Introduction

Improvements in water quality require careful monitoring of these resources, planning, and regulation (Mahmoudi et al. 2010). Water quality models facilitate the monitoring of water quality (Hosseini and Ashraf 2015). Wang et al. (2013) highlighted the importance of water quality models in predicting the changes in surface water quality in environmental management in the world. Water quality models are various depending on their unique purpose and simulation characteristics (Ali 2010).

Some water quality models simulate the movement of pollutants. Models can simulate the flows, stages, and pollutant concentrations, as a single event or continuously, in various basins including those having storm sinks and natural drainages. Mathematical simulation techniques can determine the fate and transport of pollutants in water (Ali 2010).

Proper identification of sediment resources will affect a successful sediment control program. Hydrological simulation models along with observations are useful tools for spatial ranking of runoff and sediment yield in basins (Saghafian et al. 2012). Furthermore, spatial scale data are imperative to evaluate the water quality influenced by land use. The information at the basin scale is also required to evaluate the influence of agricultural land use on water quality, and to evaluate and control suspended sediment load, particularly in basins, where construction of the dam is planned (Saghafian et al. 2012). The scale might be the whole basin or small upstream head-waters. According to the scale, different outputs will be achieved, such as the total influence of individual or local land use (Buck et al. 2004).

Water resources and water availability are the main concerns in arid and semi-arid regions. In these regions, relatively small changes in climatic conditions affect water resources significantly. Global climate change which alters local hydrological regimes significantly affects the environment, water resources, and economic systems (Samadi et al. 2013).

Water composition changes seasonally. Extreme pollution might occur during single flood events; however, seasonal water sampling is preferred to sampling at fixed time intervals. Therefore, because of seasonal changes in nutrient export, the standard monitoring protocols should be adjusted according to the season (Buck et al. 2004).

According to the literature mentioned above, water quantity is affected by changing upstream in a basin, which includes many aspects on a basin scale. The models are suitable tools to predict water quality characteristics and their changes through time and space. However, the most limiting factor for researchers to use models for developing countries is the lack of data. The long-term data are not available for many regions. Some water quality models consist of AQUATOX, AGNPS, CORMIX, DRAINMOD-N, EUTROMOD, WASP7, MIKEH11, WASP, TWQM, BIOMOC, QUAL2K, QUAN2E, WATEVAL, and HEC-5Q (Ali 2010; Banejad et al. 2013). Some developed countries instructed the guidance on environmental water quality assessment models. Therefore, the efficient environmental impact assessment strongly needs to standardize some widely used water quality models according to the developing countries’ actual conditions (Wang et al. 2013). Hosseini and Ashraf (2015) introduced some new models in water resources: LUCK (land use change modeling kit), LULC (land use and land cover). LISFLOOD (a model that explores flooding origin and the impacts of land use changes) are the other models. Also, there are WATEM/SEDEM (spatially distributed soil erosion and sediment delivery model), and SedNet (a sediment source, transport, and deposition model) among others.

The Gharasou sub-basin is the primary water supply resource for the Karkheh River Basin (KRB), the third-largest and most productive river basin of Iran. The major impetuses for studying the Gharasou River was to synthesize the information from different sources into a coherent whole; therefore, the Gharasou River was regarded as a basin-scale or sub-basin of the KRB. This paper aimed to review the results of published papers about the prediction of parameters related to the water quality of the Gharasou River. A review of the previous studies revealed that the available information is for a limited period. There are some publications about the simulation of water quality of the Gharasou River using Wavelet-artificial neural network (W-ANN), QUAL2K, SWAT, HBV, WEPP, and MPSIAC models. Moreover, researchers simulated the effect of climate change on the water quality of the Gharasou River using Statistical Downscaling Model (SDSM) and Artificial Neural Network (ANN) projections. This paper reviewed the collected data to help in determining which areas require additional research in future studies. Moreover, it helps decision-makers to develop the best strategies to control threats of sustainability in economic, social, and political aspects for the Gharasou River Basin.

Materials and methods

In this study, local databases including Irandoc and SID along with the international databases such as Google Scholar, Springer, Elsevier, Taylor and Francis, and Wiley online library were explored. In a primary step of searching the papers, it was evident that there were different spells of the Gharasou River among which a spell was chosen in this study. The results showed that there are 478 papers with English periods such as Qaraso, Qarasou, Gharasu, Gharasoo, Qarresu, and Gharaso. Also, 138 published papers or theses with Persian (Farsi) words were studied. Besides, there are two different rivers with the same name, located in Golestan and Ardabil provinces in North and Northeast Iran, respectively. Therefore, the selected papers are related to the Gharasou River in Kermanshah province. Moreover, the studies which included the Gharasou River Basin as a sub-basin of KRB were considered. Furthermore, the papers related to sub-basins of the Gharasou River Basin were examined. The key terms for this study were “water quality”, “soil erosion”, “water pollution”, “climate change”, and “modeling”. The collected 20 publications are papers published from 2007 to 2019.

