Effects of climate change on streamflow in the Dez Basin of Iran using the IHACRES model based on the CMIP6 model Open Access

Climate change does not have uniform effects worldwide as arid and semi-arid regions will become drier, whereas humid regions will become wetter (Held & Soden 2006).

There are various methods for managing the uncertainty in climate predictions and increasing confidence in future forecasts. One of these methods involves selecting the best general circulation model (GCM) for simulating climatic variables in historical periods for future predictions. Among the available methods, probability-based approaches perform better in predicting hydro-climatic processes (Raftery et al. 2005; Madadgar & Moradkhani 2014)

GCMs, also known as global climate models, mathematically represent the physical processes in the Earth system, including the atmosphere, oceans, land surface, and the icy component. Various methods, ranging from simple downscaling techniques to more complex statistical regression models and weather generators, have been successfully employed to downscale statistical outputs of GCMs (Boé et al. 2006; Ghosh & Mujumdar 2008).

The impacts of climate change on river flows can be simulated by applying an appropriate hydrological model and using statistical predictions of meteorological variables as an input to the model. Since the early 1960s, various hydrological models, both physically- and conceptual-based ones, have been developed for simulating hydrological processes in watershed areas. Some notable examples include SLURP (Jain et al. 1998), SHE (Xevi et al. 1997), Soil and Water Assessment Tools (Arnold et al. 1998), and IHACRES (Jakeman et al. 1990; Liang et al. 1994).

Numerous studies (Abushandi & Merkel 2011; Thompson 2012; Zahabiyoun et al. 2013; Hawkins 2015; Xu & Luo 2015; House et al. 2016; Goodarzi et al. 2020; Fatehifar et al. 2021; Niazkar et al. 2023) have successfully employed various types of hydrological models to simulate climate change effects on the water balance of watershed areas. Among hydrological models, IHACRES stands out due to its minimum data requirements, ease of use, and cost-effectiveness in data preparation (Croke et al. 2005). Moreover, it can be applied to many watershed areas because it has been utilized for datasets with temporal resolutions ranging from 6 min to 1 month and for watersheds of varying sizes from 490 m² in China to 1,000 km² in England (Littlewood et al. 2007). Additionally, IHACRES is a simple rainfall-runoff model that employs only temperature and rainfall as input data to simulate watershed outputs. Compared to other models, it has straightforward produces providing reasonable results.

In this study, for the first time, the assessment of the climate change process and its impact on the runoff of the Dez watershed for the period 2015–2043 have been conducted by high-resolution downscaling of GCM data. It utilizes data from two climate models, i.e., BCC-CSM-MR and ACCESS-ESM1, and two climate scenarios, SSP5.8.5 and SSP2.4.5, using the LARS-WG climate model and the IHACRES model. The aim is to provide insights into the potential changes in the hydrological response of the Dez watershed under different climate scenarios.

The primary data used in this study include daily temperature, precipitation, and runoff observations from selected stations in the study area from 1986 to 2014. Due to the insufficient recorded data within the study area, rainfall and evaporation data from nearby stations were also utilized. To monitor runoff and temperature in the study area, three hydrometric stations (Tang-e-Panj Bakhtiari, Tang-e-Panj Sezar, and Taleh Zang) and three synoptic stations (Koohrang, Aligudarz, and Borujerd) were employed. The specifications of these six stations are provided in Table 1. The locations of these stations are depicted in Figure 2.

The annual average precipitation in the northern and eastern regions of the Dez watershed is higher compared to that of other areas within the watershed. In terms of the temporal distribution of precipitation, 48.8% of the total precipitation occurs in winter, which comprises 30.6% in autumn, 20.4% in spring, and only 0.2% in summer.

The Shared Socioeconomic Pathways (SSPs) are a new group of scenarios, which are part of the Coupled Model Intercomparison Project Phase 6 (CMIP6) developed for the Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC). These scenarios are designed to provide insights into SSPs. They describe potential alternative changes in social aspects, such as population, economic, technological, social, governance, and environmental factors based on integrated analyses of climate impacts, vulnerability, adaptation policies, and mitigation (Pielke & Ritchie 2021). Furthermore, SSPs explain future conditions based on five fundamental development pathways: SSP1 (Sustainability), SSP2 (Middle of the Road), SSP3 (Regional Rivalry), SSP4 (Inequality), and SSP5 (Fossil-Fueled Development) (O'Neill et al. 2017). Additionally, they are more up-to-date compared to the Representative Concentration Pathways of the Fifth Assessment Report (AR5). Also, they are essentially a combination of socioeconomic scenarios, which consider radiative forcing levels of 2.6, 4.5, 6.0, and 8.5 w/m² (Fix & Graß 2021).

