the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Jun 2025
Soil slope monitoring with Distributed Acoustic Sensing under wetting and drying cycles
Abstract. Hydromechanical soil response to moisture variations reflects complex subsurface dynamics that are critical for geoengineering, slope stability, and other soil health-related fields. While laboratory experiments have provided insights into soil behavior under varying wetness and loading conditions, field-scale observations with high spatial and temporal resolution remain limited. In this study, we present a 2 month field monitoring approach using Distributed Acoustic Sensing (DAS), which enables high-resolution, full-coverage, and continuous monitoring of a grass-covered soil slope. DAS allows for subsurface characterization and time-lapse monitoring of soil moisture dynamics using ambient noise interferometry. Furthermore, by analyzing nanostrain-scale deformation in conjunction with stress state derived from in situ soil moisture measurements, we demonstrate that DAS can track real-time volumetric changes in response to both long-term and daily cyclic moisture variations. We suggest DAS as a valuable tool for the continuous detection of moisture-driven changes in soil mechanical properties with high resolution.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(7654 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(7654 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-1725', Anonymous Referee #1, 28 Jun 2025
- AC2: 'Reply on RC1', Jiahui Kang, 24 Jul 2025
-
RC2: 'Comment on egusphere-2025-1725', Anonymous Referee #2, 08 Jul 2025
General comments:
This manuscript presents results from a two-month field deployment using Distributed Acoustic Sensing (DAS) to monitor a soil slope under natural wetting and drying cycles. The study captures both long-term and daily hydromechanical deformation, combining surface wave inversion, coda wave interferometry, and effective stress modeling. Integration with in-situ moisture data reveals soil “breathing” and progressive stiffening during drying. The paper is well written, the methods are clearly described, and the results are well illustrated. The findings are relevant for understanding moisture-driven soil behavior and slope stability. I only have a few concerns regarding the depth sensitivity of the data and the application of the rock physics model to interpret dv/v changes, which I believe should be addressed more clearly before drawing detailed interpretations.Specific comments:
• Depth sensitivity of surface wave inversion and soil moisture sensor:
Please provide more detail about the soil moisture sensors used in the study—specifically, the measurement depths, the type of sensors, the quantities they measure directly, and whether any scaling or calibration is needed to derive VWC. Since the sensors are not co-located with the fiber-optic cable and are installed in different slope settings, what is the justification for focusing on the 0.15 m depth in the comparison? Line 170 states that 0.15 m depth is chosen because of the cable installation depth, but given that DAS measures strain from propagating waves (which integrate energy over depth and wavelength), how is the physical cable depth directly related to the depth sensitivity of the seismic measurements?
In Figure 5d, the surface wave inversion results appear to have limited resolution in the upper ~2 meters, yet dv/v is compared to moisture changes at 0.15 m depth. Can you clarify this mismatch in depth sensitivity? Also, please provide an estimate of the seismic wavelength of the surface waves used for dv/v analysis. For instance, if the dominant frequency is ~10 Hz and shear wave velocity is ~200 m/s, the wavelength would be ~20 m—much deeper than 0.15 m. Do you have sensor data at greater depths to better match the depth sensitivity of the seismic measurements?• Modeling dv/v under the rock physics framework:
The manuscript outlines how effective elastic properties such as density and shear wave velocity are computed from effective stress, but it remains unclear how the effective stress is derived from the soil moisture profile. Could the authors clarify the exact steps used to convert volumetric water content and soil water potential into effective stress, especially given the complexity introduced by unsaturated versus saturated conditions?
Additionally, what reference shear wave velocity model is used in the dv/v modeling? Is it the same velocity model derived from surface wave inversion in Section 5.1? If not, please explain the differences and justification.
In Line 250, the authors state that the reduction in effective stress dominates during rainfall events. How is this conclusion supported within the model, especially considering that Figure 1 distinguishes suction stress behavior between unsaturated and saturated conditions? How are these different regimes handled in the dv/v modeling? A clearer explanation of how suction stress is represented and transitions across saturation states would help clarify the model’s assumptions and limitations.
