the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Oct 2024
Rain-on-snow events in mountainous catchments under climate change
Abstract. The frequency and intensity of rain-on-snow events (RoS) are expected to change in response to climate variations due to changes in precipitation, increase in air temperature and subsequent changes in the snow occurrence. In this study, we attributed these changes to the simulated variations in RoS events using a sensitivity analysis of precipitation and air temperature, and subsequent effects on RoS-related runoff responses were evaluated. We selected 93 mountainous catchments located in Central Europe across Czechia (60), Switzerland (26) and Germany (7), and used a conceptual hydrological model to simulate runoff components for 24 climate projections relative to the reference period 1980–2010. Climate change-driven RoS changes were highly variable over regions, across elevations, and within the cold season. The warmest projections suggested a decrease in RoS days by about 75 % for the Czech catchments. In contrast, the Swiss catchments may respond less sensitively, with the number of RoS days even increasing, specifically during the winter months and at higher elevations. Our projections also suggested that the RoS contribution to annual runoff will be considerably reduced from the current 10 % to 2–4 % for the warmest projections in Czechia, and from 18 % to 5–9 % in Switzerland. However, the RoS contribution to runoff may increase in winter months, especially for projections leading to an increase in precipitation, demonstrating the joint importance of air temperature and precipitation for future hydrological behavior in snow-dominated catchments.
- Preprint
(2299 KB) - Metadata XML
-
Supplement
(360 KB) - BibTeX
- EndNote
Status: open (until 03 Dec 2024)
-
RC1: 'Comment on egusphere-2024-2274', Anonymous Referee #1, 05 Nov 2024
reply
In this manuscript, the authors evaluate the sensitivity of rain-on-snow (RoS) events and their associated contribution to annual runoff given incremental changes to annual temperature and precipitation magnitudes. They test this by perturbing meteorological conditions from a 30-year reference period (1980-2010) over 93 catchments focused in Czechia, Germany, and Switzerland. This study finds that changes to RoS events depended on the geographic region, with approximately 75% of the lower-elevation Czech catchments demonstrating decreases in RoS with the larges (4 C) increases in temperature. The higher-elevation Swiss catchments were less-sensitive overall, with a higher frequency of unique temperature and precipitation perturbations driving increases to the number of RoS events. Interestingly, the results also show slightly different physical relationships between RoS days and multiple climate and snow variables in the two regions, suggesting different physical drivers of RoS occurrence. Finally, the authors go further by demonstrating how climate-driven changes to RoS events will alter historical RoS contributions runoff, both seasonally and annually.
I would like to commend the authors on their hard work and interesting manuscript. It was clear to me that a considerable amount of work went into this study, and I believe that the methodology, results, and presentation would lend itself well to the scope of HESS. However, I had two main concerns about the study. First, while the study demonstrated interesting sensitivities to incremental changes to both temperature and precipitation, there were a number of methodological and modeling decisions that could be influencing the main results. These include: a precipitation thresholding approach that excludes mixed-phase precipitation, the snow-state requirements used to prescribe the occurrence of a RoS event, the likelihood of unique combinations of temperature and precipitation perturbations, and the assumption of stationary changes in climate. While I don’t think it’s necessary for an investigation and discussion of these modeling decisions to be a major part of the manuscript, more should be done to establish whether the impacts on RoS and runoff presented by this research are outside the uncertainties that could be driven from the model decisions listed above. Secondly, not enough information on the change to snow cover was included, making it difficult to determine the extent to which RoS frequency was altered by changes to snow cover duration, relative to changes in precipitation phase. A deeper discussion on both of these comments are included in the "Major comments" below.
My recommendation is that this manuscript be returned to the authors for major revisions. Again, I would like to thank the authors for their contribution, and I really enjoyed reading this study. I would be delighted to review this manuscript again if the authors choose to continue with HESS.
Major comments:
Introduction: The authors do a really nice job at gathering and citing a number of the most relevant studies on RoS. However, the introduction often alludes to findings, complexities, shortfalls, and uncertainties in these studies without providing details or examples. The authors should consider using some of the literature review to clearly indicate the specific gaps to be addressed by this study. As an example, some of these things that could have been addressed more explicitly include: the list of “unsolved problems” in line 35, the processes/interactions that make RoS events and the resulting hydrology “complex in nature” (line 47), a deeper discussion of what is meant by “compound effect” (compounded uncertainties from estimates of snow cover and predictions of rainfall?, line 48), and what explicitly is included in the “different climate variables” (line 65).
