In this study, the artificial neural network (ANN) method was applied to investigate the impacts of climate change on the water quantity and quality of the Qu'Appelle River in Saskatchewan, Canada. First, the second-generation Canadian earth system model (CanESM2) was adopted to predict future climate conditions. The Statistical DownScaling Model (SDSM) was then applied to downscale the generated data. To analyze the water quality of the river, concentrations of dissolved oxygen (DO) and total dissolved solids (TDSs) from the river were collected. Using the collected climate and hydrometric data, the ANNs were trained to simulate (i) the ratio of snowfall-to-total precipitation based on the temperature, (ii) the river flow rate based on the temperature and precipitation; and (iii) DO and TDS concentrations based on the river flow and temperature. Finally, the generated climate change data were used as inputs to the ANN model to investigate the climate change impacts on the river flow as well as DO and TDS concentrations within the selected region. Hydrologic alteration of the river was evaluated via the Range of Variability Approach (RVA) under historical and climate change scenarios. The results under climate change scenarios were compared with those under historical scenarios and indicated that climate change would lead to a heterogeneous change in precipitation and temperature patterns. These changes would have serious degrading impacts on the river discharge as well as DO and TDS concentration levels, causing deterioration in the sustainability of the river system and ecological health of the region.

  • Climate change impacts on the flow regime and water quality parameters in the semi-arid prairies.

  • A framework is established to study the climate change impacts on flow and water quality parameters.

  • The framework is calibrated and validated for climate change impacts.

Literature review

Prudent management of water resources is of paramount importance due to growing human water demands (such as agricultural use, domestic consumption, and hydropower generation) and environmental/ecological requirements (Hassanjabbar et al. 2018). Reliable and high-quality water supplies are required to satisfy both human and environmental needs. The health of river systems is indispensable for sustainable development and biodiversity (Reid et al. 2019). Water quality indicators such as dissolved oxygen (DO) and total dissolved solids (TDSs) are significant parameters in aquatic species' health and distribution. Moreover, these indicators play a prominent role in regulating water resources for human consumption and the health of aquatic habitats (Caldwell et al. 2015; Zhang et al. 2017).

DO is the amount of non-compound oxygen dissolved in water. It is a primary indicator of river water quality and plays a critical role in the determination of water quality. DO is one of the most important parameters in water quality since the DO concentration could affect other water quality variables and also other processes. DO is essential for the survival of aerobic aquatic organisms. Thus, understanding the variation of DO concentration is quite necessary for water resource managers (Heddam & Kisi 2017). TDS consists of inorganic and organic materials such as salts, minerals, ions, and metals dissolved in water (Miranda & Krishnakumar 2015; Zhang et al. 2017). A high concentration of TDS in water represents that the water may have aesthetic issues, which may be related to staining, taste, or precipitation (Sibanda et al. 2014). In general, the levels of TDS indicate the pollutant burden on the aquatic system (Jonnalagadda & Mhere 2001; Zhang et al. 2017).

Climate change may significantly alter hydrological processes affecting water quality and quantity (Jamali et al. 2013; Abera et al. 2018; Ekwueme & Agunwamba 2021; Feistel & Hellmuth 2021; Hassanjabbar et al. 2021; Shah et al. 2021). Therefore, understanding the impact of other significant drivers such as climate change helps water managers make better contributions and decisions to solve water quality degradation and conserve aquatic ecosystems. Several studies in the area of water quality simulation and management have been carried out in different regions of Canada; however, the impacts of climate change have not been adequately addressed. Novotná et al. (2014) analyzed the impact of climate change on water discharge and quality of the Bras d'Henri river, situated in Québec, Canada. The results represented that precipitation, evapotranspiration, and river discharge would increase. Furthermore, it was concluded that erosion is expected to threaten agricultural land and water quality. He et al. (2015) investigated the spatial and temporal variations of water quality and the pollutant load from non-point sources along the Bow River, Alberta, Canada. They used a mass balance method and statistical analyses to examine water quality characteristics. It was concluded that point sources (treatment plants) play a considerable role in spatial and temporal trends in water quality parameters, while the contribution of non-point sources varied along the river. Hassanzadeh et al. (2019) presented a framework to contemplate different perspectives of stakeholders in water quality modeling. Their study was carried out in the Qu'Appelle River Basin, Saskatchewan, Canada. The Beneficial Management Practices (BMPs) and Q-methodology were used to engage the stakeholders in water quality management. It was concluded that stakeholders found the model beneficial and provided a set of recommendations to improve the model. Meshesha et al. (2020) investigated the spatiotemporal pattern of water quality and its impacts on the aquatic ecosystem of the Athabasca River Basin, Alberta, Canada. They developed a watershed scale module of the DO, dissolved organic carbon (DOC), and fecal coliforms (FC) in the Soil and Water Assessment Tool (SWAT) model. They concluded that variability of the mentioned water quality parameters may change the ecosystem services of the basin, hence, adaptive management may need to be applied in the region. Zango et al. (2021) assessed the impacts of urbanization and climate change on the river runoff as well as water nitrogen and phosphorus loads in the Carp Watershed, Ontario, Canada. They concluded that the urbanization and climate change impacts vary substantially depending on the geographic location and spatial scale.