• Figure 7a: It is confusing to present temporal variation using full shear wave velocity models, as shown in Figure 7a, given that the observed changes are on the order of ~1%. This is well within the expected uncertainty of the inversion, which appears to be significantly larger. It does not seem reasonable to interpret such small variations as physically meaningful changes in the velocity structure based on these inversion results.
• Figure 7b: What kind of smoothing or filtering was applied to the dv/v time series shown in Figure 7b? Could the apparent delay in dv/v response relative to precipitation events be an artifact of the smoothing process rather than a physical lag in the subsurface response?Citation: https://doi.org/10.5194/egusphere-2025-1725-RC2 - AC1: 'Reply on RC2', Jiahui Kang, 24 Jul 2025
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-1725', Anonymous Referee #1, 28 Jun 2025
General Comments:
The manuscript presents a multi-month DAS deployment on a grass-covered slope in central Switzerland. By pairing high-frequency (>1 Hz) ambient noise interferometry with low-frequency (<1 Hz) quasi-static strain measurements, they aim to demonstrate that DAS can be used to "track real-time volumetric changes in response to both long-term and daily cyclic moisture variations". The topic is very timely and relevant to both the DAS and geohazard communities. The integration of surface wave inversion, dv/v monitoring, and low-frequency strain analysis are technically sound. The dataset is extensive and novel, considering the longer-term duration of the low-frequency DAS measurements combined with near-surface moisture sensors. This work is an important contribution and represents a comprehensive overview of the complementary techniques that can be implemented using DAS to inform slope stability monitoring.
However, there are some critical issues that need to be addressed, relating to the author's interpretation of (1) progressive soil consolidation during drying periods, and (2) daily cyclic deformation patterns driven by moisture fluctuations, as follows:
Temperature effects: The authors indicate that the cyclical deformation patterns observed in the low-frequency DAS strain are driven by moisture fluctuations between daytime drying and nighttime moisture recovery, not by temperature variations. The effect of temperature variations are neglected after estimating that the daily temperature variations (within 1°C) would induce a strain change of about 1.1 x 10-2 millistrain which is more than two orders of magnitude smaller than the daily strain variations measured by the DAS system. However, this represents an approximation based on the properties of silica-based fiber, and does not account for the response of the DAS interrogator and fibre optic cable (see next point). Further to the above, the cyclical pattern of the low-frequency strain observations occurring across all channels (Figures 3, C1 and C2) as well as the known sensitivity of low-frequency DAS to temperature, suggest that temperature is a likely dominant contributor.
Interrogator Instrument Response: The application of low-frequency DAS for monitoring soil slope processes is still emerging. Here, the authors rely on a two-month period of continuous data acquisition using a Silixa iDAS for the measurements, which provides a measurement of strain-rate. However, the reliability and performance of DAS to measure strain and strain-rate over longer periods is still poorly understood. Ouellet et al. (2024) inferred relative displacements from the LF-DAS using another type of DAS interrogator and were able to obtain reasonable comparison with insitu displacement sensors (ShapeArrays) over a ~three-day period. In this study, there are no collocated sensors that support calibration or confirmation of the strain measurements (e.g., strain gauges, inclinometers, survey prisms), which would be important both for the interpretation and justification for the neglect of temperature. The native strain-rate measurements are integrated to derive strain over the duration of the acquisition. However, this also enables the accumulation of potential noise in the strain-rate data to accumulate over time and appear as drift. The monotonic decrease that is observed in the strain data may be a result of instrument drift, and not representative of true strain. At a minimum, the authors should address this point by including a discussion of the potential of instrument drift or consider relying on the native strain-rate measurements for their analysis and interpretation. It may also be worthwhile to compare the strain-rate measurements with the gradient of the temperature (temporal derivative) over a shorter time interval, for a more careful assessment of the relationship between the two measurements. The monotonic decrease of strain across all channels over the two-month period does not seem credible, considering both the spatial variability of the cable over the slope as well as the temporal variability considering the numerous rainfall events occurring over the period. For example, considering the nanostrain-rate sensitivity of the DAS measurements, gravity-driven processes of the slope over the two-month period with a shallowly buried cable should incur some observations of visible tension and compression in the strain data, aligning with the topographic profile along the length of the cable over the two-month period.
