Application of system dynamics model for reservoir performance under future climatic scenarios in Gelevard Dam, Iran

Water resource planners raise concerns to develop a comprehensive plan for a successful strategy to address water scarcity and meet specific needs that will emerge due to climate change in the future (Yang et al. 2008). Climate change can affect air temperature, rainfall, runoff, and water demand in the domestic, industrial, and agricultural sectors. These changes must be considered in water resource planning to better meet the future needs and expectations of society. Therefore, climate change can be a determining factor in planning and managing water resources.

Many researchers evaluated the effects of climate change on water resource systems. Recent studies report that in the future we will observe the impact of climate change on water resources. According to investigations, climate change will reduce river discharge and the system stability index (Ahmadi et al. 2015; Ehteram et al. 2018; Hakami-Kermani et al. 2020). Climate change is predicted to increase agricultural water demands and reduce system reliability (Joyce et al. 2011; Ehteram et al. 2018; Salman et al. 2020). Due to the inevitable changes, additional construction of dams and increasing the capacity of reservoirs can be a solution for sustainable water supply (Chen et al. 2016). Traynham et al. examined the capacity of the Puget Sound regional water supply system to meet future demands concerning climate change and population growth over a 75-year horizon (Traynham et al. 2011). Their findings indicated that climate change would reduce system performance in the future, thus necessitating operational policies to meet future demands. Therefore, reviewing and reassessing the procedure policies of reservoir performance in the context of climate change, and various adaptation scenarios could provide useful information for decision makers to reduce the negative effects of climate change on the reliability of water resources (Nam et al. 2017; Ranzani et al. 2018; Rehana et al. 2020; Lee & Shin 2021). In addition, water demand management keeps water scarcity at an acceptable level and can be considered as a sustainable strategy for water resources management and a means to maintain economic growth and the ecological status (Xiao-jun et al. 2014). The agricultural sector is one of the largest water consumers, which means that a reduction of any size in this sector's water demands is the most appropriate way of reducing its vulnerability (Wu et al. 2013). Thus, improving irrigation efficiency and reducing water consumption in the agricultural sector is the best way to reduce deficiencies (Zarghami et al. 2016). Optimizing the cultivation pattern as a solution to the issue of high water consumption will increase the reliability of the system and reduce the vulnerability of water resources (Zamani et al. 2017).

With the increase of the water demand in the Neka River (NR) basin, GD is constructed to store and supply sustainable water on NR. According to studies, the planning of water resources for this dam was based on the prior data, but in recent years, due to reduced rainfall, the flow of the NR has decreased. For example, the average annual flow of NR was previously reported at 112 million cubic meter (MCM), but during the past 15 years, it has decreased to less than 90 MCM per year. Worse yet, in the period of 2013–2014 and 2014–2015, the annual volume of the NR has decreased by 45 and 37 MCM per year, respectively. Therefore, studying the impacts of global warming on climatic parameters and the flow of the NR is necessary.

Meteorological data used in this study consist of maximum air temperature, minimum air temperature, wind speed, sunshine hours, and rainfall, which were obtained from the Iran Meteorological Organization. To generate meteorological data for the period of 2020–2040, the time series of climate variables including monthly rainfall (mm), monthly maximum and minimum air temperatures (°C), monthly maximum and minimum relative humidity (%), monthly wind speed, and monthly sunshine hours (h) were downscaled using the support vector machine (SVM) model. SVM is a modern statistical learning theory in data-driven modeling (Zhou et al. 2017). The uniqueness of SVM is its structural risk minimization objective that balances the model's complexity against its fitting precision, instead of an empirical risk minimum (ERM) used by most intelligent algorithms that focuses mostly on fitting accuracy (Vapnik 1999). This model architecture greatly improves model generalization ability compared with ERM-based algorithms such as artificial neural networks (Zhou et al. 2017). In recent years, SVM has been used for hydrologic predictions such as precipitation (Kisi & Cimen 2012; Hou et al. 2017). For downscaling, National Centers for Environmental Prediction (NCEP) variables are used, including 26 atmospheric variables, which have a high correlation with historical climatic parameters. Then, the ability of the SVM model to simulate climatic parameters of the historical period is evaluated. Finally, the representative concentration pathway (RCP) is used to project the climatic parameters of the future period. The RCPs are based on the groupings of economic, technological, demographic, policy, and future institutional challenges of mitigation and adaptation (Babur et al. 2016). The benefit of RCPs is their better resolution that helps in performing regional and local comparative studies (Höök et al. 2010). In this study, the output from the second-generation Canadian Earth System Model (CanESM2) is utilized for downscaling. The CanESM2 is developed by the Canadian Center for Climate Modeling and Analysis (CCCma) of Environment Canada (Arora & Boer 2014). Three RCP scenarios including RCP2.6, RCP4.5, and RCP8.5 are utilized to quantify the variations in the climatic parameters over the Neka Basin for the time period of 2020–2040.

