The new ‘surface storage’ concept versus the old ‘sponge effect’ concept: application to the analysis of the spatio-temporal variability of the annual daily maximum flow characteristics in southern Quebec (Canada)

It has become increasingly clear that global warming will amplify the intensity, duration and frequency of flooding in many regions around the world. In their study based on satellite image analysis, Tellman et al. (2021) estimated that from 2000 to 2015, the population exposed to flooding rose from 58 million to 86 million – a 35% increase in 15 years. One of the recommended means of mitigating the intensity and frequency of these floods is the conservation of wetlands (Maltby 1991). Thus, certain organizations such as The International Union for Conservation of Nature (IUCN), Wetlands International and the Ramsar Convention on Wetlands of International Importance have promoted this conservation of wetlands as an effective means of reducing the intensity and frequency of floods (Acreman & Holden 2013). This reduction effect is perfectly in line with the concept of the ‘sponge effect’ (relatively high water infiltration or absorption reducing surface runoff) exerted by the wetlands on the process of water runoff. In fact, according to this concept, wetlands ‘store water during wet periods and release it during dry periods’ (Acreman & Holden 2013).

However, over time, it has become apparent that this ‘water storage’ function is not exercised by all wetlands in time and space due to the influence linked to the diversity of their intrinsic and extrinsic characteristics on runoff and infiltration processes. In their almost exhaustive synthesis, Bullock & Acreman (2003) thus revealed that about 80% of studies observed a reduction in the magnitude of floods by the floodplain wetlands (FW), while 41% observed, on the other hand, their increase by headwater wetlands. As for the slope non-floodplain wetlands (NFW), 42% of the studies listed mentioned a reduction in the magnitude of floods, while 44% mentioned, on the other hand, an increase in this magnitude (see the synthesis by Lane et al. (2018)). It follows that the concept of ‘sponge effect’ cannot be generalized for all wetlands to account for their impacts on flood magnitude due to their contradictory effects (Acreman & Holden 2013). On the hydrological level, this contradiction translates in fact a difference of the hydrological processes which generate the flows. Not all wetlands generate the same hydrological process.

Furthermore, within a watershed, particularly for relatively large watersheds, there are other topographic components of the landscape (lakes, depressions, natural and artificial ponds, natural and artificial ditches, etc.) which can amplify or attenuate the effects of wetlands on runoff and infiltration processes thus affecting the flood magnitude. The last thus results from the interaction of all these topographic components of the landscape present in these catchment areas. This interaction is not often taken into account in all studies devoted to the impacts of wetlands on floods because these impacts are analyzed separately from those induced by other topographical components or land types (lakes, depressions, reservoirs, etc.) and factors (agriculture, deforestation, urbanization, etc.), even by hydrological models.

To overcome some of these weaknesses of the sponge effect concept, other authors have introduced a new concept: surface water storage concept to explain the temporal (e.g., Quin & Destouni 2018; Rajib et al. 2020) or spatial (e.g., Assani 2022) variability of flood magnitude. The hydrological process on which the new concept is based consists of the significant storage of runoff water on the surface, thus slowing the flow toward the river channels and significantly reducing infiltration (Assani 2022). Unlike the sponge effect concept, which applies only to wetlands, the new ‘surface water storage’ concept takes into account the interaction of all topographic components (lakes, natural and artificial depressions, natural and artificial ditches, reservoirs, etc.) that can influence, to varying degrees, the process of storing runoff water in a watershed, as several works have already demonstrated (e.g., Rains 2011; Shook & Pomeroy 2011; Shook et al. 2013; Evenson et al. 2018; Quin & Destouni 2018; Rajib et al. 2020). In this regard, Rajib et al. (2020) demonstrated that the integration into hydrological models of all topographic components (wetlands, lakes, reservoirs depressions, etc.) that influence the water storage significantly improved the prediction of flows in the Upper Mississippi River Basin in the United States. This study clearly demonstrates that surface storage by different topographical components of the landscape (land types) plays a very important role in the genesis of flows in general, and that of floods in particular. However, despite the introduction of this new concept, there is still no study in the scientific literature that has already confronted the two concepts to explain the temporal and spatial variability of flows in a given region. This is a gap that needs to be filled in order to better establish the difference between the hydrological processes which underlie the two concepts in the explanation of flood and low water levels spatio-temporal variability from the perspective of managing extreme flows in the current context of warming.