Cable Instrument Response: Please include the specifications of the fiber optic cable used in this study. Particularly at low frequencies, the type of cable also plays an important role in the instrument response (e.g. tight-buffered versus loose-tube, see Ouellet et al. 2024). The impact of the cable type on the response should be included in the discussion.
Gauge length effects: A channel spacing of 1 m and gauge length of 10 m is used in this study. Why were these data acquisition settings used? A gauge length of 10 m could mask any localized changes in moisture. The author's conclusions (L405) that "This enables direct field-scale observations of soil mechanical response at sub-meter resolution" are technically incorrect, considering the settings (1-m channel spacing, 10-m gauge length) used in this study. The impact of the 10-m gauge length on the results should be included in the discussion, notably in comparing or integrating these measurements with point-based sensors, as for the effective stress-strain response.
Coda wave interferometry: The dv/v estimates are computed with daily cross-correlation waveforms. As such, they cannot resolve sub-diurnal moisture cycles and therefore the claim that the author's key observations of "daily cyclic deformation patterns driven by moisture fluctuations" is supported by the dv/v analysis, appears invalid. Further to this, the dv/v are computed in the 8 to 16 Hz frequency range. The fundamental mode sensitivity kernel (Figure A2b.) appears to indicate varying dv/v sensitivity from 0 to 12 m, extending well below the partially saturated zone in the upper metres. The insitu sensors providing moisture measurements only extend up to ~1 m. The rock physics-based model of dv/v relies on a two-layer soil profile extending to a depth of only 1.38 m. Considering the known sensitivity of dv/v to greater depths (from the sensitivity kernel) it seems important to address this discrepancy more thoroughly in a discussion, or improve the model by extending to a similar depth as the dv/v.
Specific Comments:
- Topographic Profile: It would be helpful to include a topographic profile or elevation cross-section along the cable route. This would improve interpretation of both seismic and low-frequency strain data, particularly in understanding how slope angle and local relief may influence stress distribution, hydrological changes, and strain patterns.
- Spatiotemporal Strain-Rate Images: To support interpretation of the low-frequency DAS data, it would be helpful to include a spatiotemporal plot of strain-rate over the entire acquisition period. This would help readers visually assess both temporal variability and any spatially coherent patterns.
- Figure 3c: In addition to the cyclic signals (occurring over all channels), rain-induced high amplitude signals (associated with rainfall events) and quasi-static cow signals, there appears to be a fourth type of signal (positive strain-rate signals occurring at multiple channels over short time periods that are not associated with rainfall events). Please comment on these signals and whether they are attributed to moisture changes or other processes.
- Section 2.1.1, Line 106 – Spatial Heterogeneity: The authors state that “spatial heterogeneity on the slope further complicates effective stress distributions.” However, the key observations of cyclical strain observations appear generally spatially homogenous across channels. Given DAS’s advantage in spatial resolution, it would strengthen the paper to highlight any observed heterogeneity in strain or inferred stress response. Do spatial variations in the LF-DAS signal align with known heterogeneity in vegetation, soil type, or moisture content?
- Section 7.2. Effective stress-strain response. This section would benefit from greater clarity on the input data and analytical steps. Is the effective stress calculated using Equation (1) based on measured VWC and SWP? Is the associated strain derived from the LF-DAS data, and if so, is this averaged over the full array or selected channel segments? Explicitly stating this would help improve the clarity of this section.
- Depth sensitivity of dv/v: Since the dv/v analysis is performed in the 8–14 Hz frequency band, it would be useful to include an estimation of the corresponding sensitivity depth range. This would support statements such as that on Line ~325: “This suggests that seismic waves integrate the water infiltration process throughout the soil profile rather than merely reflecting near-surface saturation.” This could be completed by referring to the fundamental mode sensitivity kernel shown in Figure A2. b).
