the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Dec 2024
Modelling the effects of climate and landcover change on the hydrologic regime of a snowmelt-dominated montane catchment
Abstract. Climate change poses risks to society through the potential to alter peak flows, low flows, and annual runoff yield. Wildfires are projected to increase due to climate change; however, little is known about their combined effects on hydrology. This study models the combined impacts of climate and landcover changes on the hydrologic regime of a snowmelt-dominated montane catchment, to identify management strategies that mitigate negative impacts. The combination of climate change and stand replacing landcover disturbance in the middle and high elevations is predicted to advance the timing of the peak flow two to nine times (depending on emission pathway) more than the advance generated by disturbance alone. The modelling predicts that the combined impacts of climate change and landcover disturbance on peak flow magnitude are generally offsetting for events with return periods less than 5–25 years, but additive for more extreme events. There is a dependency of extreme peak flows on the distribution of landcover. The modelling predicts an increasing importance of rainfall in controlling peak flow response under a changing climate, at the expense of snowmelt influence. Extreme summer low flows are predicted to become commonplace in the future, with most of the change in frequency occurring by the 2050s. Low annual yield is predicted to become more prevalent by the 2050s, but then fully recover or become less prevalent (compared to the current climate) by the 2080s, because of increased precipitation in the fall-spring period. The modelling suggests that landcover disturbance can have a mitigative influence on low water supply. The mitigative influence is predicted to be sustained under a changing climate for annual water yield, but not for summer low flow. The study results demonstrate the importance of examining complexity in three dimensions with respect to modelling changes to the hydrological regime: climate change, landcover change, and numerous hydrological indicators. Moreover, for managing watershed risk, the results indicate there is a need to carefully evaluate the interplay among environmental variables, the landscape, and the values at risk. Strategies to reduce one risk may increase others, or effective strategies may become less worthwhile in the future.
- Preprint
(5787 KB) - Metadata XML
-
Supplement
(4976 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Review of hess-2024-361', Anonymous Referee #1, 24 Jan 2025
This paper combines a large amount of data and different climate and land cover scenarios in a modelling study to determine the combined effects of land cover change and climate change on the snowpack and streamflow regime for a headwater catchment in British Columbia, Canada. This is a huge undertaking and I appreciate that the effects are analysed for many different aspects of the snowpack and hydrograph. The paper clearly shows the interaction between the effects of land cover change and climate change. For some aspects of the hydrograph or snowpack, the land cover change enhances the effects caused by climate change and for others, it mitigates it. The results furthermore highlight the importance of the location of the disturbance in the catchment (i.e., whether the vegetation is replaced in the upper or lower part of the catchment) and the time since the disturbance. These are important results and highlight the need to consider the effects of land cover and climate change jointly and to not study the effects of land cover change for only one climatic period.
Unfortunately, some of the model decisions are not so clearly described and it is not very clear how the model was calibrated. There is also no mention of the uncertainties in the results due to parameter uncertainty. Considering the potentially very large number of parameters that are optimized, it is possible that a different parameter set would lead to considerably different results. The lack of uncertainty analyses is acknowledged in the final part of the discussion, but I would argue that at least some model uncertainties need to be presented. As a result of the lack of a clear description of the calibration procedure and the lack of an uncertainty analysis, it is not clear how the presented results are influenced by the model decisions or model parameter sets (equifinality).
The graphs used to present the results are clear and very useful. But it would be good if they had error bars to represent the range of results caused by equally good fitting model parameter sets. I like it that the time series of the simulated and observed runoff are given for the individual years in the Supplementary material. The paper is long but overall, well written.
Specific comments:
- L14 and 862: Quantify this in a different way, e.g., in days or weeks. 2-9 times more is important if we talk about an advance of a week or several weeks due to disturbance but not if the advance is only 1 day.
- L26: Maybe use a different word than values (hydrograph characteristics, hydrological signatures?)
- L76: Considering all the uncertainties in these assessments, the decimals are probably not warranted here.
- L92: Give some info on the model here already. It would be good to know for the reader early on if you are using a physically based, spatially distributed model or some other model, if it was calibrated or not, etc.
