Caizi Lake (117°01′‒117°10′E, 30°43′‒30°58′N), located in the northeast of Anqing City, Anhui Province, is a representative shallow water barrier lake. It has a north subtropical humid monsoon climate. The Caizi Lake Basin covers an area of 3234 km², and the annual average rainfall is about 1389.1 mm. The average elevation at the bottom of the lake is 8.5 m, and the normal water storage level is 11.0 m. The water area of the lake fluctuates from 145.2 to 242.9 km². The corresponding water level and total volume are 8.1‒15.1 m and 2.87×10⁸‒16.1×10⁸ m³, respectively (Wang et al., 2018b). The main tributaries of Caizi Lake include Dasha River, Guache River, Longmian River, and Kongcheng River. The lake water flows into the Yangtze River through the Zongyang sluice built in 1959. Zongyang sluice is a masonry sluice built on rock foundation. The designed maximum drainage flow is 1150 m³/s, and the check drainage flow is 1288 m³/s. The designed flood control water level difference is 4.5 m (the Yangtze River water level is 17.27 m, and the corresponding internal water level is 12.77 m). When the water level of Chefuling station in Caizi Lake is relatively low (< 11.5 m), in order to alleviate the drought in the area along Caizi Lake, Zongyang gate is usually opened to divert water from the river to the lake, and the water level of Chefuling station is controlled at 12.2‒12.5 m. When the water level of the Yangtze River rises sharply in the flood season, the Zongyang sluice will be closed to prevent river water from flowing backward.

In the Caizi Lake Basin, there are two hydrological stations, namely Shahebu in the upper reach of Dasha River and Zongyang sluice at the outlet of the basin. There is only one water level station (i.e., Chefuling) in the lake area. Anqing hydrological station is about 27 km away from the Zongyang sluice. The history and current hydrological data in this study were mainly obtained from the four hydrological stations mentioned above. Among them, the data collected by Chefuling station are mainly the monthly water level from 1989 to 2018, the daily water level and rainfall in 2018; the data of Zongyang sluice station include the daily water level in upstream and downstream of the sluice and the daily water flow in 2018; the data of Anqing and Shahebu stations are the daily water level and flow in 2018, respectively. The distribution of main water systems and hydrological stations around Caizi Lake is shown in Figure 1.

Firstly, the time series analysis of the water level in Caizi Lake was carried out by trend lines, box plot and moving average to understand the characteristics of annual and inter-annual fluctuations of water levels. It is noted that for linear regression trend analysis, a positive slope indicates an increase in water level, while a negative slope represents a decrease in water level. In addition, box plots of water level in different periods can not only visually identify the abnormal values of data batches, but also reflect the trend and range of water level fluctuation in each stage. Secondly, Mann-Kendall (M-K) testing (Wang et al., 2012) and wavelet analysis were used to determine the water level mutation point and periodicity. Thirdly, the correlation among the water levels of Caizi Lake, Zongyang sluice (above and below the sluice) and Anqing station of the Yangtze River was investigated by Pearson correlation analysis. Finally, the impact of river-lake isolation on the change of water level in Caizi Lake was further detected.

Wavelet analysis is a method for time series problems with multiresolution time-domain (MRTD) technique. It can clearly reveal a variety of change periods hidden in the time series, and fully disclose the changing trend of the system in different temporal scales. At present, wavelet analysis is becoming a common tool for analyzing non-stationary time series in signal processing, seismic exploration, hydrometeorology and many other fields (Biswas et al., 2013; Liao et al., 2020). A wavelet is a wave-like oscillation with amplitude that begins at zero, increases, and then decreases back to zero. Common wavelet basis functions are Dmey, Haar, Daubechies, Meyer, Morlet, etc. In this study, the complex Morlet wavelet was selected, which is the most commonly used in hydrology (Lee and Kim, 2019), and can be defined as:

where w is a constant, and i is an imaginary number.

For the time series f(x)∈L²(R), the continuous wavelet transform (CWT) is:

where W_f(a, b) is the wavelet transform coefficient, a is a scale factor, which reflects the frequency-domain characteristics, b is a time factor, which reflects the time-domain characteristics, and $\bar{\phi }\text{(}t\text{)}$ is a complex conjugate function of$\phi \text{(}t\text{)}$. The wavelet coefficient curve can accurately identify the existence of the periodic component in hydrological sequences.

