The impacts of reservoirs, especially multiple reservoirs, on the flow regimes and ecosystems of rivers have received increasing attention. The most widely used metrics to quantify the characteristics of flow regime alterations are the indicators of hydrologic alteration (IHAs) which include 33 parameters. Due to the difference in the degree of alteration and the intercorrelation among IHA parameters, the conventional method of evaluating IHA parameters that assigns the same weight to each indicator is obviously inadequate. A revised IHA method is proposed by utilizing the projection pursuit (PP) and real-coded accelerated genetic algorithm (RAGA). Data reliability is analyzed by using the length of record (LOR) method. The projection values reflecting the comprehensive characteristics of the evaluation parameters are calculated. Based on these methods, a scientific and reliable evaluation of the cumulative impacts of cascading reservoirs on the flow regime was made by examining the Jinsha River. The results showed that with the increase in the number of reservoirs, the alteration degrees of IHA parameters gradually increased in groups 1, 2, 3 and 4 but decreased in group 5 (each group addresses the magnitude, timing, frequency, duration and rate of change in turn), and the flow duration curves showed a declining trend at the high-flow part and an increasing trend at the low-flow part. The flow regime alteration of the outlet section was more stable than before. This change had a negative impact on downstream fish reproduction and ecological protection. An attempt at ecological regulation was made to simulate the natural rising process of water, and four major Chinese carps have a positive response to the flood peak process caused by manual regulation.

Free-flowing rivers (FFRs) support diverse, complex and dynamic ecosystems
globally, providing important societal and economic services (Grill et al.,
2019). Humans have extensively altered river systems through impoundments
and diversions to meet their water, energy and transportation needs
(Nilsson et al., 2005). Only 37 % of rivers longer than 1000 km remain
free-flowing over their entire length, and 23 % flow uninterrupted into the
ocean around the world. Very long FFRs are largely restricted to remote
regions of the Arctic and the Amazon and Congo basins. From 1978 (when
China's reform and opening up began) to 2017, China experienced an
unprecedented boom in the construction of dams. From 1949 to 2017, 98 795
reservoirs and dams were built in China with a total storage capacity of

To evaluate the characteristics and ecological effects of flow regime changes, indicators are often needed to quantify the extent of hydrological alterations caused by reservoirs or dams. Olden and Poff (2003) found more than 170 hydrological indicators that can describe the different components of the flow regime and capture the ecologically relevant streamflow attributes. However, large numbers of hydrologic metrics are too complicated to use, and many metrics are intercorrelated, resulting in statistical redundancy (Gao et al., 2009; Poff and Zimmerman, 2010). Studies have sought to explore redundancy among hydrological indicators. For example, Olden and Poff (2003), Yang et al. (2008), Gao et al. (2009), and Fantin-Cruz et al. (2015) used principal component analysis (PCA) to evaluate the patterns of statistical variation for each parameter and identified a small subset of hydrological indicators as the most representative of the ecological flow regimes. Yang et al. (2017) used the criteria importance through intercriteria correlation (CRTTIC) algorithm to remove repetition and identify the weights of indicators. The weight of each hydrological indicator is assumed to be proportional to the standard deviation and inversely proportional to its correlation with other indicators. Then, high-weight indicators and some low-weight indicators that have important effects on aquatic ecology are used as representative indicators. Obviously, this selection is subjective and arbitrary. The most widely used metrics for characterizing river flow regime changes are the indicators of hydrologic alteration (IHAs), which were developed based on 33 hydrological parameters in five groups, namely, the magnitude of monthly streamflow, the magnitude and duration of annual extreme flows, the timing of annual extreme flows, the frequency and duration of high and low pulses, and the rate and frequency of flow changes (Richter et al., 1996; Mathews and Richter, 2007). Richter et al. (1997) proposed the range of variability (RVA) method for evaluating the degree of alteration of the hydrological flow regime with IHA metrics. Nevertheless, intercorrelation still exists among the 33 parameters (Olden and Poff, 2003; Gao et al., 2012). Vogel et al. (2007) proposed a small set of representative indicators, i.e., the nondimensional metrics of ecodeficit and ecosurplus, which are based on flow duration curves (FDCs) and are computed over any time period of interest (month, season or year). Ecodeficit and ecosurplus reflect the overall loss or gain of streamflow resulting from flow regulation. Some studies (Gao et al., 2009, 2012; Zhang et al., 2015) have demonstrated that the ecodeficit and ecosurplus metrics provide a simplified and adequate representation of hydrological impacts, compared with the use of the more complex IHA and RVA hydrological approaches.