General description of the study area

On a basin scale, water quantity and quality of downstream are affected by upstream changes. Therefore, for water quality assessment, a basin-scale approach is essential (Hessari et al. 2012). Karkheh River is the third major river in Iran that originates from the Zagros Mountains and flows into the Persian Gulf (Haghiabi and Mastorakis 2009) (Fig. 1). KRB is one of the major basins in western Iran. Furthermore, KRB is a vital basin for water supply in some parts of Kurdistan, Kermanshah, Hamadan, Lorestan, Ilam, Markazi, and Khuzestan provinces (Haghiabi and Mastorakis 2009; Samadi et al. 2013). KRB with an area of 51,000 km2 is located at 30° to 35° N and 46° to 49° E (Rientjes et al. 2013). The five sub-basins of the main rivers in the KRB include Gamasiab, Gharasou, Kashkan, Saymareh, and Karkheh (Ahmad and Giordano 2010). The Gharasou River is the primary resource of water supply for the Karkheh Reservoir (Fig. 1). The Gharasou River joins the Gamasiab River after running through Kermanshah city and then delivers water collected from Kermanshah and Kurdistan provinces to Saymareh River. The total length of the Gharasou River is 152 km (Atazadeh et al. 2007; Sayadi et al. 2014), and the area of the Gharasou River Basin is 5793 km2. The range of elevation of the Gharasou River Basin is from 1237 to 3350 m, with the mean elevation of 1555 m (Omani et al. 2007). The average annual rainfall of this basin is about 447 mm and ranges from 215 to 785 mm. The most rainfall takes place in February and the least in July. The annual mean of the temperature is about 14.6 °C. The average temperature of the warmest and the coldest time of year are 26.95 °C and 1.15 °C, respectively, which takes place in July and January. However, these temperatures could increase to the highest of 37.8 °C and decrease to the lowest of − 4.2 °C in these months. Besides, the mean annual potential evaporation is 2132 mm (Hosseini et al. 2016; Omani et al. 2007; Samadi et al. 2013).

Fig. 1
figure 1

The Karkhe River Basin (KRB), and the Gharasou River Basin in the northwest of KRB location in Iran (Ghobadi et al. 2015) and Gharasou River Basin in the northwest of KRB (Saadatpour 2014)

Full size image

Results and discussion

Water quality modeling

W-ANN model

Water qualitative parameters change through time. Surveying and predicting changes in the water quality parameters by models is time-consuming and costly. However, the hydrologic models can estimate these parameters accurately (Banejad et al. 2013). Various conceptual and black box models were developed during the last couple of decades. Among these models, researchers have widely used hybrid wavelet and Artificial Intelligence (AI)-based models (W-ANN) to simulate hydrologic processes. Data-driven methods of AI can model and forecast non-linear hydrological processes. Also, the management of dynamicity and noise, which is excessively concealed in datasets (Nourani et al. 2014). Figure 2 is a schematic diagram for hybrid wavelet AI models suggested by Nourani et al. (2014).

Fig. 2
figure 2

Schematic diagram of hybrid Wavelet Artificial Intelligence forecasting models suggested by (Nourani et al. 2014)

Full size image

Banejad et al. (2013) estimated total dissolved solids (TDS), electrical conductivity (EC), and sodium absorption ratio (SAR) for the Gharasou River by W-ANN hybrid model during a 24 years statistical period (1981–2004). For the estimation TDS, input data were EC, SAR, pH, SO42−, Ca2+, Mg2+, and Na+. For the estimation of SAR, the input data were TDS, pH, Na+, and HCO3−. Besides, for the estimation of EC, the input data were SO42−, Ca2+, Mg2+, SAR, and pH. Then, they compared the W-ANN model with the ANN model. The results showed that the optimized W-ANN estimated SAR the most accurately (r = 0.999). Moreover, W-ANN could estimate EC (r = 0.996) and TDS (r = 0.999) with higher accuracy and fewer errors than the ANN model. W-ANN had a root mean square error (RMSE) of 0.24, 36.62, and 7.85 for SAR, EC, and TDS, respectively. RMSE of ANN for SAR, EC, and TDS was 0.03, 54.51, and 8.45, respectively. The mean absolute error (MAE) of W-ANN was 0.004, 7.05, and 1.45 for SAR, EC, and TDS, respectively. MAE of ANN was 0.06, 9.95, and 1.54 for SAR, EC, and TDS, respectively. This model monitors results with optimized monitoring costs and high efficiency. Banejad et al. (2013) suggested that decision-makers use the W-ANN model in predicting the water quality parameters of the Gharasou River. Nourani et al. (2014) reported that by wavelet transformation, the original time series are usefully decomposed. Subsequently, the flow will be forecasted more accurately by W-ANN than regular ANN models. ANN model improves input data with various resolution levels. When there is a lack of data, and/or when the physical processes of surface–groundwater interactions are not entirely understood, the wavelet model would be a useful tool (Zare and Koch 2017). They developed an innovative coupled hybrid Wavelet-ANFIS/FCM-PSO model for developing efficient conjunctive surface–groundwater resources management systems.

