<<

MARCH 2018 L I E T A L . 625

The Time Delay of Flow and Sediment in the Middle and Lower River and Its Response to the Three Gorges Dam

YANGYANG LI,YINGXIN ZHU,LEI CHEN, AND ZHENYAO SHEN State Key Laboratory of Water Environment Simulation, School of Environment, Beijing Normal University, Beijing,

(Manuscript received 5 August 2017, in final form 3 January 2018)

ABSTRACT

As the largest hydropower project in the world, the Three Gorges Dam (TGD) has drawn extensive concern in terms of its impact on downstream areas. In this study, an improved time delay estimation and wavelet analysis were used to investigate the influence of the TGD on the streamflow and sediment in the middle and lower Yangtze River, using time series of the daily discharge and sediment concentration data from three hydrological stations downstream of the dam. The results indicated that all of the time series at the three stations have prominent annual cycles, but the cycle of daily mean sediment concentration was nearly non- existent after the impoundment of the TGD. Changes in discharge and sediment between the Yichang and the Hankou stations are larger than those between the Hankou and the Datong stations, which is mainly at- tributed to the streamflows of tributaries and Dongting Lake and the flood diversion area of Jingjiang. The transmission time of discharge for the whole Yichang–Datong river section is approximately 6 days. In ad- dition, the attenuation of discharge from the Yichang station to the Datong station is 20%–30%. In contrast, the transmission of suspended sediment is slower than that of discharge, which takes 7–7.5 days to move from the Yichang station to the Datong station. The attenuation of sediment is approximately 30% in the Yichang– Datong river section and shows a clear increasing trend after 2006, mainly because a large amount of sediment was trapped by the TGD, and the dynamic balance of sediment was disturbed.

1. Introduction Variations in discharge and sediment at stations in the main stream or tributaries of the Yangtze River during The Three Gorges Dam (TGD), the world’s largest recent decades have been documented in previous dam, was completed and began storing water in 2003. studies. Dai et al. (2008) examined the impacts of the Large dams disrupt river continuity and unavoidably in- TGD impoundment and serious droughts on river dis- duce alterations in flow, sediment, and water temperature charge reduction in 2006. A sharp decrease of sediment regimes (Chen et al. 2016; Li et al. 2011; Syvitski et al. 2005; load after the impoundment of the TGD was reported Wang et al. 2016). With the construction of the TGD, by many studies (Li et al. 2011; Q. Zhang et al. 2012, considerable attention has been focused on how the dam 2013), and the significant downward trends in sediment impacts the river regime, especially the environment flux were likely to be dominated by human activities, downstream in the middle and lower Yangtze River (Chen especially dam construction (Zhao et al. 2012). Gao et al. 2016; Stone 2008).ThemiddleandlowerYangtze et al. (2013) investigated flow regime changes in the River basin is one of the most developed and densely middle and lower Yangtze River between the pre- and populated areas in China. The flow regime, changed by the postimpoundment periods and found that the TGD the TGD, may have a great influence on water supply significantly reduced the mean flow in October, while and economic development. Furthermore, sediment de- river discharge increased in February. Compared to position in the reservoir has reduced the amount of sedi- climate variability impacts on the catchment, the ment in the middle and lower reaches and has even TGD has a much greater influence on the seasonal resulted in significant topographic changes. Therefore, it is (September–October) dryness of Poyang Lake and has necessary to understand the impact of the TGD on the flow further altered the relationship between the river and and sediment regime downstream. the lake (Guo et al. 2012; Zhang et al. 2014). Mei et al. (2015) analyzed hydrological data from the Yichang, Corresponding author: Lei Chen, [email protected] Hankou, and Datong stations and noted that the

