the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Jun 2026
Dynamic Satellite-Derived Vegetation and Radiation Inputs Advance Continental-Scale Hydrological Simulation Across China
Abstract. Global vegetation greening is reshaping water and energy cycles, challenging land surface hydrological modeling. Satellite remote sensing provides dynamic observations of vegetation and radiation, offering a pathway to improve simulations. However, models often rely on static parameters, failing to capture critical transient biogeophysical feedbacks. This study quantifies the impact of integrating remote sensing data—leaf area index, fractional vegetation cover, albedo, and downward radiation—into the Variable Infiltration Capacity (VIC) model across China. We evaluated simulations against observed runoff from 50 stations, evapotranspiration (ET) from >40 flux sites, and satellite ET products. The dynamic data-driven VIC model accurately simulated runoff and ET. In ungauged basins, simple parameter transfer achieved Nash-Sutcliffe efficiency >0.6 for runoff. Using static vegetation parameters induced substantial biases: a national-scale ET underestimation of 5 % (20 mm yr−1) and runoff overestimation of 14 % (29 mm yr−1). In the rapidly greening basin (i.e., Pearl River Basin), dynamic vegetation data corrected ET and runoff biases by ~70 mm yr−1. Remote sensing radiation data offered limited improvement, likely due to the model's inherent radiation estimation capability. This work provides conclusive evidence that dynamic remote sensing data, particularly vegetation parameters, are crucial for accurate large-scale hydrological simulation in changing environments, offering a practical framework for data-sparse regions.
- Preprint
(3463 KB) - Metadata XML
-
Supplement
(338 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2026-2505', Nima Zafarmomen, 05 Jul 2026
-
CC2: 'Reply on CC1', Dawei Peng, 23 Jul 2026
Dear Reviewer,
We sincerely thank the commenter for the constructive and insightful comments. We appreciate the positive assessment of the relevance and timeliness of our study, and we fully agree that several aspects of the manuscript should be clarified and strengthened in the revised version. Our detailed responses are provided below.
Comment 1:
The manuscript should more clearly state its main novelty. The study includes dynamic vegetation inputs, dynamic radiation inputs, VIC calibration, parameter transfer to ungauged basins, and multi-source ET validation. The authors should clarify whether the primary contribution is the dynamic vegetation–radiation forcing framework, the continental-scale evaluation, or the parameter-transfer approach.
Response 1:
We thank the reviewer for this helpful comment. The primary novelty of our study is the first continental‑scale demonstration that incorporating satellite‑derived dynamic vegetation and radiation forcings significantly improves VIC model simulations of evapotranspiration and runoff, with China as a critical case due to its global hotspot status for vegetation greening. The VIC calibration, parameter transfer, and multi‑source ET validation are supporting components that enable this core attribution analysis, rather than being the main contribution themselves. We will revise the Introduction and Conclusions to clearly state this hierarchy.
Comment 2:
The treatment of dynamic vegetation parameters should be described more explicitly. Since LAI, FVC, and albedo directly affect canopy interception, transpiration, soil evaporation, and runoff partitioning, the authors should explain more clearly how each variable influences the VIC model equations and which parameter contributes most to the final improvement.
Response 2:
We appreciate the suggestion to describe the treatment of dynamic vegetation parameters in more detail. Specifically, LAI directly influences canopy interception, canopy resistance, transpiration, and vegetation shading; FVC controls the partitioning between vegetated and bare-soil fractions, thereby affecting transpiration and soil evaporation; while albedo regulates the absorption of incoming shortwave radiation and consequently influences the available surface energy for evapotranspiration. In the revised manuscript, we will expand the methodological description of how LAI, FVC, and albedo are incorporated into VIC and how they affect key hydrological processes. We will briefly discuss the relative roles of LAI, FVC, and albedo in contributing to the simulated improvements.
Comment 3:
The parameter-transfer method for ungauged basins is promising, but the multiple linear regression approach should be justified in more detail. The authors should discuss whether nonlinear or machine-learning regionalization methods were considered, and why MLR is sufficient for transferring VIC parameters across diverse hydroclimatic regions.
Response 3:
Thank you for this helpful suggestion. In the revised manuscript, we will provide a clearer explanation for adopting multiple linear regression (MLR) as a parsimonious, interpretable, and computationally efficient regionalization approach for large-scale hydrological applications. Given the limited number of calibrated basins, more flexible nonlinear or machine-learning-based methods may be prone to overfitting, although they could offer improved transfer performance when larger calibration datasets are available. We also acknowledge that such approaches may be particularly advantageous in highly heterogeneous hydroclimatic regions. However, a systematic comparison of different regionalization methods is beyond the scope of the present study. We will also clarify the limitations of the current MLR-based transfer method and discuss possible future extensions using nonlinear regionalization approaches.
Comment 4:
The manuscript reports relatively strong runoff performance, but ET validation against flux towers shows notable bias at some sites. The authors should discuss the possible causes of this bias, including scale mismatch between flux-tower footprints and model grids, uncertainty in satellite vegetation products, and limitations in VIC’s ET partitioning.
Response 4:
Thank you for this insightful comment. We appreciate the comment regarding the ET validation against flux-tower observations and acknowledge that the notable biases observed at some sites were not sufficiently discussed in the current manuscript. These may include the spatial scale mismatch between flux-tower footprints and the VIC grid cells, uncertainties in satellite-derived vegetation products, energy-balance closure issues in eddy-covariance observations, and limitations in VIC’s representation of ET partitioning, irrigation, groundwater access, root-zone processes, and local vegetation physiological responses. We will incorporate a more comprehensive discussion of these factors to better explain the remaining ET biases and the limitations of the current modeling framework.