- L126-129: It is nice that you describe the vegetation here and give the codes that you will use for the vegetation codes throughout the text but it is hard for the reader to remember these codes, especially since there are also codes for the different scenarios. In other words, it would be a lot easier for the reader to understand the parts about the vegetation if you would just write out the names instead of using the codes.
- L139: In addition to the mean annual runoff, also mention the mean annual precipitation, either averaged over the catchment or for at least one station. This is important information about the study site.
- Section 2.2.1: It would be good to already mention how many HRUs there are in this section (now it is only mentioned on L243) and how many parameters there are per HRU. Now this section is short and a lack of knowledge on the model and its parameters early on in the paper, hampers the understanding of the other parts in section 2.2.
- L195: How well is well? Is there a reference here or a result that you can add to the supp materials?
- L267: How many parameters are there per HRU and in total? and how many of these were calibrated? Even after reading the paper, this is unclear to me. Please add this information clearly in the methods section. Ideally already in section 2.2.1.
- L268: What weighting did you use for the calibration? Equal for each of these objective functions?
- L281, 283 and ff: What exactly do you mean by constrained (or in L301 and 306 by informed)? Did you pick a parameter value a priori and not calibrate it or did you select a parameter range and calibrate within this range?
- L313: This wording is not clear. Did you use it to guess a specific value and then use this in the model? Did you calibrate within a certain range? A bit more information, or clearer wording would be useful.
- L321: This is not clear - how did you get values for each specific channel? How different were these values?
- L331: How many parameters were optimized and how many were fixed based on field knowledge? Also did you use the same parameters for all the HRUs with the same vegetation or soil? Would it be possible to add a table with all parameter values and the range used for the optimization somewhere?
- L333, 389: How did you weigh these different objective functions in the calibration? All equal weight? Or did you optimize each function individually first? From L323-326, it appears that you did it sequentially? Or did you just use different time periods for each of these objective functions and calibrate everything at the same time using some weighted function? The current description doesn’t make the calibration process very clear to me. Also, what is the reason for not using the NSE for the entire study period as well?
- L461: Already mention here if this is largely due to a change in precipitation or due to a change in evapotranspiration.
- Figure 9: The shape of the curve changes as well. What is causing this? This requires some discussion.
- Section 5.2.4: Make it clearer that this is the annual *average* discharge
- L598: A lot of the quickflow probably consists of subsurface stormflow or even groundwater flow. The majority of quickflow is unlikely to be overland flow (surface runoff) for a forested catchment.
- L780: Groundwater would be a more likely source for the streamflow in the dry period than soil water (retention).
Minor comments
- L11: Mention the name of the model or the type of model in the abstract.
- L82-89: Move to the study site description.
- L121: Explain that BEC is the biogeoclimatic ecosystem classification.
- L191: Lowest temperatures instead of coolest temperatures.
- Figure 5: Maybe still add South and North to the axis labels for subpanel b?
- L431: These differences are very small. Highlight that first before giving the values!
- L700: What do you mean by snowpack loads?
- L720-721: Explain better how this sentence fits here / what you mean by this? What is the link to the previous or next sentence?
- L797 values at risk: Do you mean the streamflow signatures / hydrograph characteristics? This could be worded more clearly.
Citation: https://doi.org/10.5194/hess-2024-361-RC1 -
RC2: 'Comment on hess-2024-361', Anonymous Referee #2, 06 Feb 2025
Dear authors of the manuscript, thank you very much for revising the manuscript. The current version has undergone substantial enhancement in terms of content and presentation.
The manuscript is generally well-written, with figures and tables that are well-placed and clearly illustrate the results. The manuscript provides insights on annual and seasonal hydrological changes, as well as on the development of extreme summer low flows and peak flows under a range of climate scenarios and landcover conditions. The conclusions drawn are supported by the findings of the model. The in-depth analysis of potential future changes is informed by the CSIRO85 model. However, I still have some minor comments that should be addressed:
L 12-14: what does “two to nine times more” mean in absolute terms? Absolute figures would provide more clarity here.
L 26-27: are the “values at risk” related to the society as mentioned in the first sentence of the abstract? This needs clarification, also in the conclusions section.
