3.1. Calibration and Validation of the Model
In this study, the parameter sensitivity for the Nakdong River Basin was examined prior to the calibration and validation of the model, and optimal parameters were selected through this process.
Table 4 summarizes the parameters used for the calibration and validation of the model.
The calibration period of the SWAT model was set to six years (2005–2010), and its validation period was set to seven years (2011–2017). For the six weirs (SJW, GMW, CGW, GJW, HCW, and HAW), three years (2013–2015) were set as the calibration period and two years (2016–2017) were set as the validation period, because they began operation in August 2012.
The statistical results for hydrology and water quality for the model calibration and validation are summarized in
Table 5 and
Figure 3. The applicability of the model was evaluated using the coefficient of determination (R
2), Nash and Sutcliffe [
23] model efficiency (NSE), root-mean-square error (RMSE), and percent bias (PBIAS). As R
2 approaches 1, the observed values fall in perfect agreement with the simulated values. NSE evaluates the efficiency of the model in the range from −∞ to 1. When it is higher than 0, the applicability of the model can be said to be high [
24]. RMSE represents the error between the measured and simulated values, and as it is close to 0, the error is small. PBIAS represents the error between the observed and simulated values as a percentage. A smaller error value means that the model has higher efficiency [
22].
The statistical analysis results of the dam inflow showed that R
2 ranged from 0.71 to 0.90 and that NSE ranged from 0.59 to 0.78 for the entire period. RMSE ranged from 1.14 to 1.69 mm/day, while PBIAS ranged from −18.04 to 7.32%. The statistical analysis results were significant for all calibration and validation points. Regarding the water quality R
2 results for the entire period, the R
2 of SS ranged from 0.58 to 0.83, while that of T-N ranged from 0.53 to 0.68. T-P exhibited a correlation between 0.56 and 0.79%. Diff was calibrated to be 35% or less, as suggested by Donigian [
25]. The SWAT constructed in this study did not consider the QULE2E module.
3.2. Analysis of Streamflow and Water Quality Interaction in 2017 According to the Dam-Weir Combined Operation Scenarios
In this study, streamflow and water quality interaction were examined by standard unit watershed for the year 2017 by constructing input data for the dam-weir combined operation scenarios. Prior to the simulation, the actual observation data from 2005 to 2016 were applied to stabilize the SWAT.
Dam operation proceeds similar to the operating conditions under which the water supply capacity of the existing dams was investigated. By operating weirs, the water level can be maintained at a constant level, in contrast to dam operation. The operating rule is applied based on the rules suggested by ME [
26].
Figure 4 summarizes the streamflow and water quality simulation results by scenario for 2017. In the figure, the simulation results of each scenario and the changes compared to observed data were shown based on the spatial distribution.
In terms of streamflow, the average annual flow rate of the entire Nakdong River Basin in observed data was found to be 28.7 m
3/s. Scenario 1, in which dam release was simulated, showed a decrease in the average annual streamflow by 0.05 m
3/s compared to the observed data, as the annual average streamflow from HCW to the downstream area of the Nakdong River decreased. Scenarios 2, 3, 4, and 5 exhibited increases in the average annual streamflow by 0.13~0.79 m
3/s compared to the observed data, which were attributed to the simultaneous release of the dams and weirs. Scenario 8, which involved weir gate full open sequential release, showed an increase in streamflow by 0.55 m
3/s compared to the observed data (
Figure 4a).
However, the water quality improvement and deterioration phenomena were different in each section. The simulation results of SS in observed data showed an annual average SS load of 120.0 ton/day. In the downstream case in ADD and IHD, the SS loads decreased, but in the SJW-CGW and NKD downstream-HAW sections, which are the areas downstream of the dams, the SS loads increased. Scenarios 4, 5, 7, and 8, in which weir gate full open release was simulated, show increases in SS loads by 8.5~11.3 ton/day compared to the observed data (
Figure 4b). Based on the relationship between SS and streamflow described in previous studies [
27,
28,
29], it is likely that high inflow will be accompanied by the transport of high SS from upstream to the estuary.
