Spatial–temporal variability of seasonal daily minimum flows in southern Quebec: synthesis on the impacts of climate, agriculture and wetlands

It is becoming clear that current global warming is increasingly impacting extreme flows in many parts of the world. With regard to minimum flows, numerous works, based on climate models under different scenarios of GHG (greenhouse gas) emissions, have already demonstrated the high sensitivity of these flows to the increase in temperature (Andreas et al. 2017; Donnelly et al. 2017). However, the impacts induced by global warming on these low water flows are not uniform. Several studies have shown that these impacts depend on many factors: the magnitude of changes in temperature and precipitation, types of climate, types of land use, physiographic characteristics of watersheds, dams and reservoirs, etc. (e.g., Assefa & Moges 2018; Azouaoui & Assani 2017; Dudley et al. 2020; Moore et al. 2020). Other work has demonstrated that these impacts induced by global warming can be amplified or mitigated by changes in land use (e.g., Zhang et al. 2018; Foroumandi et al. 2021, 2022). However, unlike the temporal variability of minimum flows, there are very few studies on its spatial variability in the international scientific literature (Poff et al. 2006). These authors compared the impacts of agriculture, urbanization, vegetation cover and dams on the spatial variability of minimum flows, among others, of watersheds located in hydroclimatic regions of the United States.

Regarding the types of climate, in cold temperate regions, rivers are characterized by two distinct types of low-flow regimes (Kinnard et al. 2022): a winter low-flow regime and a summer low-flow regime. These two types of regimes are not influenced by the same climate factors. The winter low-flow regime is primarily influenced by the amount of rainfall in the fall and winter temperatures, with low flows occurring in late winter before the snow melts. The summer low-flow regime is mainly influenced by snow accumulation during the cold season and by spring and summer temperatures. However, it has been established that global warming does not affect precipitation and temperatures in the cold and warm seasons in these temperate regions in a uniform way. As a result, the temporal variability of low flows resulting from global warming varies depending on the season, with this difference reflected in seasonal temperature and precipitation regimes. In Quebec, studies on the temporal variability of temperatures and precipitation have shown a significant decrease in snowfall but an increase in temperature during the cold season (Yagouti et al. 2008; Brown 2010; Guerfi et al. 2015), as well as an increase in temperature and rainfall during the warm season (Perrault et al. 1999; Assani et al. 2008, 2019). As for the change in land use, following the modernization of its agriculture initiated since 1950, Quebec has lost more than half of the areas under cultivation in favor of fallow land and reforested land (Ruiz 2019). This reduction in the agricultural area can thus potentially affect the temporal variability of minimum flows.

Several previous studies have analyzed the temporal variability of seasonal and annual low flows in Canada, in general, and in southern Quebec, in particular (Bonsal & Shabbar 2008; Khaliq et al. 2008, 2009; Quilbé et al. 2008; Ehsanzadeh & Adamowski 2010; Assani et al. 2011a, 2021; Daigle et al. 2011; Sylvain et al. 2015; Blanchette et al. 2018; Berthot et al. 2021; Kinnard et al. 2022) in the context of global warming. Other works have analyzed the impacts of rainfall (Assani et al. 2006), deforestation (Lavigne et al. 2004), agricultural land cover (Muma et al. 2011; Assani et al. 2021), wetland area (Fossey et al. 2016a, 2016b; Assani 2022) and dams (Assani et al. 2005, 2011a; Azouaoui & Assani 2017) on the spatial variability on minimum flows.

These different studies raise two important questions about the factors of the spatio-temporal variability of seasonal daily minimum flows in southern Quebec in the context of global warming.

• - With regard to spatial variability, all these studies separately analyze the impacts of factors related to climate, vegetation cover, land-use change and the physiographic characteristics of watersheds on the spatial variability of these flows. However, it is extremely important to determine the major factors that influence this variability in the context of the development of water resources in southern Quebec.

• - As for the temporal variability, none of these studies has ever compared the impacts induced by the increase in rainfall, the decrease in the snowfall and the decrease in the agricultural area on the seasonal daily minimum flows in order to determine how this decrease in agricultural land cover can amplify or attenuate the effects of changes in precipitation regimes in the context of global warming on the temporal variability of these flows. It is important to emphasize that, unlike in underdeveloped countries, the modernization of agriculture in developed countries has resulted in a significant decrease in agricultural land cover. This trend has continued today. But the impacts of this decrease on minimum flows are still very little studied in these countries. These impacts are never integrated into the prediction of these flows in the context of warming by climate and hydrological models (e.g., Quilbé et al. 2008; Andreas et al. 2017).