In Quebec, with regard to temporal variability, some work has already been devoted to the hydrological impacts of wetlands on flood and low flow based exclusively on the ‘sponge effect’ concept (Fossey et al. 2015; Fossey & Rousseau 2016a, 2016b; Blanchette et al. 2019). As for the spatial variability of flows, Assani (2022) instead applied the new ‘surface storage’ concept to explain the differences in the flow characteristics observed between four watersheds. But all of this work has not compared the two concepts in order to determine which concept and the underlying hydrological processes best explain the spatial and temporal variability of flow characteristics (magnitude, duration, frequency, timing and variability) in Quebec. The objective of our study is to answer this question. Unlike previous work, which was based on the analysis of flows in a few watersheds, this study will focus on the analysis of a larger number of watersheds representative of the climatic, physiographic and land use and land cover conditions of Southern Quebec. The two concepts comparison will also be based on the results already published on the low water levels spatio-temporal variability (Assani et al. 2021, 2022). The main stage of this comparative analysis will first consist in identifying the annual daily maximum flow characteristics whose spatio-temporal variability is mainly influenced by the wetlands. The final step is to determine which of two concepts (sponge effect or surface storage) best explains their spatio-temporal variability in southern Quebec.

Daily flow data was taken from the website of the Ministère d'Environnement et de la Lutte contre les changements climatiques du Québec's Centre d'expertise hydrique du Québec (https://www.cehq.gouv.qc.ca/index_en.asp, accessed 20 February 2020). Physiographic and land use/cover variables data of watersheds was obtained using methods and techniques already described in detail namely by Belzile et al. (1997), Assani et al. (2021) and Kinnard et al. (2022). Five physiographic and land use/cover variables, which significantly influence runoff and infiltration processes in a watershed, were measured: drainage density (km/km, V1), mean slope (%, V2), forest area (%, V3), agricultural area (%, V4) and wetlands area (%, V5). It is important to note that the data on the wetlands area also include those of other topographic components or water bodies (lakes, depressions, ditches, etc.). It was not possible to separate them. In this context, from a hydrological point of view, a wetland is defined as any landscape unit that temporarily or permanently stores water on surface or contains it in all depth over a hydrological year, excluding the channels (minor beds) of rivers. Surface storage (low permeability) means the absence or weak interaction with the aquifers whereas the containment (relatively high permeability) supposes, on the contrary, a relatively strong interaction with these aquifers.

As for climate data, monthly average temperature and precipitation normals were calculated over the following three periods: 1941–1970, 1971–2000 and 1981–2010. These normals were taken from the website of the federal department of Environment and Climate Change Canada (https://climat.meteo.gc.ca/climate_normals/index_f.html, accessed 18 June 2021). Seven climatic variables were defined from these averages of the climatic normals: annual total rainfall (mm, V6), annual total snowfall (cm, V7), annual total precipitations (mm, V8), winter-spring total rainfall (mm, V9), winter-spring total snowfall (cm, V10), winter-spring total precipitations (mm, V11), annual daily mean maximum temperature (°C, V12) and winter-spring mean maximum temperature (°C, V13). These data are available in Belzile et al. (1997) in particular or by a simple request addressed to the first author.

After estimating the flow (quantiles) associated with different return periods (from Q₂ to Q₁₀₀ years), their frequency (number of years a given recurrence flow was measured) was calculated over the 1930–2019 period. Regarding the duration (in days) of a given recurrence flow, the number of days that this flow was reached or exceeded in a given year over the same period (1930–2019) was determined. Finally, two indices for flow variability were calculated: the coefficient of immoderation (CI), which is the ratio between the highest and lowest value of a series, and the coefficient of variation (CV), a ratio between the standard deviation and the mean of a series shown as a percentage. The first index (CI) measures the maximum difference between the two extreme values (the highest and lowest values) of the annual daily maximum flows series during the period 1930–2019. It thus translates the maximum amplitude of the variability of these flows during the period considered. The second index (CV) measures the interannual variability of these flows during the same period.