L 267-268: which parameters have “substantial uncertainty and/or sensitivity”? Please name them. Although these parameters have been calibrated simultaneously, the issue of equifinality should be discussed (at least in section 6.4 on uncertainties).
L 303: I am still struggling with the meaning of these BEC variants. It would be nice if the BEC variants were explained in a table.
Table 5: would be helpful to split the numbers of net P into P and ET. The precipitation data given in other tables to not coincide with the aggregation used in Tab. 5
Citation: https://doi.org/10.5194/hess-2024-361-RC2 -
RC3: 'Comment on hess-2024-361', Anonymous Referee #3, 28 Feb 2025
Summary
The authors describe a study to explore the effects of climate change and landcover change on the hydrology (SWE and various discharge variables) of a forested nival watershed in the interior of British Columbia. Results are derived via simulation using hydrologic modelling in combination with various forest disturbance scenarios and mid- and end-century climate projections. The research deals with an important and timely issue, particularly considering the dramatic increase in wildfire risk experienced in western North America. However, the work suffers from the strictly qualitative, predominantly graphical, approach to both analyse and interpret the results. This results in a work that generally lacks any real scientific impact. This is particularly puzzling considering that the experimental design employed (full factorial) and the available data lend themselves well to the application of common statistical techniques (ANOVA and regression) that could give quantitative results and analysis of individual and combined effects of climate and forest cover. This is a missed opportunity. I feel that with a more quantitative assessment of results, this work would make a worthy scientific contribution; hence, my recommendation is to accept but with major revisions.
Major Issues
Experimental Design
The experimental design, which is central to the study, is described in an ad hoc fashion. The authors are encouraged to re-organize the document so that the experimental design is clearer. And in fact, it should be noted that the authors employ what is formally known as a factorial experimental design, where multiple factors (climate state/change, disturbance level, elevation, etc.) are tested for their influence on the outcome of a response variable (peak flow, max SWE, melt-out date, etc.). The experiments could be described as follows:
- Phase 1:
- a 13 x 2 design (26 categories) with 13 x climate states and 2 x disturbance levels
- Phase 2:
- SWE: a 2 x 2 x 3 x 3 design (36 categories) with 2 x slope aspects (north and south), 2 x disturbance level (forest and clearing), 3 x elevations (low, middle and high) and 3 x climate states (current, 2050s and 2080s
- Discharge: a 5 x 3 design (15 categories) with 5 x disturbance levels and 3 x climate states
Although the factorial design could have been exploited to formally explore main effects and interactions between the various factors, using such methods as ANOVA and regression, the authors opted instead for a qualitative and graphical approach and, I feel, missed an opportunity for a more impactful study. I am also not convinced that the graphical results can be used to isolate each individual effect (disturbance and climate) as stated. For example, the ‘disturbance effect’ in Figure 10 shows the effect of each disturbance conditional upon various climate states. I.e. the disturbance effect is not independent of climate state. As said prior, formal statistical approaches could be used to quantify the effect of each individual factor as well as the various combined effects. The authors could also take their existing results one step further. As each synthetic climate series is stationary, each 100-year series could be divided into non-overlapping decadal periods that would provide ten replicates per category, producing a full factorial design with replicates. Any experimental design text will explain how to more formally exploit this approach.
Phases:
I am not clear on the requirement of the Phase 1 portion of the methodology. It seems that the main purpose was to select the climate experiment to be used for Phase 2. If so, the authors should be aware that they, perhaps inadvertently, chose the ‘driest’ scenario (see Table 1); only CSIRO85 projects slightly decreased precipitation in the 2050s and has the smallest precipitation increase in the 2080s. I would highly recommend that the authors redo the Phase 2 analysis with at least one additional climate experiment (perhaps the wettest one, e.g. MIROC85 or MPI45).
Minor Issues
Line 24: Not sure ‘three dimensional’ is a suitable term for what you are describing. Perhaps ‘multi-facetted’?
Line 146: More specifically, Raven estimates cloud cover, etc. from T and P. The term ‘accounts for’ is a tad vague.