QUAL2K model

QUAL2K (or Q2K) is a river and stream water quality model. This model is typically used to assess the impact of multiple pollution sources on water quality. Wang et al. (2013) reviewed the suitability of QUAL models for rivers with branches and non-point pollution resources under one-dimensional steady-state conditions or dynamic models. The flexibility and applicability of the QUAL2K model for simulation of river water quality are the reasons for the general use of this model (Fereidoon and Khorasani 2013; Pelletier et al. 2006). Pelletier et al. (2006) described an overall mass balance for a constituent concentration (ci) in the water column of a reach (excluding hyporheic exchange) as follows:

$$ \frac{{dc_{i} }}{dt} = \frac{{Q_{i - 1} }}{{V_{i} }}c_{i - 1} - \frac{{Q_{i} }}{{V_{i} }}c_{i} - \frac{{Q_{ab,i} }}{{V_{i} }}c_{i} + \frac{{E^{\prime}_{i - 1} }}{{V_{i} }}(c_{i - 1} - c_{i} ) + \frac{{E^{\prime}_{i} }}{{V_{i} }}(c_{i + 1} - c_{i} ) + \frac{{W_{i} }}{Vi} + S_{i} $$
(1)

where Qi = flow (m3d−1, ab, abstraction); Vi = volume (m3); E′i = the bulk dispersion coefficient between reaches i and i + 1 (m3d−1); Wi = the external loading of the constituent to reach i (g d−1 or mg d−1), and Si = sources and sinks of the constituent due to reactions and mass transfer mechanisms (g m−3d−1 or mg m−3d−1)(Pelletier et al. 2006).

The Gharasou River passes through Kermanshah, a city with no effective sewage treatment plant (Atazadeh et al. 2007). There are some industrial units nearby the Gharasou River; Kamyaran sewage, Industrial town of Kermanshah, Shahid Hadadeadel sugar factory, Zamzam Company, Sahra Milk, and Dairy Company, and Oil Refinery Company. The quality changes of the Gharasou River by QUAL2K model was simulated considering nitrogen (N), phosphorus (P), total suspended solids (TSS), dissolved oxygen (DO), and biochemical oxygen demand (BOD) in four seasons in 2008. The results showed DO < 5 mg L−1 downstream of the Gharasou River because of the entrance of domestic sewage also N and P from agricultural plains (Torabi Meibodi 2011). Fereidoon and Khorasani (2013) simulated the Gharasou River quality by QUAL2K during a year considering BOD, total N (TN), and total P (TP). The results showed the rapid increases of BOD in the Sahra Milk and Dairy Company site, which has a significant effect on oxygen consumption. The trend of TN and TP along with the Gharasou River has the same trend as BOD. Therefore, based on this study, the point resource of pollution for the Gharasou River is the Sahra Company site. The simulated results by the QUAL2K model showed that the treatment of wastewater could not significantly decrease the TN. As a result, the decrease of TN could not be considered in wastewater treatment. In opposition, TP wastewater treatment has economic benefits (Fereidoon and Khorasani 2013).

The results of (Razaghian et al. 2015) study approved these findings. They also reported that the Gharasou River has the self-purification capacity after passing through Kermanshah city. Therefore, the DO was improved because of the increasing aeration coefficient (Razaghian et al. 2015).

The expert Choice Software can evaluate more specifically and gives priority to various contaminants (Hosseini and Ashraf 2015). Hosseini and Ashraf (2015) applied this software to identify and categorize the type of land use and pollutants in the Taleghan basin. The results showed that pollution of the Taleghan basin was significantly impacted by sewage, agriculture, outdoor activities, industry, tollway services, and restaurants.

Land cover/land use

Water quality predominantly depends on land use and change patterns. In this section, the two mainland uses including, agricultural and urban land uses were discussed. Predominate land uses are agricultural (67% of the basin area (Samadi et al. 2013)) and rangelands (Saghafian et al. 2012). In the Gharasou River Basin, the conversion of rangeland to rain-fed crops is the main cause of land use change (Omani et al. 2007).

