The impact of climate change is directly felt on atmospheric variables, which in turn affects the quality and dynamics of internal freshwater bodies that are crucial for human survival. The prediction of climate change impacts on internal waters, like reservoirs and lakes, is critical for their conservation (Ahmed et al., 2022) and effective management (Bárcenas-García et al., 2022). It is crucial to have a forecasted perspective on the future of aquatic environments in order to develop plans for alleviating the consequences of widespread changes, such as extreme streamflow (Mantua et al., 2010) deterioration of water quality, harmful algal blooms (Griffith & Gobler, 2020), severe deoxygenation (Mosley, 2015), loss of ecosystems (McAllister et al., 1997) and changes in fish biodiversity (Su et al., 2021).
To evaluate the potential impact of upcoming climate shifts on aquatic ecosystems, three key factors are measured in lake water quality: DO, water temperature, and PO4. Among these, temperature plays the most crucial role as it can be easily measured and has a significant influence on thermal stratification and various other parameters. These parameters include the metabolic function of aquatic life, photosynthesis, toxicity of compounds, dissolved oxygen levels, conductivity, pH, and density. Furthermore, PO4 contributes to eutrophication in reservoirs (Müller et al., 2019). Similarly, DO is a crucial metric for assessing water quality and serves as a fundamental parameter in the classification of reservoirs (Abbasi & Abbasi, 2012; Orlob, 1983). The balance of essential nutrients in reservoirs, like phosphorus, nitrogen, and iron, is regulated by the availability of oxygen. Decreases in oxygen levels in water result in a variety of adverse ecological and biological effects (Chowdhary et al., 2020). Additionally, when the DO levels in reservoirs are too low, the water quality does not meet the basic requirement for drinking water. Many aquatic organisms need enough DO to carry out their metabolic activities. When the water temperature and acidity increase, the DO level decreases, which affects the water quality and the ecosystem of the reservoirs (Breitburg et al., 2018).
Iran, similar to several other Middle Eastern nations, has been grappling with a pronounced water scarcity issue for the past twenty years (Dubreuil et al., 2013; Michel, 2017; Saemian et al., 2022). The desiccation of rivers, lakes, and wetlands serves as evident evidence of the decline in surface water resources within this country (Alborzi et al., 2018; Kaveh Madani, 2021; Saemian et al., 2020; Saemian et al., 2022). The alterations in water availability can be ascribed to the combined impacts of climate change and human activities (Ashraf et al., 2019). Given that over 80% of the nation's territory experiences a mainly arid, except the northern coastal areas and parts of western Iran, any notable shift in water availability can have significant socio-economic and environmental consequences (Madani, 2014), transforming Insufficient water supply into a threat to country's security (K Madani, 2021).
Climate studies employ general circulation models (GCMs) as extremely complete devices to predict the future of water bodies' ecosystems. These models allow the prediction and analysis of future climate elements under different scenarios of greenhouse gas emissions, giving useful information (Meinshausen et al., 2020).
To produce detailed climate data for a small area, the predictions of broad-scale climate factors from GCMs have to be scaled down to smaller grids. Methods for downscaling can be divided into two categories: dynamical downscaling and statistical downscaling. Statistical downscaling, when contrasted with dynamical downscaling, requires less computational resources and offers quicker processing times (Rastogi et al., 2022). Applying a transfer function in the form of a downscaling method via regression offers a solid statistical strategy. The goal of this method is to identify the nonlinear relationship between the broad-scale GCM results, acting as predictors, and the local measurements, serving as predictands (Retsch et al., 2022). A support vector machine (SVM) can be also employed to establish such a relationship (Jimenez et al., 2020). Chen et al. conducted a comparison between the performance of SVM, linear multiple regression, and SDSM. The outcomes of the downscaling process revealed that SVM exhibited greater accuracy in predicting daily precipitation compared to Statistical Down-Scaling Model (SDSM) and linear multiple regression (Chen et al., 2010). (Duhan & Pandey, 2015) discovered that downscaling models based on SVM demonstrate superior performance in simulating both maximum and minimum temperatures compared to models based on MLR and ANN. Different methodologies have been used to downscale precipitation and temperature in many studies, but other weather components that are usually needed for water quality modeling, like wind speed/direction, dew point, and cloudiness, have received less attention (Yaghouti et al., 2023). Nevertheless, (Alizadeh et al., 2019) have formulated a method relying on distribution parameters to downscale wind direction and speed derived from GCMs. In addition to that, (Kawasaki, 2015) utilized the SDSM to downscale daily cloudiness as well as daily temperature. Similarly, (Cheng et al., 2008) employed an approach based on regression to downscale various meteorological variables from GCM models, such as dew point (Dp), total cloud cover, air temperature (Ta), wind velocity components, air pressure and mean sea-level, achieving satisfactory levels of accuracy.
