the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Mar 2022
Snow Persistence Explains Stream High Flow and Low Flow Signatures with Differing Relationships by Aridity and Climatic Seasonality
Abstract. Snow persistence is a globally available metric of snow cover duration that has, thus far, seen limited application to the field of hydrology. This study attempts to explore the controls that snow persistence exerts on streamflow at low and high flow conditions across a diverse spectrum of climatic aridity and seasonality in the United States and Canada. We statistically analyze how snow persistence, aridity, and seasonality conditions interact to control and explain streamflow shape and flashiness at low and high flows. For low flow condition, regardless of climatic aridity and seasonality, a larger snow persistence increases baseflow, reduces low flow variability, and increases the magnitude of extreme low flow relative to average flow. Our results further show that snow persistence becomes a stronger factor in controlling baseflow as well as the magnitude of extreme low flow relative to average flow, in regions with a relatively high aridity and/or with summer-dominant precipitation regimes (or in-phase seasonality). On the other hand, in catchments that are moderately wet to very arid with winter-dominant precipitation regimes (or out-of-phase seasonality), a longer snow persistence could typically lead to a more variability at high flow and a larger magnitude of extreme high flow relative to average flow. This study concludes by demonstrating the relevancy of snow persistence as a globally available streamflow behaviour descriptor and by demonstrating the impacts that climate change may have on snow persistence and ultimately on streamflow behaviour at low and high flows.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(1477 KB)
-
Supplement
(369 KB)
-
This preprint has been withdrawn.
- Preprint
(1477 KB) - Metadata XML
-
Supplement
(369 KB) - BibTeX
- EndNote
Interactive discussion
Status: closed
-
RC1: 'Comment on hess-2022-106', Anonymous Referee #1, 19 Apr 2022
General comments
Authors analysed how snow persistence, aridity, and seasonality conditions control streamflow, specifically flashiness at low and high flows in 1187 catchments across United States and Canada. They addressed two main research questions, which are 1) Can snow persistence explain the shape of streamflow hydrographs at low and high flow conditions? and 2) How do aridity and seasonality affect the ability of snow persistence to explaining shape-based signatures at stream low and high flow conditions? Authors found that for low flow conditions, larger snow persistence increases baseflow and reduces low flow variability regardless catchment aridity and seasonality. Authors also showed that snow persistence became a stronger factor in controlling baseflow, in regions with a relatively high aridity and/or with summer-dominant precipitation regimes. Based on analyses, authors concluded that snow persistence may serve as a useful streamflow descriptor across different climates and runoff regimes.
In my opinion, authors did an interesting work. I agree with authors, that the use of snow persistence as a metric describing catchment response across a wide range of climates is novel. Besides that, I see the novelty that this metric can be derived from satellite data and thus it can be used at a global scale. I also like that author related the snow persistence to different aridity and seasonality of catchments. Although the results are not much surprising as they mostly confirm our existing knowledge, I found the study important and novel, thus appropriate for HESS. However, I have several comments listed below, which should be addressed before I can recommend the manuscript for publication.
Major comments
After reading of abstract and study objectives, I was really motivated in further reading since I was curious about what authors investigated. However, I was a bit disappointed, because in my opinion, the study conclusions are not sufficiently supported by results and illustrations. In my opinion, only two figures showing study results (Fig. 2 and 3) are too few to draw general conclusions. Therefore, I would encourage authors to support their results with further analysis and figures. In comments below, I tried to make a few suggestions which authors may consider.
In my opinion, it might be interesting to look whether the snow persistence is a good predictor for the selected runoff signatures in years with snow-poor and snow-rich winters (or dry/wet, cold/warm years). Comparing statistics of individual years instead of mean statistics of the whole study period may allow for a more direct attribution of the inter-annual variations of snowpack to variations in runoff characteristics.
I see the evidence provided by authors that snow persistence can partly explain the selected streamflow signatures. However, the snow regime belongs to the main component of the water balance in high-elevation and/or high-latitude catchments next to precipitation and its seasonality and evapotranspiration. Therefore, similar results would be maybe achieved for any of these characteristics. It means, that not only snow persistence might explain streamflow characteristics, but also aridity or seasonality indexes might bring similar results. Therefore, to further support existing results, it might be interesting to look how strong is the snow persistence as a predictor compared to aridity and seasonality indexes. Maybe, at least some correlation analysis comparing the predictive strength of all predictors might be beneficial.
In addition to my above comments, results section should be extended. As it is now, it contains only a short description of results shown in Fig. 3 and it looks unproportionally short compared to the discussion section. As I mentioned above, interpretation based on one or two figures seems unconvincingly to me and I would encourage the authors to add more analysis and related interpretation which may further support (so far interesting) results.