The wavelet variance reflects the strength of the fluctuation energy of hydrological elements, and its change with time scale can be used to determine the main period. The wavelet variance Var(a) is calculated as:

3.1.1 Annual variation characteristics

The multi-year averages of monthly water level in Caizi Lake fluctuated greatly with a single peak, showing an obvious seasonal flood-drought alternation (Figure 2). During a hydrological year, the water level of Caizi Lake can be generally divided into the following stages: dry season (December to March), rising season (April to May), flood season (June to September) and retreating season (October to November). The average water levels of these four periods were 8.95, 9.69, 12.29 and 10.35 m respectively. The average water level was 3.34 m higher in the flood season than that in the dry season. The rising or falling rates of water level were approximately 0.83 and 1.54 m/month in the rising and retreating periods, respectively. Caizi Lake water level had a large fall in a year, showing that the falling rate of water level in the retreating season was faster than the rising rate of water level in the rising season. This indicates that there were differences in the mechanism of the lake water level change for each hydrological season. The main reason is that during the rising period, the precipitation increased in the area, and the water levels of Caizi Lake and Anqing station of the Yangtze River rose. In this case, the Zongyang sluice discharged water. The Yangtze River water replenished the lake water and had a significant impact on the discharge of the lake until the arrival of the flood season. However, the basin needed to extract irrigation water from Caizi Lake during the retreating period since the regional rainfall decreased and the Zongyang sluice still discharged water.

The monthly minimum water level in Caizi Lake was 7.79 m in November 2008, with a monthly maximum water level of 16.18 m in July 1999. The average water levels in January (the lowest) and August (the highest) were 8.85 and 13.04 m, respectively. The box plot in Figure 2 shows that the monthly water level of the lake was much more stable in the dry season than those in the other seasons.

3.1.2 Inter-annual variation characteristics

From 1989 to 2018, the annual average water level of Caizi Lake was 10.42 m, with the monthly highest, average and lowest water levels of 12.34‒14.10, 10.13‒10.76 and 8.65‒8.77 m (upper and lower quartile data), respectively. From Table 1, the slopes of linear fitting of the monthly highest, average water level, and water level in rising, flood and retreating seasons were negative, indicating a general decrease in water level. However, the slopes of linear fitting of the monthly lowest water level and water level in dry season were positive, showing a general increase in water level. Only the water level in dry season produced a P value less than 0.05 (Table 1).

From the perspective of the month, the average water levels of each month basically showed distinct “high-low” cycles with the years. The water level in January and July showed a certain trend of rise and decline with the increase of the year, and further appeared the “up-down-up” and “down- up” change law after 5-year moving average, respectively (Figure 3). However, there was no significant change for the other months.

3.1.3 Mutability and periodicity

The results of the M-K mutation testing for the monthly average water level of Caizi Lake showed that the UF and UB curves had 8 intersections, whereas the highest monthly water level had 12 intersections. In addition, the UF curve fluctuated around zero, which shows that the average and highest monthly water levels were very unstable in the past 30 years. No remarkable mutation was found for them. The lowest monthly water level appeared a weak downward trend before 1999, and then gradually increased from 2000 to 2011. Finally, it showed a rapid rise after 2012. It had only one mutation point at the 0.05 significance level (Figure 4). Further comparison with the results of sliding t-test, year 2013 could be identified as a significant mutation point of the monthly minimum water level.

From Figure 5a, there are two obvious peaks at the period scales of 128 (the first main period scale) and 18 months (the second main period scale). Figure 5b further shows that the periodicity was the most obvious at the period scale of 128 months. Figure 5c shows that the oscillating period of water level change was approximately 80 months (i.e., 7 a) at a period scale of 128 month, with about 4 “up-down” cycles. At the 18-month period scale, the oscillating period of water level change was probably 12 months, with 29 “up-down” cycles.

