The LYR starts in Hukou County at the mouth of the PYL and ends at the Yangtze River Estuary (YRE), with no large tributaries whose total length is over 1,000 km merging along the way. The Datong hydrological station is located approximately 219 km below the Hukou hydrological station and 624 km above the YRE. Hydrological drought is affected by upstream runoff and downstream tides in different periods of the year. Datong is the last hydrological station on the Yangtze River with daily average flow discharge ( Q ¯ d ) and water level ( Z ¯ d ) data before flowing into the ocean. Therefore, this study used representative data from the Datong station to study the hydrological drought situation in the LYR.

Runoff at the Datong station mainly originates from the UYR, large tributaries, and water systems in the MYR, such as the Han River (HR), DTL, and PYL water systems. In this study, runoff sources in Datong were divided into five categories. The first originated from the UYR, which was represented by the Yichang hydrological station. The second originated from the DTL water system, which is the sum inflow of the Xiang, Zi, Yuan, and Li Rivers and is represented by the Xiangtan, Taojiang, Taoyuan, and Shimen hydrological stations. At the outlet of DTL, Q ¯ d measurement data are collected by the Qilishan station, which can be used to characterize the amount of water flowing out of DTL. These data also include the amount of water that flows into DTL from the main stream of the Yangtze River through the Songzi, Taiping, and Ouchi entry points. These data are also collected by the Yichang station. Therefore, to avoid duplicate statistics, this study takes the sum of the inflow of the above four rivers in the second category originating from the DTL water system. The third was from the HR, which was represented by the Xiantao hydrological station. The fourth originated from the PYL water system, which was determined as the sum of inflows of the Gan, Fu, Xin, Rao, and Xiu Rivers. The total runoff of these rivers was represented by data from the Hukou hydrological station, which is located at the exit of the PYL. The fifth comprised runoff from other small tributaries. The study area and the diversion and confluence relationships between the rivers and lakes are shown in Figure 1 .

The data used in this study were the measured Q ¯ d and Z ¯ d values from hydrological stations in each river ( Figure 1 ). To analyze the impact of tides on hydrological drought in the study area, hourly tide level (Z_h ) data from the Nanjing, Zhenjiang, Tianshenggang, Yingchuangang, Xuliujing, and Baimao tide stations in the LYR were also collected. The data for each station are listed in Table 1 .

Reliable data are crucial for trend and mutation tests. All data were sourced from the Bureau of Hydrology, Changjiang Water Resources Commission. These data underwent hydrological data compilation (China, 2020) before being released. Table 2 presents the reliability analysis results of the annual minimum Q ¯ d and Z ¯ d at Datong station. Data from all 73 years were deemed valid, with a standardized Cronbach’s alpha (Johnson, 2021) of 0.949, indicating very high reliability.

The Mann–Kendall method is typically used to analyze river hydrology (Du et al., 2022). This method was applied to conduct trend and mutation evaluations of hydrological droughts. We used a data series (x ₁, x ₂, …, x_n ) with assumed independent and random distributions. First, statistic D is calculated as follows:

where x_i and x_j are the observed data corresponding to the i and j time series, respectively, and i < j . sgn is a symbolic function:

After standardizing the D value, the test statistic Z is calculated as follows:

where V a r ( D ) denotes variance. For a given significance level α, if | Z | ≥ Z 1 − α / 2 , the original assumption is unacceptable; more specifically, there is an obvious change trend at significance level α for the time series data. In this method, the p-value was inversely calculated through the Z 1 − α / 2 . If p > α , there is no changing trend in the sequence and vice versa. Z > 0 indicates an upward trend, whereas Z < 0 indicates a downward trend.

For mutation detection, S_k is used to represent the cumulative number of the jth sample data x j − x i when ( 1 ≤   i   ≤   j ) :

The S_k is standardized as follows:

where E ( S k ) is the mean of S_k and V a r ( S k ) is the variance. For a given significance level α , | U F k | > U α indicates an obvious change trend in all datasets. UF_k can form a curve. UB_k can be calculated using this method to produce an inverse sequence. At a given significance level, for example α = 0.05 , the critical value is ±1.96. The UF_k and UB_k curves and ± 1.96 straight lines can be drawn in one chart. If U F k > 0 , the dataset exhibits an upward trend, with a downward trend for U F k < 0 . If these variables plot beyond the critical straight line, the change in the trend is significant. A zone exceeding the critical boundary was defined as the moment at which a mutation appeared. If UF_k and UB_k cross and the crossover point is within the critical line, the time corresponding to the intersection is the start time of the mutation.