- Groundwater Level Information: Is there any information available on groundwater levels at the site? Even approximate values or nearby hydrological data would help constrain interpretations of the dv/v changes and assess whether infiltration events reach the saturated zone
- L270: Ouellet et al. (2024) follow the approach described by Leggett et al. (2022) in their study. It may be more appropriate to cite the original reference of Leggett et al. (2022) here.
- Section 6.2, L285: "The intraday strain variations (Fig. 8f) contrast with previous findings ..." Figure 8f is difficult to see within the overall figure. Consider making this into a larger figure, complete with axis labels and values for clarity.
- L405: The statement, "In conclusion, DAS integrates traditional seismic wave analysis with continuous monitoring of quasi-static deformation" should be revised to, "In conclusion, we integrate traditional seismic wave analysis with continuous. monitoring of quasi-static deformation using DAS".
Technical Corrections:
- Section 5.2, Line ~225: "We focused on ch80 for each day because of its clear separation between direct arrivals and coda waves (Fig. 6a)." Which channel(s) was cross-correlated with ch80 to obtain the cross-correlations? It does not appear to be specified in the text.
- Appendix C, Table C1. Index of fraction - should be Index of refraction
As an additional consideration for the authors', it may help to improve the clarity and impact of the manuscript by separating the seismic (>1 Hz) and low-frequency (<1 Hz) analysis into two separate studies. For instance, the extending the dv/v model over a greater depth and focusing on both the near-surface (0 to 2 m) and deeper (2 to 12 m) sensitivity of the dv/v to changes in effective stress represents an important contribution to the field of environmental seismology. Similarly, improving the analysis and interpretation of the low-frequency DAS observations, with a more rigorous evaluation of the temperature effects, alongside the cable and instrument response, represents a novel study. Separating the two analyses could help improve the clarity and impact of the overall findings.
Citation: https://doi.org/10.5194/egusphere-2025-1725-RC1 - AC2: 'Reply on RC1', Jiahui Kang, 24 Jul 2025
-
RC2: 'Comment on egusphere-2025-1725', Anonymous Referee #2, 08 Jul 2025
General comments:
This manuscript presents results from a two-month field deployment using Distributed Acoustic Sensing (DAS) to monitor a soil slope under natural wetting and drying cycles. The study captures both long-term and daily hydromechanical deformation, combining surface wave inversion, coda wave interferometry, and effective stress modeling. Integration with in-situ moisture data reveals soil “breathing” and progressive stiffening during drying. The paper is well written, the methods are clearly described, and the results are well illustrated. The findings are relevant for understanding moisture-driven soil behavior and slope stability. I only have a few concerns regarding the depth sensitivity of the data and the application of the rock physics model to interpret dv/v changes, which I believe should be addressed more clearly before drawing detailed interpretations.Specific comments:
• Depth sensitivity of surface wave inversion and soil moisture sensor:
Please provide more detail about the soil moisture sensors used in the study—specifically, the measurement depths, the type of sensors, the quantities they measure directly, and whether any scaling or calibration is needed to derive VWC. Since the sensors are not co-located with the fiber-optic cable and are installed in different slope settings, what is the justification for focusing on the 0.15 m depth in the comparison? Line 170 states that 0.15 m depth is chosen because of the cable installation depth, but given that DAS measures strain from propagating waves (which integrate energy over depth and wavelength), how is the physical cable depth directly related to the depth sensitivity of the seismic measurements?