Considering the importance of the water quality parameters such as DO and TDS, it is essential to recognize the impacts of climate change on the DO and TDS concentration to better predict future water quality and quantity of a water system. In this study, the Qu'Appelle River located within the Lower Qu'Appelle River Watershed was selected as a case study in the semi-arid prairies. The main objective of this study was to evaluate the impact of climate change on water quantity and quality of the river. In the present study, concentration levels of water quality indicators were simulated directly via climate change parameters (rainfall, snowfall, and temperature) using the artificial neural network (ANN) method. Another advantage of the study was to extract the future amount of snow from the total precipitation of the watershed under climate change using the ANN method.

Very limited studies have been conducted in the Lower Qu'Appelle River Watershed and none of them applied the same methodology and viewpoint. This study was carried out since the water quality has been declining in the Lower Qu'Appelle River while the water demands have been increasing in the watershed. The present study presented a framework to (1) quantify and evaluate future snowfall, rainfall (precipitation), and temperature of the Lower Qu'Appelle River Watershed; (2) predict the future flow of the Qu'Appelle River and evaluate the impact of climate change on the discharge of the river; (3) assess hydrological alteration of the Qu'Appelle river under climate change; and (4) simulate future water quality indicators (DO and TDS) and assess the climate change impact on the indicators.

In the following, the lower Qu'Appelle River Watershed and its challenges are introduced. Then, the proposed framework and methodology of the present study were provided. The associated results are represented and discussed in detail. Finally, the summary of the study and major findings are mentioned in the Conclusion section.

The Lower Qu'Appelle River Watershed

The Lower Qu'Appelle River Watershed is currently experiencing water scarcity and quality issues due to rising domestic, agricultural, and industrial water demands. On the other hand, climate variability is expected to exacerbate and complicate the current issues. The Lower Qu'Appelle River Watershed with an area of about 17,800 km2 is situated in southeastern Saskatchewan. The watershed covers the area that begins east of the village of Craven in Saskatchewan and continues to the Manitoba border. The Lower Qu'Appelle River Watershed is specified by hot summers, cold winters, and variable weather patterns. The mean annual precipitation in the watershed varies from as low as about 360 mm to as high as 470 mm. The average annual snowfall is about 110 mm. The economic activities in the watershed are agriculture, tourism, oil and gas development, and potash mining.

The Qu'Appelle River is the dominant watercourse within the watershed. The river flows through the following six major lakes: Pasqua, Echo, Mission, Katepwa, Crooked, and Round Lakes. The Qu'Appelle River is a slow-moving system with a low gradient river channel. The river is the primary source of drinking water for about a third of the population of Saskatchewan. The river also provides water for agricultural and industrial uses (Water Security Agency 2013; Hosseini et al. 2017). Population growth, growing industrial water consumption, and agricultural intensification are expected to increase river abstractions (Water Security Agency 2013). By 2060, water demand in the basin would exceed the 2010 demand level by 38% (Saskatchewan Watershed Authority et al. 2012). In addition to potential water scarcity, the river water quality has been further deteriorating (Smol 2009). The watershed is highly productive and characterized by frequent macrophyte growth in different parts of the river, algal blooms, and periodic fish kills (Smol 2009).

A concrete control structure on Echo Lake is used to regulate the water levels at Pasqua and Echo Lakes for recreational purposes. Moreover, a control structure at Katepwa Lake regulates water levels at Mission and Katepwa lakes for commercial fishing. Moose Jaw and Regina (cities located in the watershed) release treated wastewater effluent into the river upstream of the lakes (Hall et al. 1999; Quinlan et al. 2002). The major tributaries to the river from west to east are Wascana Creek, Last Mountain Creek, Loon Creek, Jumping Deer Creek, and Echo Creek (Water Security Agency 2013). During flood events, backflow from the Qu'Appelle is diverted through Last Mountain Creek into Last Mountain Lake. Echo Lake, Katepwa Lake, and Last Mountain Creek are equipped with control structures to regulate lake levels and fulfill environmental flow requirements in the Qu'Appelle River system. The air temperature in the watershed varies from a mean of –16 °C (in the winter) to a mean of 19 °C (in the summer). The mean annual potential evapotranspiration is 600 mm (Hall et al. 1999). The Qu'Appelle River flow is modified by interbasin water transfer from the South Saskatchewan River. At the Water Survey of Canada (WSC) hydrometric station 05JF001, Lumsden, the monthly mean flow of 4.3 m3/s occurs from November to March. However, during the spring season (March to April), the monthly mean discharge increases to 24.28 m3/s as a result of snowmelt.

The declining water quality in the watershed has been of concern to decision-makers and communities in the watershed. This study would help water managers get a better picture of the current and future status of the watershed and take efficient steps towards prudent environmental management in the region.

The proposed methodology of this study is schematically presented in Figure 1. The second-generation Canadian Earth System Model (CanESM2) and newly defined climate change scenarios, namely the Representative Concentration Pathways (RCPs), were adopted to simulate future climate conditions. The Statistical DownScaling Model (SDSM) was used to downscale the large-scale precipitation and temperature data generated by the CanESM2. In this respect, daily observed precipitation and temperature data were inputted into the model to generate climate change data for the years 2022–2050. Using the 1961–2020 period observed climate and hydrometric data, as well as collected water quality data during the 1992–2020 period, the ANN was applied to train networks to simulate the following.

Figure 1

The proposed framework of this study.

Figure 1

The proposed framework of this study.

