The river hydrological regime is a key indicator to ensure the ecological function of rivers. We evaluated the changes in the hydrological regime of the Zi River by combining two methods for assessing the degree of change. Furthermore, we screened the most ecologically relevant hydraulic indicators (ERHIs) using principal component analysis (PCA) and quantified the independent impacts of human activities and climate change using the Whale Optimization Algorithm and Random Forest model (WOA-RF). The results showed that (1) 1987 was the year of the Zi River flow mutation, the overall average annual runoff in the Zi River Basin showed a decreasing trend, and the flow had four cyclic change time scales. (2) The overall hydrological variability in the Zi River Basin was 42%, and after the preferential selection of hydrological indicators, eight indicators were identified as the ERHIs of the basin, and the hydrological variability was 48.5%. The hydrological change in the basin was determined to be moderate. (3) The contribution of climate change is 28.8%, and the contribution of human activities is 71.2%. The construction of reservoirs was the most significant factor contributing to the impact of human activities, with a 65.6% contribution.

  • The study successfully screened out the eight ERHIs and solved the problem of indicator redundancy.

  • It was found that the natural river flow reconstructed by the WOA-RF model applies to the Zi River Basin.

  • It was found that the contribution of human activities to flow change was as high as 71.2%, while that of climate change was only 28.8%.

The hydrological regime of a river is a key indicator for ensuring the ecological functioning of the river. It maintains the sustainable development and utilization of water resources and is the basis for many organisms' survival, reproduction, and development (Sun et al. 2014). With increasing climate change and anthropogenic disturbances, globally important rivers such as the Mississippi River (Haider et al. 2025), the Nile (Fathi et al. 2019), the Yellow River (Zhou et al. 2017), and the Yangtze River (Guo et al. 2021) have all experienced different degrees of changes in their hydrological regime. The impacts of climate change and human activities are the two main drivers of river flow (Wei et al. 2020). In recent decades, with socio-economic development and population growth, intense anthropogenic disturbances have altered the hydrological cycle processes in the basin (Guan et al. 2021), which has changed the ecohydrological regime of rivers to varying degrees (Sharifi et al. 2024). Therefore, analyzing changes in hydrological regimes and distinguishing between the impacts of human activities and climate change on hydrological regimes is important for the healthy development of river hydrology.

In recent years, scholars at home and abroad have discussed the quantitative analysis of river hydrological regimes. Richter et al. (1996) proposed the Indicators of Hydrologic Alteration (IHA), which are the most widely used, and the index system contains 33 indicators divided into five groups (flow, event, frequency, duration, and rate of change) and further simplified the common 170 hydrological indicators. It uses changes in river flow and external ecological change to determine a basin's overall degree of change. Later, the IHA method combined with the range of variability approach (RVA) method for univariate and integrated hydrologic evaluation was proposed by Richter et al. (1998). Gunawardana et al. (2021) quantified the pre- and post-impacts of the transboundary Srepok River Basin hydropower development on the hydrologic cycle using the RVA in IHA. Although the IHA system is relatively well developed, there are problems of autocorrelation and overlapping information between IHA variables (Smakhtin et al. 2006). The most ecologically relevant hydrologic indicators (ERHIs) have received more and more attention from researchers. Jiang et al. (2022) screened seven indicators as ERHIs for the Weihe River using principal component analysis (PCA). Li et al. (2022) studied the Han River Basin by selecting strong variability indicators from ERHIs and diagnosing watershed variability with a strong variability index. Replacing the large number of IHA with ERHIs facilitates the study by retaining as much key information as possible from the original data and significantly reducing the redundancy of information from the large number of hydrological variables (Yang et al. 2024).

Meanwhile, for the attribution analysis of river flow changes, scholars in recent years have mainly used various hydrological models and elasticity coefficient methods to analyze the influencing factors of flow changes in different rivers. Mirjalili & Lewis (2016) proposed a new population intelligence optimization algorithm, the Whale Optimization Algorithm (WOA). Wang et al. (2023a, 2023b, 2023b) proposed a method for predicting monthly flow intervals based on the long short-term memory (LSTM) model optimized by WOA and nonparametric kernel density estimation. The Random Forest (RF) model proposed by Wang et al. (2022) for attributing flow changes in the upper Yongding River Basin provides a new method and idea for quantitatively identifying the contribution of flow changes in the basin. Soo et al. (2024) found that the RF model exhibited the best performance in flow simulation and prediction in the Klang River Basin.

Several scholars have studied the hydrological regime of the Zi River Basin from diverse perspectives. Wang et al. (2019) evaluated the hydrological regime using the IHA-RVA method. Meanwhile, Li et al. (2018) employed the cumulative distance level method and the sliding t-test to analyze trends and detect abrupt changes in precipitation and flow series over the past 55 years within the basin. Notably, most hydrological researchers have confined their analysis of flow changes to identifying trends and mutation points (Li et al. 2018; Wang et al. 2019). There is a pressing need for a more in-depth exploration of the variability of various hydrological indicators of flow and the extent of these variations. Furthermore, even when this issue is addressed, the overlap among hydrological indicators must be examined, a particularly acute problem in the Zi River Basin. Additionally, a quantitative analysis of the impacts of human activities and climate change on hydrological changes in the Zi River Basin remains an area requiring further research.

In this study, a strategy based on PCA to select ERHIs combined with the WOA-RF model is proposed to address the above problems, which effectively solves the problem of variable redundancy of traditional IHA method through PCA dimensionality reduction and improves the scientificity of hydrological variability assessment; meanwhile, combining with the WOA-optimized RF model, the generalization ability and reliability of flow attribution analysis are significantly enhanced. Compared with the traditional IHA or single RF method, the optimization and breakthrough of the methodology are achieved.

Therefore, given the limitations of previous studies, the study's main objectives are the following: (1) to analyze the characteristics of long-term series changes of hydrometeorological elements in the Zi River Basin. (2) Access the changes in the hydrological regime of Zi River before and after the sudden change based on IHA and select the indicators of ERHIs of the Zi River Basin by PCA to reduce the influence of autocorrelation information on the assessment results. (3) To reconstruct the natural flow in the Zi River Basin based on the WOA-RF model and to evaluate the impacts of climate change and human activities on the changes in the hydrological regime in the Zi River Basin. This study helps us systematically understand the characteristics of changes in the hydrological regime of the Zi River Basin. The results can provide a new approach to studying hydrological changes in the Zi River Basin and are of great reference value for watershed ecological protection and water resource management. Additionally, they serve as a guide for similar research endeavors.

Study area

The Zi River Basin, located in central Hunan Province, is one of the four major water systems within the Dongting Lake system of the Yangtze River Basin, ranking as the third largest river in Hunan (Zhou & Mao 2022). The source of the Zi River is divided into the left source and the right source: the left source, Nan Water, originates from Bei Qingshan of Chengbu Miao Autonomous County; the right source, Fuyi Water, originates from Yue Chengling of Guangxi Resources County, and the two waters converge at Shuangjiangkou of Shaoyang County. The river flows through several key locations, including Shaoyang, Lengshuijiang, Xinhua, and Taojiang, ultimately merging into Dongting Lake near Ganxi Harbor in Yiyang. The total length of the river is 653 km, and it encompasses a basin area of 28,142 km² (Li et al. 2018). The Zi River Basin features significant infrastructure, including large reservoirs such as Zhexi (Class I) and Majitang (Class II), primarily constructed for power generation. Zhexi Reservoir, the largest in the basin, is situated on the mainstream Zi River in Zhexi Township, Anhua County. It is characterized by a concrete single-pier dam with a substantial hydraulic head, mainly for power generation while providing flood control and navigation benefits. The basin's topography exhibits a southwest-to-northeast gradient, with higher elevations in the southwest and lower elevations in the northeast. This results in a pronounced elevation difference between the upstream and downstream sections. The basin's layout is illustrated in Figure 1.
Figure 1

Overview map of the Zi River Basin.

Figure 1

Overview map of the Zi River Basin.