DOI: 10.1175/JHM-D-17-0119.1 Ó 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses). Unauthenticated | Downloaded 10/01/21 07:51 PM UTC 626 JOURNAL OF HYDROMETEOROLOGY VOLUME 19 hydrology of the Yangtze River is mainly controlled by the Using the time series of daily mean discharge and sedi- TGD. Luan and Jin (2016) calculated and analyzed the ment concentration data from the Yichang, Hankou, and impacts of the TGD impoundment on sediment based on a Datong hydrological stations, the present study was con- water-sediment transport numerical algorithm. These ducted with the following objectives: 1) to assess the pe- studies imply that the TGD may have a significant influ- riodic variation in discharge and sediment concentration ence on river discharge and sediment transport in the caused by the TGD and investigate the dependencies middle and lower Yangtze River. Although the seasonality among stations and 2) to quantify the lag time and atten- of the altered levels prevailed throughout the downstream uation of discharge and sediment in the main stream. reaches of the Yangtze River, the time of occurrence and magnitude of level changes varied among the stations (Wang et al. 2013). Understanding the lag time and 2. Materials and methods transmission factors of river discharge and sediment be- a. Description of the study area tween upstream and downstream is critical for building precise models and evaluating and improving control The Yangtze River, the longest river in China, plays a strategies for pollutants transported by the river. Never- vital role in Chinese economic development and environ- theless, there are insufficient studies on the lag time, river mental conservation. The expansive Yangtze River basin is discharge transmission factor estimation (TFE), and sus- divided into the upper, middle, and lower Yangtze reaches, pended sediment in the Yangtze River. though we concentrate on the middle and lower Yangtze Several methods have been used to investigate the im- River in this paper. The upper reaches extend from the river pacts of dam construction, such as the Mann–Kendall test source to the Yichang station, which is located just down- (Assani 2016; Chen et al. 2016), the continuous wavelet stream of the TGD. The middle reaches flow from the transform (CWT) technique (White et al. 2005; Zhang Yichang to Hankou stations with a length of 950 km. The et al. 2012), time series analysis (Li et al. 2011), scanning lower reaches are 600 km long, flowing from Hankou to t test (Q. Zhang et al. 2013), indicators of hydrologic alter- the river mouth, though the river section between the ation (IHA), and the range of variability approach (RVA; Hankou and Datong stations is defined as being within the Alrajoula et al. 2016; Wang et al. 2016). However, previous lower reach in this study to avoid the influence of tides studies on the Yangtze River focused more on analyzing (Fig. 1). There are two large lakes in the middle and lower individual stations, and few studies have evaluated the Yangtze River: Dongting Lake and Poyang Lake. Ex- quantitative relationships between upstream and down- changes of water and material between the two lakes and stream regions. A more detailed investigation is required the main channel support flood control, the ecosystem, and into the alterations of discharge and sediment regimes re- water supply in the middle and lower basin (Gao et al. 2013). sulting from the TGD with more accurate methods. Time The Yangtze River basin is characterized by a subtropical delay estimation (TDE) between the transmissions of an monsoon climate, and the monsoon-driven precipitation array of sensors has been utilized in different areas. For causes seasonal variability in the river flow, with high water example, Hocking and Kelly (2016) quantified the time lag and sediment discharge in the wet season from May to between rainfall and recharge in groundwater, and the re- October (Li et al. 2011). The construction of the TGD has sults showed that the significant delay should be in- resulted in a series of changes in flow and sediment regimes. corporated into numerical groundwater models or that it For example, after the impoundment of the TGD, a sharp was a source of a calibration error. DeWalle et al. (2016) decrease in the sediment load was widely reported. There- estimated the lag times between atmospheric deposition fore, understanding the impact of the TGD on downstream and stream chemistry by cross-correlating monthly data areas is necessary. In this paper, we use wavelet analysis to from four pairs of stream and deposition monitoring sites, assess the periodic variation of hydrological features and the noting that understanding lag times between changes in dependencies of discharge and sediment regimes among chemical inputs and watershed responses is critical to eval- stations. We improve the time delay estimation algorithm to uating and refining pollutant control strategies. Xia et al. quantify the lag time and attenuation of discharge and (2016) proposed a high-resolution time delay estimation sediment between upstream and downstream sites. scheme to obtain the complete time sequence structure of b. Data source the geometric scattering of an underwater target. TDE is also used in optics, seismology, fault location, and bio- Data were sourced from the hydrological yearbook of medical engineering to evaluate lag time and attenuation China, consisting of discharges and sediments at the (Liao et al. 2013; Ling et al. 2015; Yuan et al. 2014; Zhou Yichang, Hankou, and Datong gauging stations in the et al. 2014). However, TDE is seldom applied in hydraulic main stream of the Yangtze River. These stations rep- engineering analysis at present. resent the hydrological regime of the upper, middle, and

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC MARCH 2018 L I E T A L . 627