Regarding the simulation results of T-N, the T-N loads decreased in all scenarios compared to the observed data. The simulation result shows that the weir release scenarios were more effective in reducing T-N loads than the dam control scenarios. In Scenarios 2 and 3, in which dam and weir control was simulated, the T-N load decreased by 3.5 to 4.7 kg/day, while Scenario 6 reduced the T-N load by 5.8 kg/day when only weir was released sequentially without dam control (
Figure 4c).
As shown in
Figure 4d, the simulation results of T-P revealed that the T-P loads in the GMW-CGW and HCD-HCW sections were increased in dam control scenarios compared to the observed data, but were reduced in the HCW-Nakdong River estuary section, as was the case with the T-N simulation results. In the cases of Scenarios 7 and 8, the T-P loads decreased in the HCW-Nakdong River estuary section. In Scenario 8, which exhibited the largest load change, the T-P load decreased by 50.5 kg/day compared to the observed data in the Nakdong River estuary.
Table 6 presents the monthly streamflow and water quality (SS, T-N, and T-P) changes at MG stations from June to December in 2017, which is when the dam-weir combined operation scenarios were applied. The calculation of the water quality reduction efficiency was ranked based on the reduced load compared to the observed data.
Scenarios 4 and 8 showed the greatest effect of reducing water quality loads. Although the two scenarios have something in common to fully open the weir gate, Scenario 4 simultaneously releases dam and weirs, while Scenario 8 opens only weirs sequentially.
For the operation of dams and weirs in the case of the full gate open (Scenario 4), the river flow was 325.35 m3/s, but the SS load was increased by 67.51%. In the case of Scenario 8, the SS load increased by 49.11% compared to the observed data, but this increase was lower than that in Scenario 4. This was because the SS load increased in the downstream of ADD and IHD by dam control. The reductions of T-N and T-P loads in Scenarios 4 and 8 were 1.00~1.12% and 1.14~1.45%, respectively.
Similar relationships between streamflow and water quality were found in previous studies [
30,
31,
32]. This is reflected in T-N and T-P loads and wash off, which were concentrated in the dam and weir operation, showing clear patterns of scenarios. The change in characteristics could be explained in that the environmental condition changes substantially when the hydrologic structures begin operating, and the streamflow changes lead to SS, T-N and T-P changes in the water body.
Figure 5 and
Figure 6 show monthly water quality simulation results of the observed data as well as Scenarios 4 and 8. The analysis results for each scenario showed improvement in the watershed in the simulated results of T-N and T-P. The analysis results were illustrated on the map by sub-basin, and the areas where rapid changes were made in the downstream where the dams and weirs were operated were derived.
Scenario 4 exhibited T-N and T-P load reduced compared to the observed data for the simultaneous release of the dams and weirs. Scenario 4 exhibited T-N and T-P load reduction effects within the shortest period of time among the scenarios by causing the largest streamflow change within that period of time and instantly increasing the stagnant flow velocity of the Nakdong River estuary. However, it was confirmed that water quality (SS, T-N, and T-P) worsened at the downstream of five dams.
Scenario 8 exhibited T-N and T-P load reduced compared to the observed data from July to September when the sequential release occurred. However, T-N and T-P loads in October after the completion of sequential release were temporally increased compared to the observed data.
In general, water quality is closely related to streamflow, and it is known to worsen during water shortage events in dry seasons; therefore, the increased streamflow according to the dry season operation method should result in water quality improvements. The highest positive effects appearing in September and November indicate the possibility of reduced T-N and T-P loads in the estuaries during those periods. This is attributed to the high inflow volume observed during those periods. In most cases, because streamflow higher than those secured through the natural water cycling system of the river are required, the increased streamflow of the water body resulting from coordinated operations of hydrological facilities will have positive effects on the environment. In addition, the increased loads in December could also be associated with low temperature, and T-N and T-P wash-off also increased due to the increase in release compared to the observed data (
Figure 5 and
Figure 6).
When the water level drops the target release, the actual operating conditions should be adjusted appropriately according to the water storage and flow rate conditions of the dams and weirs in the water systems [
19]. This study found that water quality improvement could be maintained through sequential weir operation (Scenario 8) while minimizing the operation of the hydrological facility.