the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Technical note: Comparing three different methods for allocating river points to coarse-resolution hydrological modelling grid cells
Juliette Godet
Eric Gaume
Pierre Javelle
Pierre Nicolle
Olivier Payrastre
Abstract. The allocation of points in a river network to pixels of a coarse-resolution hydrological modelling grid is a wellknown issue, especially for hydrologists who use measurements at gauging stations to calibrate and validate distributed hydrological models. To address this issue, the traditional approach involves examining grid cells surrounding the considered river point and selecting the best candidate, based on distance and upstream drainage area as decision criteria. However, recent studies have suggested that focusing on basin boundaries rather than basin areas could prevent many allocation errors, even though the performance gain is rarely assessed. This paper compares different allocation methods and examines their relative performance. Three methods representing various families of methods have been designed: area-based, topology-based and contour-based methods. These methods are implemented to allocate 2580 river points to a 1 km hydrological modelling grid. These points are distributed along the entire hydrographic network of the French southeastern Mediterranean region, covering upstream drainage areas ranging from 5 km2 to 3000 km2. The results indicate that the differences between the methods can be significant, especially for small upstream catchments areas.
- Preprint
(2399 KB) - Metadata XML
- BibTeX
- EndNote
Juliette Godet et al.
Status: open (extended)
-
RC1: 'Comment on hess-2023-165', Anonymous Referee #1, 15 Aug 2023
reply
The paper investigates different method to allocate locations of stream gauges to the correctriver cell in course resolution distributed hydrological models. Three different methods areinvestigated and compared for the French southeastern Mediterranean region. The methods arebased on 1) upstream area and distance; 2) high-resolution river topology; and 3) catchmentcontour. The methods are compared based on the overlap between the high resolution catchmentcontour of the gauge and the low resolution catchment contour of the model upstream from theallocated river cell.The topic is relevant and often overlooked. The paper is also generally well written and themethods are mostly well described. However, I have some concerns about the methods and theresults as outlined below. I therefore recommend major revisions of the paper.## main comments- For gauges which are located between two confluences within one cell, see e.g. P3 in Figure 3, theauthors state that these cannot be allocated to the correct river cell using method 2, but can beallocated using method 1 and 3. In my opinion, the only correct allocation would be to both upstreamcells (e.g., cells C3 and C5), by comparing the sum of the model discharge against the observeddischarge. With method 1 and 3, while the method does assign the gauge to a single cell, I think thatis an incorrect allocation for these cases. This is not discussed in the paper. Also, with a smallextension, method 2 would actually be able to correctly allocate the gauge to both cells.- The authors compare the different methods based on the CSI of overlapping catchment contours, whichis also optimized in the allocation process of method 3. I find this single metric for benchmarking the differentmethods too limited. For a fair comparison, it would be better to use multiple metrics including difference inupstream area (which is also easier to understand). Or if possible, use manually allocated gauges as a reference,to understand the true errors made by each method.## minor comments- It would be helpful to illustrate in Figure 2-4 to which river cell the gauges are allocated.- Line 89: Consider using a more commonly used notation for CSI (see e.g., Fleischmann et al., 2019). It is also not entirely clear to me how the CSI is calculated because of the different resolutions of the catchment contours. Is the CSI calculated based on the low resolution catchment contour of the model or the high resolution catchment contour of the gauge? This could make quite a difference for certain catchments.- Line 94: The inline formula is hard to read and the variables unclear as they refer to criteria used in other papers. Could the authors explain the variables shortly here to make interpretation easier?- Figure 9: Can you add the outflow points of all cells in Figure 9B to better understand why the cell just upstream from the gauge is not found?- Figure 10: the stacked histograms are difficult to read. Consider using a different histogram style.- Line 211: It is stated that method 2 requires a vector-based description of the river network (which I guess is the same a the high resolution river topology / flow directions?). However, if I understand correctly, method 3, would require a vector-based description of the catchment contour which is not mentioned here.- Line 222: I suggest to mention vector-based models already in the introduction to emphasize that issue and proposed methods are specific to raster-based models.Citation: https://doi.org/
10.5194/hess-2023-165-RC1
Juliette Godet et al.
Juliette Godet et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
305 | 86 | 10 | 401 | 6 | 6 |
- HTML: 305
- PDF: 86
- XML: 10
- Total: 401
- BibTeX: 6
- EndNote: 6
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1