the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 22 May 2026
Soil microwave background retrieval and snow sensitivity from multi-frequency SAR observations over an agro-forested environment in northern Ontario
Abstract. Accurate retrieval of Snow Water Equivalent (SWE) using Synthetic Aperture Radar (SAR) requires effectively decoupling the signal contribution of the snowpack from that of the underlying soil. This study evaluates a multi-frequency soil parameter inversion methodology using Snow Microwave Radiative Transfer (SMRT) model in a temperate, agro-forested environment in Powassan, Ontario. Using multi-frequency observations from Cryospheric SAR (CryoSAR) (L-band), Radarsat Constellation Mission (RCM) (C-band), and TerraSAR-X (TSX) (X-band) acquired during the 2022/2023 winter season, soil roughness and permittivity were jointly inverted to reproduce observed backscatter. The inversion strategy, which optimizes a single time-invariant roughness per site alongside time-varying permittivity, achieved strong agreement between simulated and observed signals across frequencies (Global R2 = 0.87, RMSE =1.25 dB). Sensitivity analyses reveal a clear frequency-dependant hierarchy of controls: surface roughness dominates L-band backscatter (particularly in VV polarization), soil permittivity governs C- and X-band responses, and extending the analysis to explicitly include snow properties shows that the dominant controls progressively shift to snow microstructure and depth toward Ku-band. Comparisons with in situ measurements indicate that inverted parameters represent effective values at the radar scale; specifically, inverted roughness differs from LiDAR-derived topography, suggesting the influence of basal snow layer properties. Despite complications arising from spatial heterogeneity of soil properties including freeze/thaw cycles, the results demonstrate the feasibility of retrieving soil background parameters to support future multi-frequency snow missions such as Terrestrial Snow Mass Mission (TSMM).
Competing interests: At least one of the (co-)authors is a member of the editorial board of The Cryosphere.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(19485 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-656', Christian Mätzler, 10 Jul 2026
-
RC2: 'Comment on egusphere-2026-656', Carrie Vuyovich, 11 Aug 2026
In this study, an inversion model is used to estimate soil parameters in a snow-covered agricultural site in Ontario at L-/C-/and X-band frequencies and compared to measured values during the same time period. A sensitivity analysis is performed to better understand the impact of these parameters as well as snowpack characteristics on the microwave signals. The study finds good agreement between measured and simulated backscatter using the optimized soil parameters and lays the foundation for constraining the background signals from SWE retrievals using high frequency observations. The study is an important step in defining the algorithm for a future volume scattering snow mission. However I think some aspects could be clarified and more could be done to demonstrate how this approach would be implemented. I offer some comments and suggestions below on how I believe the analysis could be strengthened before publication.
My main comment is about the mismatch between the collected soil data and the inverted parameters. If I understand right, the soil parameters used in the inversion are optimized using the collected radar data and then fed through the inversion model and compared back to the collected radar data. So the good relationship in Figure 4 is not surprising. The relationship between the inverted parameters and the measured values is not as good (Figures 5 and 6), but it’s not clear to me how well they need to match. What if you fed the actual soil data into the inversion model, how well do you then match the observed backscatter? What if you used the optimized soil parameters to simulate Ku-band and compared that to observations? What sort of accuracy do you get (need)?
Additional comments:
Page 5, Line 110 – how big are the undisturbed zones?
Page 5, Line 114 – how close were the snow pits to the undisturbed zones? i.e. how representative is the snow compared to the undisturbed site? In many agricultural areas wind redistribution can cause large spatial variability.
Page 6, Line 128 – what is the resolution of the TSX and RCM data?
Page 6, Figure 2: This is a nice flight line coverage strategy. Where does the study area fit within these flight boxes? Can you add a thin dashed line showing the yellow box from Figure 1? Maybe add the corner reflector locations to the map too?
Page 6, Line 129 – Was vegetation removed from the bare-earth DEM?
Page 7, Table 1 – Adding the Ku-band lines to this table seems confusing since they weren’t used and the data is already provided elsewhere.
Page 7, SAR data processing – it’s unclear to me whether the SAR data was processed around the “disturbed” snow pit locations or around the undisturbed locations or both. Could you clarify? Does the 20m x 20m box include the “undisturbed” locations?
Page 7, Line 146-147 – Sorry if I missed it but I don’t see at what resolution SMRT was run. Was it run as a point model for each of the 20x20 squares?
Page 8, 3.2.1 Snowpack measurements – can you provide the snowpack (and soil) measurements, maybe in a table in a supplement. It would be nice to see what sort of variability is present. A figure might work as well for the snowpack properties
Page 8, Lines 159-161 – It’s not clear what the daily average soil permittivity on Mar 1 is being compared to. The OECP measurements? Why use the daily average instead of the data at the same time as the measurement? For comparison to the inverted values are you using the average value on the same day & time of the SAR collection?
Page 8, line 171 – “apart” instead of “a part”
Page 13, Line 263-264 – Did you try inverting MSS during snow-free conditions to compare to the MSSlidar?
Figure 5 –There is a lot of variability between measurement sites? As well as between instruments. I know there is some discussion about this later, but it’s not clear which data to trust. A couple suggestions that could be useful to look at: 1. It would be interesting to see a time series of the HP values collected coincident with the SAR data along with the inverted permittivity values from each date; and 2. It would be interesting to see this data spatially, maybe an example from one date and one sensor comparing the values across the study site. Can you comment on how much variability you expect there is in each 20x20 box? From this plot some of the results look almost exactly opposite, i.e. measurement saying frozen while the inverted value is unfrozen and vice versa.
Page 18, Figure 9 – Not clear to me how the sensitivity analysis was performed for real and imaginary permittivity when those are the value held fixed. Maybe I’m missing something. What does it mean to have frozen soil (blue line in far left column) and a permittivity > 10?
Page 20, Line 370-373 – Was there an attempt to remove surface vegetation? At this point density I would think it’s possible.
Page 20, line 381: “This value is lower than those found in this study for the Powassan site, which is likely due to the agricultural versus tundra environments.” Can you talk more about the difference in surface roughness between an agricultural and tundra site?
Page 21-22, Lines 404-422 – I think this text belongs in the TSMM Perspective section since it speaks directly to how this work supports the mission concept and currently that section is fairly small.
Page 21, Lines 404-417 – It is unclear to me from this section how the two-step retrieval strategy would work. Line 405 says “the lower frequency inversions constrain the soil background, but they do not fully determine Ku-band response” but then later (line 408) it says “Lower-frequency observations first constrain the soil background. Ku-band observations can then be interpreted with stronger focus on snow properties”. It feels like details are missing. How accurate does the background characterization need to be to do the Ku-band SWE retrievals?
Page 21, Line 417 – What type of ancillary information?
Page 22, Line 432 - “First, the inversion approach relies on the accuracy of the SMRT model, which may have inherent limitations in representing complex soil and snow interactions.” – could you use this study to say something about the accuracy of the SMRT model?
Page 23, Line 459 - “The study also highlighted the importance of constraining the soil background parameters for accurate interpretation of Ku-band backscatter…” This is mentioned several times, but I don’t think this statement is well supported. I understand that future work will evaluate SWE retrievals from this Ku-band, but in this study it could help support your conclusions if you used the inverted soil parameters from lower frequency observations to simulate Ku-band and compared that to the Ku-band observations collected by CryoSAR.
Citation: https://doi.org/10.5194/egusphere-2026-656-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 202 | 98 | 24 | 324 | 23 | 23 |
- HTML: 202
- PDF: 98
- XML: 24
- Total: 324
- BibTeX: 23
- EndNote: 23
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
See added comments