Close modal

• The T2S network: Simulation of the ratio of snowfall-to-precipitation based on the temperature.

• The P2Q network: Simulation of the river flow/discharge based on the temperature and precipitation.

• The Q2DO network: Simulation of the DO concentration based on the river flow and temperature.

• The Q2TDS network: Simulation of the TDS concentration based on the river flow and temperature.

Finally, the generated climate change data during 2022–2050 were used as inputs to the ANN model to simulate future river flow as well as water quality indicators including the DO and TDS. The results under climate change scenarios were analyzed and compared with those under the historical scenario. Then, the discharge outputs of the ANNs were used as the inputs in the Indicator of Hydrologic Alteration (IHA) software to evaluate the hydrologic alteration of the Qu'Appelle River under climate change conditions.

The following historical climate data, hydrometric data, and water quality data were collected to perform the simulation: daily precipitation and temperature data during 1961–2020 were used to predict future climate conditions. Monthly hydrometric data from 1961 to 2020 were adopted to simulate river runoff and water quality parameters. Monthly DO and TDS data, collected from the Water Security Agency of Saskatchewan, during 1992–2020 were used to simulate the future water quality of the river.

The ANN

The ANN is a powerful technique to model complex non-linear relationships, especially where the explicit form of the relation between the variables is not known (Gallant 1993; Singh et al. 2009). The basic ANN model is comprised of three layers: (1) the input layer, where the required data are introduced and the weighted sum of the input is calculated; (2) the hidden layer, where the data are processed, and (3) the output layer, where the results of the model are generated. The layers are formed with the elements named neurons. The signals passing through the neurons in the process of training are modified by the weights and transfer/activation function. This process is repeated until the desired output (output layer) is reached (Govindaraju 2000; Schmidhuber 2015). In this study, the feed–forward neural networks and back-propagation learning were constructed to predict future DO, TDS, and discharge. In the feed–forward approach, the information move in one direction (forward) from the input layer to the output layer. After the output is predicted and the error obtained, the back-propagation algorithm redistributes the error back through the model, thus, the weights are adjusted accordingly while a new error value is obtained through the training process. This process is repeated based on the defined number of iterations until the minimum error is achieved. More detailed information about the ANN may be found in the literature (Chakraborty et al. 1992; DeSilets et al. 1992; Vemuri 1992; Singh et al. 2009; Sarkar & Pandey 2015; Bansal & Ganesan 2019). The characteristics of the ANNs are provided in Table 1. The number of neurons was determined based on sensitivity analysis. As already noted, four networks were defined in this study including T2S (to simulate the ratio of snowfall-to-precipitation based on the temperature), P2Q (to simulate the river flow/discharge based on the temperature and precipitation), Q2DO (to simulate the DO concentration based on the river flow and temperature), and Q2TDS (to simulate the TDS concentration based on the river flow and temperature). The inputs of the networks were selected based on the following factors: (1) former studies and information; (2) trial and error (building several networks and selecting the best ones); and (3) hydrology of the region.

Table 1

Networks’ characteristics in this study

NetworksInputOutputNumber of neurons
T2S Tt  
P2Q Tt, Rat, St, St−1, St−2, St−3, St−4, and St−6 Qt 
Q2DO Qt, Tt DOt 
Q2TDS Qt, Tt TDSt 
NetworksInputOutputNumber of neurons
T2S Tt  
P2Q Tt, Rat, St, St−1, St−2, St−3, St−4, and St−6 Qt 
Q2DO Qt, Tt DOt 
Q2TDS Qt, Tt TDSt 

Tt is the temperature during period t (°C), is the ratio of snowfall-to-total precipitation (%), Rat is the rainfall during period t (mm), St is the snowfall during period t (mm), Qt is the river flow during period t (m3/s), DOt is the dissolved oxygen during period t (mg/l), TDSt is the total dissolved solids during period t (mg/l).

To evaluate the performance of these networks, three performance criteria were considered: the coefficient of determination (R2), the root mean square error (RMSE), and the Nash–Sutcliffe efficiency (NSE) coefficient. These criteria were calculated as follows:
(1)
(2)
(3)
where n is the number of time steps, and are the observed and simulated variables, respectively; and represents the mean of the observed variable.

In this study, the ANN method was used as a powerful tool to simulate the water quantity and quality of the Qu'Appelle River. Several studies have taken advantage of the ANN to simulate water quality (Chen et al. 2010; Antanasijevic et al. 2014) and quantity (Dehghani et al. 2014; Kourgialas et al. 2015) of rivers.

Climate change modeling and scenarios

The RCPs were used as climate change scenarios in this study. The RCPs were introduced by the Intergovernmental Panel on Climate Change (IPCC) in its fifth Assessment Report (AR5) (IPCC 2013). This study adopted RCP 4.5 and RCP 8.5 as climate change scenarios. This study simulated future weather conditions using the second-generation CanESM2.

The CanESM2 encompasses ocean, atmosphere, sea ice, land, and carbon cycle models. The ocean component of the model is a version of the National Center for Atmospheric Research (NCAR) Community Ocean Model. This component with a horizontal resolution of 1.41° × 0.94° has 40 vertical levels from 10 m near the surface to 400 m in the deep ocean. The atmospheric component of the CanESM2 utilizes the spectral transform method. This component has a T63 horizontal resolution and 35 vertical levels. The sea ice component adopts cavitating fluid rheology combined with a thermodynamic formulation. The carbon response and feedback of the land and ocean system (combined carbon response and feedback parameter) are linearly related to the physical climate feedback parameter, representing how carbon-climate responses and feedback are interconnected (Yang & Saenko 2012; Williams et al. 2019).