Figure 2 shows the trend and mutation analysis results of the annual minimum Q ¯ d and Z ¯ d at Datong station over the past 70 years. The trend analysis results ( Figures 2A, B ) show that the p values of the annual minimum Q ¯ d and Z ¯ d were all close to 0, below the confidence level of 0.05, and the Z values were all >0, indicating that the annual minimum Q ¯ d and Z ¯ d exhibited an upward trend. The mutation analysis results ( Figures 2C, D ) reveal that there were no cross-phenomena between UB and UF within the confidence interval for the annual minimum Q ¯ d , indicating a non-significant mutation. However, for the minimum Z ¯ d , the UB and UF curves crossed within the confidence interval around 1989, indicating a mutation during this period.

Notably, the date of occurrence of the minimum Q ¯ d within a year did not always correspond to that of Z ¯ d at Datong. Figure 3 shows the statistical results of this mismatch. Since the 1950s, the number of years in which the two events occurred on different dates has increased each year ( Figure 3A ). In particular, since 1996, the dates on which the two occurrences appeared each year have been almost inconsistent, and their cumulative frequency has annually increased, reaching 53.4% by 2023 ( Figure 3B ). Figure 3C gives the number of days in which the minimum Z ¯ d appeared behind the Q ¯ d within the year. In rare years such as 1956, 1998, and 2011, the minimum Q ¯ d and Z ¯ d values appeared at the beginning and end of the year, respectively, with an interval of more than 320 days. In the past 20 years, there has been an upward trend in delayed days, with 5 years in which the delayed days exceeded 26 days and another 4 years ranging from 3 to 5 days. This result indicates that the minimum Z ¯ d at Datong within the year was not entirely determined by the minimum Q ¯ d but was also related to other factors.

Figure 4 shows the trend and mutation analysis results for the monthly average flow discharge ( Q ¯ m ) and water level ( Z ¯ m ). Regardless of Q ¯ m or the Z ¯ m , the p-values from January to March before the flood season and from September to November after the flood season were all <0.05, indicating a significant trend in the Q ¯ m and the Z ¯ m . The D-value from January to March was >0, indicating an upward trend in these 3 months, whereas the D-value from September to November was <0, indicating a downward trend in these 3 months.

The mutation analysis results ( Figure 5 ) indicated that the UB and UF of the Q ¯ m from January to March and from September to November both crossed within the confidence interval, indicating a mutation for Q ¯ m . Mutations in January, February, and December occurred in the 1990s, whereas mutations in March, September, and October occurred in the 2000s. The mutation characteristics of the Z ¯ m were similar to those of Q ¯ m .

In 2022, the YRB experienced an extreme hydrological drought from the beginning of the flood season onward. This drought can be divided into four stages ( Figure 6 ): a pre-flood season with abundant water, an initial flood season with a rapid decline, a main flood season with a continuous decline, and a post-flood season with the driest season.

In stage 1 in June 2022, owing to the influence of abundant rainfall in the YRB, Q ¯ d and Z ¯ d at Datong began to increase in volatility at a rate significantly higher than that of the multiyear average of the same period. The average elevated amplitude of Z ¯ d was 1.6 m, and the maximum was 2.41 m, which occurred on June 9. The annual maximum Q ¯ d (61800 m³/s) and Z ¯ d (13.38 m) appeared on June 24, after which the values began to decline rapidly.

In stage 2 in July, the Q ¯ d and Z ¯ d at Datong continued to rapidly decline, with an average daily decline of 0.11 m for Z ¯ d , and the maximum of 0.19 m occurred on July 26. Q ¯ d and Z ¯ d on July 4 changed from higher to lower than the multiyear average value during the same period. Afterwards, the depressed magnitude gradually increased, from 0.12 m on July 4 to approximately 3.6 m at the end of July for Z ¯ d , with an average depressed magnitude of approximately 1.9 m. On July 27, Q ¯ d and Z ¯ d at Datong dropped to 35,400 m³/s and 9.83 m, respectively, ranking the 10th lowest values in the same period. The lowest monthly values of Q ¯ d (31,300 m³/s) and Z ¯ d (9.06 m) that appeared on July 31 ranked the fourth lowest in recorded history.

