the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Sep 2024
Are rivers becoming more intermittent in France? Learning from an extended set of climate projections based on the Coupled Model Intercomparison Project phase 5 (CMIP5)
Abstract. This study aims to assess the changes in the intermittency of river flows across France in the context of climate change. Projection of flow intermittence are derived from the results of the Explore2 project, which is the latest national study that proposes a wide range of potential hydrological futures for the 21st century. The multi-model approach developed within the Explore2 project enable to characterize uncertainties in future flow intermittence. Combined with discrete observations of flow states, hydrological projections are post-processed to compute the daily probability of flow intermittency (PFI) on each element of the partition of France in hydroecoregions (HER2).
The post-processing consists of calibrating logistic regressions between the historical flow states of the Observatoire National des Étiages (ONDE) network and the flow data simulated by the hydrological models (HMs) involved in Explore2 projected with the Safran reanalysis as inputs. After calibration, these regressions are used to project daily PFIs for the whole of the 21st century, based on flow simulations from five HMs driven by up to 17 climate projections under RCP 2.6, 4.5, and 8.5 climate change scenarios.
The results show good agreement among the HMs regarding the increase in flow intermittency under RCP 4.5 and 8.5. The changes in mean daily PFI between July and October, and the shifts in the first and last days when PFI exceeds 20 %, suggest a gradual intensification and extension of dry spells throughout the century. The southern regions of France are likely to experience greater increases in runoff intermittency than the northern regions. Uncertainty is greater in northern France, due to the variability of rainfall. Mountainous regions such as the Alps and the Pyrenees are likely to experience changes in the dynamics of snowmelt and groundwater recharge, which could lead to changes in their runoff regimes.
- Preprint
(5869 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-2737', Anonymous Referee #1, 30 Oct 2024
GENERAL COMMENTS
The manuscript tries to understand how river intermittency across France could change over time as a consequence of climate change. To do so, the authors relate the number of dry locations within a predetermined area (PFI) to the average exceedance probability of the available flowrates within the same area (being them either measured or modeled). Then, a number of climatic and hydrological models are employed to provide future scenarios of streamflow, and from that estimate future scenarios of PFI. Despite the huge uncertainties that these types of works inevitably contain, this method shows how rivers are expected to become more intermittent in the future.
Overall, the manuscript is well written and the work seems well executed. However, a number of concerns arise, as reported in the specific comments below, mainly regarding the heterogeneity of the data used as a basis for extrapolating the future scenarios, and the tecniques that have (or have not) been used to cope with the data limitations.
SPECIFIC COMMENTS
1) The work heavily relies on the Explore2 project for future streamflow scenarios, which is explained in a manuscript that is not currently available to the public (Sauquet et al. in prep). While I gave a first review of this manuscript, access to Sasquet et al. (in prep) is a requirement in order to fully understand the soundness of the projections reported here and provide an informed decision on the manuscript.
2) I am concerned about the biases in the datasets used for this analysis. Only 20% of ONDE stations have drainage area of <10km2, while most of the river network length is in the headwaters (give the power-law scaling of network length with contributing area). As such, there is a strong underrepresentation of headwaters. At the same time, ONDE stations are selected to highlight non-perennial conditions, so non-perennial streams are over-represented in this dataset. Given these biases, how representative is PFIh of the actual river intermittence in France, and its heterogeneity?
3) On the same note, the work refers multiple times to headwater streams (e.g., in line 231 "the 3302 ONDE sites are located in headwaters"). However, only 20% of the ONDE stations have contributing areas smaller than 10km2, and 15% have areas >100km2. I'm not sure about the author's definition of headwater stream, but I would refer to non-perennial rivers rather than headwater streams.
4) In the ONDE dataset, not all stations within one Hydro-ecological region are observed in the same day. In fact, some stations have been observed 40 times, others more than 120. The number of observed stations in a given day also ranges from 10 to more than 2000. Is this taken into account when estimating PFIh? As an example, let's say in one region there are 100 ONDE stations. 40 of these stations are observed on day A, and 20 on day B (let's imagine groups A and B do not overlap). If stations observed on day A are generally wetter than stations of day B, the corresponding PFI(A) and PFI(B) are not directly comparable (i.e., the same PFI will be related to significantly different values of FQ) and cannot be modeled with the same logistic regression. Further, the other 40 stations have not been observed in either date. This occurrence can create significant inaccuracies in the regional estimation of PFI and should be taken into account.
5) The entire available discharge timeseries of each gauging station is used to estimate the flow duration curve. I imagine different gauging stations have highly different time windows of data availability. How d this affect the estimation of FQ,HER2h? Since logistic regression models are fitted on data between 2012 and 2022, you should use the same time window for the flow duration curves.
Finally, more citations could be added along the text, such as:
- Jensen et al, HP 2017, Lapides et al., HP 2021 (flow intermittence results from limited rainfall, freezing conditions, human alterations)
- Durighetto et al. RSOS 2022 could be cited in line 29 (climate is a primary driver of streamflow patterns)
- Other citations that link non-perennial streams with streamflow, e.g. Shaw et al., JoH 2017, Shaw, HP 2016
- Other models that try to estimate streamflow intermittency, e.g. Jaeger et al., JoHX 2019
- Other works on the importance of non-perennial streams beyond hydrology, such as Bertassello et al., RSOS 2022, Giezendanner et al., WRR 2021.TECHNICAL CORRECTIONS
Figure 2, panel "logistic regression": what is the vertical axis representing in this plot?