Since such models use a large-scale computational resolution, their outputs need to be downscaled to improve the accuracy of the generated data. In order to downscale the CanESM2 outputs, the SDSM (Wilby et al. 1998, 1999, 2002) was adopted.

The SDSM connects statistical relationships between large-scale predictors and local climate variables based on the multiple linear regression method. The SDSM performs seven discrete functions to downscale data: (1) quality control of the observed data and data transformation; (2) screening of downscaling predictor variables; (3) calibration; (4) weather generation; (5) statistical analysis; (6) graphing model output; and (7) scenario generation using climate model predictors (Wilby et al. 2002).

The CanESM2–SDSM generated future precipitation and temperature data based on the historical data recorded in the watershed. In this study, three scenarios were defined. The first scenario simulated water quantity and quality without considering climate change impacts. However, scenarios 2 and 3 (RCP 8.5 and RCP 4.5) are associated with climate change conditions. In order to evaluate and quantify the impacts of climate change on river flow and water quality indicators (DO and TDS), the results under climate change scenarios were compared with those under the historical scenario. The scenarios are listed in Table 2.

Table 2

Defined historical and climate change scenarios

ScenarioDefinition
Historical Simulation model under historical observed data during 1961–2020. 
RCP 8.5 Simulation model under the RCP 8.5 scenario during 2022–2050. 
RCP 4.5 Simulation model under the RCP 4.5 scenario during 2022–2050. 
ScenarioDefinition
Historical Simulation model under historical observed data during 1961–2020. 
RCP 8.5 Simulation model under the RCP 8.5 scenario during 2022–2050. 
RCP 4.5 Simulation model under the RCP 4.5 scenario during 2022–2050. 

Hydrologic alteration simulation

The RVA (Richter et al. 1997) was used to estimate hydrologic alteration of the Qu'Appelle River. The Indicators of Hydrologic Alteration (IHA) software (The Nature Conservancy 2007) was employed to implement the RVA analysis. The IHA components specify within-year variation in streamflow based on a series of hydrologic attributes (33 IHA parameters/statistics). These attributes are organized into five groups. It is possible to use parametric statistics for RVA analysis and adjust the RVA boundaries, the recommended way to run an RVA analysis is to use the non-parametric defaults. Three categories of equal size are defined: The low alteration category contains all values less than or equal to 33%; the medium/moderate alteration category contains all values falling in the range of 34–67%; and the high alteration category contains all values greater than 67%.

The performance of the networks is presented in Table 3. According to the table, all networks performed quite well based on the RMSE, R2, and NSE performance criteria values. To prevent overfitting in the networks, the dataset was divided into the three parts of training, validation, and test. Since the performances of the networks in training and test sets are quite close to each other (Table 3), it can be indicated that no overfitting occurred in the networks. To further represent the performance of the networks, the QQ plots of the simulated data (outputs of the networks) relative to the observed data are shown in Figure 2. Data distribution of calibration, validation, and test are represented in the figure. As can be observed, there is a good fit between the simulated and observed data.

Table 3

Networks’ performance

Training
Test
NetworksRMSER2NSERMSER2NSE
T2S 0.006 (MCM) 0.98 0.97 0.007 (MCM) 0.98 0.97 
P2Q 0.03 (MCM) 0.89 0.81 0.04 (MCM) 0.87 0.80 
Q2DO 0.92 (mg/l) 0.83 0.76 0.99 (mg/l) 0.80 0.74 
Q2TDS 78 (mg/l) 0.87 0.83 101 (mg/l) 0.85 0.80 
Training
Test
NetworksRMSER2NSERMSER2NSE
T2S 0.006 (MCM) 0.98 0.97 0.007 (MCM) 0.98 0.97 
P2Q 0.03 (MCM) 0.89 0.81 0.04 (MCM) 0.87 0.80 
Q2DO 0.92 (mg/l) 0.83 0.76 0.99 (mg/l) 0.80 0.74 
Q2TDS 78 (mg/l) 0.87 0.83 101 (mg/l) 0.85 0.80 
Figure 2

QQ plot: (a) simulated snow (P2Q) vs. observed snow; (b) simulated discharge (T2S) vs. observed discharge; (c) simulated DO (Q2DO) vs. observed DO; and (d) simulated TDS (Q2TDS) vs. observed TDS.

Figure 2

QQ plot: (a) simulated snow (P2Q) vs. observed snow; (b) simulated discharge (T2S) vs. observed discharge; (c) simulated DO (Q2DO) vs. observed DO; and (d) simulated TDS (Q2TDS) vs. observed TDS.

Close modal

Compared with historical observed temperature data, the mean annual temperature increased by 5.5 and 8.53% under RCP 4.5 and RCP 8.5 scenarios, respectively. According to the climate change modeling results, an increase in temperature would be intensified from April to July. The mean temperature during these months would rise by 45.37 and 52.09% under the RCP 4.5 scenario and RCP 8.5 scenario, respectively. However, the mean temperature would decline from January to March (winter season). This reduction was found to be 30 and 26.5% under RCP 4.5 and RCP 8.5 scenarios, respectively. Therefore, it can be concluded that the watershed would experience hotter summers and colder winters from 2022 to 2050. This general increase in air temperature could play a role to affect water temperature, DO, and TDS concentrations. Figure 3 depicts a change in the mean annual temperatures under climate change scenarios from 2022 to 2050 relative to the historically collected data from 1965 to 2020.

