Comprehensive assessment and scenario simulation for the future of the hydrological processes in Dez river basin, Iran

Universal climate change is one of the leading factors that directly affect hydrological processes (Zhang et al. 2016). In this regard, global warming is recognized as a significant issue for climate change during the coming century (Chien et al. 2013). Possible impacts of changes in climate including temperature and rainfall have caused variations in hydrological processes such as evapotranspiration, surface runoff, timing, and magnitude of streamflow, and flood events (Neupane & Kumar 2015). Since these impacts are expected to have diverse influences across a region, different spatial and temporal distributions are created for water resources components. Furthermore, the studies show that variation in precipitation patterns plays a vital role in streamflow trends and sediment in various regions across the United States (Novotny & Stefan 2007). Moreover, temperature variation and wind speed affect evaporation and transpiration sub-processes, which directly have an influence on the surface and subsurface water budgets (Schmid et al. 2000; Hu et al. 2007; Bajracharya et al. 2018).

It worth mentioning that researchers have indicated that climate change has an observable effect on long-term hydrologic processes, while in the short-term, Land Use and Land Cover (LULC) is one of the leading factors that have influence on catchment hydrologic processes (López-Moreno et al. 2013; Ficklin et al. 2016). LULC changes due to climate change, have influence on hydrological processes by conversion in hydrological components (Memarian et al. 2014; Deng et al. 2018). LULC changes cause variation in vegetation cover and surface roughness alteration and this affects the timing and magnitude of surface runoff and infiltration. The latter has led to changes in streamflow and consequently has influenced the magnitude and frequency of floods (Schilling et al. 2014). However, land use changes, such as urbanization and agricultural extension, may cause greater surface runoff (Pai & Saraswat 2011). Urban areas generate significant paved areas in the landscape, considered as impervious surfaces that produce greater surface that runoff as only small amounts of precipitation can soak into the soil profile. Intensive agricultural activities have led to a greater runoff because of reducing surface roughness and contributed to lower interception and less pore space availability in the soil (Busman & Sands 2002; Ghaffari et al. 2010; Baker & Miller 2013).

Accurate hydrological simulation of a basin needs developed models to consider a wide range of detailed information including the list of cultivated crops and orchards, irrigation schedules, fertilization and harvesting operations and so on (Eini 2019). This detailed information, which constitutes the inputs for distributed simulation models like Soil and Water Assessment Tool (SWAT), significantly describes percolation and evapotranspiration. A common belief is that without an accurate and complete calibration and validation of a model for local conditions of the system, no additional useful analyses in respect of the model estimates are reliable (Smarzyńska & Miatkowski 2016).

Researchers have shown that the effects of climate change on various agricultural products will not have a predictable trend due to the type of product, conditions of the case study, and climate scenarios (Shahvari et al. 2019). In some studies, increased crop yields have been reported and, in others, a drop in crop yields has been reported (Boonwichai et al. 2019; Kolberg et al. 2019). The result of a study in 10 largest producing countries showed that compared to present conditions, a group of 11 crop models found a rise in yield loss risk of 12%, 6.3%, 19.4% and 16.1% for wheat, corn, rice and soybeans by 2100, respectively (Leng & Hall 2019). Earlier studies have assessed the potential impact of water shortages under different climate situations on crop production in the USA, China, Australia, South Africa, and on a global scale (Shi & Tao 2014; Araújo et al. 2016; Madadgar et al. 2017; Matiu et al. 2017; Pirttioja et al. 2019).

This objective is fulfilled by employing semi distributed and fully distributed hydrological models, capable of simulation of the entire hydrological process under different climate and LULC changes. In this regard, the SWAT has been widely used to investigate agricultural practices, crop yields, and land management impacts on water quantity and quality considering climate change scenarios (Arnold & Fohrer 2005). Considering the capabilities of the SWAT model, different aspects of LULC alteration and climate change on sedimentation, soil erosion, and runoff magnitude have been investigated. Ngo et al. (2015) evaluated runoff and sediment in the Northwest of Vietnam using SWAT. The results showed that LULC status has a major influence on runoff and sediment yield. Land use conversion by extension of forested area and employment of soil protection practices over five years (2005 to 2010), had led to a decrease in both runoff (from 342.7 to 167.6 mm) and sediment yield (from 148.1 to 74.0 ton/ha).