In Figure 5d, the surface wave inversion results appear to have limited resolution in the upper ~2 meters, yet dv/v is compared to moisture changes at 0.15 m depth. Can you clarify this mismatch in depth sensitivity? Also, please provide an estimate of the seismic wavelength of the surface waves used for dv/v analysis. For instance, if the dominant frequency is ~10 Hz and shear wave velocity is ~200 m/s, the wavelength would be ~20 m—much deeper than 0.15 m. Do you have sensor data at greater depths to better match the depth sensitivity of the seismic measurements?• Modeling dv/v under the rock physics framework:
The manuscript outlines how effective elastic properties such as density and shear wave velocity are computed from effective stress, but it remains unclear how the effective stress is derived from the soil moisture profile. Could the authors clarify the exact steps used to convert volumetric water content and soil water potential into effective stress, especially given the complexity introduced by unsaturated versus saturated conditions?
Additionally, what reference shear wave velocity model is used in the dv/v modeling? Is it the same velocity model derived from surface wave inversion in Section 5.1? If not, please explain the differences and justification.
In Line 250, the authors state that the reduction in effective stress dominates during rainfall events. How is this conclusion supported within the model, especially considering that Figure 1 distinguishes suction stress behavior between unsaturated and saturated conditions? How are these different regimes handled in the dv/v modeling? A clearer explanation of how suction stress is represented and transitions across saturation states would help clarify the model’s assumptions and limitations.
• Figure 7a: It is confusing to present temporal variation using full shear wave velocity models, as shown in Figure 7a, given that the observed changes are on the order of ~1%. This is well within the expected uncertainty of the inversion, which appears to be significantly larger. It does not seem reasonable to interpret such small variations as physically meaningful changes in the velocity structure based on these inversion results.
• Figure 7b: What kind of smoothing or filtering was applied to the dv/v time series shown in Figure 7b? Could the apparent delay in dv/v response relative to precipitation events be an artifact of the smoothing process rather than a physical lag in the subsurface response?Citation: https://doi.org/10.5194/egusphere-2025-1725-RC2 - AC1: 'Reply on RC2', Jiahui Kang, 24 Jul 2025
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 980 | 244 | 25 | 1,249 | 25 | 38 |
- HTML: 980
- PDF: 244
- XML: 25
- Total: 1,249
- BibTeX: 25
- EndNote: 38
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Fabian Walter
Tobias Halter
Patrick Paitz
Andreas Fichtner
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(7654 KB) - Metadata XML
General Comments:
The manuscript presents a multi-month DAS deployment on a grass-covered slope in central Switzerland. By pairing high-frequency (>1 Hz) ambient noise interferometry with low-frequency (<1 Hz) quasi-static strain measurements, they aim to demonstrate that DAS can be used to "track real-time volumetric changes in response to both long-term and daily cyclic moisture variations". The topic is very timely and relevant to both the DAS and geohazard communities. The integration of surface wave inversion, dv/v monitoring, and low-frequency strain analysis are technically sound. The dataset is extensive and novel, considering the longer-term duration of the low-frequency DAS measurements combined with near-surface moisture sensors. This work is an important contribution and represents a comprehensive overview of the complementary techniques that can be implemented using DAS to inform slope stability monitoring.
However, there are some critical issues that need to be addressed, relating to the author's interpretation of (1) progressive soil consolidation during drying periods, and (2) daily cyclic deformation patterns driven by moisture fluctuations, as follows:
Temperature effects: The authors indicate that the cyclical deformation patterns observed in the low-frequency DAS strain are driven by moisture fluctuations between daytime drying and nighttime moisture recovery, not by temperature variations. The effect of temperature variations are neglected after estimating that the daily temperature variations (within 1°C) would induce a strain change of about 1.1 x 10-2 millistrain which is more than two orders of magnitude smaller than the daily strain variations measured by the DAS system. However, this represents an approximation based on the properties of silica-based fiber, and does not account for the response of the DAS interrogator and fibre optic cable (see next point). Further to the above, the cyclical pattern of the low-frequency strain observations occurring across all channels (Figures 3, C1 and C2) as well as the known sensitivity of low-frequency DAS to temperature, suggest that temperature is a likely dominant contributor.