The KC coefficient is extracted from the FAO report. IR is irrigation requirement (mm d⁻¹), Pe is effective rainfall (mm d⁻¹), and Ea is irrigation efficiency.

The cultivation pattern optimization function is performed using the modeling language and optimizer (LINGO) software.

The data obtained from Regional Water Company of Mazandaran province are used to estimate the demand for domestic water and industry. To determine the demand for drinking water in the upcoming years, according to the per capita water consumption and the rate of population growth in those years, the amount of drinking water demand in the coming years is calculated on a monthly basis. Furthermore, water demand in the industrial sector is estimated based on the information gathered for the study area for the base and future periods.

The water demand for the environmental sector is calculated using the Montana method (Tennant 1976). Environmental water demand is considered identical in different years. The long-term average monthly inflow to the reservoir is then calculated. In the first and second half of the year, respectively, 10 and 30% of the inflow are determined for environmental needs.

In this study, observed weather data (i.e., monthly rainfall, monthly maximum and minimum air temperatures, monthly maximum and minimum relative humidity (RH), monthly wind speed, and monthly sunshine hours) for the period of 1985–2005 are used to calibrate, and the period 2006–2019 is incorporated to validate the SVM model. The results of the performance evaluation demonstrate that the performance of climatic parameters downscaling by the SVM model is in good agreement with the prior observed climatic parameters (Table 1) (Babolhakami et al. 2020).

The cultivation pattern includes rice, canola, citrus, wheat, barley, and cotton crops. Reservoir performance simulation shows that GD will not be able to supply the required water to the existing cultivation pattern. The simulation reports that in the next period, the reservoir will not be able to supply the required water to the drinking, environment, industry, and agriculture sectors. Therefore, to improve the performance of the reservoir of GD and reduce the amount and severity of water shortages in different sections, the cultivation pattern is optimized. The OCP is determined using LINGO for different scenarios of climate change, according to the volume of available water, the yield of cultivated crops, and the total area under cultivation. Table 3 shows the OCP for RCP8.5 as an example. Generally, the results show that by changing the cultivation pattern, the area under irrigated crops will decrease and the area under cultivation of rainfed crops will increase. According to Table 3, the area under rice cultivation will also decrease by more than 50% and the area under citrus cultivation will increase by 100%. Also, the area under cultivation of dryland wheat and rapeseed crops shows up to 200% growth. In contrast, the area under barley cultivation will decrease and the area under irrigated cotton cultivation will remain unchanged.

Modeling is done in a way so as not to have a statistically significant difference between reality and model. Accordingly, several tests are used to compare simulated and observed values. It is important to mention that the forecast data of meteorological parameters have been calibrated and verified by the SVM model (Table 1). Also, the prediction of runoff values by the IHACRES model has been calibrated and validated (Table 2). To validate the Vensim model simulations, behavior repetition, unit compatibility, and structure evaluation tests are used. In the behavior repetition test, the output of the model for the historical period is simulated and compared with the measurement data. The results of this test for the period of 2006–2014 show that the model was able to predict the output value with an RMSE of 5.8 MCM per year and R² = 0.89. In the unit compatibility test, the measurement units for each variable in the model are checked. The structure evaluation test examines the correctness of the model in terms of its compatibility with the dynamic stock and flow diagram (Figure 2). This test focuses on establishing the physical and governing laws of the model (Sterman 2002).

The performance of GD in climate change conditions and existing cultivation pattern is simulated by the GD model. The priorities of water demand are set in the following order: domestic, environment, industry, and agriculture sectors. The results show that if the current cultivation pattern continues, water shortages will occur in all sectors in the coming period. Therefore, to reduce water stress and improve the performance of GD, the optimal cultivation model is applied.

