4 Results

4.1 Impact of reservoir on runoff

Before the baseflow recession analysis, we first analyze the impact of the construction of the Chaersen Reservoir on the runoff of the downstream Zhenxi station. The streamflow duration curves of the two comparison periods before and after the construction of the reservoir are plotted, respectively, as shown in Fig. 2. After the construction of the reservoir, the peak flow of Zhenxi station decreases significantly, while the low‑flow increases significantly (Fig. 2c). The streamflow of the two upstream hydrological stations (Suolun and Dashizhai) are not significantly different in the comparison periods (Fig. 2a, Fig. 2b), and the deviations in some stages may be mainly caused by the differences in climate characteristics. In order to quantitatively analyze the impact of reservoir construction on runoff, we have calculated some characteristic values of streamflow (maximum: Qmax, minimum: Qmin, average: Qmean, seven‑days minimum: 7Qmin) in the comparison periods, and the percentage bias between them, as shown in Table 2. The streamflow characteristic values of Suolun station in the post‑comparison period is generally smaller, with a bias of -16% to -24%. However, these values of Dashizhai station in the post‑comparison period are generally larger, with a bias of 11% to 60%. This opposite deviation may be caused by the spatial difference of climate characteristics, which will cancel each other when converging to downstream Zhenxi station. Considering the large difference between the streamflow magnitude of the two upstream hydrological stations, this study estimates the impact of climate change on the downstream Zhenxi station by weighted average of the two opposite deviations, with the proportion of average streamflow as the weight (Table 2). After excluding the impact of climate change, the construction of Chaersen Reservoir resulted in a 14% decrease in Qmean of the downstream Zhenxi station, a 35% decrease in Qmax, a 380% increase in Qmin, and a 235% increase in 7Qmin.
Figure 2. Streamflow duration curves before and after the construction of reservoir. (a) Suolun station, (b) Dashizhai station, (c) Zhenxi station.
Table 2. Characteristic values of streamflow before and after the construction of reservoir, m3/s.

4.2 Impact of reservoir on baseflow recession characteristics

According to the method in section 3, the recession coefficients of the two comparison periods are calculated, and the results are shown in Fig. 3. In general, after the construction of the reservoir, the recession coefficient a of the downstream Zhenxi station is obviously reduced, while the coefficient b is obviously increased. No significant differences are found in the recession coefficients of the comparison periods at the upstream Suolun and Dashizhai stations (Fig. 3). There are inherent differences in the coefficients calculated by different methods, Stoelzle et al. (2013) has discussed it in detail. The recession coefficients obtained by linear regression and binning methods are not significantly different, but these obtained by the lower envelope method are significantly lower. Different baseflow recession segments extraction methods always cause certain fluctuations in the coefficients. Therefore, we use the same recession analysis method combination to avoid misjudgment due to the systematic error between different recession analysis methods (e.g. using the average result of different recession extraction methods under the same parameter optimization method, Table 3).
In order to quantitatively analyze the impact of the reservoir, we calculate the percentage bias between the average results of the two comparison periods, as shown in Table 3. Under different parameter optimization methods, the average recession coefficients a in the post‑comparison period of Suolun station are smaller, with a bias of -3% to -10%, and the average values of b are larger, with a bias of 0% to 3%. The average a of Dashizhai station are also smaller, with a bias of -12% to -15%, and the average b are slightly larger, with a bias of -2% to 6%. The deviation directions of the two upstream hydrological stations are the same, so this study uses the average of the two hydrological stations to estimate the bias caused by the climate change between the comparison periods of the downstream Zhenxi station (Table 3). After excluding the influence of climate change, the recession coefficient a of different parameter optimization methods decreased by 57% ~ 63% (the average is 60%), and the b increased by 21% ~ 27% (the average is 24%). Although different parameter optimization methods will get different results, the impact degree of reservoir construction on the analysis results of different methods is basically the same.
In order to more intuitively observe the influence of reservoir construction on the baseflow recession characteristics, we have drawn the master recession curves of the comparison periods before and after the construction of the reservoir, the equation is\(\ Q_{(t+t)}=\left[a\left(b-1\right)t+Q_{t}^{1-b}\right]^{\frac{1}{1-b}}\), as shown in Fig. 4, Fig. S4 and Fig. S5. When the recession starts with average streamflow (Fig. 4a), the master recession curves of Zhenxi station before and after the construction of the reservoir intersects. And the master recession curve after the construction of the reservoir is significantly higher in the later stage of the recession, and the recession rate becomes significantly smaller. However, the master recession curves of Suolun and Dashizhai stations do not change significantly before and after the construction of the reservoir. When the recession starts with same streamflow (Fig. 4b), the master recession curves of Zhenxi, Suolun and Dashizhai stations before the construction of the reservoir are basically the same. However, after the construction of the reservoir, the master recession curve of Zhenxi station is significantly higher. In other words, the construction of the reservoir slows the baseflow recession rate and increases the drainage time, causing the master recession curve to shift upward.
Changes in the recession coefficients and the master recession curve also mean changes in the basin-scale storage-discharge (S-Q) relationship. Based on the nonlinear reservoir theoretical model (Sect. 3.2), the S-Q relationship curves before and after the construction of the reservoir are drawn, as shown in Fig. 5, Fig. S6, and Fig. S7. Before the construction of the reservoir, the S-Q relationships of the three hydrological stations are approximately linear, but the nonlinearity of the S-Q relationship of Zhenxi station increases significantly after the construction of the reservoir. In addition, after the construction of the reservoir, the storage capacity of Zhenxi station increases significantly. When the baseflow is 5m3/s (60th percentile), the storage capacity is 3.1 times before the reservoir construction; When the baseflow is 20m3/s (32thpercentile), the storage capacity becomes 2.1 times as before; When the baseflow is 40m3/s (22thpercentile), the storage capacity becomes 1.7 times as before. That is, with the increase of baseflow, the increase rate of basin‑scale storage becomes slower and slower.
Figure 3. Recession coefficients analysis results of different methods. LR: linear regression, BIN: binning, LE: lower envelope.
Table 3. The average recession coefficients of different parameter optimization methods. The bias is calculated with 1982~1986 as reference.
Figure 4. Master recession curves, with the recession coefficients calculated by linear regression (LR) method. (a) Recession starts with average streamflow, (b) recession starts with same streamflow.
Figure 5. Storage-discharge (S-Q) relationship curves, with the recession coefficients calculated by linear regression (LR) method.