the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Feb 2025
Metric-based analysis of the historical drivers of surface hydrological connectivity
Abstract. Hydrological connectivity is essential for the maintenance of important hydrological and ecological processes of catchments. Over time, human activities have altered the natural patterns of hydrological connectivity, leading to habitat loss and deterioration. Historical information from cartographic maps can be used to enhance our understanding of large-scale hydrological processes such as connectivity, by offering snapshots of past, less human-impacted landscapes and hydrological systems. The focus of this study is on historical surface hydrological connectivity and its landscape drivers (e.g., lithology, topography, land use/ land cover) in ten Swiss catchments from different biogeographic regions (i.e., Pre-alpine, Alpine, Karstic, Plateau), and with varying physiographic characteristics. We employed hydromorphological metrics derived from historical maps (~ late 19th century) as proxies of surface hydrological connectivity, with the main goal of identifying the primary landscape drivers of connectivity. As expected from theory, the historical patterns of hydrological connectivity in the studied catchments were mostly driven by landscape topography, and in particular by the slope and the morphology of the valley bottom. Unexpected relationships between connectivity and its drivers could be traced back to human practices, such as specific irrigation techniques and peat digging. Overall, our study shows how historical information can be employed to gain a deeper understanding on important large-scale hydrological processes, their primary drivers and on the history of human exploitation of the territory. Finally, this kind of approach paves the way for the characterization of how connectivity has changed through time.
- Preprint
(1925 KB) - Metadata XML
-
Supplement
(3062 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-199', Anonymous Referee #1, 18 Mar 2025
This article presents a novel and ambitious analysis of historical surface hydrological connectivity across ten Swiss catchments, using digitized versions of the Siegfried map developed between 1870 and 1926. The authors derive a set of hydromorphological metrics as proxies for surface hydrological connectivity, including drainage density, confluence density, stream order metrics, and shoreline metrics related to different land cover types. These metrics are then statistically related to landscape variables such as lithology, topography, and historical land use/land cover to identify the main drivers of hydrological connectivity in the late 19th century. The study finds that topographic features, particularly valley morphology and slope, are consistent primary drivers of longitudinal connectivity, while wetlands and forested areas play key roles in lateral connectivity. Several unexpected relationships are identified, such as the positive association between building area and hydrological metrics in some catchments, which are interpreted in the context of historical irrigation and land use practices.
I have following comments for improvement before possible publication.
Major Comments
- The methodology is well developed and generally appropriate for the research questions posed, but there are areas where improvements could increase scientific robustness and interpretive power. The use of the Siegfried map is innovative and provides a good source of historical spatial data, but the temporal spread of the maps across more than four decades introduces a source of temporal inconsistency that is only partially acknowledged. The assumption that these maps provide a coherent "snapshot" of late 19th-century conditions should be supported by more quantitative analysis or stratification. For instance, stratifying map sheets by decade or analyzing whether certain catchments are represented earlier or later in the mapping process could reveal biases in the spatial representation of features. Alternatively, some form of uncertainty band or temporal metadata analysis could be introduced to account for potential inconsistency.
- The derivation of hydromorphological metrics is clear and well justified, yet the redundancy observed between certain metrics, notably the perfect correlation between drainage and shoreline density, suggests a need for dimensionality reduction or orthogonality testing. Methods such as principal component analysis or hierarchical clustering could be used to examine the independence of metrics and reduce redundancy in the dataset. The use of stream order (Strahler) is appropriate, but it would benefit from a sensitivity test given the known limitations of applying stream order logic to historical maps where ephemeral streams might be underrepresented and headwaters potentially truncated due to mapping resolution or classification ambiguity.
- There is limited quantitative treatment of uncertainty in the metric derivation process. While the authors recognize spatial inaccuracies in the historical maps and the potential for scanning and digitization errors, these are discussed only qualitatively. A formal uncertainty propagation analysis—perhaps using Monte Carlo simulations that randomly shift hydrographic features within known error bounds—would provide confidence intervals for key metrics like drainage density and confluence density. Without this, small observed differences between catchments or sub-catchments may not be statistically meaningful.
- The classification of landscape features from the historical maps, particularly land use and land cover types such as forests, wetlands, buildings, and fields, appears to rely on machine learning, but little detail is given on the classification model used or its validation accuracy. Accuracy assessment using known control areas or expert interpretation would strengthen the validity of the LULC dataset. This is especially important given that key explanatory variables in the regression models depend heavily on this classification. Misclassification of small but hydrologically important features such as wetlands could lead to erroneous model outputs or misinterpretation of connectivity relationships.
- The multiscale approach is a major strength of the paper, as it acknowledges the hierarchical nature of hydrological systems. The use of mixed-effects models with nested spatial structure and the control of spatial autocorrelation using locational variables is appropriate and well-motivated. However, no diagnostic tests are reported to assess whether spatial autocorrelation remains in model residuals. Simple tests such as Moran’s I or residual correlograms could be used to confirm that spatial dependence has been sufficiently removed, which is crucial for the validity of regression coefficients. In addition, the model structure includes a large number of potential explanatory variables, and while multicollinearity is addressed through VIF analysis, the possibility of overfitting or omitted variable bias remains. Using penalized regression methods or cross-validation could help address this.