The aim of our study is to answer these two questions. This study will be based on the analysis of daily seasonal minimum flows measured over a very long period of more than 80 years. These two questions have never been analyzed in the international scientific literature. Their analysis thus underpins the originality of our study.

Watershed and land-use physiographic data (average slope; drainage density; and agricultural, forest, wetland and urbanized surface area, etc.) were measured by the Glaciolab laboratory at the Université du Québec à Trois-Rivières, whose methodology has been described in detail by Assani et al. (2021) and Kinnard et al. (2022). Wetland data were supplemented with data from Belzile et al. (1997). It is important to note that wetlands also include small, shallow lakes and other types of water bodies. This is, therefore, a very broad definition of these wetlands. Under these conditions, the objective of the study is not to determine the impact of a particular type of wetland. As for the agricultural area, it is the land cultivated each year thus excluding fallow and reforested areas. The temporal variability of these cultivated areas has been analyzed in detail by Ruiz (2019) during the period 1950–2011. Daily flow data were obtained from the website of the Quebec Ministère de l'Environnement et de la Lutte contre les changements climatiques's Centre d'expertise hydrique du Québec (https://www.cehq.gouv.qc.ca/index_en.asp, accessed on 2020-02-20). Climate data on temperature and precipitation were obtained from Environment and Climate Change Canada's website (https://climat.meteo.gc.ca/climate_normals/index_f.html, accessed on 2021-06-18). These are the monthly averages of the climatic normals for the periods 1971–2000 and 1981–2010.

The reason for this study is the lack of studies on winter and summer low-flow spatial variability factors, coupled with the lack of studies on long-term winter and summer low-flow trends over a relatively long period (over 80 years) that take into account short- and long-term persistence effects. The purpose of this study is to address these gaps. To construct the seasonal minimum flow statistical series, the lowest (minimum) daily flow values measured in summer (June to August) and winter (January to March) were selected for each year from 1930–2019 for each of the 17 rivers studied. The means of these series were then compared to the means of parametric (ANOVA) and non-parametric (Kruskal–Wallis) statistical tests. Finally, these means were correlated with physiographic and climatic variables (see Table 2) to determine their spatial variability factors.

Table 2

Correlation coefficients (R) calculated between physio-climatic factors and daily minimum flows in winter and summer

View Large

Three statistical methods for analyzing long-term trends were applied to examine the temporal variability of these flows. The first test was the Mann–Kendall (MK) test described by Sneyers (1990). Since this test does not eliminate the effects of autocorrelation on the trend, four other tests were applied that do. The second set of tests consisted of two tests that eliminate autocorrelation effects and are based on data filtering. These included the prewhitening method (MMK-PW) described, in particular, by Von Storch (1995) and the trend-free prewhitening method (TFPW) described, in particular, by Yue et al. (2002). The other two tests eliminated autocorrelation effects by correcting the variance of a statistical series. These are the Modified Mann–Kendall Test1 (MMKY) and the Modified Mann–Kendall Test2 (MMKH) tests described by Yue & Wang (2004) and Hamed & Rao (1998), respectively. Finally, the last statistical test applied eliminates the long-term persistence (Hurst effect) on the trend. This is the long-term persistence (LTP) test described by Hamed (2008). There is already a great deal of work in the scientific literature that has mathematically described these statistical tests (e.g., Dinpashoh et al. 2014; Serinaldi & Kilsby 2016). These tests will, therefore, not be described in detail in this study.

Correlation coefficient values calculated between the physio-climatic variables and the means of daily minimum flows for both seasons are shown in Table 2. In terms of physiographic variables, in winter, flows had a strong positive correlation with the wetland surface area only. In summer, flows were also positively correlated with the average slope of the watersheds but negatively correlated with the agricultural surface area. Note that minimum flows and wetland surface area were more strongly correlated in winter than in summer. As for climate variables, in winter, daily minimum flows were positively correlated with the temperature in March but negatively correlated with snowfall. In summer, daily minimum flows were not significantly correlated with any climate variable.

