A case study of impacts of wetland area on ood index characteristics and channel morphology in southern Quebec (Canada)

Two characteristics (magnitude and duration-frequency) of two indices of high (Qmax, annual ooding) and low intensities (Q90) oods and the morphology (bankfull width and sinuosity) of channels were compared for the Matawin (1,390 km²) and Petite Nation (1,330 km²) rivers on the Canadian Shield, which are differentiated mainly by wetland and forest areas. Wetlands cover 9% and 15% respectively in the Matawin and Petite Nation watersheds. This comparison revealed that the magnitude of high oods (Qmax) in the Matawin River was on average about twice the magnitude of high oods in the Petite Nation River from 1945–2019. No signicant difference was observed in the magnitude of low oods (Q90) between the two rivers. As for the duration-frequency of heavy oods (Qmax), it was, on average, about three times higher in the Petite Nation River watershed than in the Matawin River watershed. The opposite is true for low oods (Q90). Morphologically, this difference in the magnitude and duration-frequency of heavy oods is seen in the bankfull width and sinuosity between the channels of the two rivers. The Matawin River channel is narrower and more sinuous than the Petite Nation River channel. This study is the rst to demonstrate the impact of wetlands on channel morphology for Canadian Shield rivers in southern Quebec.


Introduction
Flooding and channel morphology are the result of many factors. Land use is considered one of the major factors in ood genesis and, consequently, in the morphological evolution of river channels. Much work has already been devoted to the impacts of land use on river oods (e.g., Acreman  urbanization and agriculture. No study has yet examined the impacts of wetlands on channel morphological evolution. However, much work has already been devoted to the hydrological impacts of these wetlands. In their almost exhaustive summaries, Bullock and Acreman (2003) as well as Acreman and Holden (2013) have shown that these hydrological impacts induced by wetlands can vary from one watershed to another depending on several following factors: landscape location and con guration, soil characteristics, topography, soil moisture status, etc. Morphological, these conclusions suggest that wetlands can induce different morphological impacts. It therefore becomes extremely important to determine the impacts induced by wetlands on the morphological evolution of the river channels according to the hydrological changes they induce. This problem has never been analyzed in the scienti c literature to our knowledge, particularly in Southern Quebec. It is therefore imperative to ll this gap with regard to Quebec, because for two last decades, like many regions in the world, this province of Canada has been facing several catastrophic oods which have causes extensive and expensive material damages. These oods are obviously attributed to climate change. One of the solution mentioned in public opinion is the restoration of wetlands, the area of which has signi cantly decreased over time in Quebec due to the intensive development of agriculture and urbanization. To validate this thesis, it is therefore important to determine the impacts of these wetlands on the high and low intensities ood ows and, by incidence, on the morphological evolution of the river channels.
In Quebec, several studies have already been done on the impacts of agriculture on river ows (e.g., Assani  Quinton and Roulet, 1990aRoulet, , 1990bRoulet, , 1991Roulet and Woo, 1986). However, no study yet exists that examines how these wetlands impact channel morphology in relation to ood characteristics. To ll this gap, this study will compare the annual ood characteristics and morphology (bankfull width and sinuosity) of two watersheds, which differ mainly in wetland area, in southern Quebec. The purpose of this study is to determine whether wetlands can mitigate or amplify the intensity and/or duration of high and low ood ows and their impacts on channel morphology.