Close modal

Data

The data used in this study include meteorological data and hydrological data. We collected daily flow data from the Taojiang Hydrological Station in the Zi River Basin, spanning from 1961 to 2017, and this data will be used to analyze the changing pattern of flow in the Zi River Basin. The Taojiang hydrological station is located at the downstream outlet of the Zi River Basin, and its long-term flow data were obtained from the Hunan Provincial Bureau of Hydrology and Water Resources Survey. In addition, the day-by-day meteorological data from 1961 to 2017 were obtained from five meteorological stations: Anhua, Xinhua, Wugang, Shaoyang, and Dongkou. The meteorological data were obtained from the raw data of day-by-day observations of 824 national basic meteorological stations provided by the China Meteorological Network (http://data.cma.cn/), covering indicators such as precipitation, average temperature, relative humidity, wind speed, sunshine duration, and potential evapotranspiration. The geographical locations and altitudes of all the above meteorological stations are shown in Table 1.

Table 1

Description of the hydrological and meteorological stations in the study area

StationStation typeLongitude (E)Latitude (N)Altitude (m)
Taojiang Hydrological station 112.12 28.53 52.10 
Xinhua Meteorological station 111.30 27.75 211.90 
Anhua 111.22 28.38 128.30 
Shaoyang 111.45 27.18 311.00 
Wugang 110.63 26.73 341.00 
Dongkou 110.60 27.03 339.50 
StationStation typeLongitude (E)Latitude (N)Altitude (m)
Taojiang Hydrological station 112.12 28.53 52.10 
Xinhua Meteorological station 111.30 27.75 211.90 
Anhua 111.22 28.38 128.30 
Shaoyang 111.45 27.18 311.00 
Wugang 110.63 26.73 341.00 
Dongkou 110.60 27.03 339.50 

Sequence trend, mutation, and periodicity tests

The Mann–Kendall (M–K) test is a typical nonparametric test that can be used to test for trends and significance in hydrological and meteorological time series (Pirnia et al. 2019). The test steps are as follows:

  • (1) Set up a time series {Xt} and a sequence of observations {xt, t= 1, 2, …, n} to construct a sequential column:
    (1)
  • (2) Under the assumption that the time series are randomly independent, the statistics are defined as follows:
    (2)

In this aforementioned formula, ‘UFk’ is the statistic of the standard normal distribution; ‘E(sk)’ is the average of sk; and ‘Var(sk)’ is the variance of sk, where UF1= 0, x1,x2…,xn is mutually independent and continuously identically distributed:
(3)
(4)
  • (3) Arrange the time series x in reverse order, repeat (1) and (2), and produce both:
    (5)
    where UB1 = 0, by analyzing statistical sequences UBk and UFk, the time node of mutation of sequence x can be further analyzed to show the mutation region. If the UBk and UFk curves intersect, and the intersection is between the two critical lines, then the moment corresponding to the intersection is when the mutation begins.

The cumulative anomaly method is mainly applied to hydrometeorological studies. It validates the ranking of cumulative anomalies by evaluating their dispersion Xt concerning the mean value of the data. In the case of Xt decreasing, it indicates values below the data mean, while its increasing indicates values above the mean. Fluctuation plots of their long-term series can show increases and decreases in the data. In general, the curve fluctuations reflect the timing of hydrological variations, and the mutation points usually correspond to the turning points of the curve or positions where the slope changes significantly (Wang et al. 2023a, 2023b, 2023b).

The time series x (x1, x2, x3, , xn) of cumulative outliers at a particular time t can be expressed as follows:
(6)
The sliding T test divides the sequence of runs at a certain point into subsequences and . The sample sizes are and , the means are and and the variances are and (Sun 2023). The test statistic is calculated using:
(7)
(8)

Equation (7) follows a t-distribution with . Given a threshold value , when , it is considered to be a hydrological mutation point.

The continuous wavelet transform method based on the Moret wavelet has been used to analyze the cyclical variability of flow on a long-term series scale (Briciu 2014). is a wavelet function and satisfies the relation:
(9)
(10)
In this equation, is the wavelet given by Equation. For a given energy limited signal , the continuous wavelet function is:
(11)

In this equation, ‘W’ is the wavelet transform coefficients; ‘f(t)’ is the signal or square productable function; ‘a’ is the scale factor, reflecting the period length of the wavelet; ‘b’ is the translation factor, reflecting the time translation. ‘’ is the complex conjugate of ‘’.

Evaluation of hydrological change degree

Richter et al. (1996) proposed that the IHA reflect rivers' hydrological regimes. They include indicators of flow, events, frequency, and rate of change, as shown in Table 2.

Table 2

Indexes of IHA

IHA statistics groupCharacteristicsParameters
Monthly water conditions (1–12) Magnitude, timing The mean value for each calendar month 
Extreme water conditions (13–23) Magnitude, duration Annual 1-, 3-, 7-, 30-, 90-day minimum and maximum; base flow index 
Timing of extreme water conditions (24–25) Timing Julian date of each annual 1-day minimum and maximum 
High and low pulses (26–29) Magnitude, duration, frequency No. and duration of low/high pulses 
Water condition changes (30–32) Frequency, variability No. of rises falls, and reversals 
IHA statistics groupCharacteristicsParameters
Monthly water conditions (1–12) Magnitude, timing The mean value for each calendar month 
Extreme water conditions (13–23) Magnitude, duration Annual 1-, 3-, 7-, 30-, 90-day minimum and maximum; base flow index 
Timing of extreme water conditions (24–25) Timing Julian date of each annual 1-day minimum and maximum 
High and low pulses (26–29) Magnitude, duration, frequency No. and duration of low/high pulses 
Water condition changes (30–32) Frequency, variability No. of rises falls, and reversals 

Richter et al. (1998) proposed the IHA-based RVA method for assessing changes in hydrological conditions. The RVA method is formulated as follows:
(12)
(13)

Po denotes the frequency with which a hydrological parameter falls within the flow management target range following a hydrological change. Pe denotes the expected value, which is the range of the ecohydrological target interval. The expected value for this study is 50%. Di is a positive value, which indicates that the parameter of the hydrological sequence after variation falls within the target interval higher than the expected value; Di is a negative value, which indicates that the parameter of the hydrological sequence after variation falls within the target interval lower than the expected value. When , the hydrological parameters are considered to be highly variable (H); when , the hydrological parameters are considered to be moderately variable (M); when , the hydrological parameters are considered to be lowly variable (L).

Black et al. (2005) proposed another broad hydrological alteration index, the Dundee Hydrological Regime Assessment Method (DHRAM), based on the five groups of 32 parameters of IHA, to measure the degree and extent of anthropogenic alteration of hydrological regimes. DHRAM classified the mean and coefficient of variation of the five groups of 32 parameters of IHA into three impact categories: Level 1 (one impact point with a low threshold), Level 2 (two impact points with a medium threshold), and Level 3 (three impact points with a high threshold) (Table 3). Summing the mean and coefficient of variation impact categories for each group of indicators resulted in a score (score range 0–30), by which the degree of impact was assessed for each of the five classes. The higher the score, the more significant the change in the hydrological situation of the river and the greater the likelihood of damage to the river ecosystem.