In summer, the six tests produced dissimilar results. A difference was observed between the results of tests that eliminate the effects of short-term persistence and the results of tests that eliminate long-term persistence. Differences were also observed in trends between the three hydroclimatic regions. On the south shore south of 47°N, daily minimum flows in summer increased significantly for all rivers. However, the LTP test that eliminates the effects of long-term persistence did not detect a significant trend for the Eaton and Etchemin rivers. North of the 47th parallel on the south shore, the long-term trend was characterized by a decrease in flows over time for all rivers. The Z scores for all six statistical tests were negative, except for the Blanche River. Nevertheless, the statistical significance of this trend varied depending on the test. The LTP test detected a significant trend for two rivers only (Trois-Pistoles and Rimouski). On the north shore, the Z scores for all six tests were also negative for all rivers, reflecting a trend toward decreasing minimum flows over time in summer in this hydroclimatic region. But unlike the other five tests, the LTP test detected a statistically significant trend for two rivers (Petite-Nation and Du Nord) only.

To answer questions on the main spatial variability factors for seasonal daily minimum flows and to determine which seasons and hydroclimatic regions are most sensitive to climate change in southern Quebec, the results obtained from these analyses were compared. The data are summarized in Table 5. The agricultural surface area was the main spatial variability factor in spring, summer and fall. In winter, the wetland surface area was the main factor. Spatial variability was not influenced by climate variables, with the exception of the winter season, where the spatial variability of daily minimum flows was mainly influenced by snowfall. Winter flows increased significantly over time in the three hydroclimatic regions. Winter was most affected by climate change. In contrast, summer flows tended to decrease overall over time. When comparing the regions, the southeast hydroclimatic region located south of 47°N was characterized by a significant increase in daily minimum flows in all four seasons. This region was the most affected of the three studied. In contrast, the Eastern hydroclimatic region, located north of the 47th parallel, appears to be the least sensitive to temperature and precipitation regime changes. Very few stations in this hydroclimatic region were affected by a significant change in flow means over the four seasons.

The comparison of winter and summer daily minimum flow means revealed a distinct regional difference. During both seasons, these flows were higher on the north shore than on the south shore. The correlation analysis revealed that these flows are correlated with several physiographic and climatic factors. Of these factors, the wetland surface area was the only one that was positively correlated with flows in both seasons. The wetland surface area in the watersheds was higher (>8%) on the north shore than on the south shore (<4%). The influence of wetlands on minimum flows has been well established (see syntheses provided by Bullock & Acreman 2003; Lane et al. 2018). According to the well-known concept of the sponge effect, wetlands store water and recharge the water tables that supply river flows during low flows. Flows in these watersheds, therefore, remain relatively high. However, this sponge effect does not fully apply in Quebec to explain the difference in the magnitude of daily minimum flows between two watersheds that differ in the area of wetlands. In a recent study, Assani (2022) showed that the presence of wetlands greatly slows the decrease in flows during the fall hydrological drought period. This decrease makes it possible to keep low flows at relatively high levels so that they occur later in the fall season, unlike low flows in an agricultural watershed. In winter, minimum flows were, therefore, much higher in non-agricultural watersheds than those in agricultural watersheds in southern Quebec. Remember that in Quebec, in winter, soils, wetlands and all bodies of water become frozen, thus preventing any infiltration of water into the ground. The minimum flows are thus influenced by the extent of the recharge of the aquifers in the autumn before this general frost. In the other three seasons, wetlands kept runoff on the surface longer, slowing down the infiltration process. As a result, wetlands had less of an influence on water table recharge and low flows.

The second physiographic factor, which negatively correlated with minimum flows, was agricultural area. Unlike wetlands, the agricultural surface area is greater on the south shore (>10%) than that on the north shore (<10%). The impacts of agriculture on minimum flows have been studied in southern Quebec (e.g., Quilbé et al. 2008; Muma et al. 2011; Assani 2022). These studies all showed that the hydrological impacts of agriculture result in a significant decrease in minimum flows caused by soil sealing, reducing the amount of water that infiltrated the soil and, in turn, recharged aquifers. However, the decrease in minimum flows caused by the hydrological impacts of agriculture seemed to be more pronounced during winter than in the other three seasons. Minimum flows in winter depended primarily on the volume of water recharging the water tables in fall. In agricultural watersheds, the presence of crops reduced soil sealing for much of the fall, although the extent to which crops promoted infiltration and hindered runoff varied. In the only study devoted to the impacts of agriculture on the spatial variability of minimum flows in particular, Poff et al. (2006) demonstrated that in the United States, the increase in agricultural land cover in watersheds had less impact on the reduction in minimum flows, unlike urbanization, which caused a significant increase or decrease in these flows according the hydroclimatic regions.