Figure 3

Change in temperature under climate change relative to the historical observed data (TCHH1: change in temperature under the RCP 4.5 scenario; TCHH2: change in temperature under the RCP 8.5 scenario).

Figure 3

Change in temperature under climate change relative to the historical observed data (TCHH1: change in temperature under the RCP 4.5 scenario; TCHH2: change in temperature under the RCP 8.5 scenario).

Close modal

Furthermore, the climate change modeling results demonstrated that the average amount of precipitation would not change considerably during 2022–2050. Compared with observed total precipitation data, climate change would increase the mean annual precipitation by 3.72% under the RCP 4.5 scenario and 0.69% under the RCP 8.5 scenario. The average precipitation increased to a great extent from January to March, with 31.5% under the RCP 4.5 scenario and 31.64% under the RCP 8.5 scenario. The major decrease in future precipitation would occur from April to June. This decline was found to be 20.7 and 24.1% under RCP 4.5 and RCP 8.5 scenarios, respectively. Figure 4 shows the change in simulated average total precipitation under RCP 4.5 and RCP 8.5 scenarios relative to that of historical observed data.

Figure 4

Change in precipitation under climate change relative to the historical observed data (PCHH1: change in precipitation under the RCP 4.5 scenario; PCHH2: change in precipitation under the RCP 8.5 scenario).

Figure 4

Change in precipitation under climate change relative to the historical observed data (PCHH1: change in precipitation under the RCP 4.5 scenario; PCHH2: change in precipitation under the RCP 8.5 scenario).

Close modal

To simulate snowfall during 2022–2050, an ANN (T2S) was trained to simulate the ratio of snow to precipitation (%) based on air temperature data obtained from climate change modeling. The output of the network is shown in Figure 5. Contrary to rainfall, snowfall cannot be changed to runoff in a short period of time. Therefore, to predict river flow more accurately, the snowfall must be simulated, especially in river basins or watersheds with considerable snowfall.

Figure 5

Relationship between the amount of snow from total precipitation and temperature.

Figure 5

Relationship between the amount of snow from total precipitation and temperature.

Close modal

The P2Q network was trained to estimate the river flow based on rainfall, snowfall (snowmelt), and temperature. A sensitivity analysis was performed to determine the dependence of river discharge in the present month on those of the previous months. It was found that river flow in a given month depended on the flow up to six preceding months (from t to t–6). Thus, six time lags were considered for the snow and inputted to the P2Q network. In the majority of months, the river flow would be increased under climate change conditions compared to that of the historical scenario. However, this increase in discharge is not significant. The mean annual river flow would be increased by 1.94 and 0.52% compared to that of the historical scenario under RCP 4.5 and RCP 8.5 scenarios, respectively. The greatest decrease in river flow occurred in June, with about 18% under the RCP 4.5 scenario and 21% under the RCP 8.5 scenario. Nevertheless, the mean annual river flow would be increased by 7.69 and 6.63% under RCP 4.5 and RCP 8.5 scenarios, respectively. Further analysis revealed that in the months with no precipitation, the effect of temperature on the river flow was quite insignificant, however, its major impacts were on the stored snow in that particular month. Figure 6 depicts the change in river discharge under RCP 4.5 and RCP 8.6 scenarios relative to that of the historical scenario.

Figure 6

Change in river flow under the climate change relative to the historical scenario. QCHH1: change in the river flow under the RCP 4.5 scenario; QCHH2: change in the river flow under the RCP 8.5 scenario.

Figure 6

Change in river flow under the climate change relative to the historical scenario. QCHH1: change in the river flow under the RCP 4.5 scenario; QCHH2: change in the river flow under the RCP 8.5 scenario.

Close modal

To assess the impacts of climate change on DO level concentration, the Q2DO network was trained to simulate DO under climate change and historical scenarios. According to the results, even though DO concentration would be increased in the majority of months, in most of the months where DO is relatively small, its concentration is expected to be declined. In this situation even if concentration levels of DO increase in the next month(s), it cannot contribute to solving the unsustainability of the river system. It is apparent that these small levels of DO can substantially harm the aquatic life in the watershed.

According to the outputs of Q2DO, the mean annual DO was found to be increased by 4.97 and 8.45% under RCP 4.5 and RCP 8.5 scenarios, respectively compared to that of the historical scenario. However, the mean annual DO is expected to be reduced from September to November such that, compared with the historical scenario, DO levels would decline by 15.7 and 15.87% under RCP 4.5 and RCP 8.5 scenarios, respectively. This reduction would adversely affect the sustainability of the river system/riverine ecosystem and may not be recovered in the next months when its concentration levels increase. Change in future DO concentration is demonstrated in Figure 7.

Figure 7

Change in DO concentration under climate change relative to the historical scenario (DOCHH1: change in DO level under the RCP 4.5 scenario; DOCHH2: change in DO level under the RCP 8.5 scenario).

