Results

Hydrologic Model

SWAT calculates hydrological processes based on HRUs in each sub-basin in the basin. HRUs were produced by the unique combination of soil, land use, and slope in each sub-basin. After the physical and chemical parameters of the soil, land use characteristics, pollution sources and management processes of the basin were defined in the model, together with the climate data, the model was simulated between 1991-2007, after 3-years warming period, using the monthly time step in the model. Asymmetric distribution was used for precipitation distribution, and SCS (Soil Conservation Service) method was used for surface runoff (USDA 1986). Bagnold method (Bagnold 1966) was applied for the sediment transport. The results of the simulated flow rate were compared to the values of the stream gauge station on the Namazgah river results of the model. Model accuracy was determined based on Nash-Sutcliffe efficiency (NSE), and coefficient determination (R2). According to the comparison between simulated and observed discharge values on the stream gauge station on the Namazgah river, NSE, and R2 are 0.48 and 0.58. Although the accuracy values of the model indicate that the model did not reflect the real world enough, the graphical comparison of the flow rate between 1991 and 2007 shows that features of the simulated flow rate are similar to observed flow rates.
While the 1991-2002 period was used for the calibration of the model, the 2003-2007 period was used for the validation. The initial model run had r2=0.58, NS=0.48 improved to r2=0.63, NS=0.56, after the calibration by using the FACT with 200 simulation numbers. The model run had r2=0.80, NS = 0.68 after the validation (Fig. 4). The accuracy of the model results is good based on Moriasi et al. (2007) model performance assessment. Unfortunately, there is no monitoring station for water quality assessment on the Namazgah river. So, the model accuracy assessment in terms of water quality measurements couldn’t be made. However, SWAT is the most popular model since ungauged watersheds can be modelled accurately by SWAT (Gassman et al. 2014). Moreover, SWAT simulates long-term impacts of land use, land management practices and buildup of pollutants with a continuous time model (Neitsch et al. 2005).

Climate Change Impacts

The General Directorate of Meteorology has developed climate projections for the 2016-2099 period using HadGEM2-ES, MPI-ESM-MR, GFDL-ESM2M global model data sets in order to reveal how climate change will affect Turkey in the future. In this study, global model data sets RegCM4.3.4 regional model and dynamic scale-down method, according to the RCP 4.5 and RCP 8.5 scenarios, the results of the projection of 20 km resolution with 1971-2000 reference period 2016-2040, 2041-2070, 2071-2099 future periods were obtained. Climate change data obtained from the study of the General Directorate of Meteorology was used to show the effects of climate change on the flow rates and NPS pollution loads in the Namazgah dam basin.
According to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC 2014), Representative Concentration Pathways are new scenarios considering global greenhouse gas and aerosol concentrations and alternative future scenarios (SRES) as a prelude. There are four RCPs defined as 2.5, 4.5, 6.0 and 8.5 depending on the total radiative forcing path and level until 2100. In this study, RCP 4.5 (540 ppm CO2) and RCP 8.5 (940 ppm CO2) scenarios were selected to reveal the effects of future climate projections on the Namazgah dam basin. While RCP 4.5 assumes a long-term level of medium greenhouse gas concentrations with broadly pre-defined strain stabilization constraints, RCP 8.5 acknowledges that greenhouse gas emissions will increase over time in the 21st century and approach very high levels by 2100 (IPCC 2014). According to the RCP 4.5 and RCP 8.5 scenarios, the estimated monthly mean total precipitation and changes in average temperature for the years 2021-2090 were compared with the measurements of the meteorological station in the basin between 1979-2014. A decrease in the monthly average total amount of precipitation is estimated between 2021-2090. Although more rainfall was observed in autumn and spring between 1979 and 2014, the highest rainfall is expected in the summer period in 2022-2090 (Fig. 5a). An increase of approximately 0.49oC is expected in average monthly temperatures 2021-2090 compared to 1979-2014 concerning the RCP 8.5 scenario, while a decrease of 0.35 oC is expected concerning the RCP 4.5 scenario (Fig. 5b).
After the hydrological model was run for 2021-2099 according to the RCP 4.5 and RCP 8.5 scenarios, the effect of climate change on flow rate and NPS pollution loads was shown. Mean flow rate values according to RCP 8.5 and RCP 4.5 scenarios, respectively; It is estimated at 0.67 and 0.68 m3/sec. When predicted flow rate values based on RCP 4.5 and RCP 8.5 were examined for 2022-2047, 2048-2072, and 2073-2099 periods, 0.64, 0.70; 0.71, 0.68, and 0.80, 0.64, m3/sec, respectively were predicted. TN values based on RCP 4.5 and RCP 8.5 for 2022-2047, 2048-2072, and 2073-2099 periods changes 43947.02, 48987.89; 50653.91, 46405.73, and 54913.97, 46246.08, respectively. TP values based on RCP 4.5 and RCP 8.5 for 2022-2047, 2048-2072, and 2073-2099 periods varies 11474.19, 13808.91; 13943.31, 11959.92, and 14174.19, 12035.26, respectively (Fig. 6, Table 2).