3.2.1 The hydrological relationship between Caizi Lake and Yangtze River

Table 2 lists the correlation coefficients between the water levels of the four hydrological stations and that in Caizi Lake. The correlation coefficient between the water levels of the sluice downstream and Anqing station was very high (r=0.996) in 2018 because the water flow below the Zongyang sluice was connected with the main stream of the Yangtze River. This indicates that the water level changes of the two stations were almost the same. Since the Changhe River upstream of the Zongyang sluice was connected with Caizi Lake, the fluctuation of the upper water level of the Zongyang sluice was similar to that of Chefuling in Caizi Lake (r=0.949). In addition, the correlation coefficient between the water level of Caizi Lake and that at Anqing station of the Yangtze River was 0.824. The correlation coefficients between the water levels of Caizi Lake and Anqing station in the dry, rising, flood, and retreating seasons in 2018 were 0.172, 0.735, 0.218 and 0.552, respectively (Figure 6).

3.2.2 Zongyang sluice discharge and water level changes in Caizi Lake

The characteristic water levels of Caizi Lake changed greatly before and after the completion of Zongyang sluice. Before the completion of the sluice, the lowest and highest water levels of the lake were 8.38 (February 24, 1958) and 17.96 (August 1, 1954) according to hydrological records of Anqing station, respectively. After the operation of Zongyang sluice, the lowest and highest water levels of the lake were 8.36 (April 23, 1983) and 17.29 (July 7, 2016), respectively. This implies that the operation of Zongyang sluice had better effect on the control of water level in flood season than that in dry season.

Figure 7 shows the water flow duration curves of the hydrological stations of Zongyang sluice and Shahebu in 2018. As can be seen from this figure, the release time of the Zongyang sluice gate was mainly from May to July (the average flow was 57.3 m³/s) and during the retreating period (from mid-September to mid-October), with the average flow 108.2 m³/s. During the main flood season from July to September, the sluice gate was closed most of the time. The gate was opened for a few days in order to let the water from the Yangtze River pour into the Changhe River, so as to alleviate the high water level of sluice downstream. To some extent, it explains the reason why the rising and retreating seasons of Caizi Lake lagged behind that of Anqing station of the Yangtze River, and the rising and falling rates of the water levels were slower as shown in Figure 6.

In 2018, the average flow of Shahebu station was 7.9 m³/s, which was relatively small and mainly concentrated from March to August, with a mean flow of 13.1 m³/s. From September 11 to October 7, 2018, Caizi Lake was in the retreating season with little rainfall. During this period, Zongyang sluice gate was opened for 27 consecutive days to release water, with an average flow of 153.5 m³/s, far exceeding the average discharge of Shahebu station (3.2 m³/s) in the same period. Table 3 shows that the runoff in the upper reach of Shahebu station accounted for about 20% of the total runoff in the Caizi Lake Basin, reflecting that the discharge of Zongyang sluice was much greater than runoff inflow into the lake in this period.

In 1959, before the flood season, Zongyang sluice has been put into operation, which broke the connection of Caizi Lake with the Yangtze River. In fact, as a freshwater lake in the outflow region of China, the primary driving factor for the change in water level is the regional rainfall characteristics affected by monsoon (Qin and Xie, 2006). Figure 8 shows that the shapes, trends and inflection points of the variation curves of cumulative rainfall and cumulative increase in water level were similar. In addition, there was a good linear relationship (R²=0.992, P<0.01) between the cumulative rainfall and the cumulative increase in water level. This proves that precipitation played a dominant role in controlling the water level change of the lake. In Figure 5c, the water level of Caizi Lake presents a periodic “up and down” conversion, which indicates that the lake experienced frequent alternation of drying and wetting.