FIG. 1. Locations of the selected hydrological stations in the Yangtze River basin. lower Yangtze reaches, respectively. The Yichang station, using horizontally oriented water bottles. Suspended sedi- constructed in 1877, is located just downstream of the TGD ment concentrations were obtained after samples dry (at with a distance of 46 km and can record direct responses to 1058C) and reach constant weights. Water discharge was such a large project. The Hankou station, constructed in determined as the product of cross-section area and mean 1865, is a site in the middle reaches of the river, representing velocity. Before these data were released to the public, they the mainstream hydrological regime change after the inflow had gone through rigorous verification (e.g., eliminating of the . The Datong station, constructed in 1922, is outliers and correcting random errors) and uncertainty approximately 640 km away from the estuary and is the analysis following government protocols to ensure that the upstream limit of tidal influence. The test section for each systemwide confidence level is above 95% (Dai and Liu station has a straight river channel. The dataset of daily mean 2013). Data published by CWRC have been widely used in 2 discharges (m3 s 1) for the three stations ranges from 1951 to geomorphology and hydrology studies at home and abroad 2009, although the dataset of daily mean sediment concen- (Gao et al. 2014; Xu and Milliman 2009). The quality control 2 trations (kg m 3) for these three stations ranges from 1980 to imposed by the surveying agencies ensured the reliability of 2009. Based on the raw data, the daily mean sediment flux of the data, and the published work that used these data all each station is estimated, as follows: attest to the trustworthiness of the data (Dai and Liu 2013). Because of the matching of original datasets and the 5 3 accuracy requirements of the wavelet analysis method, the Qs Cs Q, (1) daily mean discharges, daily mean sediment concentra- 21 where Qs is the daily mean sediment flux (kg s ), Cs is tions, and sediment flux used in the wavelet analysis were 2 the daily mean sediment concentration (kg m 3), and Q selected from 1990 to 2009. In the time delay estimation 2 is the daily mean discharge (m3 s 1). and transmission factor estimation, the daily mean dis- Measurements and quality control were carried out by charges for the period 1951–2009 and the daily mean the Changjiang Water Resources Commission (CWRC; sediment flux for the period 1980–2009 were used. available at www.cjh.com.cn) of the Ministry of Water Re- Three approaches based on the wavelet transform sources. The details about measurement can be found in method were applied in our study to assess the periodic Yang et al. (2014) and Zhou et al. (2016), arraying 10–30 variation of hydrological features and the dependencies vertical profiles for a specific gauging station (cross section), of discharge and sediment regimes between stations. and for each profile, depth and flow velocity were measured Meanwhile, the time delay estimation algorithm was using a velocimeter at the surface and at 0.2, 0.4, 0.6, 0.8, and improved to quantify the lag time and attenuation of 1 water column depth. At the same depth, where the flow discharge and sediment between upstream and down- velocity was determined, water samples were collected stream sites. The study procedure is shown in Fig. 2.

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC 628 JOURNAL OF HYDROMETEOROLOGY VOLUME 19

FIG. 2. The study procedure. c. Description of wavelet analysis different scales of observation. The Morlet wavelet, which we used as the wavelet function in this study, is Wavelet analysis has been extensively applied in the defined as area of time series analysis and prediction due to its multiresolution and localization capabilities both in the M 1 iv t 2t2/2 c (t) 5 e 0 e , (3) time and frequency domains (Maheswaran and Khosa p1/4 2012). Put simply, wavelet transforms provide useful v mathematic decompositions of original time series data where 0 is the central frequency of the wavelet at various resolutions by controlling the scaling and (Torrence and Compo 1998; Yang et al. 2016) shifting factors of a single wavelet: the mother wavelet 2) CROSS-WAVELET TRANSFORM (Nalley et al. 2012). Torrence and Compo (1998) ex- plained the wavelet transform method in detail. The cross-wavelet transform (XWT), as introduced by Three wavelet transform approaches were applied in our Hudgins et al. (1993), was developed to identify the study. We employed the continuous wavelet transform to cross-wavelet spectrum of two time series and un- assess the variations in the daily mean discharges and sed- derstand how the phase angle represents mechanisms in iment concentrations at three stations along the Yangtze the causal and physical relationships between the time River. In addition, the cross wavelet transform and wavelet series (Grinsted et al. 2004; Yu and Lin 2015). The XWT coherence analysis were used to assess the dependencies of defined for two time series X and Y, with wavelet X Y the daily mean discharges and sediment flux among up- transforms W and W , respectively, is stream and downstream stations along the river. XY 5 X Y* Wn (s) Wn (s)Wn (s), (4) 1) CONTINUOUS WAVELET TRANSFORM The continuous wavelet transform of a time-dependent where * denotes a complex conjugate. Torrence and variable y(t) for a specific location is defined as (Fang et al. Webster (1999) defined the cross-wavelet energy as jWXY s j jWXY s j 2015) n ( ) . The larger n ( ) is, the more closely re- lated the two time series are, and vice versa. The sta- ð   ‘ 1 t 2 t tistical significance of the cross wavelet is estimated W(a, t) 5 y(t) pffiffiffic* dt, (2) 2‘ a a using Monte Carlo methods with red noise to determine the 5% significance level (Torrence and Webster 1999). where c* is the complex conjugate of the mother 3) WAVELET COHERENCE ANALYSIS wavelet c, which can be selected from a variety of functions, t is the time, t is the translation factor (time As in the study conducted by Torrence and Webster shift), and a (.0) is the scale factor corresponding to (1999), the wavelet coherence (WTC) can be estimated