Table 3

Hydrological change thresholds used to assign impact points

IHA indicatorsChange in IHA group score
A point of influenceTwo points of influenceThree points influence
1a (The first set of means) 19.9 43.7 67.5 
1b (The first group of CVs) 29.4 97.6 165.7 
2a (The second set of means) 42.9 88.2 133.4 
2b (The second group of CVs) 84.5 122.7 160.8 
3a (The third group of means) 7.0 21.2 35.5 
3b (The third group of CVs) 33.4 50.3 67.3 
4a (The fourth group of means) 36.4 65.1 93.8 
4b (The fourth group of CVs) 30.5 76.1 121.6 
5a (The fourth group of means) 46.0 82.7 119.4 
5b (The fifth group of CVs) 49.1 79.9 110.6 
IHA indicatorsChange in IHA group score
A point of influenceTwo points of influenceThree points influence
1a (The first set of means) 19.9 43.7 67.5 
1b (The first group of CVs) 29.4 97.6 165.7 
2a (The second set of means) 42.9 88.2 133.4 
2b (The second group of CVs) 84.5 122.7 160.8 
3a (The third group of means) 7.0 21.2 35.5 
3b (The third group of CVs) 33.4 50.3 67.3 
4a (The fourth group of means) 36.4 65.1 93.8 
4b (The fourth group of CVs) 30.5 76.1 121.6 
5a (The fourth group of means) 46.0 82.7 119.4 
5b (The fifth group of CVs) 49.1 79.9 110.6 

Principal component analysis

PCA transforms a large number of correlated variables into a few uncorrelated variables through the idea of dimensionality reduction. It retains as much information as possible in the original data and the principal components (PCs). PCs are the new few composite variables, each a linear combination of the original variables with important applications in fields such as hydrology and ecology (Kholoosi et al. 2016; Yang et al. 2017). After dimensionality reduction, the indicators obtained by the IHA system are called the most ERHIs. These indicators represent many hydrological variables through fewer key hydrological variables and can reflect the most important ecologically relevant information. The ecologically relevant information reflected by the ERHIs screened in this paper is shown in Table 4. The specific process of the PCA implementation and the principle and methodology can be found in the literature (Shlens 2014).

Table 4

Summary of ERHIs parameters and their ecosystem influences

ERHIsEcosystem influences
Median flow in March
Median flow in May
Median flow in September 
  • Habitat availability for aquatic organisms

  • Soil moisture availability for plants

  • Influences water temperature, oxygen levels, photosynthesis in water column

 
Minimum 30-day flow
Maximum 3-day flow 
  • Structuring of aquatic ecosystems by abiotic vs. biotic factors

  • Structuring of river channel morphology and physical habitat conditions

 
Number of low-flow pulse occurrences
Number of high-flow pulse occurrences 
  • Frequency and duration of anaerobic stress for plants

  • Availability of floodplain habitats for aquatic organisms

  • Nutrient and organic matter exchanges between river and floodplain

 
Average rate of increase 
  • Entrapment of organisms on islands, floodplains (rising levels)

  • Desiccation stress on low-mobility streamedge (varial zone) organisms

 
ERHIsEcosystem influences
Median flow in March
Median flow in May
Median flow in September 
  • Habitat availability for aquatic organisms

  • Soil moisture availability for plants

  • Influences water temperature, oxygen levels, photosynthesis in water column

 
Minimum 30-day flow
Maximum 3-day flow 
  • Structuring of aquatic ecosystems by abiotic vs. biotic factors

  • Structuring of river channel morphology and physical habitat conditions

 
Number of low-flow pulse occurrences
Number of high-flow pulse occurrences 
  • Frequency and duration of anaerobic stress for plants

  • Availability of floodplain habitats for aquatic organisms

  • Nutrient and organic matter exchanges between river and floodplain

 
Average rate of increase 
  • Entrapment of organisms on islands, floodplains (rising levels)

  • Desiccation stress on low-mobility streamedge (varial zone) organisms

 

Reconstruction of natural flow

RF is a statistical learning theory based on the Classification Tree, which is representative of the contemporary integrated learning methods of classification and regression. The basic unit of the RF model is the Decision Tree. The model is highly accurate, and a bring-back random sampling method is used to extract the characteristic attributes of the samples, which can significantly reduce the probability of run-time overfitting (Gislason et al. 2006). The model is based on the best fit of the model in the baseline period to determine the number of trees and the number of random features. Finally, the optimal model is determined based on the Nash–Sutcliffe efficiency coefficient (NSE), coefficient of determination (R2), and percent bias (PBIAS). The basic model is as follows:
(14)
where gi is the predicted value; n is the total number of samples; hi(mi) is the training function of each decision tree; mi is the measured data.

The WOA consists of three phases: encircling predation, bubble net predation, and search predation.

The RF model and the Whale algorithm optimization allowed for a more accurate reconstruction of natural flow. The WOA-RF model in this study includes precipitation, mean air temperature, relative humidity, wind speed, sunshine, and potential evapotranspiration, and the output factor is monthly flow. Based on artificial parameter optimization, the monthly mean natural flow for 1961–2017, considering only the effect of climate change, was recovered using the monthly mean flow before the mutation and the precipitation, mean air temperature, relative humidity, wind speed, insolation, and potential evapotranspiration after the mutation.

NSE, R2, and PBIAS were used to evaluate the accuracy and stability of the model. The NSE could effectively assess the stability of the predicted values in the prediction model, and the larger the value, the better the result. R2 can be used to measure the degree of linear correlation between the simulated values and the measured values; the closer the value of R2 is to 1, the higher the linear fit is, and PBIAS can be used to test the accuracy of the predicted values, the closer the value is to 0, the better the model is. The formula is as follows:
(15)
(16)
(17)

In the above equation, and denote the observed and simulated flow values (m3/s), respectively, and and denote the average observed and simulated flow values (m3/s), respectively.

Quantitative evaluation of flow changes

This study was conducted using Peng et al.'s (2016) proposed model for calculating the relative contribution to quantifying the contribution of climate change and human activities to flow changes:
(18)
(19)
(20)
where: is the change in the impact period relative to the multi-year average flow in the base period (m3/s), which is the amount of change due to the combined impacts of climate change and human activities; and are the multi-year averages of the measured flows in the base and impact periods, respectively (m3/s); and are the changes in flows due to climate change and human activities between the two periods, respectively (m3/s); and and are the multi-year averages of the flows in the base period and the impact period obtained from the modeling simulation (m3/s), which are the flows obtained by the assumption of only climate change impacts in the impact period.
From this, it can be concluded that the impacts of climate change and human activities on flow changes are, respectively:
(21)
(22)
where and represent the relative contribution of climate change and human activities to flow changes, respectively.

Analysis of hydrometeorological elements

The M–K test was used to analyze the trend of mean annual precipitation, mean annual runoff, and mean potential yearly evaporation at Taojiang station from 1961 to 2017, as shown in Figure 2, in which the test statistics of runoff and precipitation were −0.1170 and −0.3098, which were between ±1.96 and did not pass the test of the 95% significance level, indicating that the overall mean annual runoff and precipitation in the Zi River Basin were on a decreasing trend, but not significant. The test value of potential evaporation was 0.1033 between ±1.96, indicating that the potential evaporation of the Zi River Basin showed an increasing trend, but it was not significant.
Figure 2

(a) Interannual variability of runoff versus precipitation. (b) Interannual variability of runoff versus potential evaporation.

Figure 2

(a) Interannual variability of runoff versus precipitation. (b) Interannual variability of runoff versus potential evaporation.

Close modal
Sudden change analysis of the annual average flow sequence at the Taojiang Station from 1961 to 2017 was conducted using the M–K test, cumulative anomaly method, and sliding t-test. The results are presented in Figure 3. Figure 3(a) shows that the UF and UB curves exhibit multiple mutation points within the 95% confidence interval. Consequently, the cumulative anomaly method and sliding t-test were employed to identify the specific mutation year further. According to Figure 3(b), significant changes in average annual flow occurred during the 1980s, with notable turning points identified in 1977, 1987, 1989, and 2006. A sliding t-test was subsequently performed on these results. Figure 3(c) indicates that the statistical value 1987 was t = −4.28, with its absolute value exceeding the threshold of 2.78, thus successfully passing the 95% significance test. This indicates that an actual mutation occurred in the Zi River basin in 1987, which is designated as this study's mutation year. The findings are summarized in Table 5. These results are generally consistent with the analysis conducted by Wang et al. (2019) regarding the flow of the Zi River. Therefore, 1987 is confirmed as the year of mutation for the Zi River flow.
Table 5

Result of the flow mutation test

Hydrological stationsMutation point
M–K testCumulative anomaly methodSliding t-testMutation year
Taojiang station 1961, 1963, 1968, 1972, 1975, 1978, 1984, 1987, 2004 1977, 1987, 1989, 2006 1987 1987 
Hydrological stationsMutation point
M–K testCumulative anomaly methodSliding t-testMutation year
Taojiang station 1961, 1963, 1968, 1972, 1975, 1978, 1984, 1987, 2004 1977, 1987, 1989, 2006 1987 1987 
Figure 3

(a) M–K mutation test. (b) The cumulative anomaly method. (c) Sliding t-test method.