Historically, Caizi Lake was a river-connecting lake. Caizi Lake water level change should be basically consistent with the Anqing section of the Yangtze River (Zhang et al., 2015). But the correlation coefficients between the water levels of Caizi Lake and Anqing station were relatively low after the sluice was built (Table 2). This may be because the Anqing station of the Yangtze River was significantly affected by the upstream flow. At a certain water level, the Zongyang sluice gate was closed and the hydraulic connection between Anqing station and Chefuling station was cut off (Li et al., 2019). In this case, the water level changes of these two stations were not consistent. Although the overall trend of the water level duration curves in the rising and retreating periods were similar, the rising and retreating periods of Caizi Lake had a time lag of about one month compared with Anqing station as affected by the Zongyang sluice. Compared with the Anqing station of the Yangtze River, the water level of Caizi Lake rose more slowly and fluctuated less in flood season (Figure 6). The completion of Zongyang sluice is beneficial to the regulation of water storage capacity of Caizi Lake and the maintenance of the relatively stable water level during the dry season. During the past 30 years, the steady rise of the water level in Caizi Lake during the dry season (e.g. January) was found (Figure 3), which may be related to the increase of precipitation in the lower reach of the Yangtze River caused by global change (He et al., 2013). However, in a flood year, due to the high water level of the main stream of the Yangtze River, Zongyang sluice cannot be opened to discharge the flood for Caizi Lake. In case of frequent rainfall events in the region, the water level of Caizi Lake will continue to rise, which is likely to cause high water level, or even extreme value. This is the reason why the highest water level of Caizi Lake occurred in July 2016.

Zongyang sluice has changed the river-lake relation, and has a great impact on the hydrological regime of Caizi Lake. Its eco-environmental effects are also complex. Except for the damage to the lake-river connectivity caused by the construction of the sluice, the continuous high water level of Caizi Lake caused by poor drainage during the flood period is the greatest risk at this stage. This can only be alleviated gradually by continuously strengthening the construction of water conservancy projects such as river reconstruction, dredging channels, digging ponds, repairing reservoirs and building dams in the entire basin. At present, YHWD is currently in full swing. The west line of the Yangtze-to-Chaohu Water Diversion Project, which is an important part of YHWD, will divert water from Zongyang gate to Caizi Lake, which would have a great influence on the water level in the dry season of Caizi Lake (Li et al., 2019). On the one hand, the long-term and large-scale water diversion during the dry season will significantly raise the water level of Caizi Lake, which will lead to the reduction of the wetland area and the destruction of the habitat of overwintering migratory birds. On the other hand, the period for water diversion is relatively short, so the water level of Caizi Lake will rise in a short period of time. According to the succession law of the lake area, the wetland will gradually sink, while the lake area will gradually become larger. This results in the aquatic vegetation becoming more and more dominant. Meanwhile, the upstream sediment will gradually accumulate. If the speed of wetland deposition is slower than that of sediment, the lake area will be reduced. Obviously, the change of hydrological regime in Caizi Lake is about to face a double influence from river-lake isolation and YHWD. In addition, the period scale of water level change was approximately 128 months in Caizi Lake. This may be partly explained by the fact that the period scale of precipitation variation was about 10 years in the middle and lower reaches of the Yangtze River during the last 50 years (Zheng et al., 2013). From 2018, the next 3‒4 years are likely to be the low water level period based on the wavelet analysis since the wavelet coefficients are changing from positive to negative (Figure 5c). This kind of “up-down” cycles may have some influence on the wetland ecosystems (Hu et al., 2010).

In order to understand the influence of YHWD on Caizi Lake, relevant departments and scholars have carried out some research work in this field, but it still seems very limited. Liu et al. (2019) showed that the daily average water level change rate during the overwintering period of migratory birds in Caizi Lake was ‒0.051 to 0.016 m/d, and the annual average rate was ‒0.034 to 0.009 m/d. Li et al. (2019) found that more than 20,000 wintering waterbirds of 49 species were recorded at Caizi Lake. Compared to the current situation, 54.3% of the grassland and 60.5% of the mudflats were predicted to be lost during winter due to expected water level rise. At present, the adaptive operation of Caizi Lake water level from November to next April during 2018-2023 is a targeted and important study for YHWD. From Figure 3a, the water level in the dry season has been rising during the past 30 years. However, according to the plan, after the operation of the YHWD project, the water level of Caizi Lake will be higher than that under natural condition (Wang et al., 2018a). In this case, the influence of water level rise in dry season should be considered in the optimal regulation of Caizi Lake water level by YHWD, and the relationship between water level and wetland ecological environment should be quantitatively evaluated in the future.