Selection of watersheds studied and description
We compared the Petite Nation River to the Matawin River for the following reasons, to analyze the in uence of wetlands on ood indices and the morphology of the Petite Nation River channel: The watersheds of the two rivers have very similar physiographic and hydroclimatic characteristics due to their spatial proximity. In fact, both watersheds lie entirely on the Canadian Shield.
Despite this similarity in physiographic and hydroclimatic characteristics, the two watersheds differ mainly in forest and wetland area (swamps, marshes, bogs, lakes, etc.). This makes it possible to determine the in uence of wetlands on the hydrology and morphology of the Petite Nation River channel, using the Matawin River as a reference.
In both watersheds, ows have been measured continuously for more than 80 years. This makes it possible to compare their hydrological behaviour in wet and dry periods.
Finally, the Matawin River watershed, which serves as a reference watershed, is also one of the reference watersheds for monitoring the impact of climate change on river ows across Canada.
The physiographic and climatic characteristics of both watersheds are presented in Table 1 and their locations are indicated in Figure 1. The data on physiographic characteristics (area, average slope, forest and wetland percentages) and climatic characteristics (total precipitation, average annual temperatures) presented in Table 1 were obtained from Belzile et al. (1997) and the Organisme du bassins versants des rivières Rouge, Petite Nation et Saumon's report. Climate data were obtained from the Environment Canada website (https://climat.meteo.gc.ca/climate_normals/index_f.html, accessed 2020-05-09). ush with the surface, they are covered by sur cial deposits of glacio uvial and maritime origin (gravel, sand, till, etc.). The morphology of the channels of these two rivers is characterized by a quasi-regular succession of large, sinuous reaches with low slopes in sur cial deposits and narrow, substraight reaches with high slopes in rocky outcrop areas. The forest massif consists mainly of sugar maple (Acer saccharum) (sugar maple-yellow birch forest and basswood-sugar maple forest). The climate is temperate-continental, characterized by warm summers and cold winters. The Petite Nation River ows from north to south into the Ottawa River. In contrast, the Matawin River ows from west to east and ows, as the main tributary, into the Saint-Maurice River. The Mont-Tremblant massif separates their watersheds ( Figure 1). The two watersheds are very sparsely inhabited upstream of ow gauging stations. There are no agricultural activities. The only human activity is limited to logging, the main activity in both watersheds. Daily ow data were extracted from the website of the Ministère d'Environnement et de Lutte contre les changements climatiques' Centre d'expertise hydrique du Québec (https://www.cehq.gouv.qc.ca/, accessed on 2020-02-2020). It is important to note that the absence of human activities in both watersheds did not change runoff conditions over time. Table 1 shows that forest area is smaller in the Petite Nation River watershed (83%) than in the Matawin River watershed (90%) due to logging. In contrast, wetland areas (lakes, marshes, swamps, bogs, etc.) are larger in the Petite Nation River watershed (15%) than in the Matawin River watershed (9%). Ultimately, the two watersheds constitute life-sized experimental watersheds to differentiate the respective in uence of deforestation and wetlands on ood index characteristics and channel morphology in southern Quebec.

De nition of ood index characteristics and statistical data analysis
Morphological evolution depends mainly on ood evolution. To determine the in uence of wetland area on channel morphology in both rivers, the ood characteristics of magnitude and duration were compared. In terms of magnitude, four hydrological series of two ood indices were de ned. The rst hydrological series consisted of maximum ow values (Qmax) measured each year from 1945-2019. This series represents heavy (high) oods, that is, oods whose ows reached or exceeded annual ooding. The last hydrological series consisted of ow values corresponding to the 90th percentile (Q90) calculated each year based on daily ows over the same period (1945-2019). These ood ows were below the minimum ow of annual ooding. They included all medium and low oods whose peaks did not reach the lowest annual ood (annual daily maximum ow, Qmax) observed since 1945. From a morphological point of view, these two ood indices include all ood ows of low to high intensity that can cause morphological changes (bank and bed erosion, sediment transport, etc.). In each of these two series, the mean daily ow values (average magnitude), as well as the highest (maximum magnitude) and lowest (minimum magnitude) daily ow values were calculated. In terms of ood duration-frequency, the minimum value (the lowest ow) for each of the two indices (Qmax and Q90) were rst considered. The total number of days during which this minimum value ow (minimum threshold) was reached or exceeded in a year during the period from 1945-2019 was then calculated. Two other hydrological series of daily ow duration-frequency (in days) were also established, each corresponding to a ow magnitude series: DQmax and DQ90. For these hydrological ow duration-frequency series, the mean ow duration value, as well as the highest (maximum duration-frequency) and lowest (minimum duration-frequency) mean values were calculated. It is important to note that the duration-frequency of Q90 was calculated by subtracting the total number of days during which the lowest (minimum) ow corresponding to the 90th percentile ow series (Q90) during the period 1945-2019 was reached or exceeded from the total number of days that the lowest (minimum) ow from the annual ooding series (Qmax) during the same period was reached or exceeded.
Statistically, the means of these four hydrological series of both rivers were compared using statistical parametric (one-way ANOVA) and non-parametric (Kruskal-Wallis) tests. Because of the small difference in the size of the two watersheds, ows measured in m³/s were converted to speci c ows (ratio of ows to watershed area multiplied by 1,000) expressed in L/s/km² in order to compare ood index ow values.