Comment 5:
The comparison among VIC-simulated ET and satellite ET products is useful, but the authors should avoid treating satellite products as fully independent truth. Since several ET products also rely on remote sensing vegetation information, some agreement may partly reflect shared input data or similar assumptions.
Response 5:
Thank you for this valuable comment. We agree that satellite-based evapotranspiration products should not be treated as fully independent ground truth. In the revised manuscript, we will adjust the wording accordingly and refer to these datasets as satellite-based ET reference products. The comparison with these products is intended only as a reginal-scale evaluation to complement the validation based on runoff observations and eddy covariance measurements, rather than as independent validation. We will explicitly acknowledge that some ET products rely on remotely sensed vegetation information and related energy-balance or empirical assumptions, so the agreement between VIC simulations and satellite ET products may partly reflect shared input information or similar modeling assumptions. Nevertheless, the overall consistency still provides complementary evidence supporting the reliability of the simulated spatial patterns. We will therefore interpret these comparisons mainly as assessments of spatial consistency and product-level agreement, while recognizing their limitations.
Comment 6:
Some figures are information-rich but visually dense, particularly Figures 3, 6, 7, 9, and 12. The authors should improve readability by increasing font sizes, simplifying legends where possible, and ensuring that all color scales, units, and scenario labels are clearly interpretable.
Response 6:
Thank you for this helpful suggestion. In the revised manuscript, we will improve these figures by increasing font sizes, simplifying legends where appropriate, clarifying color scales, units, and experiment/scenario labels, and enhancing the overall visual presentation. We will also consider moving overly detailed information to the Supplement where appropriate to improve the clarity of the main figures.
Comment 7:
The authors are strongly recommended to cite recent work on assimilating satellite-based vegetation information into coupled surface water–groundwater modeling. In particular, Zafarmomen, N., Alizadeh, H., Bayat, M., Ehtiat, M., and Moradkhani, H.: Assimilation of sentinel-based leaf area index for modeling surface-ground water interactions in irrigation districts, Water Resources Research, 60(10), e2023WR036080, 2024, is highly relevant. That study directly addresses the assimilation of Sentinel-based LAI into a hydrological modeling framework and demonstrates the value of high-resolution vegetation information for improving water-cycle simulations, including evapotranspiration, irrigation-related processes, groundwater recharge, and surface water–groundwater interactions.
Response 7:
Thanks for suggesting the valuable work by Zafarmomen et al. (2024). We agree that this work is highly relevant to the broader topic of incorporating satellite-based vegetation information into hydrological modeling. In the revised manuscript, we will cite this study and other relevant papers to better position our work within the recent literature. Moreover, we will clarify the distinction between their Sentinel-based LAI data assimilation framework for coupled surface water–groundwater modeling and our use of satellite-derived vegetation variables as dynamic forcings in VIC for evaluating continental-scale impacts on runoff and evapotranspiration simulations across China.
Overall, we are grateful for these helpful suggestions. We believe they will help us improve the clarity, methodological transparency, and positioning of the manuscript in the revised version.
Citation: https://doi.org/10.5194/egusphere-2026-2505-CC2
-
CC2: 'Reply on CC1', Dawei Peng, 23 Jul 2026
-
RC1: 'Comment on egusphere-2026-2505', Anonymous Referee #1, 13 Aug 2026
This study addresses a timely topic by integrating dynamic satellite-derived vegetation and radiation products into a continental-scale hydrological model. The model calibration appears to be carefully conducted, and the manuscript presents an extensive set of simulations across China. I appreciate the considerable effort devoted to data collection, model development, and visualization. Nevertheless, I feel that the scientific framing and overall story could be further strengthened.
Here are my comments:
1. The scientific question and novelty need to be strengthened. The manuscript is currently framed around the question of whether dynamic vegetation and radiation inputs improve hydrological simulations. However, this premise is not entirely convincing because the importance of temporally varying vegetation parameters has long been recognized in land surface hydrology. Vegetation variables such as LAI directly regulate canopy interception, transpiration, and surface energy exchange; therefore, it is theoretically expected that replacing static vegetation parameters with dynamic observations would improve model performance. Consequently, the primary conclusion “dynamic vegetation improves hydrological simulations” is largely consistent with theoretical expectations.
I encourage the authors to move beyond the question of whether dynamic vegetation improves simulations and instead address questions such as “Under what climatic or ecological conditions do dynamic vegetation inputs become essential?" Then, the ms would be elevated from a useful model evaluation to one providing broader process-based insight into vegetation-hydrology interactions.
2. The added value of dynamic remote sensing inputs is not sufficiently quantified. The manuscript evaluates the dynamic VIC model against observations, while separately reporting the differences between the dynamic and static simulations. However, it does not present the performance of the static configuration against observations. Consequently, readers cannot directly determine the actual improvement achieved by introducing dynamic vegetation and radiation inputs. For example, are the reported reductions in ET (5%) and runoff (14%) bias sufficiently large to justify incorporating multiple dynamic remote sensing products?
3. The experimental design and calibration strategy require further clarification. The current scenario design allows qualitative comparison between dynamic and static vegetation/radiation forcings, but it does not completely isolate their individual contributions. I am uncertain whether the comparison between the dynamic and static simulations is entirely fair. Model parameters were calibrated using the fully dynamic configuration (S1), and the same parameter set was subsequently applied to the static scenarios (S2 and S3). This implicitly assumes that parameters optimized for the dynamic forcing are equally optimal for the static forcing. Model calibration often compensates for deficiencies in forcing data or model structure. Therefore, part of the reported differences between S1 and S2/S3 may reflect parameter compensation rather than solely the effects of dynamic vegetation or radiation inputs.
