Evaluating GNSS-T VOD sensitivity to plant water dynamics, rainfall interception, and dew in a coniferous forest

Schellenberg, Konstantin; Paulus, Sinikka J.; Queck, Ronald; Chaparro, David; Binks, Oliver; Mencuccini, Maurizio; Paligi, Sharath S.; Hartmann, Henrik; Schmullius, Christiane; Dubois, Clémence; Jagdhuber, Thomas

doi:10.5194/egusphere-2026-1751

Preprints

https://doi.org/10.5194/egusphere-2026-1751

Preprints

23 Apr 2026

| 23 Apr 2026

Evaluating GNSS-T VOD sensitivity to plant water dynamics, rainfall interception, and dew in a coniferous forest

Konstantin Schellenberg, Sinikka J. Paulus, Ronald Queck, David Chaparro, Oliver Binks, Maurizio Mencuccini, Sharath S. Paligi, Henrik Hartmann, Christiane Schmullius, Clémence Dubois, and Thomas Jagdhuber

Abstract. Monitoring forest canopy water is essential for understanding drought response and interception losses under climate change. At short timescales, canopy water is partitioned among internal plant water storage (S_p), rainfall interception (S_i), and dew (S_d), which together regulate plant functioning, canopy evaporation, and precipitation partitioning, yet are rarely observed simultaneously. GNSS transmissometry (GNSS-T) has recently emerged as a low-cost, continuous, stand-scale method to observe L-band vegetation optical depth (VOD) from signal attenuation. However, interpreting GNSS-T VOD remains difficult because the signal integrates multiple water pools and is similarly affected by biomass, canopy structure, and measurement noise.

Here, we applied GNSS-T in a mature Picea abies stand in Tharandt, Germany, during the 2024 growing season to separate canopy water storage into S_i, S_d, and S_p. Rainfall interception was simulated with the multilayer Penman–Rutter model CanWat and used to calibrate the empirical VOD–water-storage relationship and to convert the GNSS-T noise floor into an equivalent storage detectability threshold. GNSS-T VOD tracked interception storage robustly and approximately linearly at both 30-min and event scales (R² = 0.63/0.79), and modeled S_i explained VOD variability better than gross precipitation alone. The inferred attenuation coefficient b was physically consistent with the reported L-band values but varied seasonally, indicating that time-varying calibration is preferable to a fixed relationship. Dew-related wetting signals were distinguishable in VOD and yielded plausible mean nightly amounts, but short event duration and high noise caused unrealistic extremes and limited detectability. Diurnal changes in internal plant water storage were not directly detectable at sub-daily scales, indicating that realized variations in S_p remained below the GNSS-T noise floor during the study period which we could show using trait-based estimates of expected maximum plant water loss under non-stressed conditions.

This storage-versus-noise framework provides a practical way to benchmark the hydraulic sensitivity of GNSS-T across sites that differ in biomass, hydraulic strategy, and climate, and further highlights noise reduction as a prerequisite for plant-hydraulic applications, especially in low-biomass ecosystems. This study further promotes GNSS-T VOD as a robust monitoring instrument for resolving sub-event interception storage, a potential avenue for constraining hydrological models.

Received: 28 Mar 2026 – Discussion started: 23 Apr 2026

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.

Download & links

Preprint (PDF, 6427 KB)

Supplement (2369 KB)

Download & links

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-1751', Anonymous Referee #1, 05 Jun 2026
The authors present in this work a detailed and thorough analysis of estimating plant water status with the use of satellite (GNSS-T) retrieved vegetation optical depth (VOD) measurements. By focusing on sub-diurnal timescales, they open new ways of using this method to estimate plant water status, which is a highly relevant variable for many environmental related scientific fields.
The manuscript is well-written, and both the methodology and interpretation are generally well-articulated and clearly presented. Below, I provide several general, specific, and technical comments that I recommend addressing prior to publication.
General comments
Instrumental set-up: I wonder why the authors did not deploy leaf wetness sensors. Though I understand their hesitance as described in L588, deploying such sensors might have provided an additional constraint for defining canopy wetness (see also comment #3). This is not criticism per se on the methodology, but I am interested in whether this was an active choice (i.e. if the authors have objections against the use of leaf wetness sensors for such applications) or simply deal with what is available at the measurement site.

Sect. 2.4 and 3.1: The derived attenuation coefficient b is based on Si modelled with the CanWat model. As b is a key parameter for the entire manuscript, I think it is important that here it is discussed properly whether the modelled Si, and by extension b, is expected to contain a random and/or deterministic error. In fact, as later described in Sect. 3.2, dew generally seems to be overpredicted. Could this be due to an error in b, making its extrapolation to dew events result in an overestimated Sd, and does accounting for this error make the Sd more plausible? Additionally, in Sect. 4.1 and Fig. 10 the authors show that b has a seasonal component. Could it be that seasonal bias for dew events also results in affect the overprediction of dew? Perhaps the authors could report the uncertainty on the derived b.