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC MARCH 2018 L I E T A L . 629

using the squared absolute value of the smoothed cross- where Rw1,w2(k) is the correlation function of noise, wavelet power spectrum of each selected time series. Rss(k) is the autocorrelation function of the source sig- Hence, the squared wavelet coefficient is defined as nal, and A is the transmission factor, which is defined as

jS[s21WXY (s)]j2 R (D) R2(s) 5 n , (5) A 5 xy . (8) n 21j X j2 21j Y j2 S[s Wn (s) ]S[s Wn (s) ] Rxy(0) where S is a smoothing operator. In Eq. (5), the value If noise is an independent random process, Rx,y(k)will of squared wavelet coherence ranges between 0 and 1, reach a peak value when k is equal to D and the estimated indicating no dependencies and close dependencies, value of D is obtained. However, the peak value is often respectively. In general, XWT can better reveal the not clear in reality for a variety of reasons. Thus, pre- time–frequency areas with common high power, and processing the raw data is needed, and the most common WTC is appropriate for identifying the time–frequency method is to multiply Rss(k) by a window function. Se- correlations at the time–frequency areas with the low lecting various window functions has a great influence on powers or in the nonprimary period (Maraun and the results of TDE. The maximum likelihood window Kurths 2004; Yang et al. 2016). function is the most common window function Cxy( f ): d. Analysis methods for time delay and transmission js ( f )j2 factor estimation 5 xy Cxy( f ) , (9) sxx( f )syy( f ) Time delay estimation between signals received by dif- ferent sensors has been widely utilized in different areas, where sxy is the cross spectrum of two time series, and including radar, optics, seismology, fault location, and bio- sxx( f ) and syy( f ) are autocorrelation spectrums of these medical engineering (Liao et al. 2013; Yu and Lin 2015; two time series. Yuan et al. 2014; X. Zhang et al. 2013; Zhou et al. 2014). In To eliminate high-frequency noise and enhance the this paper, we introduce TDE to obtain high precision and stability of the algorithm, we chose the Hamming win- accuracy in the analysis of hydrological time series. The dow to process the intercept segments of time series hydrological time series from the upper-stream stations are after analyzing different types of filtering window used as the source signals, those from the downstream sta- functions. The Hamming window is essentially a low- tions are used as the received signals, and variations be- pass filter and is defined as (Kumar et al. 2011) tween upstream and downstream sites (such as the inflow of   branches, exchanges of water and material between lakes 2pn W(n) 5 0:54 2 0:46 cos , (10) and river, etc.) are considered to be noise. We improve the N 2 1 TDE algorithm to estimate the transmission time of dis- charge and sediment between upstream and downstream where N is the total length of all windowed time series. sites. Further, we then use the algorithm to extract com- In this study, the length of the intercept segments of time ponents of the upstream stations from the downstream series is relatively short, and the high-frequency noise stations to calculate the loss along the main stream, which cannot be completely eliminated just by using the Hamming is a process called transmission factor estimation. window. Therefore, we proposed an improved maximum 0:5 Suppose x(k) and y(k) are two independent received likelihood window function: Cw 5 (Cw) , the peak value signals, and they satisfy equations of which is clearly decreased. When the cross spectrum sxy is  multiplied by the improved maximum likelihood window x(k) 5 s(k) 1 w (k) 1 function, certain frequencies of sxy will not be strengthened 5 2 1 , (6) y(k) as(k D) w2(k) excessively. The comparison of cross-spectrum power is shown in 5 ... 2 where k 0, 1, , N 1 and where s(k) is the original Fig. 3. The cross spectrum without the window function , , source signal. The transmission factor is a (0 a 1). cannot display the time delay correctly, while time Both w1(k) and w2(k) are unknown additive background delays are just peak values in these with window noise. The task is to find D using the N samples of x(k) functions. However, there is interference from n 5 20 and y(k). In this paper, the cross-correlation method to n 5 70 in Fig. 3b, which is mainly because the mag- based on a maximum likelihood window is proposed. nitudes of the unimproved maximum likelihood win- The cross function is defined as dow function have a sudden increase in this range while 5 2 1 the extreme values of the improved function differ by Rxy(k) ARss(k D) Rw w (k), (7) 1, 2 an order of magnitude. Thus, interference suppression

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC 630 JOURNAL OF HYDROMETEOROLOGY VOLUME 19