4. The Results and Discussion should be reorganized to better support the scientific story. The Results section devotes considerable space to model calibration, validation, parameter transfer, and comparisons with flux towers and satellite ET products. While these analyses demonstrate that the dynamic VIC model performs reasonably well, they contribute relatively little to the manuscript’s primary objective (quantifying the value of dynamic vegetation and radiation inputs). By contrast, the comparison between dynamic and static simulations occupies only a relatively small portion of the Results.
Similarly, the Discussion focuses primarily on established hydrological mechanisms rather than interpreting the new insights generated by this study. One of the most interesting results is that dynamic radiation forcing contributes much less than dynamic vegetation forcing, yet this finding receives relatively limited discussion.
Two minor comments:
1. L30. 'hydrological simulation' of what? Please specify (e.g., runoff, ET, water balance).
2. In the Introduction section, the review of first-, second-, and third-generation land surface models is relatively lengthy and only loosely connected to the main objective of the study. Consider shortening this section and focusing more directly on the scientific gap.
Citation: https://doi.org/10.5194/egusphere-2026-2505-RC1 -
AC1: 'Reply on RC1', Xianhong Xie, 21 Sep 2026
Response to Reviewer1
Dear Reviewer,
We sincerely thank you for your careful and constructive evaluation of our manuscript. We appreciate your comments on the scientific focus, the effects of dynamic vegetation and radiation forcings, the experimental design, and the organization of the Results and Discussion, as well as other aspects of the study. Your comments have helped us clarify the main focus of the study and improve the presentation of our results. Our detailed responses to each of your comments are provided below.
Comment 1:
This study addresses a timely topic by integrating dynamic satellite-derived vegetation and radiation products into a continental-scale hydrological model. The model calibration appears to be carefully conducted, and the manuscript presents an extensive set of simulations across China. I appreciate the considerable effort devoted to data collection, model development, and visualization. Nevertheless, I feel that the scientific framing and overall story could be further strengthened.
Here are my comments:
The scientific question and novelty need to be strengthened. The manuscript is currently framed around the question of whether dynamic vegetation and radiation inputs improve hydrological simulations. However, this premise is not entirely convincing because the importance of temporally varying vegetation parameters has long been recognized in land surface hydrology. Vegetation variables such as LAI directly regulate canopy interception, transpiration, and surface energy exchange; therefore, it is theoretically expected that replacing static vegetation parameters with dynamic observations would improve model performance. Consequently, the primary conclusion “dynamic vegetation improves hydrological simulations” is largely consistent with theoretical expectations.
I encourage the authors to move beyond the question of whether dynamic vegetation improves simulations and instead address questions such as “Under what climatic or ecological conditions do dynamic vegetation inputs become essential?" Then, the ms would be elevated from a useful model evaluation to one providing broader process-based insight into vegetation-hydrology interactions.
Response 1:
We thank the reviewer for the constructive comments. We agree that the paper should move beyond whether dynamic vegetation improves simulations and instead address “under what climatic or ecological conditions dynamic vegetation inputs become essential”.
We would like to emphasize that the core novelty of this study lies in quantifying “how much” dynamic vegetation and radiation parameters improve VIC simulations under vegetation greening and radiation dynamics. This quantification provides the essential basis for identifying where and when dynamic vegetation inputs matter most. Following the reviewer's suggestion, we will further diagnose the climatic and ecological conditions under which dynamic vegetation inputs become essential, thereby offering broader process-based insight into vegetation–hydrology interactions.
Comment 2:
The added value of dynamic remote sensing inputs is not sufficiently quantified. The manuscript evaluates the dynamic VIC model against observations, while separately reporting the differences between the dynamic and static simulations. However, it does not present the performance of the static configuration against observations. Consequently, readers cannot directly determine the actual improvement achieved by introducing dynamic vegetation and radiation inputs. For example, are the reported reductions in ET (5%) and runoff (14%) bias sufficiently large to justify incorporating multiple dynamic remote sensing products?
Response 2:
We agree with the reviewer. The current manuscript reports the dynamic simulation performance against observations and separately reports dynamic–static differences, but does not show the static configuration's performance against observations. Without this, readers cannot directly judge the actual improvement gained from dynamic vegetation and radiation inputs, nor whether the reported ET (5%) and runoff (14%) bias reductions justify incorporating multiple dynamic remote sensing products.
We will therefore add the static-parameter results to the model evaluation and validation section, presenting them alongside the dynamic results so that the improvement achieved by the dynamic configuration can be assessed directly.
Comment 3:
The experimental design and calibration strategy require further clarification. The current scenario design allows qualitative comparison between dynamic and static vegetation/radiation forcings, but it does not completely isolate their individual contributions. I am uncertain whether the comparison between the dynamic and static simulations is entirely fair. Model parameters were calibrated using the fully dynamic configuration (S1), and the same parameter set was subsequently applied to the static scenarios (S2 and S3). This implicitly assumes that parameters optimized for the dynamic forcing are equally optimal for the static forcing. Model calibration often compensates for deficiencies in forcing data or model structure. Therefore, part of the reported differences between S1 and S2/S3 may reflect parameter compensation rather than solely the effects of dynamic vegetation or radiation inputs.