In stage 3 in June, the average daily decline of Z ¯ d was 0.11 m, with the largest of 0.28 m on August 5. The Q ¯ d and Z ¯ d in June were lower overall than the historical average, and the depressed magnitude continued to increase, from 3.6 m at the beginning of the month to 5.6 m at the end of the month for Z ¯ d , with an average of approximately 5.0 m. On August 18, n Q ¯ d . (19,600 m³/s) and Z ¯ d . (6.56 m) reached the penultimate lowest values during the study period. The monthly lowest Q ¯ d and Z ¯ d were 15,400 m³/s and 5.65 m, respectively, which occurred on August 31. The Z ¯ d was 0.92 m lower than the lowest value recorded in history.

In stage 4, beginning in early September, the cascade reservoirs in the UYR entered a concentrated storage period, and the Q ¯ d and Z ¯ d at Datong continued to decline. The lowest annual Q ¯ d was 6,400 m³/s, which occurred on September 15. By November 16, the lowest annual Z ¯ d was 3.91 m, which was below the drought warning level (4.3 m) for 22 days, indicating a severe hydrological drought.

Since the 1950s, the YRB has experienced multiple extreme droughts based on annual runoff, as shown in Figure 7A in 1978, 2006, 2011, 2022, and 2023. Although the annual runoff in 2022 was not the smallest, after mid-August, Q ¯ d and Z ¯ d in Datong were lower than those of the former typical drought years ( Figures 7B, C ). There were 41 days with a Q ¯ d less than the drought warning flow discharge (10,000 m³/s); these days were mainly concentrated from October to December, with November receiving the most. Compared with previous drought years, 2022 constituted a typical post-flood drought.

Although precipitation is a fundamental cause of drought, it is ultimately converted into runoff for specific river reaches. Figure 8 shows variations in the annual runoff composition of Datong during different historical periods. Overall, the annual runoff source was relatively stable, and the UYR and the two lake water systems consistently remained the main sources. The proportion of annual runoff from the UYR, determined at Yichang station, fluctuated between 39.6% and 59.7%, with an average of approximately 48.4%. The proportion of the total annual runoff in the two lake water systems fluctuated between 22.2% and 46.4%, with average proportions of 18.6% and 16.7% in the DTL and PYL water systems, respectively. Compared with the two periods of 2003–2010 (before the normal storage of the TGR) and 2011–2023 (after the normal storage of the TGR), the proportion of annual runoff sources at Yichang increased slightly by 0.7%, whereas that of the DTL and PYL water systems increased by 1.9%, indicating relatively small changes.

Judging by the composition of different months ( Figure 9 ), the runoff from Yichang increased significantly from January to May from 2011 to 2023, with an average increase of 27%–72% (approximately 7–8.5 billion m³) compared to runoff before 2011. This increase was even greater in February (72%). However, in September and October after the flood, runoff decreased by 24% (15.7 billion m³) and 20% (9 billion m³), respectively.

Overall, although the sources of annual runoff at Datong did not change, the changes in monthly runoff within the year were more obvious, mainly because of the influence of reservoir water storage in the UYR. When the water level of the reservoir decreased before the flood season, the runoff from Yichang increased, whereas it substantively decreased after the flood season (Feng et al., 2020). Because of the longer duration of the reservoir subsidence period compared with the storage period, the trend of turning from flood to drought after the flood season was more obvious at Datong station. The storage of cascade reservoirs played an important role in the hydrological drought situation in Datong after the flood season.

Widespread drought has continued to occur in various tributaries of the basin in 2022. Figure 10 shows a comparison of the runoff composition at Datong from July to October between 2022 and the typical drought years. During this period, the runoff levels from the UYR, DTL, PYL, and HR were all significantly lower. The water volume from the UYR was close to that in 2006, whereas those in the PYL, HR, and other areas in the MYR were the lowest during these drought years. As a result, the relatively small inflow in the MYR in 2022 also substantially affected the drought conditions observed at Datong.

Since the application of the TGR from 2003 to 2021, the riverbeds in the MYR and LYR have mainly eroded (Xu et al., 2023). For the tidal reach from Datong to YRE, the erosion amount for the bankfull channel (corresponding water levels for Datong and Jiangyin are 10.06 m and 2.66 m, respectively) is 18.82×10⁸ m³, of which 62.9% occurred between Datong and Jiangyin, mainly concentrated in low-water river channels, and the erosion intensity increased year by year. After a riverbed is eroded and cut down, it inevitably causes a decrease in the water level under the same inflow and tidal conditions, thereby exacerbating the hydrological drought defined by the water level. Figure 11 shows the correlation between the lowest annual Z ¯ d at the Datong station and Z h at the downstream Nanjing station. There is a close correlation between them, but this correlation varies in different periods. In recent decades, under the same lowest Z h conditions at Nanjing, the lowest Z ¯ d at Datong has slightly increased, with an amplitude of generally within 0.2 m, which has a certain buffering effect on drought conditions. Although the erosion amplitude of the riverbed in the LYR is not yet severe, in future years, with the continuous operation of cascade reservoirs in the UYR, the riverbed in the LYR will inevitably suffer more severe erosion (Hu et al., 2023), and a further increase in the erosion amplitude will exacerbate the water level decrease at the Datong station. Therefore, future drought trends in the LYR will inevitably be affected by riverbed cutting and should receive sufficient attention.