The calculated and estimated values of the flow magnitudes were converted into l/s/km² so they could be compared among watersheds with different surface areas. Mean values (Q_m) of the hydrological series were compared using parametric (one-way Analysis of Variance, ANOVA) and non-parametric (Kruskal–Wallis) tests. This method was also used to compare the means of flow durations of different recurrences. Chi-square test was used to compare the flow frequencies of different recurrences. The four characteristics (magnitude, duration, frequency and variability) of annual daily maximum flows were then correlated with the five physiographic and seven climatic variables of the watersheds. Linear correlation coefficients and Sperman's rank coefficients were calculated. Both methods led to the same results. Finally, it is important to note that each series of maximum annual daily flows was fitted separately using the GEV distribution to verify its suitability. This justified using the regional distribution proposed by Anctil et al. (1998) to estimate the flows of different recurrences. The only flow series that did not fit this GEV distribution were those of the Blanche and L'Assomption rivers. As such, they were eliminated at this stage of analysis (spatial variability).

To calculate the long-term trend of the hydrological series, in the first step, the original Mann-Kendall (MK) test (Sneyers 1990) was applied to detect the long-term trend of the hydrological series. This test is widely used in hydroclimatology and described in numerous studies. There is therefore no need to further explain its mathematical description. However, this original test does not account for the effects of short-term (autocorrelation) and long-term (Hurst effect) persistence on the long-term trend of the series. The existence of these two types of persistence (autocorrelation or dependence between the values of the discharges measured from one year to another) can distort the conclusions (existence of a significant trend in a hydrological series whereas it does not exist: rejection of the null hypothesis when it is true) of statistical tests applied to hydrological series, hence the importance of eliminating them. Three types of modifications were proposed to accommodate these persistence effects. To eliminate short-term persistence effects, several authors (von Storch 1995; Hamed & Rao 1998; Yue et al. 2002a, 2002b; Yue & Wang 2004) have proposed various modified MK tests. Some involve eliminating autocorrelation effects through data filtering, including the prewhitening method (MMK-PW), trend-free prewhitening method (TFPWMK), as well as via variance correction such as the modified Mann-Kendall test (MMKY). This variance modification eliminates the significant autocorrelation effects of all orders in the hydrological series. In hydrology, the modified test proposed by Hamed (2008) is widely used to eliminate the effects of long-term persistence on the long-term trend, known as the ‘Hurst effect’ (Koutsoyiannis & Montanari 2007). This is known as the long-term persistence test (MMK-LTP). The mathematical descriptions of all these tests and their application in hydroclimatology are described in detail in Dinpashoh et al. (2014) and Kumar et al. (2009).

As for the flow frequencies of different recurrences of the annual daily maximum flow, they were generally higher on the south shore than on the north shore south of 47°N (Table 4), despite the highest value being observed in a river on the north shore (Du Nord River). However, the application of the Chi-square test did not reveal any significant difference in the frequencies of the flows of different recurrences at the 5% threshold. Lastly, there appears to be no significant difference between the rivers on either shore in terms of the values of the two flow variability indices (CV and CI) even though the highest values were observed on the south shore (Table 5).

This comparison of four characteristics (magnitude, duration, frequency and variability) of the annual daily maximum flows reveals that the north shore rivers are distinguished from those on the south shore by the occurrence of floods of low magnitude (intensity) but of relatively long (duration). However, it should be remembered that the watersheds of the north shore are characterized by larger areas of wetlands (>5%) than those of the south shore (<5%) (see Table 1).

To determine the flow characteristics whose spatial variability is significantly correlated with wetlands, correlations were calculated between these characteristics and the climatic and physiographic factors measured in the 15 watersheds. The results of correlation analysis are presented in Tables 6 and 7.