Figure 7

Change in DO concentration under climate change relative to the historical scenario (DOCHH1: change in DO level under the RCP 4.5 scenario; DOCHH2: change in DO level under the RCP 8.5 scenario).

Close modal

In addition, the Q2TDS network was trained to simulate the TDS under historical and climate change scenarios based on river flow and temperature. The mean annual TDS concentration was predicted to decline by 3.52 and 3.56% under RCP 4.5 and RCP 8.5 scenarios, respectively. Similar to DO, this change in mean annual concentration level does not mean that water quality in the river would be improved. On the contrary, since TDS concentration would increase in some months, it is expected to play its part to harm the water quality of the river and aquatic life in the watershed. Obviously, the issue would be more pronounced in the months where DO concentration is rising while the TDS is decreasing. For instance, in June and from August to November the average TDS level is expected to be increasing (although slightly) and the average DO concentration level would be decreasing. Figure 8 shows the change in the mean annual TDS concentration under RCP 4.5 and RCP 8.5 scenarios.

Figure 8

Change in TDS concentration under climate change relative to the historical scenario (TDSCHH1: change in TDS level under the RCP 4.5 scenario; TDSCHH2: change in TDS level under the RCP 8.5 scenario).

Figure 8

Change in TDS concentration under climate change relative to the historical scenario (TDSCHH1: change in TDS level under the RCP 4.5 scenario; TDSCHH2: change in TDS level under the RCP 8.5 scenario).

Close modal

Table 4 lists degrees of the hydrologic alteration as well as their classes for the defined scenarios. According to the RVA result, even though the average hydrologic alteration is expected to place within the low alteration class under climate change conditions, this alteration would increase by 3.53 and 4.38% under RCP 4.5 and RCP 8.5 scenarios, respectively.

Table 4

Degrees of hydrologic alteration and their classes under all scenarios

ScenarioHistorical D (%)ClassRCP 4.5 D (%)ClassRCP 8.5 D (%)Class
Group 1       
 October 23 37 37 
 November 23 37 37 
 December 23 23 37 
 January 23 23 23 
 February 11 13 23 
 March 11 13 23 
 April 11 13 
 May 11 
 June 11 13 13 
 July 11 13 13 
 August 23 23 23 
 September 23 27 32 
Group 2       
 1-day minimum 34 38 32 
 3-day minimum 34 38 38 
 7-day minimum 34 34 22 
 30-day minimum 22 34 34 
 90-day minimum 22 34 34 
 1-day maximum 22 22 22 
 3-day maximum 22 22 19 
 7-day maximum 22 34 22 
 30-day maximum 19 22 22 
 90-day maximum 19 20 34 
 Base flow index 48 48 48 
Group 3       
 Date of minimum 21 27 31 
 Date of maximum 27 28 31 
Group 4       
 Low-pulse count 26 35 35 
 Low-pulse duration 26 26 26 
 High-pulse count 39 39 42 
 High-pulse duration 39 39 41 
Group 5       
Rise rate 48 42 42 
Fall rate 48 49 49 
Number of reversals 50 50 50 
Average hydrologic alteration (%) 25.38 L 28.91 L 29.75 L 
ScenarioHistorical D (%)ClassRCP 4.5 D (%)ClassRCP 8.5 D (%)Class
Group 1       
 October 23 37 37 
 November 23 37 37 
 December 23 23 37 
 January 23 23 23 
 February 11 13 23 
 March 11 13 23 
 April 11 13 
 May 11 
 June 11 13 13 
 July 11 13 13 
 August 23 23 23 
 September 23 27 32 
Group 2       
 1-day minimum 34 38 32 
 3-day minimum 34 38 38 
 7-day minimum 34 34 22 
 30-day minimum 22 34 34 
 90-day minimum 22 34 34 
 1-day maximum 22 22 22 
 3-day maximum 22 22 19 
 7-day maximum 22 34 22 
 30-day maximum 19 22 22 
 90-day maximum 19 20 34 
 Base flow index 48 48 48 
Group 3       
 Date of minimum 21 27 31 
 Date of maximum 27 28 31 
Group 4       
 Low-pulse count 26 35 35 
 Low-pulse duration 26 26 26 
 High-pulse count 39 39 42 
 High-pulse duration 39 39 41 
Group 5       
Rise rate 48 42 42 
Fall rate 48 49 49 
Number of reversals 50 50 50 
Average hydrologic alteration (%) 25.38 L 28.91 L 29.75 L 

D, degree of hydrologic alteration; L, low alteration; M, medium alteration; H, high alteration.

Further analysis of statistical variables revealed that extreme values in the temperature, precipitation, river discharge, DO, and TDS were intensified under climate change conditions. This change in maximum and minimum values is quite substantial. Moreover, the values associated with standard deviation indicated that, as already mentioned, the patterns of precipitation and temperature would change significantly resulting in a heterogeneous change in the pattern of river flow and water quality indicators. Figure 9 depicts a change in statistical values under climate change compared to those of historical data and historical scenarios.

Figure 9

Change in statistical data under climate change relative to the historical data and scenario.

Figure 9

Change in statistical data under climate change relative to the historical data and scenario.

Close modal

Hosseini et al. (2017) investigated the impact of climate change on the water quality of the ‘Upper’ Qu'Appelle River, while the present study analyzed this impact on the water quality of the ‘Lower’ Qu'Appelle River using a different viewpoint and methodology. However, some results of the present study are generally in agreement with that of Hosseini et al. (2017). For instance, similar to this study, they concluded that the average temperature and discharge would be increased under climate change conditions, and the water quality of the river would be declined.