The third physiographic factor, which most closely correlated with minimum flows, was average watershed slope. This factor influenced infiltration and runoff processes in watersheds. In theory, the greater the watershed's average slope, the greater the runoff compared to infiltration, all other conditions being equal. The average watershed slope must, therefore, be negatively correlated with minimum flows. However, in southern Quebec, the average slope was positively correlated with minimum flows in summer. This positive correlation is explained by the fact that north shore watersheds, which had the greatest minimum flows, also had the greatest average slopes because of their geological and topographical characteristics. All south shore watersheds drain two geological formations (the Appalachians and the Saint Lawrence Lowlands) that are characterized by a relatively flat topography. Their watersheds generally have less of an average slope than watersheds on the north shore. North shore watersheds cut through the Canadian Shield, whose topographic features result in a succession of valleys and plateaus.

The first climate factor that was significantly correlated with minimum flows was snowfall. But this effect occurred exclusively in winter. In Quebec, aquifer recharge occurs mainly in fall and spring. Theoretically, snowfall should not have a significant influence on minimum flows in fall, because the aquifers are being recharged exclusively through rainfall infiltration. In spring, the water that infiltrates and recharges the aquifers comes primarily from snowmelt. However, this study clearly showed that snowfall exclusively affects minimum flows in winter, resulting in a negative correlation between snowfall and minimum flows. This negative correlation is explained by the fact that if snowfall increases in fall, water infiltration is greatly reduced due to a lack of surface runoff, as well as frozen ground, which seals the soil. In summer, minimum flows were not correlated with snowfall, despite the fact that the water comes from snowmelt, which is the main source of groundwater recharge. The lack of a link between snowfall and minimum flows in summer can be partially explained by the effect of evapotranspiration on these flows and the impact of rain, as previously mentioned by Kinnard et al. (2022).

In addition to snowfall, maximum temperatures in March were significantly correlated with daily minimum flows in winter only. Daily minimum flows in winter were measured before snowmelt. However, snowmelt was affected by temperatures in March and April. If snowmelt occurred in March, minimum flows were higher than when snowmelt took place in April. This explains the positive correlation between March temperatures and daily minimum flows in winter. Note that unlike minimum flows in winter, minimum flows in summer were not correlated with temperature, even though evapotranspiration can affect the amount of water in the soil.

The results of six statistical tests analyzing long-term trends revealed a widespread increase in daily minimum flows in winter, but the spatial extent of this increase varied across hydroclimatic regions. The increase was observed almost everywhere in the only hydroclimatic region located south of 47°N on the south shore. The common feature of all watersheds in this region is a larger agricultural surface area and their location within the St. Lawrence Lowlands, which are less permeable than the Appalachians. As for the watersheds on the north shore, the Canadian Shield is even less permeable than the two geological formations on the north shore. Although the geological factor (degree of substratum permeability) was an important element in the temporal variability of daily minimum flows, the amount of variability differed significantly between watersheds on the north and south shores. At the same time, the impact of differences in climate characteristics between the north and south shores can be excluded, as changes in temperature and precipitation regimes were the same.

Nevertheless, this study showed that in winter, the general trend of temporal variability in minimum flows is characterized by an increase in flows. This increase means that recharging of the aquifers that affect them has increased significantly over time. Remember that aquifer recharge occurs primarily in fall as a result of rainfall. However, many studies have already highlighted a significant increase in rainfall in summer and fall in Quebec, in particular, and in northeastern North America, in general (Perrault et al. 1999; Small et al. 2006; Assani et al. 2008, 2011b, 2019; Yagouti et al. 2008; Hodgkins & Dudley 2011; Sadri et al. 2016). As a result, regional differences in the temporal variability of flows can only be explained by differences in land use affecting infiltration, as different aquifer characteristics cannot explain these differences, as was just demonstrated. The more widespread increase in winter flows observed in the southeastern hydroclimatic region compared to the other two regions was due to changing land use (Zhang & Schilling 2006; Dudley et al. 2020). This region is the most agricultural in Quebec. Modernization of agricultural practices began in 1950, resulting in a significant reduction in cultivated areas and a subsequent increase in uncropped land and reforestation (Ruiz 2019). This change in land use has led to water infiltration and water table recharge over time. The result is an increase in minimum flows over time, while in the other two regions, very little has changed in terms of land use that would encourage water infiltration. The increase in flows observed in the other two hydroclimatic regions (east on the south shore and southwest on the north shore) was, therefore, due exclusively to the increase in rainfall. It is important to emphasize that almost all the work on the hydrological impacts of agriculture on the temporal variability of minimum flows is based on the increase in agricultural area and not on its decrease (e.g., Foroumandi et al. 2021, 2022).