FIG. 3. The comparison of cross-spectrum power: (a) without the window function, (b) with the maximum likelihood window function, and (c) with the improved maximum likelihood window function. is achieved effectively through the improved maximum indicating that the daily mean discharges have a prom- likelihood window function without affecting the main inent annual cycle. There is no obvious change in power peak value. spectra at all three stations around the year 2003, im- To verify the improved algorithm, a random time series plying that the TGD has not caused considerable vari- is generated, from which two new time series are inter- ation in the period of daily mean discharge in the middle cepted at [1, n]and[11 D, n 1 D](n 5 1024, D 5 16), and lower reaches of the river. According to water-level respectively. There is no attenuation in these two new time scheduling, water is stored in the TGD in winter and series; that is to say, the transmission factor equals 1. released from the TGD in summer. Yang et al. (2015) The improved algorithm is used to estimate the time argued that the seasonal effects of impoundment and delay and attenuation of two time series. The results are release affect short-term discharge but not overall an- shown in Fig. 4. nual flow. It is therefore possible that there are changes As seen in Fig. 4a, the result of the TDE is 16.00 6 in monthly or seasonal flows that are not detectable at 0.00, which is very precise and not very affected by high- the annual scale. Streamflow changes at the three hy- frequency noise. The result of the TFE is 89.15% 6 drological stations might be attributed to climate 0.63%, as shown in Fig. 4b, which is clearly lower than change, such as precipitation, but not to anthropo- the expected value (100%). genic factors, such as the construction of water reservoirs (Q. Zhang et al. 2013). In addition, Fig. 5 shows relatively 3. Results and discussion high frequencies at a half-year scale (6 months) around 1998, which might be related to the large flood in the a. Hydrologic features of individual stations Yangtze River during that year. It is inferred that the intense flood pulse impacted the original hydrological 1) CHARACTERISTICS OF DAILY MEAN pulse and resulted in an abnormal half-year scale. Ad- DISCHARGE AT INDIVIDUAL STATIONS ditionally, half-year scales are shown in other years to Figure 5 illustrates the raw data variations in the daily some degree, but were neither clear nor continuous. mean discharges at the three stations based on the CWT. Further, regarding the three-wavelet spectrum around The red area at the bottom (top) of the continuous 1998, it can be seen that the power of the half-year scale power spectra represents strong variation at low (high) at the Yichang, Hankou, and Datong stations decreased frequencies, while the red area on the left-hand (right stepwise, which means that the impacts of extremely hand) side indicates significant variation at the begin- large flows decreased successively between the three ning (end) of the study period. According to Fig. 5, the stations rather than increasing in sequence with the in- daily mean discharges at all three stations show signifi- fluence of inflows of tributaries and lakes. This indicates cant high variations at an annual scale (12 months), the important role of flood control and the distributary

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC MARCH 2018 L I E T A L . 631

FIG. 4. The result of the (left) TDE and (right) TFE of two time series. system in the middle and lower Yangtze River, weak- climate change (decreasing precipitation). In compari- ening the effect of flooding on the hydrological period. son, there was an abrupt reduction in sediment flux at the three stations after the construction of the Three 2) CHARACTERISTICS OF DAILY MEAN SEDIMENT Gorges Dam, with drastic reduction rates of approxi- CONCENTRATIONS AT INDIVIDUAL STATIONS mately 88%, 70%, and 65% from pre-TGD to post- Figure 6 illustrates the raw data variations of the daily TGD at the Yichang, Hankou, and Datong stations, mean sediment concentrations at the three stations respectively (Zhao et al. 2012). It is possible that the based on the CWT. sharp and massive reductions of sediment break the Before 2003, similar to the daily mean discharges dynamic balance of sediment and hugely diminish (Fig. 5), sediment concentrations (Fig. 6) also had a the time series signal, resulting in abrupt changes in prominent annual cycle (12 months) at the three stations. the annual variation laws. Changes in the variation laws However, since the TGD began to impound water, the are gradually postponed from upstream to downstream daily mean sediment concentrations show abrupt stations, indicating a time delay in the transmission of changes at all three stations. Specifically, the annual the impacts of the TGD along the main stream, mainly cycle of sediment concentration at the Yichang station due to the dredging process in the Yichang–Hankou and was nearly nonexistent in 2003, while this also occurred Hankou–Datong river sections. The decreased sediment around the years 2003–04 at the Hankou and Datong load from the upper Yangtze River might cause the stations. The difference between these stations might be erosion of the riverbed (Gao et al. 2014; Q. Zhang et al. attributed to the increasing of response time with dis- 2013). From 2003 to 2006, channel erosion occurred, and 2 tance from the TGD. The decrease in the average an- an average of 70 Mt yr 1 of sediment was eroded along nual streamflow was less than 10% at the Yichang, the river section between Yichang and Datong (Xu and Hankou, and Datong stations, mostly resulting from Milliman 2009). Because channel-eroded sediment was

FIG. 5. CWTs of the daily mean discharges. Shown are the continuous power spectra of the daily mean discharges from 1 Jan 1990 to 31 Dec 2009 at the (a) Yichang, (b) Hankou, and (c) Datong stations.

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC 632 JOURNAL OF HYDROMETEOROLOGY VOLUME 19