In this paper, the impact of climate change on the flow and water quality of the Qu'Appelle River located within the Lower Qu'Appelle River Watershed in Saskatchewan, Canada was studied. First, the CanESM2 along with the SDSM was applied to predict future precipitation and temperature data from 2022 to 2050 under RCP 4.5 and RCP 8.5 scenarios. Next, using the observed data inputted in the ANN, the networks were trained to simulate the ratio of snowfall-to-precipitation (T2S model), river flow (P2Q model), DO (Q2DO model), and TDS (Q2TDS model). Finally, using the generated climate change data, the ANN simulated future river flow as well as the DO and TDS concentrations during the 2022–2050 period. Then, the hydrologic alteration under all scenarios was quantified and compared. The results under RCP 4.5 and RCP 8.5 scenarios were analyzed and compared with those of the historical scenario. During the above mentioned period, it was found that:

  • The mean annual temperature would be increased by 5.50–8.54%. It is generally expected to have hotter summers and colder winters.

  • The mean annual precipitation would be increased by 0.69–3.79%. However, climate change would alter precipitation patterns. It is expected to have more precipitation in the months with greater precipitation amount and less precipitation in the months with smaller precipitation amount.

  • The mean river flow was found to be increased by 0.52–1.94%. Although, change in precipitation and temperature patterns results in a heterogeneous change in the river flow pattern.

  • Even though the mean DO and TDS concentration would be improved by 4.97–8.45% and 3.53–3.56%, respectively, under climate change, it does not mean that the water quality of the river would be enhanced in the future. Climate change is expected to have serious degrading impacts on DO and TDS levels in some months causing deterioration in the sustainability of the river system and ecological health of the region. This unsustainability cannot be recovered even when these indicators improve in the next months.

  • Climate change is expected to increase hydrologic alteration of the Qu'Appelle River by 3.53–4.38%. Hence, water managers must pay more attention to the environmental flow requirements of the river.

The current research utilized a framework with the ANN for the analysis of climate change impacts on flow and water quality parameters. The framework was calibrated and validated using a watershed in the Prairies of Canada, which could be further tested and applied to other regions.

We believe that this study draws water managers’ attention to environmental considerations alongside climate change impacts and is a step forward by offering a framework applicable in any river basin/watershed which is managed to satisfy human and environmental/ecological needs.

The authors thank the financial support from the IWHR Open Research Fund of State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin (IWHR-SKL-201916), the Agriculture Development Fund (ADF) of Saskatchewan, and the NSERC Discovery Grant of Canada.

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

The authors declare there is no conflict.