To estimate climate conditions in the future, the Atmosphere-Ocean General Circulation Models (AOGCMs) and Regional Climate Models (RCMs) considering various CO₂ emission scenarios have been usually employed (Vaighan et al. 2017). An AOGCM is a type of climate model. It employs a mathematical model of the general circulation of a planetary atmosphere or ocean. It uses the Navier–Stokes equations on a rotating sphere with thermodynamic terms for various energy sources (radiation, latent heat). These equations are the basis for computer programs used to simulate the Earth's atmosphere or oceans. Atmospheric and Oceanic GCMs (AGCM and OGCM) are key components, along with sea ice and land-surface components. GCMs and global climate models are used for weather forecasting, understanding the climate, and forecasting climate change. However, to assess the hydrological impacts of climate change, statistical or dynamic downscaling is essential. To achieve this objective, pre-defined statistical relationships or dynamical downscaling utilizing an RCM nested in a AOGCM is mostly employed (Marengo et al. 2009). Only a few studies have focused on the Representative Concentration Pathways (RCP) scenarios dependent on the Fifth Assessment Report (AR5), while many studies have used old AOGCMs under SRES (AR4) scenarios (Vaighan et al. 2017; Fernández et al. 2018). Chattopadhyay et al. (2017) investigated the potential climate change impacts on hydrologic processes in the Kentucky river basin, USA utilizing SWAT and CMIP5 GCMs considering RCP 4.5 and 8.5. According to their results, simulated surface runoff and monthly water yield indicated upward trends in the autumn and spring, while winter months were faced with downward trends. In similar studies, the possible effect of climate change was predicted to be more evident in the Himalayan region, where the runoff is dominated, primarily, by glacier melt and snowmelt (Bajracharya et al. 2018).

The aim of the current study was to evaluate the impacts of climate change on streamflow, sediment yield, crop yield, actual evapotranspiration, and water budget and the effect of LULC alteration on streamflow and sediment yield in the Dez river basin in the southwest of Iran. The novelty of this study is to set up a comprehensive model for multifunctional calibration of hydrological processes. Similar to other water-limited regions of the Middle East, food and water security are the main concerns in Iran (Ashraf et al. 2018). The study area is very susceptible to climate change impacts because of its high dependency on climate-sensitive agriculture (Ashraf et al. 2018). Moreover, the basin plays an undeniable role in the economic sector, where supplying agricultural water and energy production are the main concerns. According to the mentioned studies, although these scenarios have been employed in a few studies of basins throughout the world, reasonable outcomes have been reported. The results of this study will provide reliable guidelines, ensuring sustained water accessibility by changing climate and land use, for policymakers and water resources authorities within the study area.

The basin of the Dez river is a sub-basin of the Karoon Basin and is located in a more significant division in the Persian Gulf basin (Figure 1). Total surface area of the Dez watershed is approximately 17,320 km² and its perimeter is roughly 875 km. Major cities in this basin are Aligoudarz, Azna, and Broujerd. This basin is situated in the highlands of the western and southeastern Zagros Mountains and it is considered as the most densely populated area of the country, the dominant atmospheric precipitation in this region is snowy in the autumn and winter seasons (October to March). A large portion of annual runoff from the basin originates from melting snows from late winter to late spring (March to May). Sezar and Bakhtiari are the main branches of the Dez river.