Sect. 2.5: In this section, the authors define criteria during which they assume presence of dew. By setting these criteria as strict, they aim to prevent false positives in their analysis. However, I wonder if such strict criteria may not result in too many false negatives. The authors reflect on this in L583-593, but do not mention if these false negatives could also impact the comparison between dew and non-dew state earlier on in the manuscript. These false negatives might in fact be present in Fig. 6a, where the tail of the non-dew pdf extends to relatively high VOD values. Perhaps the authors could investigate an extra classification with “potential wet canopy due to dew”, or maybe just exclude these points from the comparison with dew events.

Specific comments
L137: Please specify “low variability in LAI throughout the year”

S3a: When shifting from pooled to per-sat, the density peak shifts towards a lower VOD (besides the lower noise). Can the authors comment on this?

L242: is b (or VOD for that matter) affected by the active rainfall (rate) itself (I’m not an expert on this).

L375: Perhaps the authors could also mention the canopy scale LAI somewhere, as this would provide valuable information for other studies that might not have a 3D representation of the canopy.

L458: Please clarify whether (part of) these 19 forest ecosystems have a comparable climate to Tharandt.

L467: Are these values per unit area ground surface or unit area plant or leaf surface? In fact, this could be clarified throughout the manuscript.

6: Both panels show negative dew-yield. Is this in fact evaporation after false negatives? Or just because of random noise (if so, perhaps the authors can add the detection limit in this figure for further clarification).

L526-527: Does this mean that all dew events are 6 hr or longer? Are events that have a duration of e.g. 5 hr also averaged to 6hr VOD values? What is the average length of the dew events according to the criteria stated in Sect. 2.5?

Technical comments
L324: Please explicitly define θ

L426: “Note, however, that the those ….” --> remove “the”

L461: Should reference not be to Fig. S9?

L465: Please explicitly define AUC

L473: “(…) the a fair comparison (…)” --> remove “the” or “a”

L473: “Generally, the there is a (…)” --> remove “the”

L574: “reported observations” --> “reported observations in literature”
Citation: https://doi.org/10.5194/egusphere-2026-1751-RC1
RC2: 'Comment on egusphere-2026-1751', Anonymous Referee #2, 06 Jun 2026

General comments
The paper does exactly what the title suggests, i.e. evaluates GNSS-T VOD sensitivity to plant water dynamics, rainfall interception and dew in a coniferous forest. The authors conclude that, at this site, GNSS-T VOD can be related to interception amounts, that nightly dew amounts estimated from GNSS-T VOD were plausible but that diurnal plant water storage variations at Tharandt were too close to the estimated noise floor to be detectable. GNSS-T VOD is an increasingly popular tool in biogeosciences, and the network of installations is growing. The analysis is limited to a single stand in Tharandt. However, it would be very challenging to conduct and report such a detailed analysis as many sites in a single manuscript. Furthermore, the methodology itself is a key contribution. This kind of analysis and methodological development is essential to ensure that GNSS-T VOD observations are understood before they are used in applications, so the outcomes of the study should be of interest to the readership of BG.
The manuscript is generally well structured. The abstract provides a conscise and clear summary. The language is fluent and precise, though it could be clearer in places (specific comments below).
A strength of the manuscript is the thoughtful and creative approach to quantifying the water storage terms. The analysis is informative and insightful for the community developing GNSS-T as a tool for vegetation monitoring. A limitation of the manuscript is the degree to which the results depend on the validity of the model for storage change, or relative storage index. In addition to the potential scale mismatch between the sap flow measurements of uptake and the transpiration estimate from the CanWat model, there is the uncertainty in T itself. The manuscript would be strengthened by a validation of the method outlined in 2.6.1 to account for its influence on uncertainty in the storage terms (S_i, S_d, S_p).
In addition, the description of the S_p quantity is quite confusing and difficult to follow. The conclusions in terms of interception and dew are quite clear but an improved description of the S_p-related terms would make this aspect easier to follow and accept. Finally, the authors should discuss the uncertainty and limitations of the models and assumptions described in Section 2 on the results and conclusions.