The Gharasou River passes through Kermanshah city. Urbanization leads to land use changes of margin areas of cities, their water permeability, and disturbing the hydrology cycle of the basin. The effect of land use changes in Kermanshah city on the quantity and quality of the Gharasou River runoff was simulated by the L-THIA model. For this purpose, the land use maps in 1973 and 2011 were used. The results showed that most changes in land use occurred in agricultural lands. The results also showed an increase of 126 percent in the quantity of the runoff. The nutrients’ concentrations were decreased, but the heavy metals concentrations increased in 2011 compared with 1973 (Yari 2014).

Soil erosion modeling

MPSIAC model

The MPSIAC model is a modified version of the PSIAC model presented in 1982. PSIAC (Pacific Southwest Inter-Agency Committee) was presented in 1968. This model is capable of predicting erosion and sediment yields at the basin scale. The differences between PSAIC and MPSIAC are that the MPSIAC model is more quantitative than the PSIAC and the scoring of factors is more reliable. The nine factors include surface geology, soil K factor, climate, runoff, topography, ground cover, land use, surface erosion, and gully erosion. Scoring of these factors should be done after field surveys (Heshmati et al. 2012).

Heshmati et al. (2012) studied the soil erosion rate in the Merek sub-basin in three agro-ecological zones consisting of agriculture, rangeland, and forest. Merek sub-basin is a part of the Gharasou River Basin, with an area of 23,038 ha that lies between 34° 00′ 38″ to 34° 09′31″ N and 47° 04′ 25″ to 47° 22′ 18″ E. They investigated the soil erosion rate, and the amount of eroded soil organic carbon (SOC), N, P, and potassium (K) using the MPSIAC model. They calculated the amount of nutrients’ depletion by multiplying eroded soil (ton ha−1 yr−1) by nutrient contents (g kg−1).

Heshmati et al. (2012) estimated and scored factors of the MPSIAC model for each geomorphological facies within the agriculture area, rangeland, and forest zones. To determine the surface geology, they used a geology map. They used computerized RUSEL software (RUSEL, SWCS; 1.04) to estimate the soil K factor of the universal soil loss equation (USLE). For this purpose, five sub-factors factors are required including silt plus very fine sand (%), coarse sand (%), soil organic matter (SOM) (%), soil structure, and soil permeability. They calculated the climatic factor based on rainfall intensity (mm h−1) with a 2 year return period from Kermanshah Weather Station data as the nearest weather station and the estimated runoff factor from the X4 = 0.006R + 10Qp equation. Where R is the runoff coefficient and Qp is peak discharge of overland flow (m3 s−1). To estimate Qp, they used Qp = 0.278CIA equation. Here, Qp is peak discharge, A facies or sub-basin area (km2), and I rainfall intensity (mm h−1) with a 1 year return period. They calculated the average slope (%) of each geomorphological facies by a GIS-prepared slope map. They used quadrate plots (5–10) within each geomorphological facies to estimate the percentages of bare soil and canopy cover. They estimated surface soil erosion using the X8 = 0.25 SSF equation. SSF is a surface soil factor which Bureau of Land Management (BLM), USA provided it (Heshmati et al. 2012).

The results showed that the erosion rates in the agriculture area, rangeland, and forest were 14.47, 16.60, and 18.57 t ha−1 yr−1, respectively (Fig. 3 a). In the Merek sub-basin, the annual estimated SOC, N, P, and K depleted by erosion in the agriculture area were 147.24, 15.6, 0.172, and 4.47 kg ha−1 yr−1, respectively. The annual depletion of estimated SOC, N, P, and K in the rangeland was 176.92, 18.73, 0.170, and 4.65 kg ha−1 yr−1, respectively. Moreover, the amount of depleted SOC, N, P, and K in the forest were 306.10, 23.75, 0.165, and 5.15 kg ha−1 yr−1, respectively (Figs. 3b–e). Heshmati et al. (2012) reported the lowest decline in P by soil erosion and the highest depletion of SOC, N, and K in the forest. The highest decline in P and the lowest depletion of SOC, N, and K have occurred in the agriculture area. Moreover, the presence of smectite mineral in the soil of sloping land subjected to deforestation and overgrazing results in depleting soil nutrients and SOC in the Merek sub-basin.