Table 6

Correlation coefficients calculated between physio-climatic variables and magnitude (l/s/km²) of annual daily maximum specific flows with varying recurrence intervals from 1930 to 2019

*Significant values at the 5% threshold are shown in red bold. Red = positive correlation; Blue = negative correlation.

View Large

Table 7

Correlation coefficients calculated between physio-climatic variables and three characteristics of annual daily maximum flows from 1930 to 2019

*Significant values at the 5% threshold are shown in red bold. Red = positive correlation. V1–V13 = see Table 6.

View Large

With regard to the magnitude of the flows (Table 6), the mean values (Q_m) and those of the quantiles of different return periods are mainly correlated with the surface areas of wetlands. This correlation is significantly negative at the 5% threshold. On the other hand, they are positively correlated to winter-spring snowfall and annual total precipitation (rain and snow). With regard to annual total snow amount, it is positively correlated to the magnitude of the recurrence flows between 5 and 50 years.

As for flow duration, it was also strongly correlated with wetland surface area (Table 7). But this correlation was positive. In addition to wetland surface area, flow duration was also correlated, but negatively, with agricultural surface area and winter-spring snowfall. But unlike the wetland surface area, which is significantly correlated with the durations of all the floods of different return periods, the agricultural surface area is only correlated with the durations of the 2- and 5-year return floods. As for the winter-spring snowfall, it is correlated only with the duration of the 20 and 50-year recurrence floods.

Annual daily maximum flow frequency was negatively correlated with drainage density but positively correlated with annual and winter-spring mean daily maximum temperatures. Finally, both flow variability indices were negatively correlated with the drainage density and mean slope of the watersheds.

The correlation analysis shows that the spatial variability of the two characteristics of the annual daily maximum flows, namely the magnitude and the duration, which differentiate the rivers of two shores, is mainly influenced by the wetland surface area. This is negatively correlated with the magnitude of the flows but positively with their duration. This result, therefore, justifies the confrontation of the concepts of ‘sponge effect’ and ‘surface storage’ to explain the spatial variability of these two characteristics. These two concepts are exclusively associated with the hydrological impacts of wetlands.

Results of the long-term flow trend analysis using five MK tests are presented in Tables 8 and 9. These tests were applied to the series of the magnitude of the annual maximum daily flows (flows measured each year) on the one hand (Table 8), and that of the total number of days of recurrence flows ≥2 years, on the other hand (Table 9). Temporal autocorrelation values (R1) showed that very few of the hydrological series (3) were significantly autocorrelated. Although significant, these autocorrelation coefficients were very low, with the exception of that of the Blanche River.

Table 8

Results of the analysis of the long-term trend in the magnitude of annual maximum daily flows using different Mann-Kendall tests from 1930 to 2019

**Significant values at the 5% threshold are shown in bold. *Significant values at the 10% threshold are shown in bold; Red = positive trend. R1 = autocorrelation coefficient.

View Large

Table 9

Results of the analysis of the long-term trend in the duration (in days) of the annual daily maximum flow recurrences of ≥2 years using different Mann-Kendall tests from 1930 to 2019

**Significant values at the 5% threshold are shown in bold. *Significant values at the 10% threshold are shown in bold; Red = positive trend.

View Large

However, the analysis of the long-term trend also demonstrates a clear difference between the temporal variability of the flow magnitude of the north shore rivers (NS1–NS5) and that of the south shore rivers (SS1–SS12). In fact, all the rivers on the north shore analyzed are characterized by a significant increase in the averages of the flow magnitude over time, whereas only a third of the rivers on the south shore are.

The comparison of annual daily maximum flow characteristics showed a clear difference between the south and north shore rivers. On average, flow magnitude was at least twice as high on the south shore (SS1–SS6) as on the north shore south (NS1–NS5) of 47°N. However, the average duration of these flows was three times lower on the south shore. Annual daily maximum flows were therefore more intense but shorter on the south shore than on the north shore, on average. To explain this difference between these two flow characteristics on each shore, the correlation analysis clearly showed both to be typically correlated with wetland surface area. The magnitude or intensity of flow rates was negatively correlated, while their duration was positively correlated. Surface area data for each watershed showed that wetlands occupied more than 8% of watersheds on the north shore, but less than 4% of those on the south shore.