To evaluate the impact of climate change on the flow of the Dez River Basin, the data of two climate models ACCESS-ESM1 and BCC-CSM-MR from the CMIP6 model with two scenarios SSP245 and SSP585 that were available in the study area were used. The study period for this study includes the base period (1986–2014) to the future period (2043–2015). Complete information about the models used in this research is presented in Table 2.

The raw output of GCMs with low spatial resolution is not advisable to be used as an input of hydrological models. The significant spatial scale difference between the computational grids or cells of GCMs and the local coordinates of weather stations within the study area is a challenge. In other words, the major limitations of the climate models are their limited spatial and temporal resolutions. In such circumstances, the common approach is to enhance the spatial diversity of GCM outputs by downscaling methods, both statistical and dynamic approaches (Olsson et al. 2017).

The LARS-WG model (Semenov & Barrow 1997) is a statistical tool developed in Budapest in 1990. It is used for downscaling climate data. In essence, it is one of the models that generate random weather data and is employed to produce daily precipitation, radiation, and maximum and minimum daily temperature values at a station under current and future climate conditions. By using the LARS-WG model, daily climate scenarios for a station can be generated and connected to various simulation models, such as those related to water resources and agriculture. Finally, the fifth version of the model has been utilized in various studies (Semenov et al. 1998).

The IHACRES model is introduced as a suitable tool for assessing water resources and addressing water-related issues in developing countries. Basically, it is a rainfall-runoff model with an adequate number of parameters that has been applied to various watershed regions with different climates (Jones & Hulme 1996). Therefore, it was employed to simulate daily flows in the Dez River Basin in this study. In addition, it uses climate scenarios to simulate the flow in the future period (Croke & Jakeman 2008).

While the IHACRES model provided acceptable determination coefficients during the calibration and validation stages, Figure 6 demonstrates a relatively lower capability in modeling peak flow values. This is one of the weaknesses of the IHACRES model, particularly for flood prediction. However, since the objective of the current study is not flood assessment and calculations are conducted at monthly scales, the IHACRES model has revealed the ability to estimate runoff volume with acceptable accuracy. Thus, the widespread application of the IHACRES model in water resource studies is because it estimates runoff volume with an acceptable precision.

Figure 9 depicts the box plots of the maximum monthly temperature for the baseline and future periods for the months of September, January, March, October, November, and December. Additionally, Table 5 presents the monthly average temperature simulated by various climate models under both scenarios for the future. The range of the average monthly temperature variations simulated by various climate models during the future period compared to those of the baseline period for the SSP2.4.5 scenario varies from 0.52 to 2.95 °C, and for the SSP5.8.5 scenario, it ranges from 0.65 to 3.21 °C. Generally, temperatures in the region are projected to increase in future compared to the baseline period. The annual temperature increase simulated during the future period relative to the baseline period resulting from various climate models under both SSP5.8.5 and SSP2.4.5 scenarios will be approximately 2.36 and 2.7 °C, respectively. Furthermore, the average temperature, when combining both climate models under both emission scenarios during the future period, is expected to increase by approximately 2.53 °C compared to that of the baseline period.

Table 6 displays the average percentage changes in simulated precipitation during the future period compared to the baseline period for the study basin. The range of percentage changes in the average precipitation for different months during the future period under the SSP245 scenario varies from 70.66 to 95.00%, whereas for the SSP585 scenario, it ranges between 89.78 and 90.84%, respectively.

The average percentage changes in simulated precipitation during the future period compared to that of the baseline period under the SSP585 and SSP245 scenarios are reported in Table 7. As shown, there is an annual increase in precipitation during the future period. The highest increase is associated with the SSP245 scenario and the BCC-CSM2-MR model, reaching 36.2% relative to the baseline period. Furthermore, precipitation is one of the most diverse meteorological elements, which is why there is less consistency between the results of different GCMs for precipitation compared to temperature. Also, the results of the simulated precipitation for different months achieved by various climate models under both SSP585 and SSP245 scenarios confirm that different GCMs do not produce very similar results for a specific month. Therefore, some models may suggest an increase, while others may yield a decrease in precipitation for some months in the future.