Table 3 shows the land use area (ha) and percent in Dez river basin. Texture of the soil ranges between loam, sandy loam, sandy clay loam and loamy coarse sand. The topographical elevation varies between 728 m to 2,543 m with an average elevation of roughly 1,300 m, and average slope of 51% (Noori et al. 2016). According to the discharge stations, about 50% of the Dez river discharge at the Talezang station is related to the Bakhtiari subbasin and 42% of it is from the Sezar subbasin. The long-term trend of hydrological 185 events along with 290 flood events in the Dez river basin has been evaluated for a 50-year period (1955–2005). Results indicate that the peak flows of 17 flood events (52% of the total floods) were over 2,900 m³/s from 1991 to 2005, while these values were normal from 1955 to 1990 (Samadi et al. 2019). Average precipitation in the study area is 700 mm with maximum and minimum temperatures of 27 °C and 12 °C, respectively. We used weather data (precipitation (mm), maximum and minimum temperature (°C)) and discharge data (Sezar, Bakhtiari, Dam, and Talezang discharge stations) for the calibration (1990 to 2010) and validation (2010 to 2017) periods.

The analyses were conducted in five steps (Figure 2), steps: (1) Developing SWAT model; (2) uncertainty analysis calibration and validation of the SWAT hydrologic model for the test case of the study; (3) Developing climate change model via employing five AOGCM models considering two RCP scenarios (RCP 4.5 and RCP 8.5); (4) Assessing climate change effects on hydrological process; and (5) Assessing effects of climate change and LULC on runoff and sediment.

The SWAT is a long-term, continuous, physically distributed model developed for predicting the effect of land management practices on the hydrology, sediment yield, and water quality in agricultural watersheds (Arnold et al. 2012b; Jeon et al. 2019). The model can be operated in various time scales from a sub-daily time step to monthly/yearly duration one (Eini et al. 2019). According to Neitsch et al. (2011), computational simulation costs can be minimized within the Hydrologic Respond Unit (HRU) delineating process by combining similar soil and land-use areas into a single unit (Eini et al. 2020).

Hydrologic output and pollutant losses are initially generated at the HRU level, which are then aggregated to the sub-basin level and routed to the catchment outlet. The US Department of Agriculture (USDA) Natural Resources Conservation Service (NRCS) runoff curve number (CN) method is used to simulate the partitioning of precipitation at the soil surface between runoff and infiltration for daily time step simulations (USDA-NRCS 2004; Arnold et al. 2012a; Williams et al. 2012). The Green and Ampt Mein Larson (GAML) excess rainfall method (Mein & Larson 1973) is also available to partition precipitation between surface runoff and infiltration for sub-daily time step applications (Jeong et al. 2010; Arnold et al. 2012a; Brighenti et al. 2019).

In Equation (1), denotes the volume of surface runoff related with the HRU (m³), is the peak runoff rate (m³/s), is the soil erodibility factor, is the crop management factor, is the conservation practice factor, is the topographic factor, and is the coarse fragment factor. The complete description of the MUSLE is well recognized and utilized worldwide for investigation of water erosion (Djebou 2018).

In this study, ArcSWAT2012 is used as a visual interface to prepare a SWAT model (version 2012) within ESRI ArcMap 10.2. To set up the model, a 10 m digital elevation map, the global soil map produced by the FAO (Fischer et al. 2008) with a resolution of 10 km and the GLCC (https://lta.cr.usgs.gov/glcc/globe_int) land use map with resolution of 1 km are provided. Potential evapotranspiration calculation was carried out using the Hargreaves method that only needs daily minimum and maximum temperatures as input.

According to Figure 1, the watershed test case was divided into 57 sub-basins, 318 HRUs, and the managerial data, including amount of irrigation water for cultivated crops in the current cropping pattern (involving spring wheat, winter wheat, tomato, almond, and apple) and fertilizers of agricultural units, were incorporated into the model.

In addition, observed daily maximum and minimum temperatures were used from the existing 12 meteorological stations within the watershed. The time applied to evaluate the accuracy of the climate datasets was assigned for 1990–2017. For hydrologic modeling, the time period of 1990–1992 was determined as the warm-up period, while the periods of 1993–2010, and 2010–2017 were chosen for the calibration and the validation, respectively.