Figure 3

(a) M–K mutation test. (b) The cumulative anomaly method. (c) Sliding t-test method.

Close modal
Morlet wavelet analysis was employed to investigate the periodicity of mean annual flow at Taojiang Station from 1961 to 2017, resulting in wavelet variance and contour plots (Figure 4). The wavelet variance plot (Figure 4(b)) reveals four distinct peaks corresponding to time scales of 5, 8, 14, and 29 years. The 29-year time scale is likely indicative of the primary cycle of flow variability, while the 5-year scale corresponds to the second primary cycle, the 14-year scale to the third, and the 8-year scale to the fourth. At the 29-year time scale, the average cycle of flow change in the basin is approximately 25 years, characterized by about two transitions between periods of abundance and drought. The 14-year time scale shows an average cycle of 11 years, with around five abundance-drought transitions. For the 8-year time scale, the average cycle is between 6 and 7 years, displaying a relatively smooth pattern before 1994 but becoming more ambiguous afterward. Finally, at the 5-year time scale, the average cycle is approximately 4 years, marked by frequent alternations between high and low flows, resulting in significant fluctuations. This smaller scale can be divided into three main phases: before 1994 when the cycle appears relatively ambiguous; from 1994 to 2006, when it is characterized by a smoother cycle; and after 2006, when the cycle becomes more ambiguous again.
Figure 4

Variation of the flow cycle in the Zi River Basin: (a) wavelet contour and (b) wavelet variance.

Figure 4

Variation of the flow cycle in the Zi River Basin: (a) wavelet contour and (b) wavelet variance.

Close modal

Analysis of overall hydrological changes

In analyzing the degree of change in hydrological indicators, the RVA method was employed to quantify these changes (see Figure 5 and Table 6). Of 32 hydrological indicators, 14 exhibited a medium degree of change, while another 14 showed a low degree, collectively accounting for 44% of the total indicators assessed. Additionally, four hydrological indicators demonstrated a high degree of change, representing 12% of the total. Notably, the November median flow was the indicator with the highest degree of hydrological change. Conversely, the indicators with the lowest degree of change included the September median flow, maximum 1-day flow, maximum 3-day flow, average delay of the low-flow pulse, and average delay of the high-flow pulse. Overall, most hydrological indicators in the Zi River Basin exhibited low to medium degrees of change before and after the identified sudden change in flow. According to the RVA formula, the degree of change for five groups of indicators was measured at 45, 35, 26, 42, and 56%, respectively. Four of the five groups of indicators realized a moderate degree of change, and one group realized a low degree of change, with an overall hydrological change rate of 42%. From the perspective of degree of change, specific IHA indicators in the same group have similar values of degree of change, which also side-steps the necessity of indicator screening.
Table 6

Hydrological changes by group and overall

Hydrographic stationsDegree of change for each group (%)
Overall hydrological variability (%)
Group 1Group 2Group 3Group 4Group 5
Taojiang station 45 (M) 35 (M) 26 (L) 42 (M) 56 (M) 42 (M) 
Hydrographic stationsDegree of change for each group (%)
Overall hydrological variability (%)
Group 1Group 2Group 3Group 4Group 5
Taojiang station 45 (M) 35 (M) 26 (L) 42 (M) 56 (M) 42 (M) 
Figure 5

Hydrological variability at the Taojiang station.

Figure 5

Hydrological variability at the Taojiang station.

Close modal

The degree of hydrological change was derived using the DHRAM method (Table 7). From the study results, the total DHRAM score of Taogang station is 4, and the degree of change is 3, which is low. Considering the 32 hydrological change indicators in the IHA, the overall change in the degree group is more moderate than low change, and combining the results of the two methods, it is determined that the degree of change in the hydrological situation of the reservoir is moderate.

Table 7

Degree of changes in hydrological indicator composites

Hydrographic stationsGroupChange ratio (%)Influence degreeTotal score
MeanCVsMeanCVs
Taojiang station 16.6 28 4(3) 
8.7 28.3 
375.6 27.9 
34.3 30.5 
12.7 34.3 
Hydrographic stationsGroupChange ratio (%)Influence degreeTotal score
MeanCVsMeanCVs
Taojiang station 16.6 28 4(3) 
8.7 28.3 
375.6 27.9 
34.3 30.5 
12.7 34.3 

Note: CV is the coefficient of variation. Values in parentheses in the ‘total score’ list indicate the level of ecosystem risk due to hydrological change.

Selection of key ERHIs and characterization of their evolution

PCA was conducted to optimize IHA for the Taojiang Station from 1961 to 2017. The eigenvalues and cumulative contributions of the 32 components are illustrated in Figure 6(a). The first principal component (PC1) accounts for 22.35% of the total variance, with an eigenvalue of 7.15. The cumulative contribution of the second principal component (PC2) combined with PC1 is 38.29%, with an eigenvalue of 5.10. Notably, the eigenvalues of the first nine PCs exceed 1, and their cumulative contribution reaches 81%. PCs with cumulative contributions more significant than 80% and eigenvalue greater than 1, i.e., PC1 to PC9, were retained based on the extraction criteria. The factor loading matrices for these nine PCs were calculated, as shown in Figure 6(b). The IHA corresponding to the highest factor loading in each PC was selected as the ERHIs, with one indicator overlapping. Consequently, eight ERHIs were identified for the Taojiang River hydrological stations from 1961 to 2017. These indicators are median flow in March, median flow in May, median flow in September, minimum 30-day flow, maximum 3-day flow, number of low-flow pulse occurrences, number of high-flow pulse occurrences, and average rate of increase.
Figure 6

(a) Eigenvalues and cumulative contribution of each component. (b) Heatmap of factor loadings for the first nine PCs that satisfy the conditions.

Figure 6

(a) Eigenvalues and cumulative contribution of each component. (b) Heatmap of factor loadings for the first nine PCs that satisfy the conditions.

Close modal
We further analyzed the correlation between these indicators. As shown in Figure 7, it is clear that the 28 correlation coefficients (absolute values) generally do not exceed 0.3 (absolute values), and the correlation coefficients below 0.3 reach 24, accounting for 85% of the total, and the maximum value is only 0.49. The problem of redundancy between IHA parameters has been successfully resolved. The combined degree of change of the eight ERHIs at the Taojiang hydrological station is 48.5%, which can be seen to be not much different from the overall degree of hydrological change of 42%, verifying the reasonableness of the selection of the hydrological indexes and their representativeness of the ecohydrological regime in the Zi River Basin.
Figure 7

Correlation between the selected ERHIs.

Figure 7

Correlation between the selected ERHIs.

Close modal
Figure 8 shows the time-varying characteristics of the eight ERHIs at the Taojiang Hydrological Station. The number of low-flow pulse occurrences and the average rate of increase increased by 0.31 and 0.75 per year, respectively, indicating that the frequent occurrence of low-flow pulses during the dry season may lead to the shrinkage of aquatic habitats and affect their survival. The median flow in March, the minimum 30-day flow, and the number of high-flow pulse occurrences showed an increasing trend, which may change the conditions of spawning for fish in the spring. The median flow in May steadily reduced, which may affect the growth of aquatic vegetation and the stability of the wetland ecosystem. Meanwhile, the median flow in September and the maximum 3-day flow did not change much in general, indicating that the hydrological conditions were relatively stable in the fall. These trends suggest that the hydrological situation in the basin is changing, which may have far-reaching impacts on the regional ecosystem.
Figure 8

(a) Median flow in March. (b) Median flow in May. (c) Median flow in September. (d) Minimum 30-day flow. (e) Maximum 3-day flow. (f) Number of low-flow pulse occurrences. (g) Number of high-flow pulse occurrences. (h) Average rate of increase. (The red line in the graph indicates a linear increasing trend, the green line indicates a linear decreasing trend, and the yellow line indicates a linear steady trend.)