In summer, the overall trend in the temporal variability of daily minimum flows was characterized by a decrease in two of the three hydroclimatic regions. Minimum flows in summer were greatly impacted by aquifer recharge supplied by snowmelt. Due to the widespread decrease in snowfall, aquifer recharge also decreased over time, causing a downward trend in daily minimum flows. However, despite the decline in snowfall, only the most agricultural hydroclimatic region in the southeast was characterized by an increase in minimum flows in summer due to the increase in infiltration caused by the very significant decrease in cultivated land, as previously mentioned.

Along with the decrease in snowfall and the increase in rainfall, temperatures have increased during all four seasons in southern Quebec. This temperature increase was expected to result in a significant increase in evapotranspiration and a subsequent decrease in daily minimum flows, especially during the warm season (summer and fall). However, the increase in temperature was also expected to lead to increased minimum flows in winter due to early snowmelt. The positive correlation between March temperatures and daily minimum flows in winter demonstrates this relationship. Although this study did not clearly demonstrate the impact of evapotranspiration on daily minimum flows in summer, such an impact cannot be definitively ruled out, despite the lack of a correlation between temperature and minimum flows in summer.

Minimum flows play a crucial role in the functioning of river ecosystems. In particular, they make it possible to determine the permanent or intermittent nature of river flows, which significantly influences the dynamics of aquatic and semi-aquatic flora and fauna. These flows are impacted to varying degrees by global warming, as demonstrated by the increasingly severe droughts that increasingly affect many regions, even those once considered humid, of the planet. The spatio-temporal variability of these flows results from the interaction of natural (climate, physiographic characteristics of watersheds, type of soils and land cover, etc.) and anthropogenic (deforestation, agriculture, urbanization, dams, etc.) factors. In many studies, this interaction is not taken into account to identify the major factors of the spatio-temporal variability of these flows in the context of global warming with a view to integrating them into a global water management policy to better mitigate the impacts of global warming. Almost all the works devoted to the analysis of the spatio-temporal variability of minimum flows take little or no account of the interaction of these different factors because they analyze the impacts of these different natural and anthropogenic factors separately.

With respect to southern Quebec, this study has clearly identified that the main factor in the spatial variability of seasonal daily minimum flows is, without a doubt, the agricultural area. The impacts of this factor translate into a significant drop in the seasonal minimum daily flows due to the increase in runoff resulting from the sealing of the soil by agricultural activities. However, this sealing does not explain the spatial variability of minimum flows in winter. This variability is explained by the wetland area, which slows down the decrease in minimum flows (Assani 2022) in the fall before snowfall, which inhibits any surface runoff. This also explains the significant relationship between this climatic factor and the spatial variability of minimum flows in winter, unlike the other three seasons.

As for temporal variability, this study highlighted the impact of the reduction in the agricultural area, which translates into a systematic increase in minimum flows in most agricultural watersheds of Quebec due to the increase in infiltration in land left fallow and/or reforested. This increase is also due to that of the increase in rainfall observed during the relatively hot seasons. Thus, despite the significant decrease in the amount of snowfall, the seasonal minimum flows have not significantly decreased as predicted by all climate models. However, this decrease is still observed in certain less agricultural watersheds in summer, where the supply of aquifers comes mainly from snowmelt in the spring.

The main limitation of this study lies in the fact of not having analyzed the impacts of urbanization, due to hydrological data, on the spatio-temporal variability of minimum flows. Indeed, in developed countries, land formerly devoted to agriculture is increasingly devoted to urbanization and industrial development. These anthropogenic activities induce hydrological impacts totally different from those induced by agriculture. Therefore, these impacts must be taken into consideration to better predict the evolution of minimum flows in the context of current global warming in order to develop the best management policies to protect river ecosystems. Finally, this study clearly demonstrates that it is not possible to generalize the predictions of seasonal daily minimum flows from climate and hydrological models to all the watersheds of southern Quebec. The quality of these predictions strongly depends on the evolution of land-use changes. This aspect must be taken into consideration for better water management in each watershed of southern Quebec.

This research was funded by the Natural Sciences and Engineering Council of Canada (grant no. ‘261274/2019’).

A.A.A. conceptualized the whole article, conducted a formal analysis and funding acquisition, developed the methodology, supervised the work, and wrote the original draft. A.Z. developed methodology, software, and formal analysis and wrote the original draft. C.K. and A.R. conceptualized the whole article, wrote the review, and edited the article.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.