Response 3:
Thanks for this insightful comment. In the current experimental design, the seven VIC soil parameters were calibrated under the fully dynamic configuration (S1) and subsequently applied unchanged to the static-input configurations. To ensure a fairer comparison, we will conduct additional configuration-specific calibrations for each static-input configuration using the same calibration period, parameter ranges, objective function, and validation procedure as those used for S1. The original common-parameter experiments will be retained as controlled sensitivity tests, whereas the separately calibrated simulations will be used to compare the attainable performance of the dynamic- and static-input configurations under equivalent calibration conditions. This additional analysis will determine whether the performance differences associated with the dynamic inputs remain after accounting for parameter compensation.
Comment 4:
The Results and Discussion should be reorganized to better support the scientific story. The Results section devotes considerable space to model calibration, validation, parameter transfer, and comparisons with flux towers and satellite ET products. While these analyses demonstrate that the dynamic VIC model performs reasonably well, they contribute relatively little to the manuscript’s primary objective (quantifying the value of dynamic vegetation and radiation inputs). By contrast, the comparison between dynamic and static simulations occupies only a relatively small portion of the Results.
Similarly, the Discussion focuses primarily on established hydrological mechanisms rather than interpreting the new insights generated by this study. One of the most interesting results is that dynamic radiation forcing contributes much less than dynamic vegetation forcing, yet this finding receives relatively limited discussion.
Response 4:
The Results and Discussion will be reorganized to emphasize the comparison of dynamic and static input configurations. Sections 3.1 and 3.2 will evaluate simulated runoff and ET, respectively. Section 3.3 will compare S1–S4, where S4 is a newly added baseline configuration using static vegetation inputs and VIC’s default internally estimated radiation. This additional configuration completes the vegetation–radiation factorial design. Material on calibration, parameter transfer, and general validation will be condensed, and detailed MLR results will be moved to the Supplement.
The revised Discussion will focus on the magnitude, uncertainty, spatial heterogeneity, and relative effects of dynamic vegetation and radiation inputs, together with the recalibration sensitivity. Their relative importance will be interpreted from the revised observation-based and factorial analyses rather than inferred solely from the current scenario contrasts.
Two minor comments:
- L30. 'hydrological simulation' of what? Please specify (e.g., runoff, ET, water balance).
Response:
“Hydrological simulation” refers to simulations of ET and runoff.
- In the Introduction section, the review of first-, second-, and third-generation land surface models is relatively lengthy and only loosely connected to the main objective of the study. Consider shortening this section and focusing more directly on the scientific gap.
Response:
The historical review of first-, second-, and third-generation land-surface models will be condensed, retaining only the context needed to position VIC and motivate the use of time-varying inputs. The revised Introduction will focus more directly on the added value, spatial heterogeneity, and relative effects of dynamic vegetation and radiation inputs.
Citation: https://doi.org/10.5194/egusphere-2026-2505-AC1
-
AC1: 'Reply on RC1', Xianhong Xie, 21 Sep 2026
-
RC2: 'Comment on egusphere-2026-2505', Anonymous Referee #2, 20 Aug 2026
The manuscript addresses a relevant topic and presents a large scale application of VIC with dynamic satellite derived vegetation and radiation inputs. However, I am not convinced that the methodological and scientific novelty is sufficient for publication in HESS in its current form.
1.The central approach of replacing climatological or static LAI, FVC, albedo, and radiation inputs with satellite derived dynamic products is conceptually straightforward and has been explored in previous land surface and hydrological modeling studies. Likewise, the finding that vegetation greening increases ET and reduces runoff is already well established. The main contribution here is therefore largely a continental scale application and synthesis rather than a substantial methodological or process level advance.
2.The S1 to S3 experiments quantify the sensitivity of VIC simulations to dynamic vegetation and radiation inputs, but differences among scenarios do not by themselves demonstrate that the fully dynamic configuration is more accurate. Stronger evidence would require a systematic comparison of all scenarios against independent observations.
3.Although the manuscript emphasizes high ET accuracy, the site scale validation shows only moderate performance, with a mean correlation of approximately 0.61 and a mean bias close to minus 39%. This weakens the claim that the dynamic forcing substantially improves the realism of ET simulations.
4.Several ET products used for validation share similar or overlapping remote sensing inputs with those used to drive VIC, including LAI, FVC, albedo, and reanalysis data. Therefore, the high agreement between VIC and these products cannot be treated as fully independent validation and may partly reflect common input information.
5.The parameter transfer framework is based on multiple linear regression using climate, radiation, and vegetation predictors. While the results are encouraging, the method itself is conventional, and the manuscript does not sufficiently demonstrate that this regionalization framework provides a methodological advance beyond existing approaches for ungauged basins.
6.The model uses dynamic LAI, FVC, and albedo but retains a fixed 2010 land cover map. Therefore, the framework represents temporal variation in vegetation state rather than the full effects of land use and land cover change. This limitation should be more clearly reflected in the interpretation of the results.
Citation: https://doi.org/10.5194/egusphere-2026-2505-RC2 -
AC2: 'Reply on RC2', Xianhong Xie, 21 Sep 2026
Response to Reviewer2
Dear Reviewer,
We sincerely thank you for the careful and constructive evaluation of our manuscript. The comments have helped us clarify the scientific contribution of the study, strengthen the comparison and observation-based evaluation of the different forcing configurations, and improve our discussion of the methodological limitations. We have carefully considered each comment and will make corresponding revisions to the manuscript, as detailed in our responses below.
Comment 1:
The manuscript addresses a relevant topic and presents a large scale application of VIC with dynamic satellite derived vegetation and radiation inputs. However, I am not convinced that the methodological and scientific novelty is sufficient for publication in HESS in its current form.