Figure 8

(a) Median flow in March. (b) Median flow in May. (c) Median flow in September. (d) Minimum 30-day flow. (e) Maximum 3-day flow. (f) Number of low-flow pulse occurrences. (g) Number of high-flow pulse occurrences. (h) Average rate of increase. (The red line in the graph indicates a linear increasing trend, the green line indicates a linear decreasing trend, and the yellow line indicates a linear steady trend.)

Close modal

RF model and attribution analysis

The WOA-RF model of monthly mean flow in the base period is constructed by combining the conditions of hydrometeorological data. The training and validation periods are divided according to the proportion of the base period (1961–1986) and the variability period (1987–2017) of the natural cycle to the whole cycle (1961–2017), with 1961–1973 as the training period and 1973–1986 as the validation period. The parameters related to the calibration of the WOA-RF model mainly include population size (POP), the maximum number of selective generations (Max-iter), the number of hyperparameters (dim), the lower limit (lb), and the upper limit (ub). The parameter values used in this study were 40, 10, 2, [1, 1] and [10, 10], respectively. The simulated average monthly flow rate after parameter adjustment and the actual average monthly flow rate are shown in Figure 9. At this time, the model fits better, with NSE of 0.945 in the training period and 0.889 in the validation period, the fitting coefficient R2 of 0.941 in the training period and 0.893 in the validation period, PBIAS of 0.8% in the training period and −4.8% in the validation period. The fitting accuracy can be used to simulate the whole study period, and the natural flow recovered by the model can be used as a basis for analyzing the flow changes in the Zi River Basin.
Figure 9

(a) Comparison of simulated and measured monthly mean flow values at the Taojiang station for 1961–1972. (b) Correlation plot between model and actual for the training period. (c) Comparison of simulated and actual monthly mean flow values at the Taojiang station for 1973–1986. (d) Correlation plots of model versus actual for the validation period.

Figure 9

(a) Comparison of simulated and measured monthly mean flow values at the Taojiang station for 1961–1972. (b) Correlation plot between model and actual for the training period. (c) Comparison of simulated and actual monthly mean flow values at the Taojiang station for 1973–1986. (d) Correlation plots of model versus actual for the validation period.

Close modal
After constructing the base period WOA-RF model, the impact period climate factors were input for simulation throughout the study period. The WOA-RF model was used to simulate the month-by-month mean flow during the influence period, and the yearly mean flow process line was plotted through the time scale conversion. The simulation results of the yearly average flow at the Taojiang station are shown in Figure 10. The flow process line was calculated from , , and , and Equations (19) and (20) were used to obtain , , i.e., climate change contributed 28.8% to flow, and human activities contributed 71.2%. The influence of climate change and human activities on the flow is in the opposite direction. Climate change causes the flow to show a downward trend, while human activities show an upward trend, and the contribution of human activities to the flow change is more significant.
Figure 10

Simulation results of the annual mean flow at the Taojiang station.

Figure 10

Simulation results of the annual mean flow at the Taojiang station.

Close modal
The land use data of 1980, 1990, 2000, 2010, and 2020 years were used to filter out the mainland categories leading to anthropogenic changes by PCA. Figure 11(a) shows the eigenvalues and cumulative contributions of the six components. The PC1 accounted for 65.6% of the total variance and had an eigenvalue of 3.94. The PC2 had a cumulative contribution rate of 91.04% with PC1 and an eigenvalue of 1.53. According to the extraction criteria, PCs with a cumulative contribution rate of more than 80% and an eigenvalue of more than 1, i.e., PC1 to PC2, were retained. We found that the contribution rate of the land changes caused by the construction of reservoirs could reach 65.6%, which is the most significant contributor to the shift influenced by human activities.
Figure 11

(a) Eigenvalues and cumulative contribution of each component. (b) Heatmap of factor loadings for the first two PCs that satisfy the conditions.

Figure 11

(a) Eigenvalues and cumulative contribution of each component. (b) Heatmap of factor loadings for the first two PCs that satisfy the conditions.

Close modal

Ecological impacts of changes in hydrological regimes

This study used PCA to screen eight ERHIs in the Zi River Basin. The eight metrics are median flow in March, median flow in May, median flow in September, minimum 30-day flow, maximum 3-day flow, number of low-flow pulse occurrences, number of high-flow pulse occurrences, and average rate of increase. Median monthly flows (median flows in March, May, and September) mainly affect habitat availability for riverine aquatic organisms as well as water affecting water temperature, oxygen levels, photosynthesis in the water column (CONSERVANCY T N 2009); minimum 30-day flow, maximum 3-day flow are related to channel morphology and structure and natural habitat conditions (CONSERVANCY T N 2009); the number of low-flow pulse occurrences, and the number of high-flow pulse occurrences mainly affects aquatic organisms habitat availability in the river rookery as well as the relationship between the river and the nutrient and organic matter exchanges between the river and the rambler (CONSERVANCY T N 2009); The average rate of increase, in turn, correlates with the trapping (rising numbers) of organisms on islands and flood plains (CONSERVANCY T N 2009). Gao et al. (2009) found some variation in the results of analyses across watersheds and datasets. However, it tended to include three common elements: at least one monthly mean flow statistic, at least two extreme event statistics representing high and low extremes, and at least one statistic correlating the frequency of low and high flows. It agrees with the eight ERHIs initially selected for this paper.

Ecological response analyses are an important form of assessment to address the adverse impacts of changes in river hydrological regimes on ecosystems (Döll & Zhang 2010; Kuriqi et al. 2021). Thus, discussing the ecological response to the current hydrological regime is necessary. The evaluation of ecological response to changes in the hydrological regime of the basin can be analyzed in terms of changes in the streamflow biodiversity index (Wang et al. 2023a, 2023b, 2023b). Figure 12 demonstrates the results of the analysis of the interannual changes in the biodiversity index of the Zi River Basin using the SI index from 1961 to 2017. The flow biodiversity index of the Zi River Basin showed an increasing trend throughout the study, and the natural habitat conditions in the Zi River Basin were favorable. As the largest hydropower station in the Zi River Basin, the Zhexi Hydropower Station is located in the lower reaches of the Zi River. After the reservoir is impounded, the original water flow of the water body becomes slow flow or still water, which provides a suitable environment for the channel catfish subfamily and fish family, which prefer slow flow and open water life. However, because the dam's power generation and flood discharge are affected by the seasons, the flow, water level, and bait organisms in the river below the dam vary significantly with the seasons, adversely affecting the growth and reproduction of fish (Liu et al. 2024). The construction of intensively graded dams has hurt fish growth and reproduction. The construction of intensively graded dams has dramatically changed the hydrological regime and ecological environment of the river, narrowing the living space of fish and hindering their reproduction, growth, feeding, and other activities, especially the reproduction and survival of migratory fish, which have been greatly affected (Liu et al. 2014). Comprehensive strategies are proposed to address the Zi River Basin ecosystem's sensitivity to hydrological changes, particularly seasonal flow fluctuations and extreme events. Optimize reservoir operations and implement ecological scheduling to stabilize flow during peak fish reproduction periods (spring and early summer), avoiding extremely low flow or sudden fluctuations that disrupt spawning and juvenile growth. During flood discharge, minimize frequent or intense high-flow pulses by establishing dynamic flow thresholds to adjust operations based on ecological needs.
Figure 12

SI index for the Zi River Basin.

Figure 12

SI index for the Zi River Basin.