Fig. 3
figure 3

Rates of soil erosion (a), SOC (b), N (c), P (d) and K (e) depletion in the Merek basin (reproduce from data reported by (Heshmati et al. 2012))

Full size image

WEPP model

The WEPP (Water Erosion Prediction Project) is a physical model that computes spatial and temporal distributions of soil loss and deposition. The WEPP model uses physical concepts of erosion and hydraulics of overland flow science (Flanagan and Nearing 1995; Parvizi et al. 2015). The WEPP model dynamically simulates runoff and soil erosion, considering plant growth, residue decomposition, and winter process (Parvizi et al. 2015). The WEPP model estimates time and place in a basin or on a hill slope, where erosion is occurring. Therefore, the best conservation management can be selected to most effectively control soil erosion (Flanagan and Nearing 1995). The WEPP model uses solutions to a steady-state sediment continuity equation (Eq. 2) to estimate sediment load amount and net detachment or deposition at points down a profile (Flanagan et al. 2001).

$$ \frac{dG}{{dx}} = D_{f} + D_{i} $$
(2)

where G = sediment load (kg s−1 m−1), x = distance downslope (m), Df = rill erosion rate in kg s−1 m−2, and Di = interrill sediment delivery rate in kg s−1 m−2.

Interrill sediment delivery to rills is predicted in WEPP using Eq. 3:

$$ D_{i} = K_{iadj} I_{e} \sigma_{ir} SDR_{RR} F_{nozzle} (R_{s} /w_{rill} ) $$
(3)

where Kiadj = the adjusted interrill erodibility factor in kg s m−4, Ie = effective rainfall intensity in m s−1, σir = the interrill runoff rate in m s−1, SDRRR = a sediment delivery ratio that is a function of random roughness, row side-slope, and the interrill particle size distribution, Fnozzle is an adjustment factor to account for sprinkler irrigation nozzle impact energy variation (value of 1.0 for natural rainfall conditions), Rs and Wrill are the respective rill spacing and rill width in m (Flanagan et al. 2001).

Parvizi et al. (2015) verified the WEPP model for runoff and erosion estimation in different range types in the Kabude-ye Olya sub-basin of the Gharasou River Basin. The Soil Conservation Research Station of Kabude-ye Olya is located at 34° to 15° N, and 47° to 5° E. Soil Conservation and Watershed Management Research Institute (SCWMRI) has established this station since 2003 to investigate the calibration of soil erosion and runoff by the WEPP model (Hemmati et al. 2009). Erosion plots were installed in three range types at the slopes of 25, 35, and 45%. An event-based erosion and runoff were simulated in each plot by the v2008.9 version of the WEPP model. Results indicated that the maximum model accuracy for the prediction of runoff was for the slope of 45%.

The R2 corresponding to the relationships between the observed and predicted parameters, or the Nash–Sutcliffe model efficiency coefficient (ENS) could quantify the evaluation of models (Banejad et al. 2013; Hosseini et al. 2016). ENS calculated by Eq. 4 (Srinivasulu and Jain 2006):

$$ E_{NS} = \frac{{\sum\nolimits_{i = 1}^{N} {\left( {M_{i} - \mu } \right)^{2} } - \sum\nolimits_{i = 1}^{N} {\left( {P_{i} - M_{i} } \right)^{2} } }}{{\sum\nolimits_{i = 1}^{N} {\left( {M_{i} - \mu } \right)^{2} } }} $$
(4)

where Mi is the measured and Pi is the network output value, µ is average, and N is the number of input samples.

The WEPP model estimated runoff in the slope of 25 with R2 = 0.94 and ENS=0.70. Estimated R2 and ENS for runoff in the slope of 35 were 0.73, 0.69, respectively. For the slope of 45%, the runoff was estimated as R2 = 0.84 and ENS = 0.73. The average amounts of error of the WEPP model for prediction of runoff were 0.61 L. This amount is negligible in comparison with annual changes of runoff (0–18 and 0–24 L) for the slopes of 25 and 35%, respectively. R2 for soil erosion predicted by the WEPP model was 0.89, 0.61, and 0.84 for the slopes of 25, 35, and 45%, respectively. However, negative amounts of ENS indicated the low efficiency of the model especially in two slopes of 25 and 35%. ENS were -0.47, -0.43, and 0.10 for the slopes of 25, 35, and 45%, respectively. Parvizi et al. (2015) conclude that the WEPP model predicted the runoff accurately especially in the slope of 25% and under-estimated soil erosion in all the studied slopes.

Runoff simulation

SWAT model

The Soil and Water Assessment Tool (SWAT) model is designed to run with readily available input data. It simulates water, sediment, and agricultural chemical transport at the river basin scale (Easton et al. 2008; Saghafian et al. 2012). The model was developed at the U.S. Department of Agriculture (USDA) Agricultural Research Service (ARS) and the Soil and Water Research Laboratory in Temple, Texas.