Scenario Analysis

Land-use scenarios were developed and explored to understand the sensitivity of model outputs to understand the impact of land-use/cover changes on the flow rates and NPS pollution loads of the Namazgah river. The land-use scenarios were chosen according to the Regulation on Protection of Drinking and Utility Water Basins (PDUWB 2017). Forest areas are protected and enhanced, the current agricultural areas are protected and it isn’t permitted to increase their areas. Thus, based on the regulation, two different scenarios were chosen to observe land use changes impacts on water quantity and quality, which are; conversion of shrubland to the forest and, conversion of agricultural areas to the forest
conversion of shrubland to forest
Examining the impacts of the conversion of shrubland to forest areas shows that mean flow rate values according to RCP 4.5 and RCP 8.5 scenarios were estimated as 0.8 and 0.67 m3/sec, respectively. Predicted flow rate values based on RCP 4.5 and RCP 8.5 for 2022-2047, 2048-2072, and 2073-2099 periods were 0.72, 0.79; 0.79, 0.77; and 0.87, 0.73 m3/sec, respectively. TN values based on RCP 4.5 and RCP 8.5 for 2022-2047, 2048-2072, and 2073-2099 periods changes 75200.2, 83661.618; 106754, 112049.514, and 125218, 104668.712, respectively. TP values based on RCP 4.5 and RCP 8.5 for 2022-2047, 2048-2072, and 2073-2099 periods varies 13562.3, 16497.508; 15897.7, 14410.515; 15750.6, 14627.425, respectively (Fig. 7, table 2).
conversion of agricultural areas to forest
According to the RCP 4.5 and RCP 8.5 scenarios, the average flow rates were estimated as 0.51 and 0.77 m3/sec, respectively, based on this scenario. Estimated flow rate values based on RCP 4.5 and RCP 8.5 were examined for 2022-2047, 2048-2072, and 2073-2099 periods, 0.727, 0.5298; 0.798, 0.504, and 0.876, 0.48 m3/sec were predicted, respectively. TN values based on RCP 4.5 and RCP 8.5 for 2022-2047, 2048-2072, and 2073-2099 periods changes 75200.2, 36282.471; 106754, 32743.9884 and 125218, 32592.5749, respectively. TP values based on RCP 4.5 and RCP 8.5 for 2022-2047, 2048-2072, and 2073-2099 periods varies 13562.3, 12034.256; 15897.7, 10499.158, and 15750.6, 10543.612, respectively (Fig. 8, table 2).

Statistical Assessment of Model Results and Scenario Analyses

Descriptive statistics of the results obtained from the modeling studies are presented in Table 3. Considering two climate scenarios and two land use scenarios, TN load, TP load and Q Tukey HSD were compared using multiple benchmarks. Statistical significance level of two and higher interaction terms was determined using full factorial design analysis of variance (ANOVA). LSM values (which can also be defined as adjusted means) are values predicted by the model for certain level combinations of categorical variables when all other model factors are set to neutral values. This approach was chosen due to unequal sample sizes (Table 4).
According to the Tukey HSD multiple comparison test, it was concluded that there would be no significant change in the TP load in the future depending on the climate and land use scenarios. On the other hand, it is estimated that there may be significant changes in TN load and Q between some scenarios. There is no significant difference between S1, S4 and S5 scenarios in terms of TN load. Again, it is estimated that there will be no difference between S2, S3 and S6 scenarios in terms of TN load. On the other hand, it is expected that there will be a significant difference between scenarios S1, S4 and S5 and scenarios S2, S3 and S6. In terms of flow rate (Q), a significant difference is expected between the S5 scenario and other scenarios. On the other hand, it is predicted that there will be no significant difference between scenarios S1, S2, S3, S4 and S6.
The changes that may occur in TN, TP and Q depending on the climate and land use scenarios were also compared with the data of the previous period (1991-2007) by using Dunnet’s test (Table 5). Depending on the climate and land use scenarios, when the 2022-2099 period is compared with the 1991-2007 period, it is estimated that there will be an increase in phosphorus loads and a decrease in stream flow rate. On the other hand, it is expected that the TN load will decrease only in the S5 scenario and increase in all other scenarios. When evaluated together with the data presented in Table 5, the changes that may occur in the TP load in all scenarios will not be significant when compared to S0. On the other hand, the change that may occur when compared with the flow rate of all scenarios will be statistically significant. In terms of TN load, the difference between S0 and S1, S4, S5 scenarios is statistically insignificant, whereas the difference between S0 and S2, S3 and S6 scenarios will be significant.