Specific comments
L 105: I would somewhere mention the basic statistics of the study catchments (e.g., as a range of values), such as area, elevation, annual precipitation, snow persistence, etc.
L 110: I suggest including equations of how the main characteristics (snow persistence, aridity index, seasonality index) have been calculated.
L 147: Maybe I did not understand correctly, but 30% of area difference between the two different approaches of area calculation sounds as a large difference. Why is it so much?
L149: I understand that only perennial rivers were considered for the analysis. Nevertheless, would the results be different if also river intermittency would be considered? Please discuss shortly.
L173: 30th and 70th percentiles for low- and high FDCs sound rather as arbitrary choice. Is there any reason for choosing exactly those thresholds?
Authors defined several streamflow signatures for the analysis. This is fine, but I would suggest including a few more, for example low flow duration or deficit volumes. Especially the former might be beneficial to further explain the role of snow persistence on low flow regime.
L 195: How the last day with snow presence has been calculated? Due to elevation range of individual catchments, the snow may be melted at lower elevations while some snow may be still present at higher elevations. Please clarify shortly.
L286-287: The research questions here are repetition from above, please considerer whether it is needed to introduce them again at this place.
L320: Maybe I missed it somehow, but I do not see reducing low flow variability from Fig. 3a-c as noted by authors. Please, add some more explanation.
Technical corrections
L154-155: I would omit these two lines.
L296: “snowpack” rather than “snow pack”.
Fig. 4: Consider adjusting light blue and moderate blue colours in the figure since one can hardly see the difference between both.
L388: Perhaps, the brackets in “changes in” are not necessary.
L606: If I checked correctly, Muñoz-Sabater et al. (2021) has already the final paper published.
Citation: https://doi.org/10.5194/hess-2022-106-RC1 - AC1: 'Reply on RC1', Ali Ameli, 12 Jun 2022
-
RC2: 'Comment on hess-2022-106', Anonymous Referee #2, 27 Apr 2022
1 General comments
Le et al. analysed how well snow persistence (along with aridity and precipitation seasonality) can explain a range of flow signatures in over 1000 catchments in North America, using a 19-year data set. They applied a linear model with interaction term (multilinear regression) and visually analysed the influence of snow persistence on each response variable. With a very short results section and basically only one figure (Figure 3), they come to significant results, such as that snow persistence influences low flow characteristics, and in some climate regions also high flow characteristics. Furthermore, the authors established a link between the spatial changes observed and future climatic changes, such as how a reduction in snow persistence could change flow characteristics.
In my opinion, the fitted linear models are not able to capture the variance sufficiently to draw these essential conclusions. The authors report values for R2 ranging from 0.11 to 0.25 (Table 1). Similarly (or as a consequence), the explained effect of snow persistence on the response variables is small: the largest values on the y-axes in Fig. 3 cover only 0.1% (for Q95) to 4% (for Q5) of the indicated interquartile range in Table 1. The effect is statistically significant, as mentioned by the authors, but in my view too small to be relevant. This is a common problem: as sample size increases, decreasing effects become statistically significant. The authors need to find ways to create models with greater predictive value that are able to produce effects of relevant size. In my opinion, the small effect size of the linear models makes this manuscript too weak to be considered for publication in HESS. I will explain this in more detail in the next section.
2 Specific comments on the small effect size
The authors discuss these low R2 values in their “Limitations” section and mention that this is to be expected as geological and topographical factors were not included. They cite Addor et al. (2018), for example, who considered these factors important. I disagree with this expectation of low R2 values and also with the explanation: Addor et al. (2018), a very similar but much more comprehensive study (barely cited by Le et al.), concluded “… that climatic attributes are by far the most influential predictors for signatures that can be well predicted based on catchment attributes”. Instead of simple linear models, they trained Random Forests and found that they could explain large parts of the variances of signatures such as Q95 (R2 >0.8), Q5 (R2 ~0.6) and BFI (R2 ~0.5) with climatic attributes alone (read from their Figure 5). These values are much larger compared to those reported here. Only for the slope of the flow duration curve was a similarly small R2 value reported.
The reasons for the larger R2 values reported by Addor et al. (2018) could be that they used
- more and other climatic variables
- more complex models
- a longer dataset, limited to the US.