The type of agricultural products, the amount of irrigation according to location changes of agricultural products in the basin, irrigation efficiency, amount of urea fertilizer, plant growth period, and the range of yield changes in different areas of the basin are presented in Table 1. According to the available information, irrigation took place at 10-day intervals based on the plant's water requirement; fertilization was performed during the first irrigation. Furthermore, changes in the observed performance of agricultural products in this drainage basin are indicated in Figure 6. According to Table 1, the average yield of winter wheat is 3.2, almond 19, spring wheat 4.7, apple 20.7 and tomato 22 tons per hectare.

Also according to the information provided by the Iranian Meteorological Organization (IMO) and CROPWAT: A Computer Program for Irrigation Planning and Management (www.fao.org/land-water/databases-and-software/cropwat/en/) the actual evapotranspiration of each agricultural product has been studied in the modeling process. The yields of agricultural products and actual evapotranspiration are moderate and estimated to be in the basin. Hence, it is not possible to simulate agricultural yield and actual evapotranspiration as a time series and they were evaluated on an average basis (Epelde et al. 2015). Observed actual evapotranspiration changes in the box plot concerning different points in the basin are indicated in Figure 6. The actual average evapotranspiration of winter wheat is 463 mm, almond is 902 mm, spring wheat is 614 mm, and apple is 891 mm and tomato 765 mm.

Agricultural products are calibrated based on the average basin drainage data including actual evapotranspiration and an average yield of agricultural products in SWAT-CUP software. In the calibration of agricultural yield, one of the most important parameters is the amount and type of fertilizer used for plants. Likewise, the soil parameters such as depth of root development (RDMX) and soil depth (SOL_Z) are primary parameters that should be investigated initially and simultaneously with runoff simulation. Sol_BD and Sol_AWC parameters, which are respectively related to soil porosity and water holding capacity of the soil, not only are critical in runoff simulation but also play a significant role in the calibration of plant yield. The values of these parameters highly influence the plants' water reception and water stress. Conversely, BLAI, T_BASE, T_OPT, and BIO_E are important parameters for agricultural products that are separately described in Table 10.

Calibration and uncertainty analysis of the outcomes produced by the model were implemented utilizing the SUFI2 algorithm in the SWAT-CUP software (Abbaspour et al. 2015). This technique permits setting ranges for the parameters of interest and afterward running multiple simulations with various parameter sets sampled by Latin hypercube. In this study, Nash–Sutcliffe Efficiency (NSE) objective function was assigned, and the software returned the range of predicted uncertainty within the 95% of the best simulations. Furthermore, to compare the performance of models, the statistical indices of NSE (Equation (3)) and R² (Equation (4)) were used. Table 4 demonstrates the equations utilized to compute every statistical metric, where: N is the number of samples; is observed value (surface runoff (m³/s)) for day i; is the model output (surface runoff (m³/s)) for day i; and represent average of model output and observed data (Duan et al. 2016).

In this research, five AOGCM considering two RCP scenarios, RCP 4.5 and RCP 8.5, were applied for the near period (2021–2040). Appendix A shows the selected climate model for this study. These scenarios were obtained from WDC Climate website (https://cera-www.dkrz.de/WDCC/ui/cerasearch). Table 5 shows changes in precipitation, and also maximum and minimum temperatures. For coupling of five AOGCMs, an average of monthly changes was used.

The percentage of precipitation changes under each of the scenarios and the maximum and minimum increase in temperature of each of the scenarios are presented in Figure 3. The highest drop in rainfall is predicted in the RCP4.5 scenario (24% decrease) in August, and the highest increase is expected to be in July under the RCP8.5 scenario. Furthermore, in scenario RCP8.5, the maximum temperature shows an increase of 3 °C in the summer months (June-August). Table 5 is the combination of (ensemble) variations of the five selected models which display precipitation values relative to the base period and the absolute changes in temperature.

Hydrological assessments within a watershed scale need a reliable water budget for management and evaluation of the future. Based on this concept, water budget components were calculated in the form of a historical period and future. Precipitation, snowmelt, actual and potential evapotranspiration, shallow aquifer recharge, and deep aquifer recharge were then calculated.