Specific comments
A clearer description of some critical details of the data processing need to be included. They are directly relevant to the assumptions made, the interpretation of the results and the conclusions drawn. In particular, the calculation of the baseline per satellite and sky sector (line 172).
A key assumption in this study is that noise related to the receiver and wave transmission is purely random, and that all deterministic noise is due to satellite tracks and canopy distribution. The authors should demonstrate that this is the case for their receiver and installation.
A stationary diurnal Fourier series is fit on dry periods as a rough proxy for plant-water dynamics. The details of this fit and its validity are very relevant for Equation 5, and the calculation of the noise threshold. Please provide the details of this fit in the Appendix. The validity of the fit should be demonstrated, and its impact on the noise threshold should be discussed, as well as the implications on the conclusions regarding detectability. In particular, it is unclear if this term has been removed from the VOD(t)* series used to assess the detectability of S_p.
A substantial part of the analysis relies on estimates of canopy water storage S_i, and transpiration T from the CanWat model. It is stated in line 216 that this model has been applied at Tharandt and provide a reference, but the authors should include relevant details of the model validation. In particular, the uncertainty in the S_i and T estimates should be included in Section 2, and taken into consideration in Sections 3 and 4.
Lines 243-245 are unclear. First it is stated that b is an effective parameter integrating over the distribution of GNSS-T incidence angles. Then it is stated that the incidence angle is constant. Please clarify, connecting these statements to how VOD at any timestep is obtained by aggregating data in space (including across incidence angles) and time.
Figure 8 and related discussion: How should S_p*noc values be interpreted? “S_p is a dimensionless quantity, resulting from the integration of percentages of fluxes”. “S_p is normalized agsint the pre-event baseline value”. It is difficult to follow from Section 2 what this quantity represents or how it has this range of values.
Lines 400-410: How is this connected to the use of the stationary diurnal Fourier series discussed in line 194? Am I correct to assume that this Fourier series is used purely to estimate the noise? How does it compare to the VOD*(t) in equation 16?
Technical corrections
Lines 69-76: It would be appropriate to acknowledge efforts prior to the publication of Humphrey and Frankenberg (2023) in which several groups in the USA and Europe (and maybe others) provided proof of concept that GNSS transmissivity could be used to monitor attenuation due to vegetation using stationary and mobile sensors.
Line 145: “… and three m coaxial cables” (?). Something is missing in this sentence.
Figure 1 caption: “… maximum extent …”
Line 206 - 211: It is unusual to express VWC is mm. Please explain here that this is motivated by the use of VOD for interception in which VWC refers to mm of interception rather than the more conventional (internal) vegetation water content which is generally expressed in kg m^-2.
Line 204: “To express the threshold metric in mm, so that it might be compared to interception amounts, a linear model is used: “
Line 207: the VOD=b.VWC relationship of Jackson and O’Neill, 1990 was obtained for internal vegetation water content of a growing crop, changes in which were primarily due to biomass increase associated with growth. There is a implicit assumption in Line 207 that the same relationship applies without reservation to free water on the surface of the canopy. Acknowledge and justify this assumption, even if it is just a necessary, pragmatic one.
Line 291: Should this say crown? The later discussion suggests that the VOD is also sensitive to water in the stem. Please clarify/correct.
Section heading 2.6.1. “Dynamic plant water(or storage) proxy”?
Line 337: As an alternative
Line 393: “In a comparison maximum-change framework” (?)
Line 395: “. However, due to ….”
Table 3 caption: “used for estimation of the …”
Line 405: What is meant by baseline-normalized here? If you just mean normalized according to equations 16 and 17, it would be better to avoid the term baseline to avoid any confusion between this and the baseline correction mentioned in line 170.
Lines 208-209: Should the b in brackets be L?
Line 431: … influence of the spatial domain is obvious …
Line 528: Re-write sentence.
Line 531: “This, however, contrasts empirical observations shown before”. Correct grammar, but also specify the empirical observations to which you refer.
Figure 8, caption: Where is the *dry (wet canopy) on this figure?
Line 548 and line 644: The expected range of b values from the cited papers refers to the relationship between VOD and (internal) VWC. While the values may be similar, there is no reason to assume they are expected.
Line 551: b is an effective parameter in general.
Line 638: bridge the gap
Line 641: What is storage in this sentence? Are internal and surface water storage lumped together and assumed to be equivalent?

Citation: https://doi.org/10.5194/egusphere-2026-1751-RC2

Supplement

https://doi.org/10.5194/egusphere-2026-1751-supplement

Viewed

Total article views: 582 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
403	156	23	582	36	20	21

HTML: 403
PDF: 156
XML: 23
Total: 582
Supplement: 36
BibTeX: 20
EndNote: 21

Views and downloads (calculated since 23 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	135	60	9	204
May 2026	233	82	11	326
Jun 2026	35	14	3	52

Cumulative views and downloads (calculated since 23 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	135	60	9	204
May 2026	233	82	11	326
Jun 2026	35	14	3	52

Viewed (geographical distribution)

Total article views: 580 (including HTML, PDF, and XML) Thereof 580 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 25 Jun 2026

Short summary

We evaluated Global Navigation Satellite System (GNSS) signals to monitor forest water storage, helping to close an important observational gap in plant hydraulics and micrometeorology at stand-scale. We show that signal losses in the canopy are strongly linked to three-dimensionally modeled rainfall interception. However, detecting subtle fluxes, such as internal water dynamics, dew and intercepted light rain is inhibited by signal noise, and the hydraulics properties of the ecosystem.


Total:	0
HTML:	0
PDF:	0
XML:	0