The first item is important for the aim of the Le et al. manuscript, namely to show the predictive value of snowpack persistence (SP). As the authors indicate, snowpack persistence is easier to determine compared to snowfall fraction (which was used by Addor et al., 2008) and is therefore a very interesting and globally available predictor variable. To show the predictive value of SP, I would suggest repeating the Addor et al. (2018) study for the US and Canadian datasets and only use their climatic variables, then replace the snowfall fraction with SP and then remove step-wise all other climatic variables until the three used here remain (i.e. SP, seasonality of precipitation and aridity). With this setup, one can find out what the authors were aiming for, namely (line 418ff): "how far we can go in explaining detailed streamflow characteristics with a simple, widely available and accurate satellite-based snow-related metric" (along with seasonality and aridity).
Citation: https://doi.org/10.5194/hess-2022-106-RC2 - AC2: 'Reply on RC2', Ali Ameli, 12 Jun 2022
Interactive discussion
Status: closed
-
RC1: 'Comment on hess-2022-106', Anonymous Referee #1, 19 Apr 2022
General comments
Authors analysed how snow persistence, aridity, and seasonality conditions control streamflow, specifically flashiness at low and high flows in 1187 catchments across United States and Canada. They addressed two main research questions, which are 1) Can snow persistence explain the shape of streamflow hydrographs at low and high flow conditions? and 2) How do aridity and seasonality affect the ability of snow persistence to explaining shape-based signatures at stream low and high flow conditions? Authors found that for low flow conditions, larger snow persistence increases baseflow and reduces low flow variability regardless catchment aridity and seasonality. Authors also showed that snow persistence became a stronger factor in controlling baseflow, in regions with a relatively high aridity and/or with summer-dominant precipitation regimes. Based on analyses, authors concluded that snow persistence may serve as a useful streamflow descriptor across different climates and runoff regimes.
In my opinion, authors did an interesting work. I agree with authors, that the use of snow persistence as a metric describing catchment response across a wide range of climates is novel. Besides that, I see the novelty that this metric can be derived from satellite data and thus it can be used at a global scale. I also like that author related the snow persistence to different aridity and seasonality of catchments. Although the results are not much surprising as they mostly confirm our existing knowledge, I found the study important and novel, thus appropriate for HESS. However, I have several comments listed below, which should be addressed before I can recommend the manuscript for publication.
Major comments
After reading of abstract and study objectives, I was really motivated in further reading since I was curious about what authors investigated. However, I was a bit disappointed, because in my opinion, the study conclusions are not sufficiently supported by results and illustrations. In my opinion, only two figures showing study results (Fig. 2 and 3) are too few to draw general conclusions. Therefore, I would encourage authors to support their results with further analysis and figures. In comments below, I tried to make a few suggestions which authors may consider.
In my opinion, it might be interesting to look whether the snow persistence is a good predictor for the selected runoff signatures in years with snow-poor and snow-rich winters (or dry/wet, cold/warm years). Comparing statistics of individual years instead of mean statistics of the whole study period may allow for a more direct attribution of the inter-annual variations of snowpack to variations in runoff characteristics.
I see the evidence provided by authors that snow persistence can partly explain the selected streamflow signatures. However, the snow regime belongs to the main component of the water balance in high-elevation and/or high-latitude catchments next to precipitation and its seasonality and evapotranspiration. Therefore, similar results would be maybe achieved for any of these characteristics. It means, that not only snow persistence might explain streamflow characteristics, but also aridity or seasonality indexes might bring similar results. Therefore, to further support existing results, it might be interesting to look how strong is the snow persistence as a predictor compared to aridity and seasonality indexes. Maybe, at least some correlation analysis comparing the predictive strength of all predictors might be beneficial.
In addition to my above comments, results section should be extended. As it is now, it contains only a short description of results shown in Fig. 3 and it looks unproportionally short compared to the discussion section. As I mentioned above, interpretation based on one or two figures seems unconvincingly to me and I would encourage the authors to add more analysis and related interpretation which may further support (so far interesting) results.
Specific comments
L 105: I would somewhere mention the basic statistics of the study catchments (e.g., as a range of values), such as area, elevation, annual precipitation, snow persistence, etc.
L 110: I suggest including equations of how the main characteristics (snow persistence, aridity index, seasonality index) have been calculated.
L 147: Maybe I did not understand correctly, but 30% of area difference between the two different approaches of area calculation sounds as a large difference. Why is it so much?
L149: I understand that only perennial rivers were considered for the analysis. Nevertheless, would the results be different if also river intermittency would be considered? Please discuss shortly.
L173: 30th and 70th percentiles for low- and high FDCs sound rather as arbitrary choice. Is there any reason for choosing exactly those thresholds?
Authors defined several streamflow signatures for the analysis. This is fine, but I would suggest including a few more, for example low flow duration or deficit volumes. Especially the former might be beneficial to further explain the role of snow persistence on low flow regime.