The central approach of replacing climatological or static LAI, FVC, albedo, and radiation inputs with satellite derived dynamic products is conceptually straightforward and has been explored in previous land surface and hydrological modeling studies. Likewise, the finding that vegetation greening increases ET and reduces runoff is already well established. The main contribution here is therefore largely a continental scale application and synthesis rather than a substantial methodological or process level advance.
Response 1:
We thank the reviewer for this thoughtful comment. We agree that replacing static or climatological LAI, FVC, albedo, and radiation inputs with satellite-derived dynamic products is conceptually straightforward, and that vegetation greening increasing ET and reducing runoff is well established.
We would like to clarify the novelty of this study. Its contribution is not the replacement approach itself, but the systematic quantification of how much dynamic vegetation and radiation parameters improve VIC simulations under vegetation greening and radiation dynamics, and under what climatic and ecological conditions these dynamic inputs become essential. This moves the work beyond a continental-scale application and synthesis toward process-based insight into vegetation–hydrology interactions.
In the revision, we will explicitly highlight this novelty and strengthen the diagnosis of the climatic and ecological conditions that control the value of dynamic inputs.
Comment 2:
The S1 to S3 experiments quantify the sensitivity of VIC simulations to dynamic vegetation and radiation inputs, but differences among scenarios do not by themselves demonstrate that the fully dynamic configuration is more accurate. Stronger evidence would require a systematic comparison of all scenarios against independent observations.
Response 2:
We agree that differences among S1–S3 scenarios alone do not demonstrate that the fully dynamic configuration is more accurate.
Our purpose with S1–S3 is different: to quantify, under the same set of model parameters, how much runoff and ET diverge when different vegetation indices and radiation inputs are used. This isolates the model's sensitivity to vegetation and radiation inputs from parameter effects, complementing the observation-based evaluation. We agree that a systematic comparison of all scenarios against independent observations would be stronger evidence, and we will add the static configuration's performance against observations; given the well-documented biases of static vegetation and radiation inputs, the static results are expected to perform worse.
Comment 3:
Although the manuscript emphasizes high ET accuracy, the site scale validation shows only moderate performance, with a mean correlation of approximately 0.61 and a mean bias close to minus 39%. This weakens the claim that the dynamic forcing substantially improves the realism of ET simulations.
Response 3:
We acknowledge that the site-scale ET performance (mean correlation ~0.61, mean bias ~−39%) is moderate rather than strong. This partly reflects an inherent challenge in evaluating regional-scale ET simulations against flux tower observations: the eddy covariance footprint (typically <1 km²) is 2–4 orders of magnitude smaller than model grid cells, and this spatial mismatch introduces representativeness errors that are difficult to eliminate. Such scale inconsistencies have been shown to significantly degrade apparent model–observation agreement even when the model itself performs reasonably. We will further investigate the sources of the high bias and optimize model parameters in the revision.
Comment 4:
Several ET products used for validation share similar or overlapping remote sensing inputs with those used to drive VIC, including LAI, FVC, albedo, and reanalysis data. Therefore, the high agreement between VIC and these products cannot be treated as fully independent validation and may partly reflect common input information.
Response 4:
We acknowledge the reviewer's concern. We would like to clarify the independence of the validation datasets used in this study.
The VIC model is driven by GLASS LAI, FVC, albedo, and downward shortwave radiation. Among the four ET products used for validation:
GLEAM uses its own forcing from satellite and reanalysis data, with LAI from a separate source and a soil-moisture-constrained stress formulation. It does not use GLASS LAI, FVC, albedo, or radiation.
PML-V2 (China) uses MODIS LAI and albedo together with CMFD/GLDAS meteorological forcing, and is independent of the GLASS products used here.
GLASS ET is a multi-source ET fusion product. It does not use the GLASS LAI, FVC, albedo, or radiation products that drive our VIC simulations.
ETMonitor shares partial inputs with our VIC forcing: it uses GLASS albedo and LAI, but not GLASS FVC or downward shortwave radiation.
More importantly, even where ETMonitor shares some input data with our VIC forcing, the two systems are fundamentally different in structure. VIC is a complete land surface model that simultaneously solves coupled water and energy balance processes. ETMonitor is a remote-sensing-driven ET model. Their agreement therefore reflects consistency in ET estimates rather than a common-input artifact.
We will add this clarification to the manuscript and explicitly state that the VIC-simulated ET is evaluated against four largely independent ET products, supporting the use of these products for validation.
Comment 5:
The parameter transfer framework is based on multiple linear regression using climate, radiation, and vegetation predictors. While the results are encouraging, the method itself is conventional, and the manuscript does not sufficiently demonstrate that this regionalization framework provides a methodological advance beyond existing approaches for ungauged basins.
Response 5:
We agree that the multiple linear regression framework itself is conventional, and we do not claim it as a methodological advance.
Our intention is the opposite: to show that a simple, conventional method—multiple linear regression combining climate, radiation, and vegetation predictors—can already achieve satisfactory parameter transfer results when vegetation information is explicitly incorporated. In other words, the contribution lies not in the transfer method, but in demonstrating that adding vegetation predictors to a standard regionalization framework is sufficient to achieve good performance, without requiring more complex approaches. We will clarify this positioning in the revision. We also note in the Discussion that combining more advanced approaches, such as nonlinear or machine-learning methods, may yield further improvements.
Comment 6:
The model uses dynamic LAI, FVC, and albedo but retains a fixed 2010 land cover map. Therefore, the framework represents temporal variation in vegetation state rather than the full effects of land use and land cover change. This limitation should be more clearly reflected in the interpretation of the results.