The downcutting of the riverbed can also cause a decrease in Z ¯ d at the same Q ¯ d . Figure 12 shows the changes in the relationship between Z ¯ d and Q ¯ d at Datong station over the past 20 years. It can be seen that the relationship has been relatively stable, but the Z ¯ d still slightly decreases when the Q ¯ d is <40,000 m³/s, and the bigger the Q ¯ d , the smaller drop of the Z ¯ d . When the Q ¯ d are 15,000 m³/s, 25,000 m³/s, and 35,000 m³/s, compared to 2001, the Z ¯ d drops in 2021 are 0.36 m, 0.28 m, and 0.16 m, respectively. When the Q ¯ d further increases, the Z ¯ d remains basically unchanged. This is consistent with the above-mentioned riverbed erosion law, that is, erosion is mainly concentrated in low-water river channels. As a result, Z ¯ d decline is also mainly concentrated in smaller Q ¯ d .

Tidal reach is affected by both the input runoff from the upstream river and downstream tidal currents. The tidal level at the Baimao hydrological station, located at the entrance of the YRE, was selected to analyze the effects of tidal currents. Figure 13 shows the correlation between the characteristic Z_h at the Baimao station and those distributed along the LYR from 2010 to 2020. A close linear correlation exists between Baimao and other stations in the LYR for the annual average low Z_h ( Figure 13A ). Except for the correlation coefficient with Tianshenggang, which was 0.61, the linear correlation coefficients with the other stations were above 0.9. This result indicates that, on average, low Z_h in the YRE will have a significant impact on hydrological drought in the LYR. However, for the lowest annual Z_h , the correlation between Baimao and the stations in the LYR above Xuliujing was extremely low, with a linear correlation coefficient generally within 0.2 as shown in Figure 13B . In summary, the extremely low Z_h in the YRE had a relatively small impact on the overall extreme drought in the LYR. Therefore, although the average low Z_h had an impact on drought in the LYR, because the downstream low Z_h is not controllable by human activities, future research on the impact of downstream tidal currents on droughts in the LYR should be conducted from a global climate perspective.

The water intake in tidal river sections is not only affected by the lowest Z ¯ d but also the Q ¯ d . This is because the main purpose of water intake is to utilize freshwater resources, and when the Q ¯ d is low, saltwater intrusion is extremely detrimental to the acquisition of freshwater resources. In recent years, under the comprehensive influence of continuous erosion of the riverbed and the replenishment scheduling of the TGR during the dry season, the frequency of extremely low Q ¯ d at Datong station has decreased, and the extremely low Z ¯ d , has been raised. Therefore, for general dry water conditions, the impact of riverbed cutting is greater than the replenishment effect of reservoirs, especially since 2020, in the flow range of 11,000–25,000 m³/s, under the same inflow Q ¯ d , the measured average and minimum Z ¯ d have both decreased by approximately 0.35 m.

Considering the relationship shown in Figures 11 , 12 , the Z ¯ d was found to decrease by 0.22 m for every 1,000 m³/s decrease in Q ¯ d at Datong station, and the low Z_h at Nanjing station decreased by 0.53 m for every 1 m decrease in Z ¯ d , approaching a ratio of 2:1. To maintain the low Z_h of Nanjing station with a 95%–99% guarantee rate without decreasing, the Q ¯ d of Datong station needs to gradually increase from 10,000 m³/s to 12,500 m³/s. To ensure the safe operation of the existing water intake project and maintain the Z_h at Nanjing station above 2.3 m, the minimum Q ¯ d of Datong should be increased to 11,300 m³/s.

The continuous decline in Z ¯ d caused by hydrological drought will deteriorate the navigation environment, which will significantly impact shipping. The narrowing of the waterway after Z ¯ d decreases will make it difficult for ships to rendezvous and overtake. Due to Z ¯ d limitations, cargo ship loads will need to decrease to meet these navigation conditions. In addition, the rapid drop in Z ¯ d after the flood season can also make it difficult to maintain the corresponding depth of the sea vessel channel, leading to closures and reduced shipping efficiency.