Assessing runoff sensitivity of North American Prairie Pothole Region basins to wetland drainage using a basin classification–based virtual modeling approach
- 1Environment and Climate Change Canada, Saskatoon, Saskatchewan, Canada
- 2Centre for Hydrology, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
- 3School of Environment and Sustainability, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
- 4Natural Resources Canada, Ottawa, Ontario, Canada
- 1Environment and Climate Change Canada, Saskatoon, Saskatchewan, Canada
- 2Centre for Hydrology, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
- 3School of Environment and Sustainability, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
- 4Natural Resources Canada, Ottawa, Ontario, Canada
Abstract. Wetland drainage has been pervasive in the North American Prairie Pothole Region. There is strong evidence that this drainage increases hydrological connectivity of previously isolated wetlands and, in turn, streamflow response to precipitation. It can be hard to disentangle the role of climate from the influence of wetland drainage in observed streamflow records. In this study, a basin classification-based virtual modelling approach is described that can isolate these effects on runoff regimes. Three knowledge gaps were addressed. First, it was determined that the spatial pattern in which wetlands are drained has little influence on how much the runoff regime was altered. Second, no threshold could be identified below which wetland drainage has no effect on the streamflow regime, with drainage thresholds as low as 10 % by area were evaluated. Third, wetter regions were less sensitive to drainage as they tend to be better hydrologically connected even in the absence of drainage. Low flows were the least affected by drainage. During extremely wet years, runoff depths could double as the result of complete wetland removal. Simulated median annual runoff depths were the most responsive, potentially tripling under typical conditions with the high rates of wetland drainage. As storage capacity is removed from the landscape through wetland drainage, the size of the storage deficit of median years begins to decrease and to converge on those of the extreme wet years. Model simulations of flood frequency suggest that because of these changes in antecedent conditions, precipitation that once could generate a median event with wetland drainage can generate what would have been a maximum event without wetland drainage. The advantage of the basin classification-based virtual modelling approach employed here is that it simulated a long period that included a wide variety of precipitation and antecedent storage conditions across a diversity of wetland complexes. This has allowed seemingly disparate results of past research to be put into context and finds that conflicting results are often only because of differences in spatial scale and temporal scope of investigation. A conceptual framework is provided that shows, in general, how annual runoff in different climatic and drainage situations will likely respond to wetland drainage in the Prairie Pothole Region.
Christopher Spence et al.
Status: open (until 13 Jul 2022)
-
RC1: 'Comment on hess-2022-102', Anonymous Referee #1, 10 Jun 2022
reply
Summary Comment: The topic of the paper is important and timely, however, the methods were extremely hard to follow. There was a lot of important information and clarifying details that were missing from the Methods section. The model calibration and evaluation was weak, and a discussion of the sources of uncertainty in the approach needs to be added.
Abstract – please add the spatial extent of the study
Line 22 – change “were” to “being” evaluated, or revise.
Line 153 – Since Spence et al. 2022 is still in review, please clarify how this effort is distinct from this prior effort. Is the application of the model to a new catchment class the only difference?
Lines 171-172 – is the data that is shown in Figure 1 an output from Spence et al 2022 or from Wolfe et al 2019 or another source? It is hard to tell.
Figure 1 – Indicate or clarify what the actual extent of the model and simulated data is? Or clarify in the text that the theoretical model was forced with 4 different climate datasets. If this was the case, what was the size of the simulated basin?
Lines 183-185 – add some basic information on the land cover, slope, elevation, and soil type, what is the spatial resolution and source of these datasets? Was land cover assumed to be stable or non-changing over the modeling effort, other than changing drained wetlands to agriculture?
Line 193 – clarify what size HRUs are used.
Lines 197-201 – Why not use actual wetland datasets instead of artificially created ones? There are some existing efforts by Amani et al. and Mahdianpari et al. for instance.
Line 202 – What DEM was used to route runoff? Were any manipulations/changes made to the DEM to condition it hydrologically?
Lines 207 – Please just list the actual source of the wetland extent data, instead of pointing to another publication, which did not generate a wetland dataset from what I can tell.
Lines 211-216 – So no actual data on drainage was used? How did the simulated drainage account for drainage already present? Or was this assumed to not be relevant since the wetlands were simulated as well? It was also difficult to tell what the term “drainage” was referring to. In the U.S. PPR, drainage is usually installed under the ag fields, although historically ditches were also used to drain wetlands. Was drainage simulated by “removing” wetlands or just increasing the rate at which water left the wetland. This is important to clarify.
Table 1 – A lot of these parameters are not explained. The routing length is the length from where to where? Is this the average length? The LAI values seem weirdly low…why are the 0.001 in a grassland, cultivated field and shrubland? And the woodland also seems low, a LAI of ~1.5-3 would make more sense here. Look at an example paper to more appropriately parameterize these such as Asner et al. 2003, Global synthesis of leaf area index observations: implications for ecological and remote sensing studies: Global leaf area index. Depending on how the model is set up, this could influence the model findings.
Line 230 – what is the spatial resolution of the precipitation data, is it station data given that it is collected only at 4 locations?
Line 252 – if 1965-2006 was used to assess model behavior, what years were used to train and calibrate the models? And if actual discharge was used, how did you guys account for the influence of existing land use and wetlands on the discharge values?
Table 3 – where are these gages? Can these be added to figure 1?