Response 6:
We agree with the reviewer. Using a fixed 2010 land cover map introduces uncertainty, and our framework therefore captures temporal variation in vegetation state rather than the full effects of land use and land cover change.
We chose the 2010 land cover map because China's ecological restoration programs were largely implemented in the early 2000s, and land use types had become broadly stable around 2010. In this context, the dynamic LAI, FVC, and albedo inputs compensate to a considerable extent for the absence of dynamic land cover. We will clarify this point and its implications in the Discussion.
Citation: https://doi.org/10.5194/egusphere-2026-2505-AC2
-
AC2: 'Reply on RC2', Xianhong Xie, 21 Sep 2026
-
RC3: 'Comment on egusphere-2026-2505', Anonymous Referee #3, 24 Aug 2026
Overall assessment
This study investigates the effects of dynamic satellite-derived vegetation and radiation information on continental-scale VIC hydrological simulations over China. The topic is within the scope of HESS. The manuscript is generally well written, and the overall research framework is relatively complete.
However, there are still some issues with the logical connection between the research question, results, and conclusions. In particular, although the manuscript aims to assess the extent to which dynamic vegetation information can improve model performance, the current results mainly demonstrate that “dynamic vegetation can improve model performance” , which is already shown in the introduction, while the magnitude and spatial distribution of the improvement, as well as the relative contributions of dynamic vegetation and dynamic radiation, are not sufficiently analyzed. Therefore, the conclusions remain largely qualitative and do not fully address the scientific questions raised in the Introduction. I suggest that the authors further strengthen the quantitative comparison among different scenarios and the manuscript should be reconsidered for publication after improving the overall logic and organization of the results.
Major comments
1. The comparison among different scenarios is insufficient to support the conclusion that dynamic vegetation improves model performance.
Sections 3.1 and 3.2 mainly present the model performance under the combined dynamic vegetation + dynamic radiation scenario, while the improvement among different scenarios is mainly presented in Figure 11. This organization does not sufficiently support the central scientific question of the manuscript.
The authors should make the comparison among different scenarios a more central part of the Results section, rather than summarizing the comparison mainly in Figure 11. In addition to the overall performance metrics, the authors should analyze the spatial distribution of the improvement, showing where the model performance is substantially improved, where the improvement is limited, and whether there are regions where the performance deteriorates. The changes in runoff and ET should also be discussed separately. Furthermore, the respective contributions of dynamic vegetation and dynamic radiation should be more clearly separated.
2. The benefits of introducing additional dynamic inputs need to be quantified.
Improved model performance after introducing additional dynamic inputs is not particularly surprising. Moreover, such improvement comes at the cost of increased model complexity (or increased complexity of the model inputs). Therefore, a more important question for this study is: How large is the improvement associated with dynamic vegetation, and is this improvement sufficient to justify the additional model complexity?
3. The research questions stated in the Introduction are not fully consistent with the final conclusions.
The logic of the Introduction suggests that the manuscript aims to determine the extent to which dynamic vegetation information can improve large-scale hydrological simulations, rather than simply demonstrating that dynamic vegetation is beneficial. However, the main conclusion is essentially that “dynamic vegetation improves model performance.” This conclusion has already been supported to some extent by previous studies.
Specific comments
4. The Introduction discusses LSMs extensively, whereas VIC is more commonly regarded as a hydrological model. The authors should clarify how VIC is positioned in this study and be cautious when generalizing the results to other LSMs.
5. The authors should clarify how dynamic vegetation information is coupled into VIC, including how LAI, FVC, and albedo are incorporated, and which inputs are changed or kept consistent among the different scenarios.
6. Figure 2 needs improvement in resolution and readability.
The symbols for the ET stations and hydrological stations could be enlarged. Figure 2 should ideally can be stand-alone to be understood without relying heavily on the main text. At present, the different basin colors and the highlighted sub-basins in Figure 2b are not easy to understand. In addition, the Yellow River, Huai River, and Hai River are represented using similar warm colors, making their boundaries difficult to distinguish.
7. The criteria used to define the calibration and validation periods in Figure 3 should be clarified.
Although the available data periods may differ among stations, the authors should at least specify how the calibration and validation periods were determined and whether a warm-up period was used.
8. The titles of Sections 3.1 and 3.2 should be revised.
“Model parameter calibration, validation, and transfer” and “Validation of model-simulated ET” appear to overlap conceptually. I suggest organizing Sections 3.1 and 3.2 separately according to runoff/baseflow and ET, respectively, so that the structure is more consistent.
9. More details on the MLR results should be provided.
If the results are too extensive to include in the main text, the authors should at least provide the regression coefficients, R², statistical significance, and relevant predictors in the Supplementary Material.
10. Figure 6f does not need to retain two decimal places.
The large number of digits makes the figure difficult to read. Reducing the number of decimal places would improve readability.
11. The method used to calculate the trends shown in Figure 6 is not described in the Methodology section.
12. Some numerical labels in Figure 12d appear to have incorrect signs.
Citation: https://doi.org/10.5194/egusphere-2026-2505-RC3 -
AC3: 'Reply on RC3', Xianhong Xie, 21 Sep 2026
Response to Reviewer3
Dear Reviewer,
We sincerely thank you for the careful and constructive evaluation of our manuscript. The comments have helped us strengthen the quantitative comparison among the different forcing scenarios, clarify the logical connection between the research questions, the results, and the conclusions, and improve the presentation of the methodological details and figures. We have carefully considered each comment and will make corresponding revisions to the manuscript, as detailed in our responses below.