The average long-term rainfall in this watershed is about 683 mm, which is known as the input of the hydrological cycle. Other components of the hydrological cycle are infiltration 87 mm (13%), runoff 184 mm (27%) and AET 412 mm (60%) (Table 6).

To study and evaluate the effects of the LULC extreme change, three land use change scenarios were assigned for the developed model. The scenarios are: (i) change of crops (spring wheat, tomato, and winter wheat) to orchards (almond and apple), (ii) green pasture (grassland) to bare lands (shrub land), and (iii) spring wheat to winter wheat (rainfed). Also, changes in runoff and sediment were calculated for each scenario for the near future (2021–2040) and under all RCPs.

To ascertain that the calibrated models are representative of the hydrological response in the watershed simulated monthly runoff and sediment in discharge stations were carried out. Moreover, crop yield and AET of five major agricultural products were calibrated. According to Figure 1, four discharge stations are located in the basin. For runoff calibration and validation, the entire stations were included and for sediment just one station, an outlet of the watershed, was included due to lack of appropriate data. First, by selecting various parameters in SWAT-CUP and performing the model by 500 iterations applying the SUFI2 algorithm, sensitive parameters have been selected by P-value and t-Stat (Table 7). To complete both the runoff and sediment simulation, NSE and R² indices were selected as goodness of fit factors. In most of the hydrological simulations, NSE and R² should be more than 0.6 to have acceptable results for prediction of future and management practices (Arnold et al. 2012b; Anand et al. 2018).

In this research, the SWAT model showed reasonable performance in runoff and sediment calibration and validation in the entire discharge stations. Table 7 shows calibrated parameters values and their sensitive ranking. The lowest P-value with the highest absolute value of t-Stat indicates the most sensitive parameter and vice versa. According to the results, CN2.mgt, TLAPS.sub, and GW_REVAP.gw were recognized as the most sensitive parameters for runoff simulation in this watershed. Also, in Figure 4 time series of calibration and validation have been shown. Table 8 shows NSE and R² values in calibration (1990–2010) and validation (2010–2017) periods.

As mentioned earlier, the parameters given in Table 7 were used in the simulation of sediment. The results of statistical indices indicated that the developed SWAT model for the study area has a high accuracy in both calibration and validation phases. In accordance with the obtained results in Figure 5 and Table 9, both of the simulated time series and the statistical indices achieved the high accuracy of the developed model.

In a study by Arnold et al. (2012b), the most sensitive and widely used parameters of the SWAT model presented are similar to the parameters selected in this study. In addition, in the Nilawar & Waikar (2019) study in India, calibration parameters selected for runoff and sediment are similar to the selected parameters of the present study which have achieved excellent results (NSE and R2 ∼ 0.91 to 0.62) in runoff and sediment simulation. Also, in another research conducted by Anand et al. (2018) in India with a SWAT model, good results, i.e. NSE and R² were more than 0.75, were obtained in 26 discharge stations. Finally, the similarity of results of this study and its selected parameters with the mentioned articles and other studies such as Ayivi & Jha (2018), conducted in North Carolina, and Visakh et al. (2019), conducted in India, reflects the high precision of the SWAT model developed in this study in the simulation of runoff and sediment and prediction of the future. It can be seen that in all of the mentioned studies, the SWAT model is accepted as a powerful model in hydrological simulation.

According to the variation interval of each plant parameter of the SWAT model, the yield of agricultural products was calibrated. Subsequently, the average yield of each agricultural product in all HRUs was calculated and compared with the observed values. Similarly, considering that AET is a function of the growth and yield of agricultural products, this balance component was also examined and compared to the observed data like the yield of agricultural products. Finally, based on location variations in products yield and AET, the NSE and R2 indexes were calculated. The values ​​of plant parameters and their descriptions are separately presented in Table 10. Furthermore, the mean observed values and the results of the SWAT model for agricultural yield and AET are presented in Table 11.