Scholars have become increasingly concerned with the cumulative effects of
multiple dams deriving from individual dams on hydrological processes
(Santucci et al., 2005; Deitch et al., 2013; Wang et al., 2017a, 2018; Wen et al.,
2018; Huang et al., 2018). The combined effect of cascade
reservoirs on hydrological processes is cumulative and greater than that
associated with individual reservoirs (Huang et al., 2010; Dos Santos et
al., 2018). In comparison, Santucci et al. (2005) found little evidence of
the cumulative effects of low-head dams (

In summary, previous studies on method improvement were based on the statistical reduction in the dimensionality of multi-index data and evaluated the hydrological alterations of rivers. The disadvantage is that retaining most of the information also leads to the loss of some information. For example, a PCA usually only maintains 80 % of the data information. In this study, a very different idea was employed. Data mining and data optimization methods were used to identify the characteristics of the indicator system and identify the difference weight of each indicator to overcome the deficiency in the comprehensive evaluation given the same weight for each indicator. At the same time, global optimization also reduced the deviation in the evaluation caused by intercorrelation among indicators.

Based on previous studies, the objectives of the present study are as follows: (1) to develop an updated weight determination method for IHA indicators and precisely evaluate hydrological alteration; (2) to analyze cumulative effects on the flow regime of the construction of cascade reservoirs; and (3) to simulate the process of natural flooding, implement reservoir flow regulation and provide recommendations for water resource management.

The Jinsha River comprises the upper reaches of the Yangtze River in China
and originates from the northern foot of the Tanggula Mountains in the
Tibetan Plateau. The Jinsha River flows along a distance of approximately
3500 km and has a drainage area of 502 000 km

Generalized map of study region. TGD: Three Gorges Dam.

Large reservoirs built in the Jinsha River basin.

There is a national nature reserve in China that protects rare fish from 1.8 km downstream of Xiangjiaba Reservoir to the main stream of the Yangtze
River in Masangxi, in the municipality of Chongqing (355 km in length). The nature
reserve's main protection targets include three rare fish, i.e., paddlefish
(

The daily streamflow data of the Panzhihua, Huatan, Pingshan and Xiangjiaba hydrological gauges were collected (Table 2). The climate data used in this study are daily precipitation data at 28 stations from 1966 to 2017, and the daily inflow and outflow data of Xiangjiaba Reservoir and Xiluodu Reservoir were collected to analyze the effects of flow regulation. Pingshan Station is located approximately 28 km upstream of Xiangjiaba Reservoir and covers 99.96 % of the controlled drainage area of the reservoir, but it has been out of service since 2012 due to the construction of Xiangjiaba Reservoir. Therefore, data from Xiangjiaba Station, which was newly built in 2012 and is located close to Xiangjiaba Reservoir, were supplemented with Pingshan Station data (Huang et al., 2018).

List of hydrological stations and their features.

Xiangjiaba Station was also called Pingshan Station in this study.

Projection pursuit (Friedman and Tukey, 1974; Wang et al., 2017b, 2019) uses data mining and data optimization methods to project high-dimensional data into low-dimensional space and analyze the characteristics of high-dimensional data through the distribution structure of low-dimensional projection data. The main steps are as follows.