Accurate forecasting for the transport of water, sediment, and agricultural chemicals using SWAT would be possible. For this purpose the hydrological cycle (i.e., overall water circulation in the basin-scale) is simulated by Eq. 5 (Devia et al. 2015):

$$ SW_{t} = SW_{0} + \sum\nolimits_{t = 1}^{t} {(Rv - Qs - W_{seepage} - ET - Q_{gw} )} $$
(5)

where, SWt = the humidity of the soil, SW0 = base humidity, Rv = rainfall volume in mm water, Qs = the surface runoff, Wseepage = seepage of water from the soil to underlying layers, ET = evapotranspiration, Qgw = groundwater runoff, and t = time in days (Devia et al. 2015).

Land management practices influence water sediment and agricultural chemical yields. SWAT is capable of estimating them in large complex basins with a variety of soil types, land use, and management practices at the long-time periods (Neitsch et al. 2011). Hosseini et al. (2016) evaluated various land use management practices on a large basin by SWAT. Hosseini and Ashraf (2015) established a modified SWAT model to predict sediment yield and water supply for the future in the Taleghan region, Iran. Their results revealed that the modified model could predict the flow rate and sediment yield for ungagged basins in semi-arid areas. They also suggested the SWAT model for regions with similar climatic conditions (Hosseini and Ashraf 2015). Saghafian et al. (2012) reported that the SWAT model could simulate monthly runoff in the Gharasou River Basin. However, monthly sediment concentration was estimated low (R2 = 0.53), especially in high concentration values. However, they reported that while sediment concentrations higher than 2000 mg L−1 were omitted, R2 increased to 0.7. Shahoei et al. (2017) reported that the SWAT model predicted weakly the daily runoff of the Ravansar Sanjabi sub-basin of the Gharasou River Basin as confirmed by ENS = 0.14 and 0.16 related to calibration and validation data, respectively. Ravansar Sanjabi sub-basin is located from 46° 21′ 30′′ E to 46° 49′ 30′′ E and 34° 20′ 0′′ N to 34° 60′ 0′′ N with an area of 1260 km2.

SWAT model can also estimate annual and monthly water budget (Hosseini et al. 2016). Managers also can simulate the existing water quantity and sediment yield with different land use variation scenarios (Hosseini and Ashraf 2015).

The results of papers, which used the SWAT model with various purposes in the Gharasou River Basin, are presented as follows:

  1. (i)

    Omani et al. (2007) simulated the effect of management conditions on runoff and sediment load in the Gharasou River Basin using the SWAT2000 model for the period of 1980–2000. The SWAT model could reasonably simulate hydrologic components and erosion in the Gharasou River Basin. The results of calibration and validation for monthly streamflow simulation were as (R2 = 0.72–0.89, ENS = 0.71–0.90), and (R2 = 0.71–0.94 and ENS = 0.50–0.93), respectively. The results of calibration and validation for TSS simulation were as (R2 = 0.71–0.96, ENS = 0.59–0.80), and (R2 = 0.82–0.99, ENS = 0.82–0.90), respectively. The predicted sediment yield of the Gharasou River Basin was 5.1–11.8, with an average of 3.5 ton ha−1 (Omani et al. 2007).

  2. (ii)

    Hoseini (2014) evaluated the hydrological budget in the Gharasou River Basin from 1982 to 2005 by the SWAT model. For this purpose, the daily data, including rainfall, minimum and maximum temperatures, relative humidity, and runoff was used. The results of calibration and validation for monthly runoff simulation were as (R2 = 0.60, ENS = 0.56), and (R2 = 0.65, ENS = 0.60), respectively.

  3. (iii)

    Hosseini et al. (2016) evaluated the hydrological budget in the Gharasou River Basin, from 1995 to 2005 by the SWAT2012 model. The hydrologic budget included components of soil water content, evapotranspiration, and surface, lateral, and groundwater flows. The findings showed that SWAT optimized the hydrologic budget reasonably well, and the monthly flow simulation became possible. In the Gharasou River Basin, the highest water loss (49.3%) occurred through evapotranspiration from April to the end of May, when snowmelt and intense storms occur. Groundwater flow constituted 30% of the hydrologic budget to supply water for agriculture. Soil moisture varied as 26.2 mm (equal to 5.2% of mean average rainfall) during the simulation period. Hosseini et al. (2016) suggested a 15% surface runoff (about 436 MCM in volume) should be considered in the agricultural planning of the area. In this research R2 values for average monthly discharges in Golchehr, Gharabaghestan, and Gharasou, three mainstream gauges, were 0.40, 0.71, and 0.61 for calibration and 0.37, 0.87, and 0.65 for validation periods. ENS at the outlet were 0.43 to 0.73 in the three mentioned outlets for both periods.

