| 27 Mar 2026
A Raster–Vector Framework for Multi-Scale Hydrological–Hydraulic Modeling Across Large Domains
Abstract. Hydrological models are widely used for large-scale flood modeling but typically rely on simplified routing schemes and a grid-based discretization that poorly capture river flow dynamics and geometry. Conversely, hydraulic models enable more accurate representation of flow dynamics through physically based routing schemes and vector-based geometry, but their application to large domains is constrained by significant parametrization challenges. Bridging these limitations requires integrated hydrological-hydraulic (H&H) modeling approaches capable of reconciling the spatial scale and structural mismatch between the two models while ensuring seamless coupling and computational efficiency for large-scale applications. We present an integrated raster-vector H&H modeling framework that leverages sub-grid representation of both river networks and drainage areas derived from high-resolution topography. The framework is implemented within the open-source smash modeling platform and integrates an end-to-end workflow from DEM preprocessing to internally coupled H&H simulations. The framework was evaluated over the Garonne river basin (France, 50 100 km2) through one-directional coupling of a grid-based conceptual hydrological model with a vector-based 1D hydrodynamic model solving the zero-convective inertia approximation of the shallow water equations. Geometric preprocessing analysis performed across spatial resolutions ranging from 0.5 km to 10 km demonstrates that sub-grid information enable to maintain high spatial accuracy of DEM-derived sub-grid networks across scales in relation to the mapped hydrography, and reduces significantly catchment area errors compared to the default grid-based area delineation method. H&H simulations conducted at 1 km, 5 km, and 10 km without recalibration show robust preservation of flow timing across scales and demonstrate more stable streamflow bias across resolutions when using sub-grid drainage areas, while grid-based areas exhibit scale-dependent volume biases reflecting drainage area misrepresentation. The proposed H&H framework demonstrates scalable and efficient simulations at large domains within a unified modeling environment, offering promising perspectives for assimilation of multi-source water surface observations to infer key model parameters, addressing critical parametrization challenges in data-sparse regions.
Main comments
The manuscript presents a timely and novel study, proposing a method to transforming high-resolution vector river networks into coarse-resolution raster river grids addresses a methodological challenge that has not yet been well resolved in the existing research. This aspect represents a potentially contribution to large-scale hydrological hydrodynamic modeling. However, the aspects of the framework present several limitations regarding maturity, scalability and demonstrated benefits.
The overall workflow appears rather complex, involving multiple intermediate processing steps. The river-network transformation procedure has not been adequately integrated into a unified and streamlined framework, which reduces both the transparency and reproducibility of the methodology.
The selected study area is comparatively small (approximately 50,000 km²), which limits the manuscript's ability to demonstrate the stability and transferability of the proposed method for broader large-domain applications. The framework does not yet appear to be readily applicable to large-scale basins or multi-river complex systems, as the computational burden may become prohibitive. This brings concerns regarding its scalability and practical applicability beyond the relatively limited study domain.
While the study emphasizes methodological novelty, it remains unclear whether the proposed framework leads to a substantial improvement in hydrological simulation skill or process representation compared to existing methods. In particular, the overall presentation of the manuscript, including figures and visualizations, requires substantial improvement to meet the standards of clarity. And the results and discussion require stronger evidence to support the claimed methodological advantages.
Overall, although the framework appears technically consistent, the manuscript does not yet convincingly demonstrate clear improvements in predictive performance. For these reasons, I do not recommend publication in GMD in its current form.
Detailed comments:
Abstract
Line 9: define DEM at its first occur.
1 Introduction
Lines 60-63: the authors clearly identify the primary H&H challenge as the scale mismatch between coarse-grid hydrological models and fine-resolution vector river networks. However, from line 78 onwards, the manuscript shifts to a detailed discussion of channel geometry and bathymetry parameterization (e.g., depth, friction), which is not clearly linked back to this core scale-mismatch problem. The authors should better integrate this discussion into the main methodological narrative.
2 Methodology
Section 2.2.1-2.2.2: These describe standard DEM preprocessing steps which some of these are not part of the methodological novelty and there is redundancy with Section 2.2.3. Remove redundant descriptions.
Section 2.2.3-2.2.4: These contain the core methodological contribution. Emphasize and clarify.
Line 219: define GR4 at first instance.
3 Study sites and data
3.1 Garonne basin
The selected basin (~50,000 km²) is relatively small and predominantly flat. What is the rationale for selecting such a low-relief basin? How would the framework perform in larger and more complex basins? The river network appears simplified, with limited representation of small tributaries.
Lines 65-66: the author refer to the preprocessing step of H&H framework remains computationally intensive and presents ongoing efficiency challenges. The discussion of scalability and computational efficiency is essential.
3.2.1 Mapped river network
Figure 4: Add north arrow and scale bar.
3.2.2 Digital Elevation Models (DEMs)
Line 144 & 296: Define MERIT and BD ALTI at first instance.
Clarify consistent spatial resolution (MERIT resampled to 100 m, BD ALTI resolution unclear).