Forecasting the future dynamics of the water circulation system in hydrological systems across various climate variability scenarios presents a substantial and challenging task (Iranmanesh et al., 2021). The investigation of climate change's influence on runoff requires the utilization of diverse input variables in the developed methodologies and models (Das & Nanduri, 2018). Hence, the availability of the observed data plays a crucial role in the process of choosing appropriate models (Raju & Kumar, 2020). Among the extensive range of hydrological models, scholars emphasize the importance of SWAT (Fan & Shibata, 2015) and IHACRES (Azadi et al., 2021). Within water planning and assessment frameworks, Moghadam et al. combined IHACRES with the Groundwater Modeling System (GMS) to model runoff results and predict water levels. This integration facilitates the examination of diverse water management scenarios (Moghadam et al., 2022).
Various indicators have been employed to characterize the effects of climate pattern changes on the water quality of reservoirs. These indicators include temperature of water (Quan et al., 2022), DO levels (Zhang et al., 2015), total phosphorus (TP) (Zhang et al., 2019), and initial process of generating organic matter (Shao et al., 2023). When examining the impacts of climate change in distant future scenarios, numerical models act as the primary method of assessment (Moe et al., 2016). Before utilizing models, it's crucial to verify them using either on-site observations (Piccioni et al., 2022; Shahzad et al., 2021) or data from remote sensing (Yáñez-Morroni et al., 2023). One study utilized the MINLAKE96 model, a vertical lake water quality model, to predict thermally stratified periods under a climate scenario. They discovered that atmospheric carbon dioxide concentrations doubled (Stefan et al., 2001). Several studies, such as (Adrian et al., 2009) have confirmed the worldwide trend of increased temperatures in lakes and reservoirs. Yaghouti et al., (2023) later utilized the CE-QUAL-W2 to evaluate a methodological framework for examining the impacts of climate change on water temperature and dissolved oxygen levels under three prospective climate scenarios in Mississippi Lake, Southeastern Ontario, Canada. The outcomes indicate substantial alterations in temperature and oxygen availability for Mississippi Lake. Morales-Marin et al. (2021) demonstrated, using CE-QUAL-W2, that the future climate change scenarios, which result in increased temperature in Lake Diefenbaker, would lead to a rise in the concentration of phosphorus in the lake, consequently intensifying algal blooms. Furthermore, both an empirical model and SWAT were used to examine water quantity and quality at a catchment scale within a river in Japan under the influence of climate change. The primary focus of the authors was on changes in precipitation and temperature (Fan & Shibata, 2015).
The primary objective of the current investigation is to develop a comprehensive understanding of the changes in water quality parameters under various future scenarios by assessing the impacts of climatic variables on thermal stratification, DO, and PO4 levels in the Karun IV Reservoir located in Charmahal & Bahtyari Province, Iran, during the climate change period spanning from 2081 to 2100. To achieve this, the study employs the CanESM5 model (under emission scenarios SSP1-1.9 and SSP5-8.5) in conjunction with the CE-QUAL-W2 model. The obtained results are then compared with data from a baseline period ranging from 1995 to 2014. The main differences between the present research and the previous studies are as follows:
-
Enhancing climate data downscaling using the optimized machine learning techniques.
-
Evaluating the pre-suggested modeling for a large dam (an artificial lake with a maximum depth of greater than 100 m) with regulated water abstraction rather than natural shallow lakes (with a maximum depth of less than 20 m) without outlet control.
-
Using real operational data applied for power plant of the dam.
-
Simultaneously modeling water temperature, DO, and PO4, as the most important characteristic of eutrophication.
-
Applying a rainfall-runoff approach for estimating inflow of the lake.
-
Exploring the impact of climate change specifically in arid and semi-arid climates.