Channel morphology and sedimentology analysis
To analyze the in uence of wetlands on the morphology of the two channels, bankfull width and sinuosity were measured. The method used is described in detail in our previous work (e.g., Aubry et al., 2013;Vadnais et al., 2012). The rst step of the mapping analysis was to map the banks for the two rivers under study over a maximum distance of 10 km (as the crow ies from the ow gauging station).
This distance was largely su cient to compare the mean bankfull width and sinuosity of the two rivers, since the surface area of the two watersheds upstream of the gauging stations was almost identical. To map the banks, aerial photographs from the Inventaire écoforestier du Québec méridional (produced between 2002 and 2017) were used with ArcGIS software. These images were provided by the Ministère des Forêts, de la Faune et des Parcs (MFFP). The resolution is 30 cm and planimetric accuracy is 2 m.
Classes of linear entities (polylines) were created using photointerpretation to accurately map the upper limit of the channels where there is a break in the vegetation and the mark left by the highest natural waters on each side of the shoreline (bankfull).
After vectorizing the shorelines, a central centreline corresponding to the watercourse channel was generated using the "TIN" function in Esri's ArcGIS software. This function made it possible to create a linear interpolation between the channel boundary vectors, assigning an equal distance value to them. In our case, values of 1 and 3 were assigned to each of the banks (left and right) and subsequently, a line was produced at the median (value of 2) of the interpolation between the banks using the "Contour" function. After the aerial photographs were processed, the river banks were digitized (at bankfull level as de ned by the vegetation edge) using the ArcMap tools from the ArcGIS software. To limit error on the bank outline, a three-dimensional digital stereoscope was used, as this allows for three-dimensional viewing of the channel. Channel width was measured by automatically drawing straight lines perpendicular to the channel using an ArcGIS extension developed at the Laboratoire Interdisciplinaire d'application en géomatique environnementale (LIAGE) and integrated into ArcGIS. This automated approach eliminated all human error associated with the perpendicular lines and thus with measuring the bankfull width of the channel. The software also allowed for automated drawing of perpendicular lines at the same location between two points separated by the same distance on aerial photos taken at different times, something that is very di cult to do manually. Once the banks were delineated, the perpendicular lines were drawn by rst nding the midpoint of each cross-section (between the two banks), then drawing the centerline joining all midpoints, and nally drawing lines perpendicular to this centreline at all midpoints. Over 250 measurements (according to the total 10-km length of the analyzed section) of the channel bankfull width were thus done with the software at spacing of approximately 50 m. Due to differences in sinuosity between the two rivers, the number of bankfull width measurements was higher for the Matawin River (more sinuous channel) than for the Petite Nation River. Ground observations were used to validate the channel shore delineation done with aerial photography. The channel shore limits used to calculate the channel bankfull width were de ned fairly accurately with this technique. Several eld visits were done to test the validity of bank delineation from aerial photographs against eld measurements. To determine the maximum error on bankfull width measurements from photo interpretation, air photo interpretation (measurement of width from aerial photos) was carried out by two independent operators. The maximum difference in bankfull mean width at a given station between the values obtained by the two operators was < 1 m. This maximum difference was therefore deemed to represent the maximum error on channel width measurements from photo interpretation at a given station. Mean bankfull width values for the two rivers were compared using the Kruskal-Wallis nonparametric and ANOVA parametric tests. Similar to ows, because of the difference in the size of two watersheds, the bankfull width measured in meters for the channels was changed to a speci c width (ratio between the bankfull width and the watershed area) at bankfull expressed in mm/km². This change is explained by the fact that the width of a river channel increases proportionally to the surface area of its watershed, as demonstrated by numerous studies on hydraulic geometry theory (e.g., Ferguson, 1986). To calculate sinuosity, we calculated the ratio of the real total channel length over a distance of 10 km (including meanders) to the length of the channel as the crow ies over the same distance of 10 km (ignoring meanders).
With respect to sedimentological data, for the Matawin River, 15 sediment samples were collected from two banks (seven on the left bank and eight on the right) in August (low-water event) in 2018 in the section whose morphology was analyzed. At each sampling site, a sediment sample (approximately 200 g) was taken near the foot and top of the bank. The same approach was used for the Petite Nation River, where samples were taken from ten sites ( ve on the left bank and ve on the right). The average distance between two sampling sites was approximately 2 km for the Matawin River and 1 km for the Petite Nation River. The sediment collected in the eld was oven-dried at 105°C for two days, then sieved in the lab in a bank of wire mesh sieves. After sieving, sediments were divided into three granulometric classes: sand, silt and clay.