L 195: How the last day with snow presence has been calculated? Due to elevation range of individual catchments, the snow may be melted at lower elevations while some snow may be still present at higher elevations. Please clarify shortly.
L286-287: The research questions here are repetition from above, please considerer whether it is needed to introduce them again at this place.
L320: Maybe I missed it somehow, but I do not see reducing low flow variability from Fig. 3a-c as noted by authors. Please, add some more explanation.
Technical corrections
L154-155: I would omit these two lines.
L296: “snowpack” rather than “snow pack”.
Fig. 4: Consider adjusting light blue and moderate blue colours in the figure since one can hardly see the difference between both.
L388: Perhaps, the brackets in “changes in” are not necessary.
L606: If I checked correctly, Muñoz-Sabater et al. (2021) has already the final paper published.
Citation: https://doi.org/10.5194/hess-2022-106-RC1 - AC1: 'Reply on RC1', Ali Ameli, 12 Jun 2022
-
RC2: 'Comment on hess-2022-106', Anonymous Referee #2, 27 Apr 2022
1 General comments
Le et al. analysed how well snow persistence (along with aridity and precipitation seasonality) can explain a range of flow signatures in over 1000 catchments in North America, using a 19-year data set. They applied a linear model with interaction term (multilinear regression) and visually analysed the influence of snow persistence on each response variable. With a very short results section and basically only one figure (Figure 3), they come to significant results, such as that snow persistence influences low flow characteristics, and in some climate regions also high flow characteristics. Furthermore, the authors established a link between the spatial changes observed and future climatic changes, such as how a reduction in snow persistence could change flow characteristics.
In my opinion, the fitted linear models are not able to capture the variance sufficiently to draw these essential conclusions. The authors report values for R2 ranging from 0.11 to 0.25 (Table 1). Similarly (or as a consequence), the explained effect of snow persistence on the response variables is small: the largest values on the y-axes in Fig. 3 cover only 0.1% (for Q95) to 4% (for Q5) of the indicated interquartile range in Table 1. The effect is statistically significant, as mentioned by the authors, but in my view too small to be relevant. This is a common problem: as sample size increases, decreasing effects become statistically significant. The authors need to find ways to create models with greater predictive value that are able to produce effects of relevant size. In my opinion, the small effect size of the linear models makes this manuscript too weak to be considered for publication in HESS. I will explain this in more detail in the next section.
2 Specific comments on the small effect size
The authors discuss these low R2 values in their “Limitations” section and mention that this is to be expected as geological and topographical factors were not included. They cite Addor et al. (2018), for example, who considered these factors important. I disagree with this expectation of low R2 values and also with the explanation: Addor et al. (2018), a very similar but much more comprehensive study (barely cited by Le et al.), concluded “… that climatic attributes are by far the most influential predictors for signatures that can be well predicted based on catchment attributes”. Instead of simple linear models, they trained Random Forests and found that they could explain large parts of the variances of signatures such as Q95 (R2 >0.8), Q5 (R2 ~0.6) and BFI (R2 ~0.5) with climatic attributes alone (read from their Figure 5). These values are much larger compared to those reported here. Only for the slope of the flow duration curve was a similarly small R2 value reported.
The reasons for the larger R2 values reported by Addor et al. (2018) could be that they used
- more and other climatic variables
- more complex models
- a longer dataset, limited to the US.
The first item is important for the aim of the Le et al. manuscript, namely to show the predictive value of snowpack persistence (SP). As the authors indicate, snowpack persistence is easier to determine compared to snowfall fraction (which was used by Addor et al., 2008) and is therefore a very interesting and globally available predictor variable. To show the predictive value of SP, I would suggest repeating the Addor et al. (2018) study for the US and Canadian datasets and only use their climatic variables, then replace the snowfall fraction with SP and then remove step-wise all other climatic variables until the three used here remain (i.e. SP, seasonality of precipitation and aridity). With this setup, one can find out what the authors were aiming for, namely (line 418ff): "how far we can go in explaining detailed streamflow characteristics with a simple, widely available and accurate satellite-based snow-related metric" (along with seasonality and aridity).
Citation: https://doi.org/10.5194/hess-2022-106-RC2 - AC2: 'Reply on RC2', Ali Ameli, 12 Jun 2022
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 924 | 332 | 47 | 1,303 | 79 | 34 | 32 |
- HTML: 924
- PDF: 332
- XML: 47
- Total: 1,303
- Supplement: 79
- BibTeX: 34
- EndNote: 32
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Cited
Edward Le
Joseph Janssen
John Hammond
This preprint has been withdrawn.
- Preprint
(1477 KB) - Metadata XML
-
Supplement
(369 KB) - BibTeX
- EndNote