Does higher spatial resolution improve snow estimates?
Abstract. Given the tradeoffs between spatial and temporal resolution, questions about resolution optimality are fundamental to the study of global snow. Answers to these questions will inform future scientific priorities and mission specifications. Heterogeneity of mountain snowpacks drives a need for daily snow cover mapping at the slope scale (≤ 30 m) that is unmet for a variety of scientific users, ranging from hydrologists to the military to wildlife biologists. But finer spatial resolution usually requires coarser temporal or spectral resolution. Thus, no single sensor can meet all these needs. Recently, constellations of satellites and fusion techniques have made noteworthy progress. The efficacy of two such recent advances is examined: 1) a fused MODIS - Landsat product with daily 30 m spatial resolution; and 2) a harmonized Landsat 8 - Sentinel 2A/B (HLS) product with 2–3 day temporal and 30 m spatial resolution. State-of-art spectral unmixing techniques are applied to surface reflectance products from 1 & 2 to create snow cover and albedo maps. Then an energy balance model was run to reconstruct snow water equivalent (SWE). For validation, lidar-based Airborne Snow Observatory SWE estimates were used. Results show that reconstructed SWE forced with 30 m resolution snow cover has lower bias, a measure of basin-wide accuracy, than the baseline case using MODIS (463 m cell size), but higher mean absolute error, a measure of per-pixel accuracy. However, the differences in errors may be within uncertainties from scaling artifacts e.g., basin boundary delineation. Other explanations are 1) the importance of daily acquisitions and 2) the limitations of downscaled forcings for reconstruction. Conclusions are: 1) spectrally unmixed snow cover and snow albedo from MODIS continue to provide accurate forcings for snow models; and 2) finer spatial and temporal resolution through sensor design, fusion techniques, and satellite constellations are the future for Earth observations.
Edward H. Bair et al.
Status: final response (author comments only)
RC1: 'Comment on tc-2022-230', Anonymous Referee #1, 19 Mar 2023
- AC1: 'Reply on RC1', Edward Bair, 01 May 2023
RC2: 'Comment on tc-2022-230', Anonymous Referee #2, 26 Mar 2023
- AC2: 'Reply on RC2', Edward Bair, 01 May 2023
Edward H. Bair et al.
Edward H. Bair et al.
Viewed (geographical distribution)
This article has an attractive title which unfortunately does not reflect its content. The authors present a comparison of the results obtained by backward reconstruction of the SWE from three different remote sensing products. The rationale for this study is actually more specifically given at the end of the introduction and is "considerable advances have been made in SWE reconstruction techniques as well as snow cover and albedo mapping, hence the justification for revisiting the effects of spatial and temporal resolution". But the advances in question are not specified. And the question of the albedo is not really studied in the rest of the paper.
Significant work has been done to perform these simulations, but the analysis of the results remains superficial and does not explore the mechanisms that explain the effects of the resolution on the modeling of snow cover. However, the conclusion that seems to emerge is that the resolution has no impact on the estimate of the resource, which is counterintuitive when compared to studies that immediately come to mind because I contributed to them (Baba et al. 2020, Bouamri et al. 2021) or earlier by Schlögl et al. (2016). The fact that the source products also have different revisit times (and different processing algorithms) complicates the analysis of the effect of spatial resolution. In fact, the discussion concerns the artefacts linked to the delimitation of the watershed, which does not seem to me to be a central issue. I think it is necessary to help the reader to interpret the results, perhaps through an analysis of the energy balance or semi-variograms of topographic variables.
Another thing puzzled me when reading the manuscript. The authors introduce a method for reconstructing the SWE before the accumulation peak which is a rescaling of the GLDAS SWE. This SWE is therefore produced from different forcings, which further complicates the interpretation of the results. Equation (1) is incomprehensible to me*.
The final conclusion of the article "increased spatial and temporal resolution (...) are the future of Earth observations." could have been written before carrying out this study and concerns many other fields of application than snow. However, it does not seem to me that the results and the very design of the study support this conclusion.
In terms of presentation, the authors introduce additional analyzes in the results section which have not been presented in the method section as recommended for scientific articles.
In the end, all this leads me to think that this manuscript was prepared a little too quickly, which is regrettable given the work and calculation behind the production of these datasets. I'm sorry to give such a negative review, maybe another reviewer will disagree with me.
* The ⋀ operator is an "n-ary logical and" so the result should not be a SWE value but a boolean (vector) variable. Besides I don't understand if the pixels are selected by considering the time series of SWE and fsca (the time index does not appear).
NB) I was unable to get the data from the ftp server indicated at the end of the manuscript. The connection is possible but not the download (I tried from two different networks)
L89: the bowtie effect of MODIS acquisitions was known before this reference
L98: any reason why HLS v2 was not available? What is the difference with v1 and would it change the results?
L105: any clue why the revisit is not 2-3 days?
L114: why eliminate certain images after visual inspection? this seems incompatible with a global application ("global snow").
L124: This should be explained ("a second cloud filtering step using Superpixels and Gabor filtering was used")
L131: "SPIReS, SCAG, and all other accessible snow mapping algorithms" I have checked this article and this assertion is incorrect.
L173: why not use all ASO acquisitions? There are many more on this basin since 2017.
L184: can you specify or indicate the tool? "using a mean-preserving technique with a weighted resampling covering the image". Imagine that a reader would like to use the same approach (I would).
L185: the geolocation accuracy of S2 is about 1 pixel of 10m, not 1-2 pixels of 30m. See the data quality reports by ESA. Note that recent GRI reprocessing should result to subpixel accuracy (<10m). Also, Storey et al. (2016) report that Landsat OLI has a geolocation accuracy of 18 meters (CE90), not 1-2 pixels of 30m.
L204: I may have missed something but why not make this comparison for other products? as it stands, this part on the albedo does not add much to the study.
L305: S2C should replace S2A hence it will not improve revisit time (except for a short period). https://labo.obs-mip.fr/multitemp/some-news-from-esa-regarding-the-coming-sentinels-1-and-2/
Baba, M. W., Gascoin, S., Kinnard, C., Marchane, A., and Hanich, L.: Effect of Digital Elevation Model Resolution on the Simulation of the Snow Cover Evolution in the High Atlas, Water Resources Research, 55, 5360–5378, https://doi.org/10.1029/2018WR023789, 2019.
Bouamri, H., Kinnard, C., Boudhar, A., Gascoin, S., Hanich, L., and Chehbouni, A.: MODIS does not capture the spatiotemporal heterogeneity of snow cover induced by solar radiation, Front. Earth Sci., 9, https://doi.org/10.3389/feart.2021.640250, 2021.
Schlögl, S., Marty, C., Bavay, M., and Lehning, M.: Sensitivity of Alpine3D modeled snow cover to modifications in DEM resolution, station coverage and meteorological input quantities, Environmental Modelling & Software, 83, 387–396, https://doi.org/10.1016/j.envsoft.2016.02.017, 2016.