In Table 11, average variations in yield of agricultural products and evapotranspiration show that simulated values are greater than observed values. This difference can be attributed to the fact that simulations have been carried out for a long period of time, and the average yield of the products has changed over time as a result of variations in temperature and precipitation, but observed data are available only for a few numbers of years. As observed from the correlation and error statistical indices in Table 11, these simulated values are reliable and the model has shown a good performance.

In Figure 6, variations in the yield values of plant products as well as AET are shown in the observation period in the box-plot and the results of the SWAT model are indicated using black circle. The results of AET simulation and crop yield indicated that the SWAT model has a high ability to simulate plant. The results of the simulation of plant yield in spring wheat and winter wheat were very close to the observed values and the evapotranspiration of each of these products was higher than the average observed value. The two horticultural products used in the simulation have been estimated to be higher than 2.5 tons per hectare. Also, R² and NSE indices for apple, almond and wheat showed that these products had the best accuracy in simulation.

Comparing the results obtained in this study indicated that, in other studies, the SWAT model has a good accuracy in the estimation of agricultural yields. The results of Epelde et al. (2015), which simulated four major products (wheat, barley, potato, beet) in Spain, pointed out that the yield of agricultural products in the SWAT model is not well correlated annually, and it is necessary to use the average yield of agricultural products for evaluation of the simulation. In reality, due to changes in the behavior of farmers' communities by time, the advancement of agricultural technology, pests, government policies in water and agriculture sectors, the final cost and the benefits of agriculture have a direct impact on agricultural performance, but the SWAT model does not have the ability to simulate and understand many of these factors. In addition, the simulation of climate change effects on agricultural products with the SWAT model in a basin in China was investigated in a study conducted by Niu et al. (2017). In this study, the main products of the region such as corn, spring wheat, spring barley, and spring canola-polish were investigated. The simulated agricultural products showed high correlation with the mean location observations using the coefficient R² (from 0.84 to 0.93). Similarly, the results of other studies such as Sinnathamby et al. (2017) and Daggupati et al. (2015) regarding the method of assessing the yield of agricultural products and the important parameters in the calibration of those products using the SWAT model approve the results of the present study.

Monthly changes of precipitation and max–min temperatures were also applied to the developed SWAT model after calibration and validation. Runoff variations in RCP4.5 scenario did not show any unreasonable changes in the runoff amount, and changes will take place as time displacement. In addition, in the spring, there is an increase in runoff in all scenarios. For instance, in June 154 m³/s in base period will change, for RCP 4.5 and RCP 8.5, to 205 m³/s and 199 m³/s respectively. Runoff variations in the RCP8.5 scenario indicated an increase in runoff magnitude during the spring and summer seasons. However, with an increase of about 3.5 degrees of melting temperature, this will happen earlier with more intense evapotranspiration.

As sediment load is soluble in the river, sediment load changes will change with runoff. However, it worth mentioning that either sediment producing sources or sediment active factors (including crop type, soil, and topography and so on) did not change with climate change.

By increasing runoff in winter times within the RCP4.5 and RCP 8.5 scenarios, sediment load will increase about 13%–19% in February. Likewise, in the spring and summer, the mentioned changes will have increasing trends. For instance, sediment load will be increased under RCP scenarios around 30%–37%; 42%–49%; 54%–62%; respectively at July, August, and September. The latter occurred because of runoff increasing during these months. However, in the autumn, no reasonable significant changes will occur in sediment load, of course except the one in November by −12% decrease.

In general, it can be said that average runoff in the coming period will rise from 208 cubic meters per second in the baseline period to about 228 cubic meters per second in both scenarios (about 9% increase in each scenario). The results also showed that runoff variations in each scenario would be similar to each other, with a slight increase in the RCP8.5 scenario. Similar to these variations, sediment has also increased by 10% in each scenario. Figure 7 shows monthly changes in runoff and sediment.

