation for each time window, and then apply a and b back to MSC for gap filled y.

L175. Out of curiosity why do not use quality flag of MODIS here, any issue with the flag? By using statistical method only to remove the so-called outliers, are we risking removing some true extreme values?

L210. See my comment above regarding the description of the scaling method.

L336. From Fig. 6a it is not accurate to say LST is consistently 30% higher, it is only the slope that is around 1.3.

L338. Do we really see the “slope decreases markedly for the highest temperature”? The figure only shows that slope increases a bit with the height.

L389. For those sites with footprint less than 1km (which I think many sites are), how to define this aggregated snow flag. Are they either 0 or 1?

L401 – 404. I am also wondering the rationale for choosing mean seasonal cycle and median seasonal cycle in different datasets. I also have to say in FLUXCOM mean seasonal cycle of remote sensing data was used but here the use of median seasonal cycle seems to be prevailing.

L418. Valid snow cover < 60 days = snow does not occur at the site? I have a feeling the threshold is a bit large, e.g., a site with almost two months of valid snow cover might be considered to have no snow by this filter.

L450. To double check, do you mean for each time window we get a m and n?

L455. There is a redundant “[]” in the equation. Perhaps also would be helpful to explain the terms in the equations.