چكيده لاتين :
Introduction
Simulation of quantity and quality of surface runoff in mountainous watersheds is one of the most challenging topics in modeling due to its unique features, such as unusual topography and complex hydrological processes. One of the lesser-known aspects of modeling such catchments is the uncertainty analysis of water quality predictions, especially about the vital phosphorus parameter. Phosphorus is one of the important quality variables in water, and its increase in water resources can cause eutrophication phenomena in streams and reservoirs of dams. Due to the importance of the phosphorus parameter and the fact that water quality modeling has not been employed in the Karaj catchment area so far, in this research, total phosphorus has been modeled as a water quality parameter along with the flow and sediment discharge. This study aims to identify the most sensitive parameters of the model to flow, sediment, and total phosphorus discharge and calibrate, validate and analyze the parametric uncertainty of the SWAT model in predicting these three variables in a mountainous catchment. The case study was the catchment area of the Karaj River upstream of Bileqan pond, which is one of the mountainous watersheds in Iran. There are two critical water structures along the Karaj River, namely Amirkabir dam and Bilqan pond. Amirkabir dam (Karaj) is a multi-purpose project that is constructed to supply drinking water to Tehran and regulate water for irrigation and agriculture in the suburbs of Karaj. The Bileqan pond is also the essential point of supply and transfer of drinking water in Tehran. Given the importance of this region in supplying water for different uses, providing a calibrated model for quantitative and qualitative variables of water can be the basis for decisions to apply future management scenarios in this basin.
Materials and Methods
The case study was the Karaj River catchment area upstream of Bilqan Basin, which with an average height of 2880 meters, is one of the mountainous areas located in the Alborz Mountains. This basin with an area of 1076 square kilometers in the north, includes parts of Mazandaran province. In the east and south of the catchment area includes parts of Tehran province and most of it is located in Alborz province. The average annual temperature and rainfall in this basin are 12.1 °C and 480 mm, respectively, and the average of 117 glacial days during the year is observed in this area. The long-term daily data of synoptic stations adjacent to the study area from the beginning of 1998 to the end of 2018 (21 years in total) was introduced to the model. Also, daily data of relative humidity, rainfall, minimum and maximum temperature, solar radiation hours, and wind speed as meteorological parameters measured at stations in the study area were introduced to the model. It should be noted that there was a lot of missing data in meteorological information, especially for daily temperature data. In addition to the above information, daily flow data discharged from Amirkabir dam and technical specifications of the dam were introduced to the model. In addition, orchard management information, including irrigation periods and information related to phosphate fertilizers used in regional orchards, were presented to the model. The global sensitivity analysis method was used to determine the sensitive parameters of the model. Furthermore, the SUFI2 algorithm was used in SWAT_CUP software to calibrate and analyze the parametric uncertainty of the SWAT model. This algorithm quantifies the output uncertainty by 95% prediction uncertainty boundaries.
Results and Discussion
According to the results of sensitivity analysis, the parameters Baseflow alpha-factor (ALPHA_BF), Manning’s “n” value for overland flow (OV_N), and Precipitation Laps rate (PLAPS) were the most sensitive parameters to flow, sediment, and total phosphorus, respectively. The best Nash-Sutcliffe (NS) coefficients for runoff, sediment, and total phosphorus simulation obtained in all stations and after full calibration and validation periods were equal to 0.76, 0.56, and 0.92, respectively. Simulation of the peak points of the diagram of all three quantities was also associated with increased uncertainty and decreased model prediction accuracy, but due to the placement of more than 70% of the measured runoff and sediment values and nearly 60% of the measured total phosphorus values in the prediction uncertainty boundaries generated by SUFI2 algorithm the final value of the parameters used in the calibration process can be appropriate for simulating future scenarios in similar mountain catchments. The main weakness of the model is simulating the peak points of flow and sediment discharge. In the case of flow and sediment discharge, the liability of modeling can be generalized due to the lack of accurate prediction of the snowmelt inflow to the river in spring, which begins to increase in February and reaches the peak point in May. A considerable number of missing data in meteorological stations can effectively reflect the lack of accurate model prediction at the peak points. In this region, missing daily temperature data compared to other meteorological parameters has been significant. The dependency of the SWAT model on many experimental and quasi-experimental models such as SCS-CN and MUSLE can be another factor affecting the weakness in predicting the peak points of the sediment discharge, as well.
Conclusion
According to the uncertainty analysis results, most of observed flow, sediment and total phosphorus discharge values were within the uncertainty prediction boundaries generated by the SUFI2 algorithm. The NS coefficient for all three variables has met the satisfactory modeling threshold. Therefore, it seems that the sensitive parameters identified and used in the calibration process in this study and their final values can be appropriate for modeling future scenarios for this study area and similar mountain catchments. One of the limitations of the present study was a large number of missing data in meteorological stations, especially for the temperature variable. Thus, providing required measured meteorological data to the model may emhance the simulation, especially at peak points.