Close modal

Impact of human activities on changes in the hydrological regime

Previous studies have established that human activities are the primary factor influencing changes in the hydrological regime of the basin. These activities directly or indirectly affect the quality, quantity, and flow processes through the construction of hydraulic infrastructure and modifications to subsurface conditions (Fan et al. 2023). Li et al. (2018) excluding evapotranspiration from their analyses, quantitatively assessed the impacts of precipitation and human activities on flow changes during the variability period in the Zi River Basin. They concluded that human activities were the predominant contributors to these changes. Further research has underscored that the alterations in flow within the Zi River Basin are primarily attributable to human interventions, including reservoir construction, urbanization, artificial water consumption, and other integrated factors (Li 2018). While the present study's findings align with these conclusions, notable differences may stem from the exclusion of evapotranspiration in prior studies, variations in the definition of base and variant periods, and differences in the data series and methodologies employed. Nonetheless, the overarching consensus remains that human activities are the dominant factor driving flow variability. The downstream of the Zi River Basin in Yiyang City has constructed terrace power stations, such as Zhexi and Majitang (Li et al. 2018). Under the effect of the reservoirs and dams, the average flow in May of the flood season was reduced, and the average flow in March of the dry season was slightly increased. The analyses in the ERHIs section of this study are in agreement. After the construction of a reservoir on a river, there is a tendency for the flow volume to decrease significantly, especially when it is first put into operation. However, this tendency will gradually slow down or rebound after regular operation. Therefore, the direct effect of dam construction on total flow is limited (Chen et al. 2021). However, in the 1990s, the Zhexi Reservoir needed more flood control capacity and high discharge. The Zi River Basin often suffered from floods, which, to some extent, increased Zi River flow (Xiong & Liu 1999).

Changes in land use affect the hydrological cycle, altering hydrogeological conditions and water availability and thus affecting flow changes in the basin (Liu et al. 2023). Xiong & Liu (1999) mentioned that soil erosion in the Zi River Basin increased in the 1990s in the hilly areas and that the Zi River Basin had an area of 6,655 km2 of soil erosion, which accounted for 31.9% of the area of the hillslopes and an average multi-year sand loss of 2,860,000t. The Zi River Basin had an increasing trend of flow, and the possible causes of this phenomenon were insufficient soil and water conservation measures: irrational land use may lead to soil erosion, which increases flow. When vegetation is removed or destroyed, the soil is unable to hold water efficiently, and water will flow more easily into the river, thus increasing the amount of flow (Wang et al. 2024); urbanization and land development: urbanization and land development activities usually change the cover and permeability of the ground surface, which results in the water not being able to infiltrate well into the soil and increases the amount of flow. Many artificially hardened surfaces, such as buildings, roads, and car parks, can lead to a rapid flow of rainwater into the river, thus increasing the amount of flow (Wei et al. 2023). Figure 13(a) shows the land use changes in the Zi River Basin for different periods (1980, 1990, 2000, 2010, and 2020). In 2020, the distribution of land use in the Zi River Basin was as follows: forest land (65.34%), cropland (27.9%), grassland (3.32%), construction land (1.92%), water (1.51%), and unused land (0.01%). Figure 13(b) shows the changes in land use before and after the construction of the reservoirs; the more significant land use shift in 1980–2000 may be related to rapid urbanization and infrastructure construction, while the minor land shift in 2000–2020 may be due to changes in the development pattern and the emphasis on sustainable use. Reservoir construction played an essential regulatory role in this process, influencing land use patterns by changing water resource allocation and the ecological environment.
Figure 13

(a) Land use changes over time in the Zi River Basin. (b) Land use changes before and after reservoir construction.

Figure 13

(a) Land use changes over time in the Zi River Basin. (b) Land use changes before and after reservoir construction.

Close modal
Figure 14 shows the land use changes from 1980 to 2020. The study shows that the proportion of forested land and cropland area is more significant in the land cover change of the Zi River Basin from 1980 to 2020. Compared with 1980, in 2020, the area of water and the area of construction land all showed an upward trend, respectively, from 1.19 to 1.51%, 0.81 to 1.92%; the area of forested land and unutilized land showed a weak upward trend, respectively, from 65.25 to 65.34%, 0.007 to 0.01%; the area of grassland showed a stable trend (3.32%). Meanwhile, the area of cropland shows a decreasing trend, from 29.43 to 27.90%. A reduction in the area of cropland may lead to a decrease in the soil's water retention capacity, which reduces groundwater recharge and the ability to regulate surface flow. During rainfall, surface flow increases, and soil erosion is exacerbated. Increased water helps improve the hydrologic cycle by increasing evapotranspiration and transpiration, thus providing more water during the dry season. Increased construction land reduces infiltration capacity, leading to increased surface flow and potentially exacerbating flood flows (Wang et al. 2023a, 2023b, 2023b). Over the past 40 years, the transfer of construction land and cropland has been the most complex, followed by the transfer of grassland, water, and forest land. The most direct manifestation of the complex land use changes in the Zi River Basin is soil erosion. Therefore, land use change analysis is the key to analyzing the factors affecting flow changes.
Figure 14

Land use change ratio chord map for the Zi River Basin.

Figure 14

Land use change ratio chord map for the Zi River Basin.

Close modal

Applicability and limitations

In this study, we propose a structured and robust framework for assessing the hydrological regime change and its ecological impacts. It can be adapted to a wide range of watersheds with different hydrological, climatic, and anthropogenic disturbances. PCA can be applied across watersheds to screen for ERHIs. Still, initial hydrological indicators need to be selected in accordance with the characteristics of the watersheds, e.g., in arid zones, priority should be given to the extreme flow indicators. The WOA-RF model can effectively differentiate the effects of human activities and climate change on flow and is applicable to basins with different climatic conditions. However, the implementation of the WOA-RF model needs to pay attention to the availability of data, the main drivers, and the model's calibration. Alternative methods or small-scale datasets can be explored for data-scarce basins, and basins dominated by climate change should incorporate more climatic variables. The model should be recalibrated so that each basin can adapt to its hydrological processes.

The WOA-RF model demonstrates exemplary performance in flow simulation but has limitations affecting accuracy, particularly during extreme events like floods and droughts. Parameter tuning, though critical, lacks standardization and relies on experience rather than field surveys, introducing uncertainty. Global sensitivity analysis highlights the dominant role of energy balance-related factors (e.g., potential evapotranspiration, insolation) in flow formation. At the same time, rainfall and air temperature have more minor but non-negligible effects (Figure 15). The model's sensitivity to input factor interactions, especially during extreme events, increases prediction uncertainty. For example, soil moisture and vegetation dynamics, not explicitly included in the model, significantly influence flow responses during floods and droughts. Although rainfall's primary effect is low, its total effect is high due to interactions with other factors, making it critical during extreme events. Future studies should enhance model performance by incorporating complex hydrological modules (e.g., soil moisture, vegetation) and higher-resolution input data to capture nonlinear responses during extreme conditions better.
Figure 15

Sobol sensitivity analysis (Features 1–6 represent input factors as precipitation, mean air temperature, relative humidity, wind speed, sunshine, and potential evapotranspiration, respectively).

Figure 15

Sobol sensitivity analysis (Features 1–6 represent input factors as precipitation, mean air temperature, relative humidity, wind speed, sunshine, and potential evapotranspiration, respectively).

Close modal

In order to quantitatively analyze the characteristics of hydrological regime changes in the Zi River Basin and the degree of influence of climate change and human activities on the hydrological regime changes in the Zi River Basin, we started with the inter-annual characteristics of hydrological regime changes. On this basis, the key ERHIs were selected to reduce the influence of redundant information on the hydrological regime change. The WOA-RF model was also used to quantitatively separate the degree of changes in the flow of the Zi River due to human activities and climate change. The main conclusions are as follows:

  • (1) The mutation results show that the flow in the Zi River Basin underwent a mutation in 1987. The overall mean annual flow and mean annual precipitation in the basin showed a decreasing trend, and the potential evaporation showed an increasing trend. The results of wavelet analyses indicated significant cycle changes on time scales of 29 and 5 years, with average cycles of 25 and 4 years.