Caldwell
P.
,
Segura
C.
,
Gull Laird
S.
,
Sun
G.
,
McNulty
S. G.
,
Sandercock
M.
,
Boggs
J.
&
Vose
J. M.
2015
Short-term stream water temperature observations permit rapid assessment of potential climate change impacts
.
Hydrological Processes
29
(
2196
),
2211
.
Chakraborty
K.
,
Mehrotra
K.
,
Mohan
C. K.
&
Ranka
S.
1992
Forecasting the behaviour of multivariate time series using neural networks
.
Neural Networks
5
,
961
970
.
Chen
D.
,
Lu
J.
&
Shen
Y.
2010
Artificial network modelling of concentrations of nitrogen, phosphorus and dissolved oxygen in a non-point source polluted river in Zhejiang Province, southeast China
.
Hydrological Processes
24
,
290
299
.
Dehghani
M.
,
Saghafian
B.
,
Saleh
F. N.
,
Farokhnia
A.
&
Noori
R.
2014
Uncertainty analysis of streamflow drought forecast using artificial neural networks and Monte-Carlo simulation
.
International Journal of Climatology
34
(
4
),
1169
1180
.
DeSilets
L.
,
Golden
B.
,
Wang
Q.
&
Kumar
R.
1992
Predicting salinity in the Chesapeake Bay using backpropogation
.
Computer Operation Research
19
(
3/4
),
277
285
.
Ekwueme
B. N.
&
Agunwamba
J. C.
2021
Trend analysis and variability of air temperature and rainfall in regional river basins
.
Civil Engineering Journal
7
(
5
), 816–826.
Feistel
R.
&
Hellmuth
O.
2021
Relative humidity: a control valve of the steam engine climate
.
Journal of Human, Earth, and Future
2
(
2
), 140–182.
Gallant
S. I.
1993
Neural Network Learning and Expert Systems
.
The MIT press
,
Massachusetts, USA
.
Hall
R. I.
,
Leavitt
P. R.
,
Quinlan
R.
,
Dixit
A. S.
&
Smol
J. P.
1999
Effects of agriculture, urbanization, and climate on water quality in the northern Great Plains
.
Limnology and Oceanography
44
,
739
756
.
Hassanjabbar
A.
,
Saghafian
B.
&
Jamali
S.
2018
Multi-reservoir system management under alternative policies and environmental operating conditions
.
Hydrology Research
49
(
6
),
1817
1830
.
Hassanjabbar
A.
,
Saghafian
B.
,
Sane
I.
&
Jamali
S.
2021
Reservoir management under different operating water levels, operation policies and climate change conditions
.
Proceedings of the Institution of Civil Engineers-Water Management
.
Hassanzadeh
E.
,
Strickert
G.
,
Morales-Marin
L.
,
Noble
B.
,
Baulch
H.
,
Shupena-Soulodre
E.
&
Lindenschmidt
K. E.
2019
A framework for engaging stakeholders in water quality modeling and management: application to the Qu'Appelle River Basin, Canada
.
Journal of Environmental Management
231
,
1117
1126
.
He
J.
,
Ryan
M. C.
&
Valeo
C.
2015
Changes in Water Quality Characteristics and Pollutant Sources Along A Major River Basin in Canada
.
Hosseini
N.
,
Johnston
J.
&
Lindenschmidt
K.-E.
2017
Impacts of climate change on the water quality of a regulated prairie river
.
Water
9
(
3
),
199
.
Intergovernmental Panel on Climate Change (IPCC)
2013
Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
.
Cambridge University Press
,
Cambridge
,
United Kingdom
.
New York, NY, USA
.
Jamali
S.
,
Abrishamchi
A.
&
Madani
K.
2013
Climate change and hydropower planning in the Middle East: implications for Iran's Karkheh hydropower systems
.
Journal of Energy Engineering
139
(
3
),
153
160
.
Jonnalagadda
S. B.
&
Mhere
G.
2001
Water quality of the Odzi River in the eastern highlands of Zimbabwe
.
Water Research
35
(
10
),
2371
2376
.
Quinlan
R.
,
Leavitt
P. R.
,
Dixit
A. S.
,
Hall
R. I.
&
Smol
J. P.
2002
Landscape effects of climate, agriculture, and urbanization on benthic invertebrate communities of Canadian prairie lakes
.
Limnology and Oceanography
47 (2),
378
391
.
Reid
A. J.
,
Carlson
A. K.
,
Creed
I. F.
,
Eliason
E. J.
,
Gell
P. A.
,
Johnson
P. T.
,
Kidd
K. A.
,
MacCormack
T. J.
,
Olden
J. D.
,
Ormerod
S. J.
&
Smol
J. P.
2019
Emerging threats and persistent conservation challenges for freshwater biodiversity
.
Biological Reviews
94
(
3
),
849
873
.
Richter
B. D.
,
Baumgartner
J. V.
,
Wigington
R.
&
Braun
D. P.
1997
How much water does a river need?
Freshwater Biology
37
(
1
),
231
249
.
Saskatchewan Watershed Authority (SWA)
,
Kulshreshtha
S. N.
,
Bogdan
A.
&
Nagy
C.
2012
Present and Future Water Demand in the Qu'Appelle River Basin. Saskatchewan Water-shed Authority
.
Schmidhuber
J.
2015
Deep learning in neural networks: an overview
.
Neural Network
61
,
85
117
.
Shah
L. A.
,
Khan
A. U.
,
Khan
F. A.
,
Khan
Z.
,
Rauf
A. U.
,
Rahman
S. U.
,
Iqbal
M. J.
,
Ahmad
I.
&
Abbas
A.
2021
Statistical significance assessment of streamflow elasticity of major rivers
.
Civil Engineering Journal
7
(
5
), 893–905.
Sibanda
T.
,
Chigor
V. N.
,
Koba
S.
,
Obi
C. L.
&
Okoh
A. I.
2014
Characterisation of the physicochemical qualities of a typical rural-based river: ecological and public health implications
.
International Journal of Environmental Science and Technology
11
(
6
),
1771
1780
.
Singh, K. P., Basant, A., Malik, A. & Jain G. 2009 Artificial neural network modeling of the river water quality - A case study. Ecological Modelling 220 (6), 888–895
.
Smol
J. P.
2009
Pollution of Lakes and Rivers: A Paleoenvironmental Perspective
.
John Wiley & Sons
.
The Nature Conservancy
2007
Indicators of Hydrologic Alteration, Version 7, User's Manual
. The Nature Conservancy, Charlottesville, Virginia.
Vemuri
V. R.
1992
Artificial Neural Networks: Concepts and Control Applications
.
IEEE Computer Society Press
,
California
.
Water security Agency
2013
Lower Qu'Appelle River Watershed Plan, Technical Report
.
The Water Security Agency, Saskatchewan, Canada
.
Wilby
R. L.
,
Hassan
H.
&
Hanaki
K.
1998
Statistical downscaling of hydrometeorological variables using general circulation model output
.
Journal of Hydrology
205
(
1–2
),
1
19
.
Wilby
R. L.
,
Dawson
C. W.
&
Barrow
E. M.
2002
SDSM – a decision support tool for the assessment of regional climate change impacts
.
Environmental Modelling & Software
17
(
2
),
147
159
.
Williams
R. G.
,
Katavouta
A.
&
Goodwin
P.
2019
Carbon-cycle feedbacks operating in the climate system
.
Current Climate Change Reports
5
,
282
295
.
Yang
D.
&
Saenko
O. A.
2012
Ocean heat transport and its projected change in CanESM2
.
Journal of Climate
25
(
23
),
8148
8163
.
Zango
B. S.
,
Seidou
O.
,
Sartaj
M.
,
Nakhaei
N.
&
Stiles
K.
2021
Impacts of urbanization and climate change on water quantity and quality in the Carp River watershed
.
Journal of Water and Climate Change
13 (2), 786–816, jwc2021158
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).