FIG.6.AsinFig. 5, but for the daily mean sediment concentrations. generally coarser than suspended sediment, the median be noted that sediment deposition was generally most grain size increased from 5 mm at Yichang to 11 mm serious in the initial years of the reservoir operation, downstream at Hankou in 2005 (Bulletin of Yangtze and a gradually decreasing sediment deposition rate can Sediment 2005). Because of the active channel erosion be anticipated as time goes (Li et al. 2011). The influence downstream of the Three Gorges Reservoir (TGR) after of water reservoirs on the hydrological regime, partic- June 2003, the bed level on the mainstem channel was ularly the sediment load and subsequent effects, can be lowered, and the process of sediment escaping from the far reaching (Q. Zhang et al. 2013). The TGD has been main stem to Dongting Lake essentially ceased in 2004 in operation for less than 20 years, while the dynamic (0 Mt; Xu and Milliman 2009). The erosion of the river balance of erosion and deposition between the reservoir channel, with considerable amounts of sediment re- and catchment and the interactions between the river ceived from tributaries merging at Poyang Lake and the and its tributaries and lakes needs considerably more Han River, supplements the sediment concentrations at time. Therefore, the balance of sediment is not achieved the Hankou and Datong stations to some degree and at this time. contributes to a moderate decrease. This erosion, how- b. Relationship of discharges between stations ever, does not offset the loss of sediment in the TGR (Yang et al. 2007). Consequently, the annual variation To investigate the relationship of the daily mean dis- laws of the daily mean sediment concentrations at all charges among upstream and downstream stations along stations along the main stream change consequently, the Yangtze River, the XWTs for the pairs are sum- except for the differences in time. The spatial variations marized on the left-hand side of Fig. 7, and the WTCs in the effects of the TGD may have complicated the for the pairs are summarized on the right-hand side. sediment transport processes along the river (Guo et al. According to Fig. 7a, the common period of the daily 2012). Studies that provide insight into the sediment mean discharges at the Yichang and Hankou stations is change patterns may not be sufficient until a balance in one year (except 1998), as both stations have a prom- this dynamic is achieved in the future. Thus, the question inent annual cycle. There is an apparent high-frequency of how the TGD could have affected the stability of the area around 1998; however, the interference in the XWT river morphology deserves further investigation in for the Yichang–Hankou station (Fig. 7a) is slightly the future. larger than that in the CWT for the Hankou station Along the middle and lower reaches, the Yangtze (Fig. 5b). This is because the influence of flooding was River exchanges water and sediment with numerous smaller at the Hankou station than at the Yichang sta- tributaries and lakes (e.g., Poyang Lake and the Han tion. The interpretation of Fig. 7c is similar to Fig. 7a. River), which are rather complicated interactions. The Dependencies of the daily mean discharges at the river has lost some sediment (due to siltation and rec- Hankou and Datong stations are evidently similar to lamation) through several passages but has received those at the Yichang and Hankou stations in terms of considerable sediment from tributaries. After the im- hydrological cycles. Specifically, the high-frequency poundment of the TGD, huge amounts of sediment built interference around 1998 in the XWT for the Hankou– up behind dams, which led to sediment starvation in the Datong stations is weaker than that at the Yichang– mainstem, significantly changed sediment dynamics, and Hankou station because the impacts of extremely large may have even caused erosion of the riverbed. It should flows decrease stepwise from upstream to downstream

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC MARCH 2018 L I E T A L . 633

FIG. 7. (a),(c) XWTs and (b),(d) WTCs of the daily mean discharge from 1 Jan 1990 to 31 Dec 2009. stations. More tributaries and lakes inflow into the massive streamflow inputs are present in the lower main stream, and the influence of upstream input is Yangtze River basin (Q. Zhang et al. 2013). There is no increasingly moderate; thus, the change in the hy- obvious interference in all power spectra around the drological cycle is not so extreme at downstream sta- year 2003 in Fig. 7, implying that the TGD did not cause tions. The WTC plot shown in Fig. 7b reveals the clear significant variation in the period of daily mean dis- associations at a period of 12 months for the daily mean charge in the middle and lower Yangtze River. discharges at the Yichang and Hankou stations, with In this study, the time series of daily mean discharges the R2 being mostly larger than 0.9 throughout all the at the Yichang, Hankou, and Datong stations between study periods. However, some breakdowns in co- 1951 and 2009 are used to estimate the time delay and herence can be observed for time scales more than attenuation among these three stations. 12 months, indicating a great difference between these According to Fig. 8a, it generally takes approximately two stations. It is inferred that the result is caused by 3 days for discharge to flow from the Yichang station to streamflow from the Hanjiang River and Dongting the Hankou station, except for the abnormally long time Lake and the flood diversion area of Jingjiang be- (up to 6 days) in 1956 and 2002. The abnormal values tween the Yichang and Hankou stations, not by the might be caused by the instability of the algorithm. construction of the TGD. Comparatively, Fig. 7d Figure 8b illustrates the results of TFE for discharge demonstrates a good agreement between the Hankou flows from the Yichang station to the Hankou station and Datong stations, indicating a similarity of daily and demonstrates that the attenuation of discharge mean discharges at the two stations, which might be is approximately 15% from the Yichang station to attributed to the fact that no large tributaries and no the Hankou station. Specifically, before 1963, the

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC 634 JOURNAL OF HYDROMETEOROLOGY VOLUME 19