- The assumption that topography, especially the valley bottom extent and slope, has remained unchanged since the 19th century underlies several parts of the validation and metric alignment. While reasonable in many alpine regions, this assumption may not hold in dynamic sedimentary basins or in areas with known anthropogenic modifications such as channelization or urban expansion. In such cases, the comparison between historical hydrography and contemporary DEM-derived valley bottoms may introduce bias. Some discussion or sensitivity testing around topographic stability over the study period would enhance the reliability of the validation procedure.
- The regression framework is sound and includes both fixed and mixed-effect models, with appropriate use of generalized models and two-part modeling for zero-inflated variables. However, the relationships between hydrological metrics and landscape variables are often complex and potentially nonlinear. The current linear framework may not fully capture these complexities. Generalized additive models (GAMs) could be employed to allow for more flexible, non-parametric relationships between predictors and response variables. This may be particularly useful for variables such as stream order or wetland connectivity, which may exhibit threshold or saturation behaviors.
- While the ecological relevance of hydromorphological metrics is discussed, the connection between these metrics and specific ecological processes or outcomes remains somewhat abstract. The paper could benefit from more explicit hypotheses or examples linking SHC to ecological functions such as habitat continuity, fish migration, or nutrient flux. Even though the historical nature of the data makes direct ecological validation difficult, drawing stronger conceptual links to ecological theory would help bridge the hydrological and ecological dimensions of the study.
Minor comments:
- Line 12: “Human activities have progressively altered...”
- Line 23: “into large-scale hydrological processes”
- Line 41: instead of ‘its vegetation cover’ “the vegetation cover.”
- Line 62: “Siegfried map,”
- Line 64: “primary drivers of surface hydrological connectivity...”
- Table 1 caption (line 70): “Hypothesized relationships between the landscape drivers...”
- Line 88: “...likely represent the perennial component”
- Line 165: “Please note that the percentages do not sum up to 100%...” consider placing this in a footnote or table caption to reduce main text clutter.
- Consistency in units and formatting in Table 3 (e.g., km² vs. km2), this applies to other parts in the manuscript.
- Line 218: should be “Figures were generated...”
- Line 274: “...to a lesser extent, by the percentage of area covered by buildings...”
- Line 286: “...the average slope mostly along the second axis.” → Consider specifying the direction (positive/negative).
- Table 5: Ensure consistent use of “Stream order (max)” vs. “Max stream order” and so on across tables and text.
Citation: https://doi.org/10.5194/egusphere-2025-199-RC1 -
RC2: 'Comment on egusphere-2025-199', Anonymous Referee #2, 18 Mar 2025
This paper by Antonelli et al. presents the methodology and results of the morphological analysis of hydrographic network derived from historical land use maps in Switzerland, with the purpose of linking catchment geographic descriptors with hydrological connectivity.
While I understand the general framework of the study (eg assessing the hydrological connectivity in the past as a reference to assess how it has been altered by anthropogenic modifications) and its interest for assessment of ecological quality of rivers, I don’t really see how the methodology and results presented here contribute to this general framework.
In particular :
1/ The concepts of surface hydrological connectivity, and also longitudinal and lateral connectivities are not defined. I would have wished to have at least a correspondence with hydrological processes (eg lateral connectivity = groundwater-river interaction?).
2/ The historical maps that were used date from the late XIXth century. At this time, the river morphology was not free of human influences, a lot of rivers were already significantly modified / engineered. How do the authors relate this to the more general framework of the introduction about using historical maps as « references » for river restoration ?
3/ The metrics that were chosen are very classic / general geomorphology metrics. I don’t see how they can be informative on river connectivity (how can Strahler order be informative?) and the authors provide no explanation. I would have expected for example something about dams. This is not discussed at all.
4/ The explanatory factors chosen are very general and calculated at the catchment scale. I would have expected more precise / local factors. In particular I am very surprised by the coarse geology classification, with only 2 classes (permeable, not permeable), which is not justified nor discussed. Geology variability can have a lot of effects in particular on groundwater-river interaction, at local scales. I find also very suspicious that karst is never mentioned, although the presence of karst plays a great role in river morphology (and I believe karst is present in Switzerland).
5/ The results are presented separately for two groups of catchments according to the scale of the land use maps (1:25000 and 1:50000). However the effect of the scale on the results is never discussed although these results appear to be different for both groups.
6/ The paper presents the results of many statistical analyses according to various explanatory factors. Some of these results are obvious, some are more surprising. However there is no or very little interpretation of these results in terms of physics / hydrological processes. Without physical interpretation, the impact of the paper falls short.
7/ The authors mention in the discussion (p 25, l 425-435) that the hydrographic network that was considered already incorporates artificial elements. Then I don’t understand the purpose of the study that is presented in the introduction as establishing a “reference” of connectivity (see comment 2).
8/ Following the same idea, I would have expected a comparison with “connectivity” based on a current map. Why was it not done?
More technical comments
Tables and Figures in the Results sections are too busy and difficult to read. Particular mention to Table 5 that is completely illegible.
Citation: https://doi.org/10.5194/egusphere-2025-199-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 75 | 21 | 5 | 101 | 15 | 4 | 3 |
- HTML: 75
- PDF: 21
- XML: 5
- Total: 101
- Supplement: 15
- BibTeX: 4
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 57 | 46 |
| China | 2 | 16 | 13 |
| Switzerland | 3 | 7 | 5 |
| France | 4 | 6 | 4 |
| Netherlands | 5 | 6 | 4 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 57