  4. (iv)

    Shahoei and Porhemat (2019) simulated the monthly runoff of Ravansar Sanjabi sub-basin by the SWAT model. The rainfall, minimum and maximum temperatures, wind velocity, sunshine duration, and relative humidity in daily data between 2002 and 2010 were used. The results of calibration and validation for monthly runoff simulation were as (R2 = 0.80, ENS = 0.70), and (R2 = 0.90, ENS = 0.81), respectively.

Rientjes et al. (2013) calculated water balance by satellite-based data. They used satellite images combined with routinely collected meteorological and river discharge data. They could determine the spatial and seasonal patterns of water availability, and consumption by evapotranspiration across the KRB. They used the preference-based multi-variable calibration approach for water balance calculations (detailed results will be discussed in Sect. HBV model).

SRM and AWBM models

Snowmelt Runoff Model (SRM) to simulate the daily runoff in mountain basins uses the snow cover area in addition to rainfall and temperature. Shahoei et al. (2017) reported that the SRM model resulted in acceptable in the simulation of daily runoff of the Ravansar Sanjabi sub-basin for both calibration and validation periods (ENS = 0.9 and 0.95, respectively).

The Australian Water Balance Model (AWBM) is a lumped model simulating the runoff in basins using rainfall and evaporation variables. This model has been considered by some researchers to investigate the simulation of runoff in the Gharasou River Basin. Khalili Naghtchali and Pourreza-Bilondi (2016) used AWBM using Shuffled Complex Evolution (SCE) and Genetic Algorithm (GA) to simulate the rainfall-runoff in the Gharasou River Basin. The results showed that GA better simulated monthly runoff than SCE, as evidenced by ENS = 0.67 and RMSE = 8.75 for validation data. The simulated monthly runoff of Ravansar Sanjabi sub-basin through the AWBM model resulted in ENS = 0.63 and ENS = 0.50 in calibration and validation periods, respectively (Shahoei and Porhemat 2019). Pourreza-Bilondi et al. (2019), instead of using models individually used the simple combination models Simple Model Average (SMA), Weighted Average Method (WAM), Multi-Model Super Ensemble (MMSE), and Modified Multi-Model Super Ensemble (M3SE) for the hydrological simulations in the Gharasou River Basin. The results indicated an improvement in the results of simulated flow by multiple combination techniques compared with each simulation model. They reported that the multi-model simulation generated by M3SE is more efficient than the others. Besides, M3SE can be better, at least comparable to the best-calibrated single-model simulations.

Modeling of sedimentation in the river reach

Prediction of sedimentation and erosion patterns in the river reach will be more precise by considering the effect of the suspended load in addition to the sediment yield (Ghobadian and Rahimifar 2016). They developed a computer model to simulate sediment movement in the Gharasou River to predict bed variation at the time steps. The bedload prediction model was verified using Meyer, Parker, and Wilson equations. Then, they calculated bed variation for the 18.5 km length of the Gharasou River for 5 and 10 years. Also, they calculated suspended load concentration and the effect of suspended load on sedimentation and erosion patterns. For this purpose, they used a numerical solution of a 1-D depth average differential equation for a 5 year simulation. The results showed that soil erosion is dominated pattern at the Gharasou River reach. The results also showed that when the effect of the suspended load was considered, the amounts of sedimentation and erosion increased by 7.1% and 46.5%, respectively (Ghobadian and Rahimifar 2016).

The results of sediment prediction by the GSTRS3 model indicated an acceptable level of prediction of the sediment trends assessment in the Gharasou River, as evidenced by R2 = 0.85 between observed and predicted longitudinal profiles. For this purpose, the sediment grading curves at 60 cross-sections of the Gharasou River with 18 km length were considered (Byzedi and Karami 2017).

HBV model

HBV (Hydrologiska Byrans Vattenavdelning model) is a conceptual hydrological model. Operational hydrological studies use HBV for forecasting and water balance purposes (Abebe et al. 2010). The general water balance equation is (Eq. 6):

$$ P - E - Q = \frac{d}{dt}\left( {SP + SM + UZ + LZ + {\text{lakes}}} \right) $$
(6)

where, P = precipitation, E = evaporation, Q = runoff, SP = the snowpack, SM = the soil moisture, UZ, and LZ are the upper and lower groundwater zone and lakes = the volume of the lake. The data used for rainfall, air temperature, and evaporation can be considered as daily and monthly data. Snow accumulation can be calculated by air temperature data (Devia et al. 2015).