FIG. 8. (a),(c) TDEs and (b),(d) TFEs of the daily mean discharge from 1951 to 2009. attenuation was more than 20% then rapidly dropped to was acquired by multiplying the attenuation of both approximately 15%. Decreasing attenuation can be river sections together. observed from 1980 to 1985. The attenuation was char- c. Relationship of the sediments between stations acterized as increasing from 2000 to 2005, up to 20%, while a decreasing trend was found after 2005. Figure 8c Figure 9a demonstrates that the common primary displays the result of the TDE for the discharge flow period of the daily mean sediment flux at the Yichang from the Hankou station to the Datong station. We can and Hankou stations is 1 year (12 months) before 2003 see that the transmission time of flow for this segment is (except the year 1998) because both stations show an- approximately 3 days, except in certain years, including nual cycles. The primary-period energy attenuates 1993 and 1998, in which the time delay decreased to sharply after 2003, clearly showing the effects of the 1.5 days. The outliers appearing in 1953 indicate that the Three Gorges Dam on sediment load. Moreover, the algorithm used in this study still needs further en- apparent high-frequency area around 1998 shows an hancement. The result of the TFE for the discharge expanding phenomenon, which is clearly the coupling of flows from the Hankou station to the Datong station is the daily mean sediment flux in Fig. 9a from 1997 to presented in Fig. 8d. It can be seen from Fig. 8d that the 2001. In addition to the rainy reason (from June to attenuation first increased and then decreased before September), precipitation in the Yangtze River basin 1963, reaching a stable level of approximately 8% during has a second peak in spring (from March to May). the late 1960s and 1970s. After 1980, there was large Therefore, the discharges of the tributaries and main fluctuation in individual years (such as in 1990 and 1996). stream increase considerably and even exhibit bi- The total discharge transmission time in the Yichang– modality in the XWT for sediment flux, which reflects Datong river section is approximately 6 days, which is the semiannual cycle. This phenomenon is especially calculated by adding the transmission times of both river noticeable in dry years when river flow is relatively low sections together. This result basically agrees with the in the rainy reason. Although the semiannual energy is findings of Wang et al. (2013), who showed a time lag of not typically statistically significant at the 0.05 level, the 7–8 days from the TGD to the Datong station by cal- coupling of this cycle with the high-frequency in- culating the cumulative flow-path travel time. In addi- terference in 1998 strengthened the energy in this spe- tion, our results indicated that the attenuation of cific period, leading to a clear half-year period in Fig. 9a. discharge from Yichang to Datong is 20%–30%, which The interpretation of Fig. 9c is similar to that of Fig. 9a.

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC MARCH 2018 L I E T A L . 635

FIG.9.AsinFig. 7, but for the daily mean sediment flux.

With regard to the common primary period, the Hankou discharges. Although sediment flux at the Yichang and and Datong stations also have an annual period, similar Hankou stations decreased sharply after the operation to the Yichang and Hankou stations. However, abrupt of the TGD, there is still a strong correlation between energy attenuation at the Hankou and Datong stations the two stations. Figure 9b demonstrates the clear as- occurred around 2005, which is two years later than that sociations with a period of 12 months for the daily mean at the Yichang and Hankou stations and one year later sediment flux at these stations throughout all study pe- than that of their respective changes in period. This re- riods. However, as to longer time scales of more than sult indicates that, although sediment concentrations 12 months, there is a great difference between these two and sediment flux at the Hankou and Datong stations stations, which is independent of the construction of decreased quickly after the TGD commenced opera- the TGD. Compared to Fig. 9b, Fig. 9d shows a good tions (2003), the transitive relation of these two stations agreement between the Hankou and Datong stations, was not broken until 2005. There is an evident time lag indicating a similarity in the daily mean sediment flux at between upstream and downstream stations. The re- the two stations on both annual and longer scales. sponse time increased with the distance between the In this study, the time series of daily mean sediment TGD and the study site because the sediment regime of flux at the Yichang, Hankou, and Datong stations be- the station at a distance from the reservoir was not only tween 1980 and 2009 are used to estimate the time delay affected by sediment from upstream but also by the and attenuation among these three stations. sediment from the local catchment (Li et al. 2011). According to Fig. 10a, the time delay of the suspended Channel erosion downstream of the TGD plays a key sediment concentration from the Yichang station to the role in sediment transfer in the middle and lower rea- Hankou station is approximately 3.5 days, longer than ches (Xu and Milliman 2009). The WTC for the daily the transmission time of discharge. Moreover, the fluc- mean sediment flux is similar to that for the daily mean tuation of suspended sediment is larger than that of the

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC 636 JOURNAL OF HYDROMETEOROLOGY VOLUME 19