Interrogator Instrument Response: The application of low-frequency DAS for monitoring soil slope processes is still emerging. Here, the authors rely on a two-month period of continuous data acquisition using a Silixa iDAS for the measurements, which provides a measurement of strain-rate. However, the reliability and performance of DAS to measure strain and strain-rate over longer periods is still poorly understood. Ouellet et al. (2024) inferred relative displacements from the LF-DAS using another type of DAS interrogator and were able to obtain reasonable comparison with insitu displacement sensors (ShapeArrays) over a ~three-day period. In this study, there are no collocated sensors that support calibration or confirmation of the strain measurements (e.g., strain gauges, inclinometers, survey prisms), which would be important both for the interpretation and justification for the neglect of temperature. The native strain-rate measurements are integrated to derive strain over the duration of the acquisition. However, this also enables the accumulation of potential noise in the strain-rate data to accumulate over time and appear as drift. The monotonic decrease that is observed in the strain data may be a result of instrument drift, and not representative of true strain. At a minimum, the authors should address this point by including a discussion of the potential of instrument drift or consider relying on the native strain-rate measurements for their analysis and interpretation. It may also be worthwhile to compare the strain-rate measurements with the gradient of the temperature (temporal derivative) over a shorter time interval, for a more careful assessment of the relationship between the two measurements. The monotonic decrease of strain across all channels over the two-month period does not seem credible, considering both the spatial variability of the cable over the slope as well as the temporal variability considering the numerous rainfall events occurring over the period. For example, considering the nanostrain-rate sensitivity of the DAS measurements, gravity-driven processes of the slope over the two-month period with a shallowly buried cable should incur some observations of visible tension and compression in the strain data, aligning with the topographic profile along the length of the cable over the two-month period.
Cable Instrument Response: Please include the specifications of the fiber optic cable used in this study. Particularly at low frequencies, the type of cable also plays an important role in the instrument response (e.g. tight-buffered versus loose-tube, see Ouellet et al. 2024). The impact of the cable type on the response should be included in the discussion.
Gauge length effects: A channel spacing of 1 m and gauge length of 10 m is used in this study. Why were these data acquisition settings used? A gauge length of 10 m could mask any localized changes in moisture. The author's conclusions (L405) that "This enables direct field-scale observations of soil mechanical response at sub-meter resolution" are technically incorrect, considering the settings (1-m channel spacing, 10-m gauge length) used in this study. The impact of the 10-m gauge length on the results should be included in the discussion, notably in comparing or integrating these measurements with point-based sensors, as for the effective stress-strain response.
Coda wave interferometry: The dv/v estimates are computed with daily cross-correlation waveforms. As such, they cannot resolve sub-diurnal moisture cycles and therefore the claim that the author's key observations of "daily cyclic deformation patterns driven by moisture fluctuations" is supported by the dv/v analysis, appears invalid. Further to this, the dv/v are computed in the 8 to 16 Hz frequency range. The fundamental mode sensitivity kernel (Figure A2b.) appears to indicate varying dv/v sensitivity from 0 to 12 m, extending well below the partially saturated zone in the upper metres. The insitu sensors providing moisture measurements only extend up to ~1 m. The rock physics-based model of dv/v relies on a two-layer soil profile extending to a depth of only 1.38 m. Considering the known sensitivity of dv/v to greater depths (from the sensitivity kernel) it seems important to address this discrepancy more thoroughly in a discussion, or improve the model by extending to a similar depth as the dv/v.
Specific Comments:
Technical Corrections:
As an additional consideration for the authors', it may help to improve the clarity and impact of the manuscript by separating the seismic (>1 Hz) and low-frequency (<1 Hz) analysis into two separate studies. For instance, the extending the dv/v model over a greater depth and focusing on both the near-surface (0 to 2 m) and deeper (2 to 12 m) sensitivity of the dv/v to changes in effective stress represents an important contribution to the field of environmental seismology. Similarly, improving the analysis and interpretation of the low-frequency DAS observations, with a more rigorous evaluation of the temperature effects, alongside the cable and instrument response, represents a novel study. Separating the two analyses could help improve the clarity and impact of the overall findings.