In the SSP585 scenario, which is considered the most pessimistic scenario with high greenhouse gas emissions and limited efforts to mitigate climate changes, the percentage change in monthly runoff during the future period compared to the baseline period ranges from −30 to +8%, with the highest changes occurring in the month of September and May, respectively. According to Figures 12(b) and 13(b), it is evident that the occurrence of the highest runoff levels during the baseline period occurred in September 1992. Conversely, in the future period, the peak runoff event predicted in July 2043. This indicates changes in precipitation patterns and temporal distribution of rainfall due to climate changes, which have led to alterations in the runoff pattern in the basin.

Tables 8 and 9 present the monthly percentage changes in simulated streamflow values for the future period relative to those of the baseline period and the annual percentage changes in streamflow during the future period compared to the baseline period. As shown, it is evident that there is an annual increase of 7% under the SSP245 scenario and 25% under the SSP585 scenario using the BCC-CSM2-MR model. However, for the ACCESS-ESM1 model, we expect an annual decrease (−16, −14%) in the amount of runoff. According to Table 7, the ACCESS-ESM1 model predicted lower precipitation values for the future period related to the baseline period compared to the BCC-CSM2-MR model. In other words, it resulted in a reduction in the flow rate unlike the BCC-CSM2-MR model. The increase in the streamflow, as predicted by the BCC-CSM2-MR model, can be attributed to its greater sensitivity to temperature compared to the ACCESS-ESM1 model, which has led to a greater decrease in precipitation for the future period compared to the baseline period using the ACCESS-ESM1 model.

According to the results of surveys conducted in selected stations, in most months of the year, the increase in temperature is expected in the coming periods, which will intensify with the passage of time. Also, the investigation of rainfall in the basin shows changes in the time pattern and amount of rainfall in future periods compared to the base period. In the months of January, February, March, April, and December, it will face an increase in precipitation compared to the base period, while the changes in the amount of precipitation in the months of June, July, August, and September have been insignificant. In the base period, the average maximum amount of monthly precipitation occurred in March, while in the future periods, due to the effect of climate change phenomenon on precipitation patterns in the region, the maximum amount is expected to be shifted to another month, so that in most months which is associated with an increase in precipitation, this increase is greater in August.

In this research, in order to investigate the changes in runoff in the coming period (2015–2043), firstly, the result of the amount of runoff from the Caesar, Bakhtiari, and Dez River sub-basins upstream of the Dez dam was calculated, and finally, the runoff of the entire basin upstream of the dam was calculated. The Dez is modeled at the bell trap station (watershed outlet). The investigation has shown that in the months with an increase in precipitation, an increase in runoff is also observed in the IHACRES model, and as a result, changes in the pattern of precipitation in the region have a significant effect on runoff. Also, due to the changes in the precipitation pattern, it can be expected that in the future period (2015–2043), the runoff pattern will also change compared to the base period (1986–2014). The maximum average monthly runoff is expected to move from March in the base period to May in future periods. In the future period (2015–2043), due to an increase in temperature, we will face a decrease in precipitation in the form of snow, and an early melting of snow in the year. This will affect the flow of the river in winter and spring. As these changes coincide with the increase of winter rains, the occurrence of winter floods will not be far from expected.

Understanding current and future river flow variability is required for the proper management of surface water resources. In the present study, the effect of climate change on the flow of the Dez River was investigated using the LARS-WG model, the output of two models from the sixth climate change report, ACCESS-ESM1 and BCC-CSM2-MR, and the IHACRES hydrological model. The study showed that the studied basin has more rainfall (on average 20.8% in the ACCESS-ESM1 model and 33.2% in the BCC-CSM2-MR model) and the flow of the main river is 15% on average in the ACCESS-ESM1 model, and in the BCC-CSM2-MR model, it increases by 16% compared to the base period, which indicates the changes in the precipitation pattern and the temporal distribution of precipitation due to the effect of climate change, which has caused a change in the amount of runoff in the two climate change models in the basin. Management options should be considered to reduce the effects of climate change on the main source of water supply in the study area. Following suggestions for further studies are presented as follows:

• It is suggested to use different scenarios of the sixth report in the next research and compare the results with each other.

• In this research, two models of the general circulation of the atmosphere have been used, but in the next research, it is suggested to use several climate models.

• In order to reduce the uncertainty of the hydrological model, it is recommended to use several models simultaneously. However, the most important parameter in reducing the uncertainty of the hydrological models is the method of calibrating and validating the model. This process, of course, requires sufficient time, experience, and access to reliable and long-term historical data.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.