Then, a random number with

The IHA system, consisting of 33 hydrological indicators, is employed to
assess hydrological alteration. The 33 IHAs are categorized into five groups
addressing the magnitude, timing, frequency, duration and rate of change
(Shiau and Wu, 2010), and each group has a different ecological
significance (Table 3). For the IHA statistics of the pre-impact period, its
range of variation between the 75th and 25th percentiles is considered as
the ecological target range. The alteration degree (

The 33 indicators of hydrologic alteration.

A flow duration curve is simply a plot of the ordered daily streamflow
observations

The length of record (LOR) method is used to provide quantitative advice on the length of record required for each IHA parameter. The result of the LOR for a station is considered as a reference for other stations with similar hydrologic regimes; therefore, the station with the smallest anthropogenic impact and longest record length in the study area was chosen for LOR analysis. For each IHA parameter, we calculate its statistics for each year in a data set along with the long-term mean as the reference for LOR analysis. Then, the statistics of each parameter are ordered randomly and grouped into record-length increments ranging from 2 years to the full LOR. The mean of each increment is calculated for a comparison with the long-term mean. This process is repeated 50 000 times, from which 95 %, 90 % and 85 % confidence intervals (CIs) are calculated. Finally, we calculate the LOR required within 5 % and 10 % long-term mean errors at a specified confidence interval for the river in the study. For the convenience of discussion, the LOR result within 10 % of the long-term mean with a CI of 85 % is abbreviated as 10/85 (Timpe and Kaplan, 2017).

When we evaluate the effect of reservoirs on the hydrological regime, it is
necessary to consider the potential impacts of climate change on
hydrological data, since there may be different climatic conditions in the
pre- and post-stages (Wang et al., 2017a). Therefore, the Mann–Kendall
(hereafter referred to as MK) test method was used to analyze the trend in
the annual-precipitation time series of 1966–2017 (52 years) for 28 stations in
the Jinsha River basin. The MK test is based on statistics

In the study area, annual precipitation showed no significant trend or trend
below the 10 % significance level at 21 stations, and only two and three
stations showed increasing trends at 1 % and 5 % significance levels,
respectively. A decreasing trend was found at only one station with the 1 %
significance level (Fig. 2). The moment estimation method was used to
calculate the characteristic values of precipitation from two short time
series (1966–1998 and 1999–2017) and one long time series (1966–2017).
Compared with the value of the long time series, the relative errors of the
mean of two short time series did not exceed

Precipitation changes in the Jinsha River basin. Trend of annual
precipitation at 28 stations between 1966 and 2017 where upward (downward)
triangles indicate positive (negative) trends from the MK test. The size of the
triangles depicts the significance levels 10 % (small), 5 % (medium)
and 1 % (large). Black dots show stations with no trends or trends below the
10 % significance level. The value in the legend is the standardized
statistic

The IHA statistics software developed by the US Nature Conservancy
(

Value of weights of 32 IHA parameters in

The larger the optimal projection direction value is, the greater the contribution to the flow regime evaluation is, that is, the higher the weight of the indicator is. As shown in Fig. 3, at the Pingshan and Huatan stations, the weight allocations are similar (Fig. 3b and c), and the parameters with similar high-weight values (greater than 0.04) are mean flow in January, April and May; annual minimum discharge (1 d mean, 7 d mean and 90 d mean); base flow index; duration of high flow pulse; rise rate; fall rate; and number of reversals, while at Panzhihua Station (Fig. 3a), the high-weight parameters are mean flow in January, February, March, June, September and November; annual minimum discharge (3 d mean and 30 d mean); base flow index; date of minimum; and count and duration of high pulse (Table 4). This result indicates that the data structure of the characteristics of the flow regime characterized by IHA parameters has both similar and different parts upstream and downstream, and it also implies that different weights may be related to tributary imports, reservoir constructions or interval water supplies.

Alteration degree and weight of 32 IHA parameters.

The values in bold mean high weights.