The results conclude that with the continuation of the cultivation pattern's current trend, in an optimistic climate (RCP2.6) for providing water supply, domestic, environment, industry, and agriculture will be supplied without shortage for 7 years, and in the most pessimistic climate (RCP8.5) regarding the water demand of all sectors, the supply will be sufficient only in the year 2022, and in other years, there will be water shortages in all sectors. If the optimal crop pattern model is applied, in the optimistic climate, the demand for all sectors will be met until the year 2031 without any shortages, and in the pessimistic climate, the demand will be met for 4 years and the water required for all sectors will be fully supplied.

The results of the Vensim model on the performance of GD reservoir in supplying the water required by different sectors show that if the existing cropping pattern continues, for the different climatic scenarios of RCP2.6, RCP4.5, and RCP8.5 in the next 7, 2, and 2 years, respectively, there will be no supply problem, and then there will be a water shortage crisis in the region immediately. If the OCP is applied, based on the different climatic scenarios of RCP2.6, RCP4.5, and RCP8.5, providing water for the region will not be difficult in the next 11, 5, and 4 years, respectively, and after that, water shortage will occur gradually. Therefore, applying an OCP reduces the severity of water shortage.

Table 4 depicts the values of water supply evaluation indexes to different sections and the total water demand of GD in the conditions of continuing the existing cultivation pattern and applying the OCP in climatic scenarios. These indexes are calculated based on Equations (15)–(19). Findings show that if the optimal model is applied, the five indicators of reliability, vulnerability, resilience, sustainability, and deficiency in all sectors of water consumption will be relatively improved. If the existing cropping pattern continues, the index of the reliability of water supply from the reservoir under RCP2.6, RCP4.5, and RCP8.5 climatic scenarios would be 0.67, 0.53, and 0.52, respectively, which will be improved by applying the optimal cropping pattern by 11, 11, and 10%, respectively. The vulnerability index of the requested water supply from the reservoir in the case of applying the existing cultivation pattern based on the RCP2.6, RCP4.5, and RCP8.5 climatic scenarios will be 0.16, 0.15, and 0.15, respectively, which will be possibly improved by applying the OCP by an increase of 2% in all climate scenarios. The resilience index of water demand from the reservoir will be 0.07, 0.06, and 0.06, respectively. By continuing the existing cultivation pattern based on the RCP2.6, RCP4.5, and RCP8.5 climatic scenarios, which will be improved by applying the OCP. It will reach 0.06, 0.06, and 0.05, respectively. The sustainability indexes of the GD reservoir system in the water supply of different sections based on RCP2.6, RCP4.5, and RCP8.5 climatic scenarios are 10, 7, and 7%, respectively, which will possibly have an increase of 13, 10, and 10% by applying the OCP, indicating a 3% increase. The results conclude that if the current cultivation pattern continues in the future, the average percentage of water supply demands being provided by the reservoir of GD under climatic scenarios of RCP2.6, RCP4.5, and RCP8.5 will be 74, 65, and 63%, respectively. By applying the OCP, this index will increase by 11, 14, and 13%, respectively.

The results indicate that with the increasing air temperature and decreasing precipitation, the inflow to the dam reservoir will possibly decrease, so that in the period of 2020–2040, the inflow to the reservoir of GD under climatic scenarios of RCP2.6, RCP4.5, and RCP8.5 will be reduced by 15, 36 and 37%, respectively. Reducing the inflow to the reservoir of this dam will reduce the reliability of the reservoir in water supply, while increasing vulnerability of not meeting water demands. Also, the decrease in the river flow under climatic uncertainties reduces the capacity of the system to adapt to changing conditions, which reduces the resiliency of the system. Applying the optimal cultivation model to some extent improves the performance indicators of the dam reservoir in meeting the water needs of different sectors, but cannot stabilize the performance of the GD reservoir system. This indicates that the performance of GD in the future period will be unstable due to climate change and reduced inflow to the reservoir. So, it is suggested to review the planning for the operation of the GD reservoir or to consider other policies to reduce water demands in different sectors.

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

The authors declare there is no conflict.