As noted by the authors, RoS occurrence and severity depends on both snow cover existing, and warm temperatures coinciding with a precipitation event. Given this, more should be done throughout the manuscript to establish that the results are statistically-significant provided the uncertainties from the experimental setup. For example:
- My understanding of the modeling setup is that although the threshold used to partition rain and snow could change based on the basin, it didn’t allow mixed-phase precipitation meaning that all of a given timestep’s precipitation fell as snow or rain given the smallest of changes in temperature across any given temperature threshold. Since mixed-phase precipitation can often occur across large spreads in temperature, do the authors know how the decision to use a static temperature threshold impacted the frequency of RoS relative to a threshold that allowed mixed-phase rain and snow? My hypothesis is that allowing mixed-phase while keeping the decision to filter drizzle (daily precip < 5 mm) may result in less-sensitive changes to RoS in response to changes in temperature.
- I like the authors’ decision to filter regions where RoS occurred based on a SWE state threshold of 10 mm. However, it’s worth noting that larger footprints of shallow snow experiencing a RoS event may still contribute significantly to the hydrologic pulse. This may be particularly true in future climates, and across shallow-snow regions which melt more-readily and rapidly with the heightened turbulent, latent, and sensible heat fluxes during RoS events. Did the authors test other SWE thresholds?
- The combinations of the precipitation and temperature perturbations are presented as if each of these are equal-likelihood. While I like this structured investigation of impacts from incremental changes to both temperature and precipitation, there are some edge-case scenarios that may be less likely given projected changes to climate (e.g., T4_P08). The authors should consider adding some text to ground which combinations of temperature and precipitation changes, and resulting changes to RoS frequency and severity, may be more and less-likely to emerge in future climates.
- My overarching largest concern is the assumption of stationary changes in climate. For example, I believe this modeling setup applied a constant multiplier to the historical precipitation record, preserving the timing, severity, and frequency of precipitation, and how it aligned with swings in temperature. Provided the fact that both winter precipitation magnitude and frequency is expected to change in future climates (and more so in the winter than the summer), and that these may be more likely to coincide with moist and warm temperatures, it’s likely that an increase to the frequency of precipitation events may overwhelm some of the changes to RoS frequency and severity driven by stationary changes to temperature and precipitation. The authors should investigate this.
While some of the points from above are mentioned briefly in the study discussion, the authors should consider expanding on them to investigate where, when, and in what cases the RoS sensitivities reported by this study fall outside of the noise expected from the experimental setup and procedure. The authors could consider test cases using the full range of years and catchments, or case-studies based on catchments comparing the least (e.g., coldest and driest) and most-sensitive (e.g., catchments within the transition zone) locations.
Except for Figure 8, there is little presented about the impact that changes to temperature and precipitation have on simulated snow cover. This is particularly important since changes in RoS frequency can be driven to a first-order by changes to snow cover. Additionally, the model used in this study simulated 100 m elevation bands, thereby assuming full or absent snow cover for the full band while some level of fractional snow cover likely existed. Many of the projected decreases to RoS frequency are consistent with the signals expected for a reducing snow cover. However, this isn’t presented explicitly. There are also some results that suggest that there may be RoS increases driven by an increase in the snow cover duration. For example, Figure 5 shows an increase (relative to the historical) in the percent of RoS days from T2_P1 to T2_P12 in the Western Sudetes. Given 1) that only the magnitude and not the frequency/timing of precipitation events aren’t changing between T2_P1 and T2_P12, and 2) temperature is not changing between T2_P1 and T2_P12, the ~25% increase in RoS frequency between these two models must have been driven by changes in snow cover. Is that correct? Are these increases in RoS frequency happening earlier or later in the snow season, and across what elevation bands?
Section 3.7: I really like this analysis on the runoff response! This is a great advancement on the field of RoS studies.
Minor comments:
Line 8: Change “increase” to “increases” in order to match the tense of “changes”.
Line 8: “Occurrence” is used throughout to reference RoS events. The authors could consider revising “snow occurrence” to “snow cover” in this context (also line 51).