Like other regions of the world (e.g., Vanderhoof et al. 2016; Ali et al. 2017; Ameli & Creed 2017, 2019; Perez-Valdivia et al. 2017; Evenson et al. 2018; Lee et al. 2018; Bertassello et al. 2019; Wu et al. 2020, 2023), several studies have already analyzed the impacts of wetlands on flood and low water levels in southern Quebec (Fossey & Rousseau 2016a, 2016b; Blanchette et al. 2019). Regarding floods from snowmelt, all these studies, based on the concept of sponge effect, have clearly demonstrated that both types of wetlands (non-floodplain wetlands and floodplain wetlands) cause a decrease in the magnitude of flood flows. The first criticism that can be addressed to these studies is the fact that they analyzed the impacts of these wetlands separately without comparing them with the impacts induced by other factors such as other water bodies (lakes, depressions, etc.), agriculture, deforestation, etc. Thus, these studies did not take into account the possible interactions of these different factors to be able to attribute this decrease in the magnitude of flood flows exclusively to wetlands according to the concept of sponge effect. The second criticism that can be addressed to these studies is the fact that they did not analyze the impacts of these wetlands and other water bodies on the other characteristics of the annual daily maximum flows. Thus, these studies have not demonstrated whether the concept of sponge effect can explain the increase in the duration of these floods as we have just demonstrated. It is important to remember that there are practically no studies on the impacts of wetlands on the duration of floods in particular, as revealed by the synthesis of the hydrological impacts of wetlands on flood flows (Bullock & Acreman 2003; Acreman & Holden 2013; Lane et al. 2018). However, in an experimental in situ study, coupled with hydrological modeling, in a watershed in Houston, Texas (USA), Tang et al. (2020) demonstrated the increase in the wetland surface area in this watershed resulted in a significant decrease in the flood magnitude and duration because, despite the storage of water in these wetlands, which reduced the magnitude, its infiltration into the are reduced the duration of flood flows according to the concept of sponge effect.

It follows that the sponge effect exerted by the wetlands is characterized by a decrease in the flood magnitude and duration due to the fact that part of the water stored by these wetlands infiltrates the soil to feed groundwater, thus causing a significant increase in low water flows. This hydrological process associated with the concept of sponge effect (storage and infiltration) cannot explain the increase in the duration of floods observed in Quebec on the north shore despite the existence of a greater wetland surface area in the watersheds than on the south shore. On the other hand, if we take into account all the other factors (lakes and other water bodies) which can store more water on the surface without favoring its infiltration, we can thus explain the drop in the magnitude of the flows but, on the other hand, the increase in their duration as revealed by the results of our study, thus confirming those obtained by Assani (2022) on the basis of a comparative analysis of four watersheds differing mainly by wetlands and other water bodies in Quebec.

It is this concept of surface water storage by the different topographical components (land types) including wetlands that explains the spatial variability of the magnitude and duration of flood flows generated by snowmelt, the two characteristics of annual daily maximum flows being strongly correlated to wetland surface area. The impacts of this surface storage on the flood characteristics vary according to the seasons. Assani has clearly demonstrated that this surface water storage effect also causes a significant decrease in the magnitude of flood flows generated by rain in summer and autumn (warm season) following the decrease in their frequency. On the other hand, the duration of these floods decreases due to the high evapotranspiration of the water stored on the surface which infiltrates very little into the soil to feed the groundwater. This low infiltration of water stored on the surface results in the absence of impacts of wetlands, including other water bodies, on low water flows in Quebec (Assani et al., 2022). As the evapotranspiration is much lower during spring snowmelt due to low temperatures, the loss of water stored on the surface is less and a large quantity ends up joining the river channels thus causing the magnitude to decrease but the increase in their duration. This transfer of water from wetlands and other water bodies to river channels appears to be relatively slow due to the complexity of the connectivity between these systems (Yu & Harbor 2019).