The projection values of the flow regime from the 1966–2017 hydrological series (Fig. 4) were obtained by substituting the optimal projection directions into Eq. (3) at three stations. The results of the trend analysis on the projection values by the MK test suggest that at Panzhihua Station, the projection values fluctuated periodically without any significant change trend (Fig. 4a), while the significant decreasing trends were found at the 1 % significance level at the Huatan and Pingshan stations (Fig. 4b and c), especially at Pingshan Station, where the decreasing trend was more intense after 2012. The projection value is a comprehensive evaluation result of flow regime changes and takes into account the monthly flow condition, the magnitude and duration of extreme discharge conditions, the occurrence time of extreme discharge conditions, the frequency and duration of high or low flow pulse, and the frequency and rate of hydrological-process changes. The larger the projection value is, the more distinct the intra-annual cyclical change in wet and dry situations is and the smaller the interference caused by human activities is. As shown in Fig. 4b and c, the projection values both began to show a significant decline in 1999, and a more significant decreasing trend was found at Pingshan Station from 2013 to 2017 (Fig. 4c). The timing of the two drastic changes coincides with the time when Ertan Reservoir in the Yalong River (tributary) and the Xiangjiaba and Xiluodu reservoirs in the lower reaches of the Jinsha River (main stream) were first put into operation. This finding also implies that the impact of giant reservoirs on the flow regime is substantial and that the degree of impact is further aggravated with the continuous construction of reservoirs.

Project values in

According to the commissioning time of the first generator set in Ertan Reservoir, the period of 1966–1998 is considered as the natural state of hydrological series unaffected by human activities (pre-impact period), and the period of 1999–2017 is considered as when the series is affected by human activities (post-impact period). The alterations and weights of the 32 IHA parameters are shown in Table 4. As shown in Table 4, the alteration degree calculated by the revised IHA method is larger than that by the traditional method. For the Panzhihua, Huatan and Pingshan stations, the overall alteration degrees calculated by the revised method are 0.29, 0.57 and 0.54, and those by the traditional method are 0.28, 0.50 and 0.49 by the traditional method with relative changes of 3.57 %, 14 % and 10.20 %, respectively. The traditional IHA method, analyzing overall hydrological alteration with the same weight for each parameter, constantly underestimates or overestimates actual hydrological changes, since many parameters are intercorrelated (Yang et al., 2017).

Correlation coefficients among the IHA statistics for the observed
data sets in

Figure 5 illustrates boxplots of the correlation coefficients between each IHA parameter and the remaining 31 IHA parameters at the three hydrological stations mentioned above. At Panzhihua Station, the absolute values of the correlation coefficients among the IHA parameters range from 0.0016 to 0.9976, with a mean of 0.2852 (Fig. 5a). The absolute values of the correlation coefficients at Huatan Station range from 0.0 to 0.9876 with a mean of 0.2931 (Fig. 5b). The absolute values of the correlation coefficients at Pingshan Station range from 0.0012 to 0.9972 with a mean of 0.2920 (Fig. 5c). Figure 5 shows that the correlations among parameters at the Huatan and Pingshan stations are stronger than that at Panzhihua Station, which suggests that the correlations among parameters has an impact on the evaluation of the hydrological alteration when combined with the results of the above two methods. The stronger the correlation among parameters is, the greater the impact is. Panzhihua Station is located 10 km upstream of the junction of the Yalong River and the Jinsha River; therefore, operation of Ertan Reservoir does not affect its hydrological streamflow series. This is also confirmed by the fact that the overall hydrological alteration at Panzhihua Station is low.