The climate change effect on runoff and sediment variation by SWAT model in Purna river basin in India considering two scenarios of RCP4.5 and RCP8.5 showed an increase from 7% to 17% for runoff and from 2% to 16% for sediment. These changes are similar to those of the present study (Nilawar & Waikar 2019). Moreover, Tan et al. (2017) estimated that surface runoff in all tropical regions of Malaysia would increase under all RCP scenarios in future periods between 14 and 27%. Conversely, in a basin in Alberta, Canada, it was found that sediment under the effect of climate change will face an increase of 5% to 25% in different regions (Shrestha & Wang 2018). Different results have been presented in other studies using the SWAT model and RCP scenarios. A study in Ethiopia has indicated that, under the influence of climate change and RCP 8.5 and RCP 4.5 scenarios, runoff will face a minor reduction (Serur & Sarma 2017). In China, Zhang et al. (2016) estimated that under RCP scenarios, significant changes will not occur in runoff in the upcoming periods.

Due to the effects of temperature increase and changes in precipitation pattern, the yield of agricultural products has reduced and the yield of horticultural products was slightly increased. The yield of almonds and apples in different locations increased on average about 3%–5% (1 ton per hectare) and 1%–7% (1.2 tons per hectare) in the scenario of RCP4.5 and 4%–9% (1.2 tons per hectare) and 5%–9% (1.4 tons per hectare) in the RCP8.5 scenario, respectively. Conversely, variations in winter wheat which was simulated as rain-fed wheat, showed a decreasing trend. As rain-fed wheat is planted in winter and harvested in the summer, temperature rise has had a negative effect on this product and a mean decrease of 7% to 25% (0.7 t/ha) was observed in different parts of the basin under RCP4.5 and RCP8.5 scenarios. Also, the variations in spring wheat yield (9%–19%) and tomato (4%–22%) declined and an average of 1 ton per hectare for spring wheat and 3.5 tons per hectare for tomato was estimated. Also, changes in evapotranspiration in the upcoming period indicated a decrease for all agricultural products (about 2% for horticultural products and about 6% for agricultural products) in different locations.

As irrigation and other operations of agricultural products have not changed, the long growth period of horticultural products and plant's remaining in soil throughout the year have led to withstanding the increase in temperatures and subsequent increase in evapotranspiration. Due to displacement in terms of time and increased precipitation, horticultural plants have been able to supply their water needs and their growth course begins a little earlier than before.

In studies conducted on the change in yield of agricultural products influenced by climate change under the SRES scenarios in Iran, generally products yields have decreased in dry regions, while yields have increased in temperate, cold, and dry areas (Nazari 2015). In a study carried out by Parry et al. (2004) in Iran, yield of rain-fed products, especially wheat, is estimated to decrease about 5% to 40% by 2080. Also, in another study conducted by Nassiri et al. (2006), it was predicted that a decrease of about 18% to 24% will be observed in the wheat yield as the temperature rises from 2.7 to 4.7 °C. In 2015, it was estimated that in the central Zagros, in a part of which the basin is located, the yield of rain-fed products will decrease from 7% to 13%. Studies showed that peas, corn, and potatoes would have a yield reduction of 11% to 50% by 2100 (Lashkari et al. 2012; Moradi et al. 2013; Hajarpoor et al. 2014). According to the results of this study and comparison with other studies under SRES scenarios, we will face a declining trend in agricultural products in the future, while changes in horticultural products will relatively increase.

After the hydrological simulation of the basin in the base and future periods, the water budget of the watershed was calculated. The parameters of precipitation, surface runoff, infiltration into the aquifer and AET have been calculated from the components of discharge in the catchment area. In the future, the changes will show a slight increase of about 5 mm in precipitation in both scenarios. In addition, the infiltration, which is a function of snow and snowmelt, will also be reduced from 11% to 9%. In each scenario, a surface runoff will be increased in the range of 26% to 29%. In addition, AET will increase relative to the base period (63% to 62%) but portion of AET will decrease relative to other water budget components (Table 12).