In addition to wetland surface area, flow magnitude was also positively correlated with the amount of snow. Annually, the overall amount of snow was greater on the south shore (>250 cm) than on the north shore (<250 cm). Although more snow should, in theory, have resulted in an increase in the duration of floods on the south shore compared to those on the north shore, these two variables were negatively correlated, confirming the significant influence of wetlands on the duration of flows. Agricultural area, which was greater on the south shore (>10%) than on the north shore (<5%), also had a certain influence on the duration of flows (negative correlation). The modernization of agricultural practices in southern Quebec since 1950 caused the significant drainage of wetlands, increasing drainage density in agricultural watersheds (e.g., Ruiz 2019). This densification of the drainage system promoted the rapid evacuation of runoff, which helped decrease the duration of floods while increasing the peaks (magnitude) of annual daily maximum flood. A comparison of annual flood hydrographs of the rivers on both shores showed a clear difference in their shape: the flood hydrographs of rivers on the north shore were flatter (low magnitude but long duration of flows) than those of the rivers on the south shore. The flatter shape suggested a slow but sustained transfer of runoff from the slopes to the channels.

Unlike the duration of floods, agricultural area is not significantly correlated with their magnitude (Table 2). This absence of a significant linear link does not necessarily mean that agriculture does not influence the spatial variability of the magnitude of floods in Quebec. As proof: the magnitude of floods is much higher in the most agricultural watersheds of the southeastern hydroclimatic region on the southern shore (SS1–SS6) than in those of the southwestern hydroclimatic region on the northern shore (NS1–NS5) due to higher runoff in the first than the second hydroclimatic region. This constitutes a limitation of the application of the linear correlation in the flood characteristics spatial variability factors analysis. Indeed, the link between certain factors and flood characteristics is not necessarily linear. However, Assani et al. (2021, 2022) demonstrated that agricultural area is the best linearly correlated variable with minimum flows in southern Quebec during all seasons. This result thus confirms that its influence on the spatial variability of the flood magnitude is much weaker than that of the wetland surface area. Consequently, this last factor must be considered indisputably as the main factor of the spatial variability of the annual daily maximum flow magnitude and duration in southern Quebec.

Wetland surface area was not significantly correlated with the frequency and variability of flows. These two flow characteristics were therefore no different on the two shores. Frequency was positively correlated with maximum temperature but negatively correlated with drainage density. While the temperature south of 47°N was higher on the south shore than on the north shore, the opposite was true for drainage density. Although the difference in the magnitude and duration of annual flood may not be significant between the two shores, it was nevertheless observed that overall their frequency was relatively high on the south shore as opposed to the north shore, likely due to greater runoff from snowmelt.

Both flow variability indices (CI and CV) were negatively correlated with the drainage density and mean slope of the watersheds. These two physiographic variables had a significant influence on the transfer of runoff to the channels and its subsequent evacuation outside the watersheds. Given the influence of wetlands on the shape of flood hydrographs (flattening), this negative correlation with the mean slope of the watersheds was due to the fact that the mean slope of north shore watersheds in the Canadian Shield was greater than those of the south shore watersheds, whose rivers flow on the St. Lawrence Lowlands, which are characterized by a relatively flat topography. With respect to drainage density, this correlation should theoretically be positive, given that the drainage density of rivers on the north shore was lower overall than those on the south shore. This negative correlation would likely be linked to flow duration.

In terms of flood management, the significant reduction in the floods magnitude (intensity) by wetlands is a beneficial contribution that reduces socio-economic and environmental costs. However, this benefit may be attenuated or counteracted by the increase in the duration of these floods. However, the socio-economic and environmental impacts induced by the floods duration are not yet documented in southern Quebec. Therefore, it becomes important to determine these impacts for better development of flood management policies.