Few efforts are reporting the use of satellite-based actual evapotranspiration (ETa) in multivariable calibration for conceptual rainfall-runoff modeling (Rientjes et al. 2013). Rientjes et al. (2013) presented a novel approach using satellite-based ETa in a rainfall-runoff model that uses multi-variable calibration. They input ETa as a second calibration variable. They applied the HBV model to the Upper KRB including the Gharasou River Basin using streamflow data at gauging stations. The results showed that the HBV model is a capable tool to reasonably simulate ETa. They reported the differences between monthly estimated ETa and potential crop evapotranspiration (ETp) in large scale irrigated wheat areas. These differences were 12.5% and 11.7% in the Upper and Lower KRB, respectively.

Climate change

Samadi et al. (2013) predicted future climatic conditions in the Gharasou River Basin by using regression-based statistical downscaling techniques. They used 30 years of observed and downscaled data to define future values by SDSM and ANN projections. For this purpose, they predicted an increasing daily temperature of + 0.58 °C (+ 3.90%) and + 0.48 °C (+ 3.48%) for SDSM and ANN projections, respectively. Decreasing daily rainfall was of − 0.1 mm (− 2.56%) and − 0.4 mm (− 2.82%) for SDSM and ANN projections, respectively. SDSM and ANN presented a reduction in the mean annual flow as − 3.7 and − 9.47 m3 s−1, respectively. The results of both downscaling projections showed a significant reduction of streamflow, particularly in winter (Samadi et al. 2013). Models could predict the increases in temperature and decreases in rainfall. However, the combined impact of temperature and rainfall on the pattern of the monthly change in ET, water yield, and discharge is very complicated. Besides, the irrigated, cultivated areas in Middle KRB are the most stable portions (Solaymani and Gosain 2015). The models predicted an increase in the risk of drought and heat stress throughout the KRB. Since rainfall enhances (with expectation in both spring and autumn), the risk of soil erosion and flooding will be much higher in KRB (Solaymani and Gosain 2015).

Downscaling methods do not consider extreme topography and the limitations of uneven distribution. Delivering high-resolution climate scenarios at a specific site or a specific region is possible by selecting the most suitable method or group of methods. A thorough analysis, experience, and insight can help to select these models (Samadi et al. 2013).

Zahabiyoun et al. (2013) investigated the effect of climate change on the runoff of the Gharasou River during the periods 2040–2069 by the SWAT model. They simulated the rainfall‐runoff for the period 1971–2000. Then the impact of climate change on the basin hydrology for the period 2040–2069 was evaluated by HadCM3‐AR4 global climate model data under the A2 scenario–from the SRES scenario set‐haves been downscaled. The future runoff changes have been simulated by introducing downscaled climate data in the SWAT model. The results showed that the increases in temperature for most of the months, and ± 30% of the change in the precipitation. They also reported that the monthly runoff will be changed from − 90 to 120%.

Conclusion

According to the review of the publications about the modeling of the Gharasou River water quality, the modeling of water quality indicated that W-ANN could reasonably simulate TDS, EC, and SAR for the Gharasou River. The QUAL2K model could determine the precise location of water pollution of Gharasou River by simulating BOD, TN, and TP originating from some industrial units. The SWAT model predicted weakly the daily runoff of the Ravansar Sanjabi sub-basin. However, it could predict sediment yield and the hydrological budget of the Gharasou River Basin for all sub-basins. The SRM, AWBM, and M3SE models simulated the monthly runoff of the Ravansar Sanjabi sub-basin efficiently. For the water budget, satellite data are capable tools to estimate ETa instead of routinely collected meteorological data. The data for the amount of soil erosion rate and eroded SOC, N, P, and K estimated by the MPSIAC model and the runoff data predicted by the WEPP model might be used for considering the eutrophication which evidenced by (Atazadeh et al. 2007). The rainfall, temperature, water budget, and runoff of the Gharasou River Basin were forecasted in the future considering climate change scenarios.

Study limitations

The small numbers of publications and their duration (2007–2019) for water quality estimation of the Gharasou River Basin are the most limiting factors for this study.

Recommendations

Modeling for prediction of water quality in the Gharasou River Basin similar to most of the other basins of Iran deal with some difficulties: the lack of data, high complicated water environmental conditions, and partly known complexity which leads to the researchers not being able to predict the water quality in the sub-basins owing to the unavailability of data. Hydrological models used in Iran are the models designed to the conditions of other countries, therefore, there is a need to introduce models regarding data scarcity and compatibility with the above-mentioned conditions. Moreover, using different machine learning models was suggested. Machine learning algorithms can address water quality problems in complex natural systems simultaneous with spatial heterogeneity (Giri 2021). Furthermore, the learning models can identify the key water parameters that further reduce prediction costs and improve prediction efficiency (Chen et al. 2020).