Section 2.2.2.1: If available, why wasn’t data from the Penticton Airport station used directly as a model forcing?
Line 173: You are using incorrect nomenclature. What you are describing as “emissions pathways” are what should be called climate projections (i.e. a projection is produced from a combination of an emission scenario and a global climate model). Your design only includes two, not five, emissions pathways/scenarios, RCP4.5. and RCP8.5.
Line 174: That bound “90% of projections” for what variable(s) over what region?
Line 194: Confirm the adjustment is made seasonally, and not monthly.
Line 199: What parameters are used, and then presumably adjusted, to describe the distribution of tasmin, tasmax, and precipitation.
Lines 249-254: The process being described seems to be better explained as a process if intersecting various layers as opposed to imprinting. In other words, discretizing HRUs is the process of intersecting five individual layers: sub-basins, BEC variants, disturbance history, vegetation type, and 2K x 2K grid (to limit HRU size to <= 4-km2).
Lines 279-285: The use of the word “constrained” implies that these parameter values were adjusted during calibration. Do you really mean the values were estimated a-priori using observations?
Lines 301-306: This section implies the LAI and crown closure are closely related, however, in my mind they describe different vegetation characteristics. LAI is a vegetation property (i.e. leaf area density for individual plants) whereas crown closure is a stand property (the density of individual trees/plants). The authors seem to have conflated the two properties. Is LAI, then, a combined value of vegetation and tree density?
Line 329: Which years were used for calibration and validation?
Line 332: How was the composite function constructed from the various indicators (I’m assuming it was the arithmetic mean). Were the individual metrics weighted? Show the mathematical description of the composite function.
Line 350: Replace “emissions pathways” with “climate projections”.
Line 353-354: Would recommend instead saying “For each 100-year simulation the landcover state was static and the meteorology derives from a stationary climate state”
Line 356: May use “dynamic equilibrium” instead of “wet”.
Section 2.5: There are two experiments being conducted to assess climate and landcover change on SWE, a point-scale experiment and a catchment-scale experiment. It’s not clear, however, how the two are used, or which is experiment is being referred to in the results sections (5.1.2 and 5.1.3). For the point-scale experiment, where are the sample sites located? Which experiments supply the results for Table 5 and Figure 5? Are both experiments necessary?
Line 377: Day numbers are hard to interpret. It would be helpful to add the corresponding calendar dates, 172 = June 21 and 264 = Sep 21. Or just give the dates instead.
Line 385: Incorrect reference.
Line 478: Both crown closure and LAI are used interchangeably to describe stand density. Terminology needs to be clarified and unified across the text.
Line 503: This conclusion is not obvious from the results.
Lines 515-516: This conclusion if also not obvious from the available figures.
Line 533: I don’t think the figures support this conclusion that clearly.
Tables
Table 2: Do the LAI values in this table reflect the spatial variation in crown closure? Are LAI and crown closure correlated (see earlier comment)?
Table 4: Indicate in the table header weather the variable is a mean (Winter, Summer) or a maximum (Spring).
Table 5: Which experiment are these results from?
Figures
Figure 5: Which experiment (point or catchment) do these results derive from? The categories (y-axis) on the panels take a while to interpret as they change between columns and are not explicitly identified. Also note that the ‘Disturbance effect” and “Climate effect” are not strictly independent (disturbance effect is conditional on climate state and climate effect is conditional on disturbance state). Do the bars show the mean or the median (assuming catchment scale results) or single sites (point scale results).
Figure 10: The categories (y-axis) on the panels take a while to interpret as they change between columns and are not explicitly identified. Also note that the ‘Disturbance effect” and “Climate effect” are not strictly independent (disturbance effect is conditional on climate state and climate effect is conditional on disturbance state).
Figure 12: Decimals missing on secondary y-axis labels. Recommend showing the summer period on the graph (i.e. as background shading).
Citation: https://doi.org/10.5194/hess-2024-361-RC3 - Phase 1:
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 186 | 33 | 9 | 228 | 19 | 13 | 17 |
- HTML: 186
- PDF: 33
- XML: 9
- Total: 228
- Supplement: 19
- BibTeX: 13
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1