  • (2) Combining the IHA-RVA and DHRAM methods, the 32 ecohydrological indicators in the Zi River Basin were mainly medium and low changes, with an overall change degree of 42%, and the total DHRAM score was 4. After the hydrological indicators were preferred, eight ERHIs were identified, with a combined change degree of 48.5%, a medium change. Compared with the traditional RVA method and DHRAM method, the degree of change of the hydrological regime of Zi River is more intuitively shown.

  • (3) The results of the WOA-RF model showed that climate change caused a downward trend in flow, while human activities caused an upward trend in flow. Meanwhile, the results of attribution analysis showed that the contribution of human activities and climate change to the change of flow of the capital water was 71.2 and 28.8%, respectively, and that human activities were the main factor affecting the change of flow of the capital water. The construction of reservoirs was the most significant factor contributing to the impact of human activities, with a 65.6% contribution.

The authors thank their brothers at North China University of Water Resources and Electric Power for their comments and help with this study.

This study was supported by Basic Research Project of Key Scientific Research Projects of Colleges and Universities of Henan Province (24ZX007).

X.Z. conceived the study and wrote the first draft, X.Z. and X.B. collected the data and performed the analysis, L.H. and X.J. analyzed the methodology, W.G. and H.W. supervised the paper, and all authors provided comments and assistance on the first few versions of the manuscript. All authors read and approved the final manuscript.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.