Line 10: Delete “were evaluated”
Lines 34 – 36: Without more context provided for the “unsolved problems”, this paragraph doesn’t provide a lot of new information, especially considering the literature review provided in the following paragraphs. I think the authors could consider removing these lines. I would have the same comment for the last line of the following paragraph: “Much of the current research … under ongoing climate change”.
Line 57: What is meant by “spatial and temporal distribution”? Is this referencing the changes to snow cover in time and space, in addition to the changes to rainfall frequencies?
Lines 100 – 101: In instances where a station came from outside of the catchment bounds, how far away were these stations, on average? What was the maximum and minimum distances, and what sort of uncertainties could be expected for the stations that are the furthest away?
Line 109: Is there a citation for the MeteoSwiss gridded meteorology product? What is the spatial resolution of this forcing?
Line 144: Replace “e.g. Hotovy et al. (2023)” with “(Hotovy et al., 2023)”.
Lines 154 – 156: I’m finding this passage a little confusing. The authors specify that RoS events are “multi-day” events, but then follow by saying that RoS events can include both RoS days and non-RoS days. Does this just mean that the runoff can peak after the date of the actual rainfall?
Line 166: Why does P08 correspond with -20%? Can these just be named based on their percent-perturbations (e.g., P-20% or P80%)?
Figure 3: I’m having a difficult time distinguishing the different colors, and especially in instances when the points are separated further in space. The authors could consider a few things: 1) testing different color bars, 2) adapting both the point color and size to correspond with the average RoS day occurrence, and/or 3) breaking this plot into two separate subplots -- one with a y-axis corresponding to RoS days per year and the other with the average RoS day occurrence.
Figure 3: How large is the spread in the timing of the RoS day occurrence? If the authors choose to break this into a different subplot, they could consider including whiskers to show this.
Line 213: This is a huge number of RoS events! Is there evidence to back up that this is grounded in reality?
Figure 7b: It looks like there are two solid yellow (T0_P1) lines. My guess is that the rightmost line actually represents T0_P12.
Citation: https://doi.org/10.5194/egusphere-2024-2274-RC1 -
RC2: 'Comment on egusphere-2024-2274', Anonymous Referee #2, 09 Nov 2024
reply
This manuscript examined rain-on-snow (ROS) frequencies, runoff responses, and relationships to hydroclimate variables across Czechia, Switzerland, and Germany. This was accomplished using a hydrologic model and perturbed runs were also examined to simulate climate change impacts to ROS.
This was a well written paper and fits well into this journal. The results are summarized nicely, and the discussion does well highlighting caveats and uncertainty. My only major comment is around more clarification needed for some of the methods. Otherwise, mostly minor comments. I’m giving it a minor revision as I don’t think the “major comment” will take all that much work.
Major comments:
- Several aspects of the data and methods around the HBV model were unclear and could be improved by adding more details. One concern I have that I don’t think was addressed was using station-based data as model inputs for the Czech catchments but then using gridded data for the Swiss catchments. Can the authors comment on what impact the station-based vs. gridded inputs might have on the simulations? Also, what was the spatial resolution of the gridded data used? It also was not clear to me what the HBV model inputs were. The historical dataset consists of temperature, precipitation, runoff, and SWE. Were all of these input into the HBV model?
Other Comments:
- Figure1: I think this could be improved. As is, the polygons are very small and hard to see. I might suggest and panel plot with several zoomed maps that also include more detailed terrain.
- I would suggest using the term “perturbations” instead of “projections” throughout the paper. Projections are usually associated with GCM outputs in the future, and the future periods are not directly accounted for in your method.
- Line 190: What is the median objective function? Is a value of 1 a perfect score? Please explain.
- Figure 3: Great figure! Is the annual number of ROS days the mean? Median? Please specify in the caption.
- Figure 7: It looks like the T0_P12 line is drawn wrong in 7b. It shows a solid line and based on the text in lines 276-277 it should be small-dashed.
Citation: https://doi.org/10.5194/egusphere-2024-2274-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 189 | 78 | 11 | 278 | 34 | 3 | 3 |
- HTML: 189
- PDF: 78
- XML: 11
- Total: 278
- Supplement: 34
- BibTeX: 3
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1