Comment 1:
This study investigates the effects of dynamic satellite-derived vegetation and radiation information on continental-scale VIC hydrological simulations over China. The topic is within the scope of HESS. The manuscript is generally well written, and the overall research framework is relatively complete.
However, there are still some issues with the logical connection between the research question, results, and conclusions. In particular, although the manuscript aims to assess the extent to which dynamic vegetation information can improve model performance, the current results mainly demonstrate that “dynamic vegetation can improve model performance” , which is already shown in the introduction, while the magnitude and spatial distribution of the improvement, as well as the relative contributions of dynamic vegetation and dynamic radiation, are not sufficiently analyzed. Therefore, the conclusions remain largely qualitative and do not fully address the scientific questions raised in the Introduction. I suggest that the authors further strengthen the quantitative comparison among different scenarios and the manuscript should be reconsidered for publication after improving the overall logic and organization of the results.
The comparison among different scenarios is insufficient to support the conclusion that dynamic vegetation improves model performance.
Sections 3.1 and 3.2 mainly present the model performance under the combined dynamic vegetation + dynamic radiation scenario, while the improvement among different scenarios is mainly presented in Figure 11. This organization does not sufficiently support the central scientific question of the manuscript.
The authors should make the comparison among different scenarios a more central part of the Results section, rather than summarizing the comparison mainly in Figure 11. In addition to the overall performance metrics, the authors should analyze the spatial distribution of the improvement, showing where the model performance is substantially improved, where the improvement is limited, and whether there are regions where the performance deteriorates. The changes in runoff and ET should also be discussed separately. Furthermore, the respective contributions of dynamic vegetation and dynamic radiation should be more clearly separated.
Response 1:
Thanks for the constructive comment and suggestion.
We agree with the reviewer. The current organization, with scenario comparisons concentrated in Figure 11, does not sufficiently support the central scientific question.
In the revision, we will make the scenario comparison a central part of the Results section, and:
- Reorganize the figures and text so that comparisons among scenarios (static vs. dynamic vegetation, radiation, and their combination) are presented and analyzed directly in Sections 3.1 and 3.2, rather than summarized mainly in Figure 11.
- Analyze the spatial distribution of model improvement, identifying where performance is substantially improved, where improvement is limited, and whether any regions deteriorate.
- Discuss changes in runoff and ET separately.
- More clearly separate the respective contributions of dynamic vegetation and dynamic radiation.
Comment 2:
The benefits of introducing additional dynamic inputs need to be quantified.
Improved model performance after introducing additional dynamic inputs is not particularly surprising. Moreover, such improvement comes at the cost of increased model complexity (or increased complexity of the model inputs). Therefore, a more important question for this study is: How large is the improvement associated with dynamic vegetation, and is this improvement sufficient to justify the additional model complexity?
Response 2:
We agree that quantifying the benefit of introducing additional dynamic inputs is important, and we will address this explicitly in the revision.
On the second point, however, we would like to offer a clarification. The dynamic vegetation and radiation inputs in this study are satellite-derived products that are already readily available, and incorporating them requires no change to the VIC model structure—only the replacement of static input files with time-varying ones. The added computational cost is therefore minimal, and the increased "model complexity" is largely in input data rather than in model formulation. Given this, the more relevant question is not whether the improvement justifies additional model complexity, but how large the improvement is, and under what conditions it is largest. We will quantify this benefit clearly in the revision.
Comment 3:
The research questions stated in the Introduction are not fully consistent with the final conclusions.
The logic of the Introduction suggests that the manuscript aims to determine the extent to which dynamic vegetation information can improve large-scale hydrological simulations, rather than simply demonstrating that dynamic vegetation is beneficial. However, the main conclusion is essentially that “dynamic vegetation improves model performance.” This conclusion has already been supported to some extent by previous studies.
Response 3:
We agree with the reviewer. The Introduction aims to determine the extent to which dynamic vegetation information improves large-scale hydrological simulations, whereas the current conclusion is essentially that "dynamic vegetation improves model performance." We will revise the conclusions so that they directly answer the questions posed in the Introduction.
We would also like to reiterate the novelty of this study: it quantifies how much dynamic vegetation and radiation parameters improve VIC simulations under vegetation greening and radiation dynamics, and identifies under what climatic and ecological conditions these dynamic inputs become essential. This goes beyond demonstrating that dynamic vegetation is beneficial, and provides process-based insight into vegetation–hydrology interactions.
Comment 4:
The Introduction discusses LSMs extensively, whereas VIC is more commonly regarded as a hydrological model. The authors should clarify how VIC is positioned in this study and be cautious when generalizing the results to other LSMs.
Response 4:
We agree with the reviewer. In this study, VIC is positioned as a macroscale hydrological model, consistent with its conventional characterization in the literature.
We would like to clarify that VIC is also widely recognized as a land surface model. It is one of the land surface models used in NASA's Global Land Data Assimilation System (GLDAS), alongside Noah, CLM, and Mosaic, and is likewise one of the models in the North American Land Data Assimilation System (NLDAS). VIC shares the basic features of other LSMs, including coupled water and energy balance, sub-grid vegetation tiles, and land–atmosphere flux simulation.
Given this dual positioning, we will simplify the LSM discussion in the Introduction and clarify that our conclusions pertain primarily to VIC and should be generalized to other LSMs with caution.
Comment 5:
The authors should clarify how dynamic vegetation information is coupled into VIC, including how LAI, FVC, and albedo are incorporated, and which inputs are changed or kept consistent among the different scenarios.