Black
A. R.
,
Rowan
J. S.
,
Duck
R. W.
,
Bragg
O. M.
&
Clelland
B. E.
(
2005
)
DHRAM: a method for classifying river flow regime alterations for the EC Water Framework Directive
,
Aquatic Conservation Marine and Freshwater Ecosystems
,
15
,
427
446
.
https://doi.org/10.1002/aqc.707
.
Briciu
A. E.
(
2014
)
Wavelet analysis of lunar semidiurnal tidal influence on selected inland rivers across the globe
.
Sci Rep, 4, 4193. https://doi.org/10.1038/srep04193
.
Chen
Y. B.
,
Zhang
X.
,
Yang
W. F.
&
Yang
Y. F.
(
2021
)
Quantitative analysis on influence of cascade reservoir on downstream runoff during construction and operation period
,
Yangtze River
,
52
,
116
123
.
https://doi.org/10.16232/j.cnki.1001-4179.2021.04.018
.
CONSERVANCY T N
(
2009
)
Indicators of Hydrologic Alteration, Version 7.1 User's Manual, 8–9
.
Döll
P.
&
Zhang
J.
(
2010
)
Impact of climate change on freshwater ecosystems: a global-scale analysis of ecologically relevant river flow alterations
,
Hydrology and Earth System Sciences
,
14
,
783–799. https://doi.org/10.5194/hess-14-783-2010
.
Fan
Y. Q.
,
Zheng
H.
,
Liu
Y. Y.
&
Wan
W. H.
(
2023
)
Study on hydrological regime alteration of lower reaches of Dongjiang River Basin
,
Water Resources and Hydropower Engineering
,
54
,
77
87
.
https://doi.org/10.13928/j.cnki.wrahe.2023.07.007
.
Fathi
M. M.
,
Awadallah
A. G.
,
Abdelbaki
A. M.
&
Haggag
M.
(
2019
)
A new Budyko framework extension using time series SARIMAX model
,
Journal of Hydrology
,
570
,
827
838
.
https://doi.org/10.1016/j.jhydrol.2019.01.037
.
Gao
Y.
,
Vogel
R. M.
,
Kroll
C. N.
,
Poff
L. R.
&
Olden
J. D.
(
2009
)
Development of representative indicators of hydrologic alteration
.
Journal of Hydrology, 374, 136–147. https://doi.org/10.1016/J.JHYDROL.2009.06.009
.
Gislason
P. O.
,
Benediktsson
J. A.
&
Sveinsson
J. R.
(
2006
)
Random forests for land cover classification
,
Pattern Recognition Letters
,
27
,
294
300
.
https://doi.org/10.1016/j.patrec.2005.08.011
.
Guan
X.
,
Zhang
Y.
,
Meng
Y.
,
Liu
Y.
&
Yan
D.
(
2021
)
Study on the theories and methods of ecological flow guarantee rate index under different time scales
,
Science of The Total Environment
,
771
,
145378
.
https://doi.org/10.1016/j.scitotenv.2021.145378
.
Gunawardana
S. K.
,
Shrestha
S.
,
Mohanasundaram
S.
,
Salin
K. R.
&
Piman
T.
(
2021
)
Multiple drivers of hydrological alteration in the transboundary Srepok River Basin of the Lower Mekong Region
,
Journal of Environmental Management
,
278
,
111524
.
https://doi.org/10.1016/j.jenvman.2020.111524
.
Guo
W. X.
,
Jin
Y. G.
,
Li
Y.
,
Shi
M. Y.
&
Wang
H. X.
(
2021
)
Quantitative assessment of the impact of climate change and human activities on the hydrological regime in the middle and lower reaches of the Yanatze River during the past 60 years. Resources and Environment in the Yangtze Basin 30, 2464–2470
.
Haider
M. R.
,
Dee
S. G.
,
Doss-Gollin
J.
,
Dunne
K. B. J.
&
Muñoz
S. E.
(
2025
)
Impact of 21st century climate change on Mississippi River Basin discharge in CESM2 large ensemble projections
,
Global and Planetary Change
,
104742
.
https://doi.org/10.1016/j.gloplacha.2025.104742
.
Jiang
S. H.
,
Liu
Y. T.
,
Ren
L. L.
,
Wang
M. H.
,
Zhou
L.
&
Ren
M. M.
(
2022
)
Attribution analysis of eco-hydrological regime evolution in the Weihe River Basin under changing environmen
,
Water Resources Protection
,
38
,
9
14
+ 70
.
Kholoosi
M. M.
,
Arabzadeh
R.
&
Bazrafshan
J.
(
2016
)
Regional hydrological drought monitoring using principal components analysis
,
Journal of Irrigation and Drainage Engineering
,
142
,
04015029
.
https://doi.org/10.1061/(ASCE)IR.1943-4774.0000925
.
Kuriqi
A.
,
Pinheiro
A. N.
,
Sordo-Ward
A.
,
Bejarano
M. D.
&
Garrote
L.
(
2021
)
Ecological impacts of run-of-river hydropower plants – current status and future prospects on brink of energy transition. current status and future prospects on the brink of energy transition
,
Renewable and Sustainable Energy Reviews
,
142
,
110833
.
https://doi.org/10.1016/j.rser.2021.110833
.
Li
T.
(
2018
)
Runoff Evolution Regularity and Attribution Analysis in Zijiang Watershed (Master)
.
Changsha, China
:
Changsha University of Science & Technology
.
Li
T.
,
Hu
G. H.
,
Gu
Q. F.
&
Zhou
H.
(
2018
)
Impact of precipitation and human activities on runoff in zishui river in recent 55 years
,
Journal of China Hydrology
,
38
,
54
58
+ 88
.
Liu
L. G.
,
Wang
W. B.
,
Yang
C. Y.
,
Luo
Y. S.
&
Yang
P. H.
(
2014
)
Survey on status of fish resources in mainstream of Zijiang River of Dongting Lake water system
,
South China Fisheries Science
,
10
,
1
10
.
Liu
W. L.
,
Li
X.
,
Wu
B.
,
Cao
X. G.
,
Huang
Y. P.
&
Liu
L. N.
(
2023
)
Impact of land Use change on runoff in the middle and upper reaches of xiuhe river basin
,
Research of Soil and Water Conservation
,
30
,
111
120
.
https://doi.org/10.13869/j.cnki.rswc.2023.03.010
.
Liu
Y.
,
Jiang
Y.
,
Zhang
S.
,
Wang
D.
&
Chen
H.
(
2024
)
Framework for comprehensive assessment of ecological water conveyance based on long-term evolution forecast of riparian forests in arid floodplains: synergistic consideration of flood inundation and groundwater
,
Process Safety and Environmental Protection
,
184
,
1357
1368
.
https://doi.org/10.1016/j.psep.2024.02.056
.
Mirjalili
S.
&
Lewis
A.
(
2016
)
The whale optimization algorithm
,
Advances in Engineering Software
,
95, 51–67. https://doi.org/10.1016/j.advengsoft.2016.01.008
.
Peng
D.
,
Qiu
L.
,
Fang
J.
&
Zhang
Z.
(
2016
)
Quantification of climate changes and human activities that impact runoff in the taihu lake basin, China
,
Mathematical Problems in Engineering
,
2016
,
1
7
.
https://doi.org/10.1155/2016/2194196
.
Pirnia
A.
,
Golshan
M.
,
Darabi
H.
,
Adamowski
J.
&
Rozbeh
S.
(
2019
)
Using the Mann-Kendall test and double mass curve method to explore stream flow changes in response to climate and human activities
,
Journal of Water and Climate Change
,
10
,
10
.
https://doi.org/10.2166/wcc.2018.162
.
Richter
B. D.
,
Baumgartner
J. V.
,
Powell
J.
&
Braun
D. P.
(
1996
)
A method for assessing hydrologic alteration within ecosystems
,
Conservation Biology
,
10, 1163–1174. https://doi.org/10.1046/j.1523-1739.1996.10041163.x
.
Richter
B.
,
Baumgartner
J. V.
,
Braun
D. P.
&
Powell
J. L.
(
1998
)
A spatial assessment of hydrologic alteration within a river network
,
Regulated Rivers-Research & Management
,
14
,
329
340
.
https://doi.org/10.1002/(SICI)1099-1646(199807/08)14:4<329::AID-RRR505 > 3.0.CO;2-
.
Sharifi
A.
,
Baubekova
A.
,
Patro
E. R.
,
Klöve
B.
&
Torabi Haghighi
A.
(
2024
)
The combined effects of anthropogenic and climate change on river flow alterations in the Southern Caspian Sea Iran
,
Heliyon
,
10
,
e31960
.
https://doi.org/10.1016/j.heliyon.2024.e31960
.
Shlens
J.
(
2014
)
A tutorial on principal component analysis
,
International Journal of Remote Sensing
,
51
,
1
12
.
https://doi.org/10.13140/2.1.1593.1684
.
Smakhtin
V. U.
,
Shilpakar
R. L.
&
Hughes
D. A.
(
2006
)
Hydrology-based assessment of environmental flows: an example from Nepal
,
International Association of Scientific Hydrology Bulletin
,
51
,
207
222
.
https://doi.org/10.1623/hysj.51.2.207
.
Soo
E. Z. X.
,
Chin
R. J.
,
Ling
L.
,
Huang
Y. F.
,
Lee
J. L.
&
Lee
F. W.
(
2024
)
Streamflow simulation and forecasting using remote sensing and machine learning techniques
,
Ain Shams Engineering Journal
,
15
,
103099
.
https://doi.org/10.1016/j.asej.2024.103099
.
Sun
H.
(
2023
)
Research on Ecological Flow Estimation in the Ganjiang River Basin Under Hydrological Alteration Conditions (Master)
.
Xian, China
:
Northwest University
.
https://doi.org/10.27405/d.cnki.gxbdu.2022.000022
.
Sun
Y. W.
,
Li
Q. Y.
,
Liu
L.
,
Xu
C. D.
&
Liu
Z. P.
(
2014
)
Hydrological simulation approaches for BMPs and LID practices in highly urbanised areas and the development of hydrological performance indicator system
,
Water Science and Engineering
,
7
,
12
.
https://doi.org/10.3882/j.issn.1674-2370.2014.02.003
.
Wang
H. X.
,
Zhao
Y. Y.
,
Zhuo
Z. Y.
&
Shi
Z. F.
(
2019
)
Hydrological regime assessment of zishui basin based on lHA-RVA method
,
Journal of North China University of Water Resources and Electric Power (Natural Science Edition)
,
40
,
16
21
+ 51. https://doi.org/10.19760/j.ncwu.zk.2019016
.
Wang
Y. X.
,
Liu
X.
&
Shen
Y. J.
(
2022
)
Applicability of the random forest model in quantifying the attribution of runoff changes
,
Chinese Journal of Eco- Agriculture
,
30
,
864
874
.
Wang
H.
,
Cheng
S.
,
He
N.
,
Huang
L.
,
Yang
H.
,
Hong
F.
,
Ma
Y.
,
Chen
W.
&
Guo
W.
(
2023a
)
Multi-scale flow regimes and driving forces analysis based on different models: a case study of the Wu river basin
,
Water Supply
,
23
,
3978
3996
.
https://doi.org/10.2166/ws.2023.199
.
Wang
H. X.
,
Wang
B. L.
,
Chen
H.
,
Yang
H.
&
Guo
W. X.
(
2023b
)
Analysis of evolution of hydrological regime and ecological response in Min river
,
Resources and Environment in the Yangtze Basin
,
32
,
1960
1970
.
Wang
W.
,
Wang
B.
,
Chau
K.
&
Xu
D.
(
2023c
)
Monthly runoff time series interval prediction based on WOA-VMD-LSTM using non-parametric kernel density estimation
,
Earth Science Informatics
,
16
,
2373
2389
.
https://doi.org/10.1007/s12145-023-01038-z
.
Wang
Q. H.
,
Zhang
P.
,
Wang
W.
,
Xiao
P. Q.
&
Wu
X. L.
(
2024
)
Regulation of land Use on runoff in typical watersheds of the middle reaches of the Yellow River
,
Research of Soil and Water Conservation
,
31
,
61
68
+ 78. https://doi.org/10.13869/j.cnki.rswc.2024.03.007
.
Wei
R.
,
Liu
J.
,
Zhang
T.
,
Zeng
Q.
&
Dong
X. H.
(
2020
)
Attribution analysis of runoff variation in the Upper-Middle Reaches of Yalong River
,
Resources and Environment in the Yangtze Basin
,
29
,
1643
1652
.
Wei
H. B.
,
Sun
Z. H.
&
Liu
J.
(
2023
)
Analysis of the lmpact of urbanization on runoff change in Jialu river
,
Water Resources and Power
,
41
,
16
18
.
https://doi.org/10.20040/j.cnki.1000-7709.2023.20221679
.
Xiong
G. H.
&
Liu
D. S.
(
1999
)
Genetic analysis of catastrophic flood disaster in the middle and lower reaches of Zishui river and calamity mitigation measures
,
Yangtze River
,
30
,
17
18
.
https://doi.org/10.16232/j.cnki.1001-4179.1999.10.007
.
Yang
B.
,
He
B. H.
&
Wang
D. B.
(
2017
)
Hanfeng Pre-reservoir commissioning time variation feature of the hydrology and water quality in three gorges reservoir
,
Environmental Science
,
38
,
1366
1375
.
https://doi.org/10.13227/j.hjkx.201607173
.
Yang
J.
,
Lin
Z. Q.
,
Zhao
Q. X.
,
Xie
J.
&
Liao
W. H.
(
2024
)
Research on the change of hydrological regime in the Yarlung Zanbo River Basin based on ERHls method
,
Journal of China Institute of Water Resources and Hydropower Research
,
22
,
108
119
.
https://doi.org/10.13244/j.cnki.jiwhr.20220413
.
Zhou
Y. L.
&
Mao
D. H.
(
2022
)
Evaluation and analysis of water production in Zishui basin and its response to changing environment based on INVEST model
,
Hubei Agricultural Sciences
,
61
,
16
23
.
https://doi.org/10.14088/j.cnki.issn0439-8114.2022.09.004
.
Zhou
Y.
,
Cui
T.
,
Zheng
X.
,
Yang
T.
&
Liu
M.
(
2017
)
Hydrological regime and environmental flow change in the source region of Yellow River based on lHA
,
Yellow River
,
39
,
61
64
+ 70
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).