Comparison of magnitude and duration-frequency of ood indices between the two rivers
Flow magnitude values for the two ood indices calculated in the two watersheds are indicated in Table   2. Figure 2 compares the interannual variability of mean annual maximum daily ows (Qmax) measured in the two watersheds. This table shows that the magnitude of annual ooding ows (Qmax) for the Matawin River is higher than for the Petite Nation River. The magnitude is on average about two times higher in the Matawin River watershed than in the Petite Nation River watershed. For the Q90 ood index, the ow magnitude means for the two rivers were not signi cantly different ( gure 3). It is nevertheless important to note that the maximum and minimum values for this index are higher in the Matawin River watershed than in the Petite Nation River watershed. In terms of the frequency-duration ood index, unlike ow magnitude, annual ooding ows (Qmax) were, on average, about three times shorter in the Matawin River watershed (26 days) than in the Petite Nation River watershed (76 days) per year ( gure 4). This trend was completely reversed for the durationfrequency of the Q90 index (low and medium oods). For the Q90 index, ow duration-frequency was greater in the Matawin watershed (55 days) than in the Petite Nation watershed (15 days). Figure 5 shows a concrete example of the very typical variation in daily ows (expressed in l/s/km²) during the 2018. It clearly shows that the Petite Nation River daily ows hydrograph is atter than that of the Matawin River daily ows hydrograph. In addition, it also shows that the difference in the ows magnitude and duration between the two rivers are higher for the freshet (Qmax) in spring than those for rain-induced oods (Q90) in summer and fall. As such, during freshets, ows increase and decrease more slowly in the Petite Nation River than in the Matawin River, which explains why freshets in the rst river are smaller in magnitude but higher in duration than in the second. It is important to note that freshet (Qmax) have a much greater morphological impact than those of summer-fall oods (Q90).

Comparison of the characteristics of bankfull width and sinuosity of the two rivers
Granulometric analysis of bank sediment are presented in Table 3. Both river banks (excluding rocky or pebble outcrop sections) consist mainly of ne sand. This clearly demonstrates that the lithological characteristics are identical in both watersheds given that they belong to the same geological formation (Canadian Shield). Bankfull width and sinuosity values for the two rivers are presented in Table 4 and gure 6. The mean bankfull width of the Petite Nation River was about twice the mean bankfull width and gure of the Matawin River, despite the fact that the area of the Matawin River watershed at the ow gauging station was slightly greater than the watershed area of the Petite Nation River. The Petite Nation River bankfull width (Wmax) highest value was still about four times higher than the Matawin River bankfull width value. However, this difference was relatively smaller for the minimum bankfull width value. The Matawin River channel is much more sinuous than the Petite Nation River channel. It follows that, morphologically, the Matawin River channel is narrower and more sinuous than the Petite Nation River, despite the fact that the Matawin River watershed is a little larger than the Petite Nation River watershed.