In an investigation conducted by Tan et al. (2017), in a basin in Malaysia it was estimated that under RCP scenarios, runoff components (14.6–27.2%) and AET (0.3–2.7%) would decrease. Also, in the Malaprabha basin in India, the combination of the five climatic models (GCMs) showed that AET increased by 4.1% and water demand for agriculture increased by 7.7%, while penetration would decrease by 7.3% (Reshmidevi et al. 2018). Also Nazari (2015) estimated that under the SRES scenarios, 10.3% of available water is expected to decrease in the same basin. Another study in the Weyib Basin in Kenya revealed that under RCP scenarios, runoff and evapotranspiration will decline (Serur & Sarma 2017). According to the performed studies, it can be concluded that the effects of climate change in different locations can be different; there will be no significant changes in the balance parameters in this basin.

Three management scenarios regarding LULC changes were applied to investigate the effect of climate change on runoff and sediment. It is worth mentioning that the LULC scenarios were defined according to land use distributions of the study area.

Considering the major crops used in the model, the first scenario assumed that spring wheat and tomato have changed into apple and almond trees. In the second scenario, it is assumed that green grasslands, which contain grassy plants in spring and summer and contain about 58 percent of the area, have become bare lands. In the third scenario, assumption is that spring wheat (irrigated crop) has been changed to winter wheat (rainfed farming).

According to Figure 8, under the first scenario, LULC change in the SWAT model shows a reduction in runoff in comparison with changes in the existing LULC of the watershed. These changes occurred due to increased irrigation from the river, because the water requirement for the orchard's trees is more than the other crops in the region. For instance, major changes (about 5%) happened in May to July in RCP 8.5. In addition, sediment variations in this scenario has already shown a decrease in most months of the year. Under RCP 8.5 and RCP 4.5, sediment load has decreased by around 14% in summer.

In RCP scenarios, runoff has increased, in a range of 13%–19% under the second scenario of LULC changes. The increase happened compared to the base period due to replacing the water-repellent and water-consuming areas (i.e. plants and shrubs) by bare lands. Furthermore, sediment variations in this scenario have been shown to rise in most months of the year so that, under RCP 8.5 and RCP 4.5, the sediment load has increased by around 13% in summer.

Considering the change of spring wheat to winter wheat (rainfed) under the third scenario of the study, area has not been changed significantly relative to the assigned climate scenarios. LULC changes on base period indicate that monthly runoff has been affected in most of the year. The latter occurred due to the irrigation reduction in this scenario, which led to increasing transpiration, evaporation, vegetation in winter and spring as well. Because the planting date for winter wheat is about four months earlier than spring wheat, during this time it decreases due to plant growth as well as changes in the surface cover of the runoff area. In addition, sediment in this scenario has been shown to rise in most months of the year. Under RCP 8.5 and RCP 4.5, sediment load has been increased by around 19% in summer.

In Figure 8, the runoff and sediment changes are shown relative to the base period. Moreover, all LULC scenarios that had been applied under climate change are given in appendices B and C.

In the study by Ngo et al. (2015), it was determined that, under the impact of increase in agricultural lands, the expansion of urban areas will increase runoff and sediment; in addition, increase in forest cover and implementation of soil conservation scenarios will reduce runoff and sediment.

In his study, the SWAT model was utilized to simulate the Dez river dam basin in Iran. In general, the climate change scenarios (RCP4.5, RCP8.5) did not predict unfavorable situations in the future.

The model indicated a high capability to simulate the runoff and sediment, (average accuracy of 0.75). The results of the climate change scenarios showed that climate change increases the runoff values, for example, an increase from 154 m³/s to 205 and 199 m³/s was projected for runoff, and an increase from 13 to 19% was projected for sediments. Also, agricultural yield changes under the influence of two scenarios showed that horticultural crop yields would increase while agricultural crop yields would decrease. Furthermore, the AET will face a slight decline of 2% to 6%.

Results showed that, in the future, the watershed will face a slight increase in runoff and AET. Combined LULC and climate change scenarios showed that with amplification of orchards, sediment load would decrease.

In this research, multisite downscaling method in the SWAT model was applied. We would like to suggest that researchers use different multisite downscaling methods, like quantile mapping and bias correction. In addition, in the current research, it was not possible to add groundwater simulation. So, it is highly recommended to use SWAT-MODFLOW to simulate the interaction between groundwater and surface water.

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