The large reservoirs of Ertan, Xiluodu and Xiangjiaba were successively built along the Jinsha River. Three periods were utilized for studying the cumulative impacts of cascade reservoirs. Ertan Reservoir was put into operation in 1999, and the Xiluodu and Xiangjiaba reservoirs were both put into operation in 2013. Therefore, the first period is 1966–1998 with natural flow regime conditions; the second period is 1999–2012 with the effects of individual reservoirs; and the third period is 2013–2017 with the effects of the cascade reservoirs of Ertan, Xiluodu and Xiangjiaba. A total of 32 IHA statistics at Pingshan Station were calculated, and the weights of each parameter were obtained by PP and RAGA. For the three different periods, the alteration degrees of each parameter and the overall alteration degrees are shown in Tables 5 and 6, respectively. The cumulative impacts on the flow regime are very obvious with the successive construction of the reservoirs. During the period of 1999–2012, the number of high-alteration-degree parameters was 8, with an overall alteration degree of 47 %, while during the period of 2013–2017, the number of high-alteration-degree parameters increased to 13, with an overall alteration degree of 70 %. In particular, for the parameters of the mean flow in May (5), base flow index (23) and low-pulse count (26), the alteration degrees of the three parameters were low during the period of 1999–2012, but they became high alteration degrees during the period of 2013–2017.

Alteration degree of 32 indicators in Pingshan Station in different periods.

Overall degree of alteration of five groups of IHA parameters.

The weight of each parameter has been considered.

Huang et al. (2018) found the cumulative effect of cascade reservoirs was not a simple additive effect. Since the joint operation of cascade reservoirs for guaranteed output, flood regulation and hump modulation were complicated; a decreasing trend in the impact of reservoir operation on downstream hydrological conditions along the direction of flow of the Jinsha River could be found, while the runoff averaging effect became more evident. Zhang et al. (2020) found the impacts of anthropogenic factors on water discharge were more significant than the impacts of precipitation, especially for the high discharge during flood season. In this paper, the increasing trends in the alteration degree were shown in groups 1, 2, 3 and 4, especially for the indicators of maximum or minimum flow in group 2; the change of extreme flow was the main reason for the averaging effect of the river. However, group 5 showed a decreasing trend in alteration degree, and this could mainly result from the joint operation of cascade reservoirs. For the single reservoir, its limited regulation ability made the rate of change and reversals show dramatic and uncertain changes, while for the cascade reservoirs, their joint operation with strong regulation ability enhanced the averaging effect and also made the rate of change stable, especially for the ecological regulation that has been implemented in recent years, trying to maintain the flow changes under similar natural conditions during the spawning period of fish.

As shown in Table 7, compared with the period of 1966–1998, the changes in
winter precipitation during the periods of 1999–2012 and 2013–2017 were very
small,

Changes in the annual minimum and annual maximum flows between the periods 1966–1998, 1999–2012 and 2013–2017 in Pingshan Station.

To better compare and analyze the cumulative effect of cascade reservoirs on
the flow regime, 2 years were selected at Pingshan Station, that is, 2004
and 2016, based on the annual flow of periods of 1999–2012 and
2013–2017 (Fig. 6). These years represent the years with an annual flow in
the 50th percentile, and their annual mean discharges are 4796 and
4063 m

Annual flow duration curves in 2004 and 2016 in Panzhihua Station.

The characterization of natural and altered flow regimes using IHA requires
adequate flow data (Zhang et al., 2019). Richter et al. (1997) suggested
using > 20 years of pre- and post-impact data to characterize the
hydrologic regime. Timpe and Kaplan (2017) found that fewer than 20 years of
data could be used to yield statistically significant IHA results for a
number of rivers across the Amazon by using the length of record (LOR)
method. Given this uncertainty, further research is needed to determine the
reliability of the data required by IHA. We chose 47 years from 1952 to 1998
at Huatan Station as the LOR calculation period with the smallest
anthropogenic impact and the longest record length in the study area. Table 8 shows the length of record required for the 32 IHA parameters within 5 %
and 10 % long-term mean errors at different specified confidence intervals
at Huatan Station. Comparing the results between different groups, it is
observed that the data volume requirement in group 4 is the highest, while
when comparing and analyzing within the same group, it is observed that the
amount of data required to describe the parameters for low flow is less than
that required to describe the parameters for a relatively high flow. For
example, the amount of data required for monthly mean flow in the flood
season is higher than that for the monthly mean flow in the dry season in
group 1. Zhang et al. (2019) found that the amount of data required has a
consistent relationship with the amount of average monthly flow and the
variability in hydrological data. Both the Huatan and Pingshan stations are
located downstream of the Jinsha River with similar hydrologic regimes
(Fig. 3). Referring to the results of Table 8, the 33-year daily streamflow
data from 1966 to 1998 at Pingshan Station fully satisfy the highest
requirement (31 years) to produce a

Length of record (LOR) results for each IHA parameter in Huatan Station.