Discussion
The two watersheds that were analyzed have very similar physiographic and climatic characteristics.
This similarity helped determine how wetlands in uence ood indices and the Petite Nation River channel morphology. In fact, the primary difference between the two watersheds is wetland and forest area. The area of wetlands is twice as large in the Matawin River watershed (15%) as in the Petite Nation River watershed (9%). Comparing the ow magnitude for the two ood indices (Qmax and Q90) for both rivers revealed a signi cant difference in the annual ooding ow magnitude (Qmax) between the two rivers.
The magnitude of heavy oods (Qmax) was on average about two times higher in the Matawin River watershed than in the Petite Nation River watershed. The means of the ow magnitude for low and medium oods (Q90) of both rivers was not signi cantly different. As for the duration-frequency of ows, heavy oods, associated with annual ooding (Qmax), lasted less time (about three times less on average) in the Matawin River watershed than in the Petite Nation watershed. In contrast, low and medium intensity (magnitude) oods lasted less time in the Petite Nation watershed than in the Matawin River watershed. From a hydrological point of view, heavy oods associated with annual ooding (Qmax) are generated exclusively by spring snowmelt in both watersheds. In contrast, low and medium intensity oods, associated with the Q90 index, are mainly generated by rain in summer and autumn. However, they also include ows from snowmelt but lower (< Qmax) than the discharge from the annual ood. A recent study of the frequency of these rainfall oods found that the frequency was much lower in the Petite Nation River watershed than the Matawin River watershed due to the "sponge effect" of wetlands in the Petite Nation watershed (Assani, 2020, unpublished work). This sponge effect means that the wetlands absorb more runoff, preventing it from transferring directly to the channel (e.g., Acreman & Holden, 2003;Bullock & Acreman, 2003). This results in a more signi cant decrease in the occurrence of low and medium summer and fall oods in the Petite Nation River watershed than in the Matawin River watershed. The sponge effect of the wetlands also explains the signi cant decrease in the ow magnitude of annual ooding (Qmax) in the Petite Nation River watershed. These wetlands temporarily store snowmelt, reducing the intensity of spring ood peaks. Water that is temporarily stored in these wetlands is released into the river gradually, increasing the duration of the annual ooding (see gure 6).
Finally, the difference in forest cover between the two watersheds does not seem to in uence either the magnitude or the duration-frequency of ooding. The decrease in forest cover in the Petite Nation River watershed is expected to result in an increase in ow magnitude (not a decrease) for the annual ooding as observed in the agricultural watershed of the L'Assomption River, adjacent to these two watersheds (Sylvain et al., 2015), due to increased surface runoff. Nevertheless, the L'Assomption River's agricultural watershed features impermeable soil that promotes runoff, unlike in the Petite Nation River watershed.
When the snow melts in spring, the decrease in forest cover accelerates the snowmelt, which leads to a rapid increase in ows but a decrease in their duration. This hydrological behaviour is completely different from that of the Petite Nation River watershed, which has a smaller forest area (83%) than the Matawin River watershed (90%).
Morphologically, the longer duration-frequency of heavy ood ows (Qmax) in the Petite Nation River watershed led to a relatively large widening of the river channel, despite the fact that the magnitude was much smaller than the magnitude observed in the Matawin River. As a result, the mean bankfull width of the Petite Nation River has become at least twice the Matawin River's mean bankfull width. The longer duration of heavy ood ows favours the intersection of meanders, reducing channel sinuosity. The Matawin River channel is therefore narrower and less sinuous than the Petite Nation River channel. The lithological context is the same in the two watersheds, which are located on the Canadian Shield, so this morphological difference cannot be explained by sedimentological and geological factors.
In addition to these freshet ow characteristics, the role of the amount of sediment (suspended load carried by both rivers) must not be ignored. Temporary storage of runoff in wetlands also promotes the sedimentation of suspended particles in runoff. This signi cantly reduces the amount of sediment carried by runoff to the channels. A small suspended load promotes shoreline erosion (channel widening) as has been shown in watersheds whose suspended loads have gone down signi cantly since reforestation (e.g., Assani et al., 2003; overtook agricultural activities in watersheds, which is consistent with Schumm's formula (1969).

Conclusion
Land use plays a crucial role in ood dynamics and changes in channel morphology. Comparing the impacts of different land use between the Petite Nation and Matawin watersheds made it possible to determine the hydromorphological impacts induced by the difference in wetland area. In the Petite Nation River watershed, which has a larger wetland area (15%) than the Matawin River watershed (9%), the magnitude of heavy ows associated with the annual ooding (Qmax) was almost two times lower than the magnitude observed in the Matawin River watershed. In contrast, the duration-frequency of these ows was on average about three times higher than the duration-frequency of Matawin River ows. The decrease in ow magnitude associated with an increase in duration-frequency was the result of the wetlands' sponge effect on the runoff and in ltration process of the hydrological cycle. Morphologically, the hydrological behaviour induced by wetlands resulted in a widening and decreased sinuosity of the Petite Nation River. The difference in forest area between the two watersheds had no noticeable impact on the characteristics of heavy ood ows and channel morphology, highlighting the predominant role of wetlands in the hydromorphological evolution of Canadian Shield watercourses in Quebec. In terms of ooding, this role translates as a signi cant reduction in intensity, the environmental effects of which clearly appear to be hindered by their relatively long duration.

Declarations
Con ict of Interest: No con cts of interest are declared.
Ethics Approval: Not applicable.
Consent to Participate: Not applicable.

Consent for Publication: Not applicable
Data availability: The author may provide the data used in the manuscript upon request.    Example of daily speci c ows hydrographs for two rivers in 2018: Matawin (red curve) and Petite Nation (blue curve) Rivers. This gure clearly shows that the hydrographs of the Petite Nation River is more attened than that of the Matawin River. Figure 6