Figure 3, line 2: how many of the gauging stations define nested catchments? How are the overlapping areas accounted for? I guess you count the overlapping areas more than once, as ratio also goes above 1. I think a more descriptive ratio should not double count overlapping areas in nested catchments.
Figure 3, line 3: the unit of the density is actually km^{-2}.
Figure 4, panel "bias": it seems there are no biases > 1 or < -1. I suggest you to reduce the span of the colorbar accordingly, in order to better represent the bias variability on the map.
Line 148: the $ sign is visible (I guess it should indicate an equation within the text)
Line 201: how are the four HER2 regions used for illustration been selected?
Line 410: missing verb
Line 413: how could groundwater levels improve projections? Can you quickly elaborate on what are the expected results if groundwater levels are included in the simulation (with proper citations if available)
Data availability: provide also the link to the ONDE dataset
Citation: https://doi.org/10.5194/egusphere-2024-2737-RC1 -
RC2: 'Comment on egusphere-2024-2737', Anonymous Referee #2, 11 Nov 2024
This paper details modelling and projections of the proportion of dry headwater streams in France at the scale of hydro-ecoregions. The topic is relevant, and the analyses appears to be sound. However, this paper is somewhat hard to comprehend for readers outside the specific research field or unfamiliar with France. I have the following comments to improve the paper:
MAJOR COMMENTS
- The second, third and fourth paragraphs of the introduction delves into details of the datasets and methods used in this study. This content would ideally be in the data and methods sections, and I suggest covering it there. In the introduction, I suggest that the authors provide more context and background about the literature on flow intermittence. Please include insights into the various methods that have been used to study intermittence in various regions of the globe (including outside France), and a summary of the outcomes. I suggest ending the introduction with the clear statement of the research gap this paper is addressing.
- The reader is referred to (Sauquet et al. in prep). for the hydrological model simulations that underpin the analyses, so it difficult to understand the implications of these hydrological projections on the results presented here. I suggest including the implications of hydrological modelling assumptions also in the discussion.
- The last two sentences of the abstract touch on the causes of uncertainties in northern France and the regime changes in mountainous regions. This gives the incorrect impression that these aspects are part of the analyses. While these points are touched on in the discussion, they are not the results of the analyses presented in this study. This study solely focuses on validation, projections, uncertainty partitioning and agreement in PFI, and I suggest the results summary sentences in the abstract to be consistent with that. If the authors wish to convey the additional aspects, it should be clear that those are the implications and not results of the analyses.
MINOR COMMENTS
Lines 107-108: “On average, each simulation point has discharge projections simulated by four HM (Interquartile Q1 and Q3 (IQ): 3-5).” It is unclear what the text inside the parenthesis means.
Lines 119-120: “In the end, the bias-corrected climate projections were used as inputs for the hydrological models: 10 GCM-RCM projections (resp. 9 and 17) are used..” Again, unclear what the text inside the parenthesis means. Also are there 17 GCM-RCM simulations, or 10 of them?
Figure 2 and Section 3.1: The explanation of how non-exceedance probabilities (of what flow?) are used as the explanatory variable is unclear. I think the authors mean that the flow duration curves at each gauge are used to obtain the non-exceedance probability of the observed/modelled daily flow for each day during the seven-day window. If so, please clarify this in the text as well as in the schematic. In Figure 2, it can be shown how the probabilities are read from the flow duration curve for an example campaign date which is then used as one of the data points for the logistic regression fit.
Figure 3: It is hard to make out any difference between the columns presented for the gauging stations, and the CTRIP, GRSD, ORCHIDEE, and SMASH models (columns 2, 3, 4, 6 and 7). If the information is the same, consider presenting one set of plots for all five. If they are not the same, consider how the figure can be changed to convey that. If the differences between them is not the main point, maybe you need only one of the columns to be in the main text, and the rest can be in the appendix?
Use of abbreviations H0, H1, H2: Please consider if it is necessary to name the time periods as it seems simpler to just use the years instead. The paper has quite a lot of abbreviations, and I’m having to go back and check what H0, H1, H2 stands for.
The QUALYPSO method used to characterise uncertainty in not detailed well. The authors could consider using some illustrative equations to explain how the uncertainty is partitioned, and also explain what the different uncertainties/variabilities mean.
Line 289-290: Are some words missing from this sentence?
Line 317: “In contrast, under RCP8.5…”. Unclear what the contrast is.
Lines 355-357: HM also contributes to total variability?
Lines 450-452: Citation for the changes in rainfall and flows described.
Citation: https://doi.org/10.5194/egusphere-2024-2737-RC2
Data sets
Explore2 Eric Sauquet et al. https://entrepot.recherche.data.gouv.fr/dataverse/explore2
Model code and software
PostDocINRAE Jaouen Tristan https://github.com/tjaouen/PostDocINRAE
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 179 | 31 | 11 | 221 | 3 | 2 |
- HTML: 179
- PDF: 31
- XML: 11
- Total: 221
- BibTeX: 3
- EndNote: 2
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1