If the workflow involves extracting fine-resolution vector rivers and aggregating them to coarse grids, would it be more appropriate to use pre-upscaled DEM products (e.g., MERIT Hydro IHU) to reduce resampling-induced uncertainty?
3.2.3 Hydro-meteorological data
Precipitation (1 km) and PET (8 km) have inconsistent resolutions. Clarify final resolution of all forcing datasets.
Provide detailed description of runoff validation datasets in this section.
4 Numerical experiments
4.1.1 Spatio-temporal configuration
The author only uses the two years, the year of 2018 for spin, the year of 2019 for formal simulations, the author should use longer time series forcing data for the validation of the simulation of runoff by H&H framework or justify this choice.
4.1.2 Aggregation of meteorological forcings
This section is a methodological data description and should be merged into Section 3.2.3.
4.1.4 Hydraulic model parameterization
Consider scale-dependent or spatially variable Manning coefficients, as they strongly influence discharge simulations.
4.2.1 Geometric assessment of the preprocessing
The using of visual inspection is not scalable for large domains. It is unclear how the H&H framework handles: (1) closely connected tributaries; (2) overlapping or merged coarse-grid representations. Provide a systematic or automated evaluation metric, and clarify how river connectivity is preserved during rasterization.
Line 381: clarify definitions of “sub-grid” and “grid-based” methods and ensure consistent usage throughout in the manuscript.
5 Results
5.1 Quality assessment of DEM-derived networks across scales
Lines 427-428: provide figure or table references to support reported percentages.
Lines 428-430: current scatter plots (Figure 6a) do not clearly show differences. Consider alternative visualization (e.g., boxplots).
Figure 5: All sub-plot adds north arrow and scale bar.
Figure 6: Align the term of “coarse” and “sub-grid” in the figure with that used in the manuscript. Clarify the sample size used in subplots (a), (b), and (d); the number of points appears unexpectedly low. Why not metrics are computed at the sub-catchment or tributary level?
Subplot (c): the stacked bar chart is not an effective way to compare internal vs. headwater differences between DEMs. Clarify the meaning of “omitted reference network length”.
All subplots should avoid grey backgrounds and instead use a white or very light background for clarity.
5.2 Catchment size errors across delineation methods
Lines 439-441: the definitions of sub-grid and grid-based methods are unclear. Clarify the distinction to avoid confusion.
5.3 H&H Model performance across scales
Line 451: Locations of the 18 gauging stations are not clearly shown.
Unify the term naming convention. The term used in Table 1 (Grid and Sub-grid), Figure 7 (grid-based and sub-grid), and Figure 6 (coarse vs. sub-grid) is inconsistent and may confuse readers.
6 Discussion
6.1 Coarse vs. sub-grid networks spatial accuracy across scales
Line 487: The omission of headwater segments in zoom2 likely has limited impact on discharge. In contrast, the internal omissions observed in zoom1 should be more thoroughly discussed. Specially, the manuscript should explain why BDALTI DEM fails to capture meandering river network, while MERIT DEM appears to better represent them (Figure 2). The origin of these differences between DEMs after sub-grid processing should be deeper analyzed.
6.2 Mitigating catchment size errors
Lines 495-497: add the figure reference to support the sentence.
Lines 504-505: The authors should further discuss why MERIT DEM shows larger errors at coarse resolution than BDALTI. Given that MERIT DEM is generally expected to perform better in hydrological applications, this result is counterintuitive.
A likely explanation is that the original MERIT DEM (~90 m, 3 arc-second) was resampled to 100 m using bilinear resample, which may introduce additional errors in flow routing and catchment delineation. In contrast, BDALTI has a native resolution of 25 m, which may explain its superior performance in this study. Therefore, the observed differences may not reflect intrinsic DEM quality but rather preprocessing effects. Would using preprocessed DEM products (e.g., MERIT Hydro IHU) reduce this bias?
Given the relatively small study area, the conclusion that BDALTI outperforms MERIT DEM may not be applicable to larger basins.
6.3 Understanding multi-scale performance patterns
Figures D1 and 8 do not convincingly demonstrate improved discharge simulation performance of the HH framework. Across resolutions (1 km, 5 km, 10 km), the distributions of KGE and Pearson correlation (r) are nearly identical, with similar medians. Differences in bias are minor and do not clearly indicate superior performance of the sub-grid method. The only noticeable improvement appears in KGE, while the overall spread of boxplots remains similar, with differences largely driven by the removal of extreme values.
Figures D2 and D3: Would classify results by sub-basin and DEM improve interpretation (Figures D2, D3)? Current plots do not sufficiently show these distinctions.
7 Conclusion
The manuscript emphasizes a “unified raster-vector H&H framework“ with improved physical consistency and scalability. However, the presented simulation of discharge results mainly demonstrates stable correlation (r) and redistribution of bias, without convincingly showing a significant improvement in simulation performance compared to existing methods.
Moreover, the reported advancement in catchment delineation appears limited given the limited basin size, lack of testing in complex river systems, and inconsistencies in DEM resolution and preprocessing.