Response 5:
We agree with the reviewer, and we will clarify the coupling procedure and the input settings of each scenario in the revision.
Briefly, the dynamic vegetation and radiation information is coupled into VIC as follows. First, daily-scale model forcing data are generated from the satellite-derived LAI, FVC, albedo, and radiation products and incorporated into the VIC forcing data files. Then, in the global file of VIC 4.2d, corresponding entries are added to specify the dynamic inputs, and the relevant parameters in the model data library are turned off so that the prescribed static values are not used.
For the other two scenarios, the dynamic inputs are turned off separately: in the scenario without dynamic vegetation, the vegetation parameters are held constant, whereas in the scenario without dynamic radiation, the radiation is internally simulated by the model. We will describe this procedure and specify which inputs are changed or kept consistent across the different scenarios.
Comment 6:
Figure 2 needs improvement in resolution and readability.
The symbols for the ET stations and hydrological stations could be enlarged. Figure 2 should ideally can be stand-alone to be understood without relying heavily on the main text. At present, the different basin colors and the highlighted sub-basins in Figure 2b are not easy to understand. In addition, the Yellow River, Huai River, and Hai River are represented using similar warm colors, making their boundaries difficult to distinguish.
Response 6:
Thanks. We will revise Fig. 2 by enlarging the ET-site and hydrological-station symbols, increasing the output resolution, defining every basin abbreviation and symbol in a self-contained legend, explaining the highlighted sub-basins in panel (b), adding the necessary map elements, and using clearly distinguishable hues for the Yellow, Huai, and Hai River basins. The caption will describe each panel fully. These changes are consistent with the broader figure-readability revisions already committed to in our response to CC1.
Comment 7:
The criteria used to define the calibration and validation periods in Figure 3 should be clarified.
Although the available data periods may differ among stations, the authors should at least specify how the calibration and validation periods were determined and whether a warm-up period was used.
Response 7:
Section 2.4 will clarify the criteria for selecting calibration and validation periods, including minimum record length, treatment of data gaps and unequal records, and separation of calibration and validation data. The 1995–1999 period will be specified as model spin-up before the 2000–2020 simulation. A supplementary table will report data availability and the exact calibration, validation, or evaluation periods for all 50 gauges, including 41 calibrated basins and nine parameter-transfer basins.
Comment 8:
The titles of Sections 3.1 and 3.2 should be revised.
“Model parameter calibration, validation, and transfer” and “Validation of model-simulated ET” appear to overlap conceptually. I suggest organizing Sections 3.1 and 3.2 separately according to runoff/baseflow and ET, respectively, so that the structure is more consistent.
Response 8:
The Results will be reorganized into three sections: runoff evaluation, ET evaluation, and the S1–S4 factorial comparison. S4 is a newly added scenario combining static vegetation inputs with VIC’s default internally estimated radiation. Section 3.3 will focus on observation-based scenario performance, spatial heterogeneity, main and interaction effects under the common-parameter experiment, and recalibration sensitivity. This structure removes the current overlap and places the forcing comparison at the center of the Results.
Comment 9:
More details on the MLR results should be provided.
If the results are too extensive to include in the main text, the authors should at least provide the regression coefficients, R², statistical significance, and relevant predictors in the Supplementary Material.
Response 9:
A supplementary table will report, for each of the seven calibrated parameters, the intercept and regression coefficients, retained predictors, R², p-values, standardization and variable-selection procedures, and parameter-transfer performance. Section 2.4 will briefly describe the MLR setup and clarify that it serves as a parsimonious regionalization tool supporting the forcing experiments, rather than as a methodological innovation.
Comment 10:
Figure 6f does not need to retain two decimal places.
The large number of digits makes the figure difficult to read. Reducing the number of decimal places would improve readability.
Response 10:
We will display the numerical labels in Fig. 6f to one decimal place consistently. This will reduce visual clutter without changing the underlying calculations or conclusions.
Comment 11:
The method used to calculate the trends shown in Figure 6 is not described in the Methodology section.
Response 11:
The Methods will state the exact estimator used for the trends in Figs. 6 and 12, the units of the slope, the test used for statistical significance, the treatment of serial correlation and missing values, the p-value thresholds represented by the asterisks, and the method used to construct the confidence bands. We will also recheck that the same procedure is applied consistently to all compared series.
Comment 12:
Some numerical labels in Figure 12d appear to have incorrect signs.
Response 12:
An error was identified in Fig. 12d. The figure and the corresponding description will be corrected in the revised manuscript.
Citation: https://doi.org/10.5194/egusphere-2026-2505-AC3
-
AC3: 'Reply on RC3', Xianhong Xie, 21 Sep 2026
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 142 | 35 | 24 | 201 | 27 | 11 | 9 |
- HTML: 142
- PDF: 35
- XML: 24
- Total: 201
- Supplement: 27
- BibTeX: 11
- EndNote: 9
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript presents a continental-scale hydrological modeling study that evaluates how dynamic satellite-derived vegetation and radiation inputs affect VIC model simulations across China. The authors incorporate remote sensing products for LAI, fractional vegetation cover, albedo, and radiation into VIC and compare model outputs against runoff observations from 50 stations, ET observations from 41 flux sites, and multiple satellite-based ET products. The main finding is that dynamic vegetation inputs have a much stronger effect than dynamic radiation inputs: using static vegetation leads to ET underestimation of about 5.11% and runoff overestimation of about 14.13%, while dynamic radiation has a comparatively smaller influence. The manuscript is relevant and timely because it addresses the need to represent vegetation dynamics in land surface and hydrological models under changing environmental conditions.