The application of the original MK test and the modified tests stemming from it to eliminate the effects of autocorrelation (short-term persistence) showed that the magnitude of annual maximum daily flows was affected by a statistically significant long-term trend in all of the rivers (NS1–NS5) on the north shore, contrary to those on the south shore, where less than half (33%) were affected (Table 8). This long-term trend was positive on both shores. These different tests, therefore, suggested that the magnitude of annual maximum daily flows increased significantly over time. This increase was more widespread on the north shore of the St. Lawrence River than on the south shore. The same was true of the long-term trend of the duration of maximum flows in the ≥2 years recurrence interval, for which the long-term persistence test detected no significant change in means of the hydrological series on both shores of the St. Lawrence. But the increase in the mean duration of flows is also not widespread.

All of these tests revealed an overall common trend in the magnitude and duration of annual maximum flows. This overall trend is characterized by an upward trend in these two flow characteristics in the three hydroclimatic regions, despite the significant decrease in the amount of snow in Quebec (Brown 2010; Guerfi et al. 2015). Recall that Blöschl et al. (2019), among others, observed a significant decrease floods following the decrease snow cover and snowmelt, resulting from warmer temperatures in Eastern Europe, whose climate is almost similar to that of Quebec. For Quebec, this trend is also predicted by all climate models (e.g., Boyer et al. 2010).

However, unlike Eastern Europe, the rainfall did not decrease significantly over time in Quebec. In fact, an upward trend has even been observed. This increased rainfall, particularly during the spring snowmelt, therefore compensated for the decrease in snowfall, maintaining or increasing the magnitude of annual freshet flows. As several studies have shown (Perrault et al. 1999; Assani et al. 2008), the increase in rainfall was also observed in summer and fall, thereby contributing to the increase in groundwater levels during the spring snowmelt (Assani et al. 2021). This increase in rainfall in summer and fall was also observed in the northeastern part of the United States (e.g., Small et al. 2006; Sadri et al. 2016), which led to an increase in the magnitude of seasonal minimum flows. This increase in minimum flows was also observed in the southeastern hydroclimatic region of Quebec south of 47°N in the four seasons (Assani et al. 2022).

Remember that on the north shore, the surface area of wetlands in the watersheds is higher than on the south shore. According to the concept of sponge effect, despite the increase in the rainfall, the increase in the magnitude of flows over time should in principle be less significant and less generalized on the north shore than on the south shore. All the studies devoted to the impacts of wetlands on the magnitude of flows predict a reduction of the magnitude of flood in the context of global warming due to the sponge effect in Quebec (Fossey & Rousseau 2016a, 2016b). This results in a more or less significant absorption (infiltration) of the excess runoff water brought by the increase in rainfall. However, this conclusion does not correspond to the results of our analysis of the long-term trend in the magnitude and duration of the annual maximum flows. These results clearly show that it is the watersheds of the north shore that experienced a significant increase in the magnitude and duration of flows despite the fact that the increase in rainfall was observed in all the watersheds of southern Quebec. Thus, it is the watersheds with a greater wetland surface area (>5%) that experienced a significant increase in the magnitude and, to a lesser extent, that of the duration of flows. This increase is explained by the fact that the excess runoff water brought by the increase in rainfall is not entirely stored because the storage capacity of wetlands and other water bodies does not increase over time as the rainfall increases. This surplus of water, not infiltrating into the ground, thus contributes to the significant increase in the magnitude of flood flows. It follows that, unlike the sponge effect, which attenuates the magnitude of flood flows over time by absorbing excess runoff water, the surface storage effect helps to amplify them due to the limited storage capacity of wetlands and other water bodies over time. This is a fundamental difference between the two concepts on their impacts on the temporal variability of flood magnitude and duration in the context of climate change. This increase in the magnitude of the flows over time on the north shore was also observed in floods generated by rain in warm season (summer and autumn) (Assani 2023). On the other hand, with regard to the temporal variability of the lower water levels magnitude, it increases significantly over time on the south shore in the most agricultural watersheds (due to the significant decrease in agricultural area which promotes water infiltration) than on the north shore in the least agricultural watersheds (Assani et al. 2021, 2022).