Line 279-280 – where is the St. Denis NWA in relation to the study area?
Lines – 286 – I still can’t quite tell where you guys ran this simulation – across the entire pothole till? just within or near the catchments in table 3? So I can’t tell what the distance is between these ponds and the modeled area, but since the climate stations used are spread across Canada, I have to assume that there are thousands of miles between some of the simulated areas and the ponds, consequently the pond depths are only compared to 1 climate forcing, but this doesn’t seem adequate given that the focus of the paper is on changes in simulated wetland extent and corresponding changes in discharge.
Section 2.6 – please clarify how the 4 scenarios differ….so bottom-to-top for example…wetlands were drained sequentially from the basin outlet toward the headwater streams? Was any attention paid to whether the wetland was near- or connected to a stream, or geographically isolated from the stream network in the scenarios?
Line 328 – add a comma between mean and minimum
Line 383 – what was the distance between the N. Battleford climate and the ponds?
Lines 395-399 – what figure are these results reflecting?
Figure 3- Did the 4th location not have gage data? And since the contributing areas were of different sizes, would it make more sense to normalize the x-axis, so annual discharge per contributing area? Otherwise the Brandon comparisons really look all over the place. Also please add these value in as a table so there is some quantitative way to compare.
Figure 7 – make the font size larger, it was difficult to read this figure.
Discussion – the discussion does a nice job of contextualizing the results with the findings of others but please add a paragraph discussing the limitations of the modeling approach and sources of uncertainty given the input datasets, and limited manner that the model performance was evaluated.
Comment – please revisit how the term “drainage” is used throughout the paper (e.g. line 659) since the term drainage typically refers to the movement of water through a watershed, but here it is mostly used to refer to the action of draining wetlands.
-
RC2: 'Comment on hess-2022-102', Anonymous Referee #2, 22 Jun 2022
reply
Overall, the subject matter, methodological approach, and results of this paper address relevant scientific questions within the scope of this journal. The authors present an interesting modeling application towards better understanding complex landscape hydrological processes.
It is unclear the actual spatial extent of the modeling study. I suggest including more maps in text and in supplemental material to orient the reader through the complex methods. It is clear the Cold Regions Hydrological Modelling platform (CRHM) has been developed and published in many different publications, but it would be helpful to include a spatial visualization of routing pathways of surface water, spatial orientation, and size of wetlands relative to stream networks and how these watershed characteristics change under different drainage scenarios. These details are currently lacking or hard to find, despite the major conclusions of the paper directly relating to these aspects.
Another point of clarity is to standardize use of stream discharge, streamflow depth, and runoff. These different terms are used throughout interchangeably and makes the findings hard to follow.
Overall, interesting conclusions are made but more effort could be made to distill the major findings and move some method and results information into supplemental material.
Below are specific editorial suggestions regarding text, tables, and figures.
L63 this half of the sentence does not say anything. Consider removing or re-writing
L71 replace “numerous depressions” with an order of magnitude (i.e., millions) estimate of number of basins
L80-109 this paragraph can be distilled and shortened to focus on how this information describes the system and problems
L110 This sentence is confusing and repetitive, consider re-writing
L143 These objectives are a perfect spot to set the tone on standardizing language in regards to stream response variables of interest (streamflow, runoff, discharge) and then keep consistent after that
L195-210 this section would be much easier to follow and would allow for better interpretation of results if there were at least one map showing the distribution of wetlands and runoff preferential flowpaths under different drainage scenarios. I find it hard to visualize what this looks like in virtual model space.
L293 More details about the drainage scenarios are needed to better understand what is being manipulated in the model relative to the hydrologic responses.
L307-315 This text would be easier to understand with a visual figure as well. Could be in the supplemental material
Fig 1. The weather station sites are very hard to see. Make bigger and more contrasting to the background
Fig 2. Include slope and intercept of regression model in caption. Also, make sure it is noted why the simulated depression storage is an order of magnitude larger than the observed pond depth. Consider changing units on one y-axis so you are comparing mm to mm or cm to cm
Fig 3. This figure does not say that much and could be moved to the supplemental material
Table 4. in L416-418 you show that there is low deviation between drainage scenarios. To simplify this information consider moving this whole table to the supplemental material and only present the average and sd for each site
Figure 4. similar to Table 4 suggested edits. Take average of all 4 scenarios and only visualize that in the main text. Move whole figure to supplemental. This figure is hard to pull details out of. Also, in the caption include the time period that is modeled and the units of discharge depth. Discharge depth is mentioned in the caption, but the figure shows Annual streamflow (mm).
Figure 5. add “median” “wet” and “dry” labels above the top panel next to each corresponding black vertical line.
Table 5. This information is confusing. Maybe because I do not think about 1:42 as a flood size very often.
Figure 6. This could be consolidated into one panel with the bottom panel’s y-axis range. Right now the y-axis labels seem to be wrong and missing median and max labels.
Figure 8. Different y-axis labels. Standardize runoff or streamflow and be consistent
L548 needs a new section title since this analysis and Fig. 9 talk about a different topic.
Christopher Spence et al.
Christopher Spence et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
275 | 69 | 10 | 354 | 4 | 4 |
- HTML: 275
- PDF: 69
- XML: 10
- Total: 354
- BibTeX: 4
- EndNote: 4
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1