Correlation between runoff and precipitation for the periods of pre- and post-impact in Panzhihua Station.

Hydrograph for daily average inflow, outflow and reservoir water
level in the

Li et al. (2006) found that the total number of days with rising water from
May to June in the Yangtze River is an important environmental driving
factor that determines the amount of spawn produced by fish with pelagic
eggs. Four major Chinese carps are typical fish with pelagic eggs and
the most widely distributed species in protected areas. The cumulative
impact of the construction and operation of cascade reservoirs on the flow
regime in the downstream nature reserve of the Jinsha River has aroused
widespread concern. During the period from 15 to 18 May 2018, the
management institution conducted a joint ecohydrological regulation test of
the Xiluodu and Xiangjiaba reservoirs for 4 d (Fig. 8). On 15 May, the
outflow discharge of Xiluodu Reservoir was 2770 m

In this study, a revised IHA method is presented by combining projection pursuit (PP) with a real-coded genetic algorithm (RAGA) to obtain the weight of each IHA parameter. The method is applied to assess the cumulative impacts of cascade reservoirs on the flow regime in the Jinsha River. The main points can be summarized as follows:

The impacts of the construction and operation of the cascade reservoirs on the flow regime are huge. Using the revised IHA method to analyze the cumulative effects of the Ertan, Xiangjiaba and Xiluodu reservoirs on the flow regime of the outlet section of the Jinsha River, we found that with the continuous construction of the reservoir, the alteration degrees of IHA parameters in groups 1, 2, 3 and 4 are gradually increasing but are decreasing in group 5 (rise rate, fall rate and number of reversals). Due to reservoir water storage and release, the FDC shows decreasing trends in high flow and increasing trends in low flow. The whole curve shows the characteristics of a head drop and a tail lift. The maximum flow is reduced, and the minimum flow is increased. The rate and frequency of discharge changes tend to be subtle. As the cascade reservoirs are completed, the flow regime alteration at the outlet section is more stable. This change has a negative impact on downstream fish reproduction and ecological protection.

The traditional IHA method employs 33 parameters to quantify the characteristics of flow regime changes and analyze the overall hydrological alteration with the same weight for each parameter. The revised IHA method gives each parameter its own weight by applying a projection pursuit model to project high-dimensional data into a low-dimensional space and optimize the projection direction of each parameter. This method achieves a more reasonable evaluation of hydrological alterations and overcomes the problem of underestimating the hydrological alterations in the study area due to the difference in the degree of alteration and the intercorrelation among IHA parameters.

Previous studies have suggested using > 20 years of pre- and
post-impact data to characterize the hydrologic regime. In this study, we
chose 47 years from 1952 to 1998 at Huatan Station as the LOR
calculation period. As a reference, the 33-year daily streamflow data from
1966 to 1998 at Pingshan Station fully satisfy the highest requirement (31
years) to produce a

The hydrological data used in this study were provided by China Yangtze Power Co., Ltd. The precipitation data were downloaded from

XZ and XH suggested the idea and formulated the overarching research goals and aims. XZ, HZ and KM corrected and managed the data. XZ and XH employed statistical methods to analyze study data. XZ prepared the paper with contributions from all coauthors.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

The opinions expressed here are those of the authors and not those of other individuals or organizations.

The authors would like to thank the China Meteorological Data Service Center for providing the precipitation datasets. Comments from the editor and five anonymous reviewers were greatly appreciated.

This research has been supported by the National Natural Science Foundation of China (grant nos. 51579161 and 51779160).

This paper was edited by Dominic Mazvimavi and reviewed by five anonymous referees.

The authors declare that they have no conflict of interest.