It follows from all these results that in southern Quebec, the spatio-temporal variability of floods is mainly influenced by the wetland surface area, while that of the low water levels, by the agricultural surface area. This spatio-temporal variability of floods and low flows is, therefore, better explained by the concept of surface storage. From a climatic point of view, this significant increase in flood and low water flows over time is due to the increase in precipitation in the form of rain, probably due to the general increase in temperature observed in Quebec.

Presented about three centuries ago, the concept of the sponge effect has long been used in the scientific literature to explain the impacts of wetlands on river flows. But many studies over time have shown that this concept does not fully explain the hydrological behavior of rivers in relation to these wetlands. In fact, with regard to flood flows, many studies have shown that the impacts of wetlands do not all result in a reduction in the magnitude of flood flows due to the influence of many factors related to the intrinsic and extrinsic characteristics of these wetlands. Moreover, this concept has never been applied to explain the impacts of wetlands on the other four fundamental characteristics of flood flows: duration, timing, variability and frequency. This concept is therefore hydrologically incomplete.

In Quebec, this concept has been applied to analyze the impacts of two types of wetlands (non-floodplain and riparian wetlands) on the magnitude of flood flows generated by melting snow and rain. These impacts have been shown to result in a decrease of this magnitude as predicted by the concept. However, these impacts were analyzed separately without being confronted with those induced by other topographic components of the landscape (lakes, depressions, ditches, etc.) and other factors (agriculture, deforestation, etc.) whose effects can amplify or attenuate, to varying degrees, those induced by wetlands. In order to take into account the interaction of these different topographical components and factors, a new, more global concept has been applied: surface storage. This new concept is justified by the fact that several recent studies have demonstrated the influence of these other topographic components on surface water storage. This new concept differs mainly from the concept of the sponge effect by the fact that the storage of runoff water on the surface is not accompanied by a significant infiltration of water. This hydrological process of surface storage has very little effect on low water flows, unlike the sponge effect process.

The application of this new concept of surface storage has made it possible to explain the spatio-temporal variability of the characteristics of the annual daily maximum flows generated mainly by snowmelt. Regarding spatial variability, the correlation analysis revealed a strong link between wetland surface area (including other topographic components) and the magnitude and duration of these flows. The wetlands are negatively correlated with the magnitude but positively with the duration of flows. They are not significantly correlated with either the frequency or the variability of these flows. This negative correlation with the magnitude but positive with the duration of the flows results from the process of storage water surface of the snowmelt and the rains. In terms of management, the beneficial environmental effects that result from the reduction in magnitude can be counteracted by the negative effects associated with the increase in the duration of floods. This aspect must be taken into account in the management of watersheds.

As for the temporal variability, the application of several MK tests revealed a significant increase in the magnitude and, to a lesser extent, the duration of the flood flows in the watersheds of the north shore characterized by a greater surface area of wetlands. This increase is explained by the fact that the surface storage capacity of these wetlands, which does not change over time, does not make it possible to store the excess runoff water caused by the increase in rainfall during snowmelt. This result demonstrates that the storage water surface is accompanied by a low infiltration contrary to the process related to the concept of the sponge effect.

In the current context of global warming, the storage of water on the surface will promote its evapotranspiration which will feed the intensity of the rains according to the thermodynamic principle of Clausus–Claperyron. In addition, this storage risks amplifying the intensity and duration of floods in these watersheds with a larger wetland surface area, contrary to predictions based on the concept of the sponge effect. This aspect must be taken into account when planning watersheds to mitigate the effects of flooding in southern Quebec.

This work was supported by the Natural and Engineering Science Research Council of Canada (Grant n°: 261274/2019).

A.A.A.: Conceptualization, Formal analysis, Funding acquisition, Methodology, Supervision, Formal analysis, Writing – Original Draft; A.Z.: Methodology, Software, Formal Analysis, Writing – Original Draft; C.K.: Conceptualization, Writing – Review and Editing; A.R.: Conceptualization, Writing – Review and Editing.

All relevant data are available by sending a request to the corresponding author.

The authors declare there is no conflict.