FIG. 10. As in Fig. 8, but for the daily mean sediment flux from 1980 to 2009. discharge from the same period. The maximum TDE for Hankou station to the Datong station. Overall, the time suspended sediment appeared in 1994. It is possible that delay gradually decreased from 4 to 3.5 days. An abrupt the start of construction on the TGD in 1994 disturbed drop appeared in 1989, but the cause of this is not readily the relatively stable transmission of sediment. The clear. An abrupt drop also appeared in 1998, when an minimum value was observed between 2003 and 2004, extreme flood occurred in the Yangtze River, and a third which may have been because the TGD was completed abrupt drop occurred in approximately 2006 after the and began storing water in 2003. After a period of time, second-phase water storage began in the Three Gorges the regular transmission of sediment began to recover. Reservoir. The instability of the algorithm still caused The outliers occurring in 2006 may have been caused by outliers in 1984. The TFE plot shown in Fig. 10d dem- the instability of the algorithm. Figure 10b shows the onstrates the obvious change in the attenuation of sus- result of the TFE for suspended sediment from the pended sediment from the Hankou station to the Yichang station to the Hankou station. The year 2006 Datong station over time. Specifically, from 1981 to was identified as the significant changepoint for this 1986, the attenuation decreased from 20% to 5%. Later, value. Before 2006, attenuation was steady at 15%–20%, it increased to approximately 10%. A decreasing trend but the attenuation has increased sharply to 30% since can be observed again from 1995 to 2001, while after this changepoint, which might be the result of the mas- 2001, attenuation increased from 5% to 15%. The sive sediment impoundment effect of the TGD and be- transmission of suspended sediment is slower than cause the dynamic balance of sediment downstream of the discharge, taking approximately 7–7.5 days from the the dam is disturbed. After 2006, though the trans- Yichang to Datong stations. The attenuation of sus- mission factor increases a little, the attenuation of sus- pended sediment is approximately 30% in the Yichang– pended sediment still exceeds 25%. Figure 10c Datong river section, which is more severe than the illustrates the TDE for suspended sediment from the discharge.

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC MARCH 2018 L I E T A L . 637

4. Conclusions Natural Science Foundation of China (51421065), the National Basic Research Program of China (Grant In this paper, we analyzed the daily river discharge 2010CB429003), the National Natural Science Founda- and sediment load data, and major findings are as tion of China (Grants 51409004 and 51409003), and follows: the Interdiscipline Research Funds of Beijing Normal 1) The annual cycle of daily mean sediment concentra- University. tions was nearly nonexistent after the impoundment of the TGD. Both daily mean discharges and sedi- REFERENCES ment fluxes in the Yichang–Hankou river section exhibit larger changes than in the Hankou–Datong Alrajoula, M. T., I. S. Al Zayed, N. A. Elagib, and M. R. Hamdi, river section, a finding that is not the result of the 2016: Hydrological, socio-economic and reservoir alterations of Er Roseires Dam in Sudan. Sci. Total Environ., 566–567, impacts of the TGD. 938–948, https://doi.org/10.1016/j.scitotenv.2016.05.029. 2) The transmission time of discharge in both river Assani, A., 2016: The usefulness of the Lombard method for ana- sections is approximately 3 days in preimpoundment lyzing the hydrological impacts of dams: The case of the periods and postimpoundment periods. Throughout Manouane River diversion dam, Quebec, Canada. Water, 8, the Yichang–Datong river section, the attenuation of 410, https://doi.org/10.3390/w8090410. Bulletin of Yangtze Sediment, 2005. Press of Ministry of Water discharge is 20%–30%. In addition, the transmission Resources of the People’s Republic of China, http://www.cjh. time of suspended sediment is slower than that of com.cn/. discharge. It takes approximately 3.5 days for sedi- Chen,J.,B.L.Finlayson,T.Wei,Q.Sun,M.Webber,M.Li,andZ.Chen, ment to flow from Yichang to Hankou station, and the 2016: Changes in monthly flows in the Yangtze River, China—With fluctuation of sediment is larger than the fluctuation of special reference to the Three Gorges Dam. J. Hydrol., 536, discharge during the same period. From Hankou to 293–301, https://doi.org/10.1016/j.jhydrol.2016.03.008. Dai, Z., and J. T. Liu, 2013: Impacts of large dams on downstream Datong stations, the time delay exhibits a decreasing fluvial sedimentation: An example of the Three Gorges Dam trend overall, decreasing from 4 to 3.5 days. (TGD) on the Changjiang (Yangtze River). J. Hydrol., 480, 10–18, https://doi.org/10.1016/j.jhydrol.2012.12.003. The attenuation of suspended sediment is approxi- ——, J. Du, J. Li, W. Li, and J. Chen, 2008: Runoff characteristics of mately 30% in the Yichang–Datong river section and the Changjiang River during 2006: Effect of extreme drought increased significantly after 2006 (up to 40%), a greater and the impounding of the Three Gorges Dam. Geophys. Res. increase than discharge. The inflow of branches and Lett., 35, L07406, https://doi.org/10.1029/2008GL033456. exchanges of water and material between lakes and river DeWalle, D. R., E. W. Boyer, and A. R. Buda, 2016: Exploring lag times between monthly atmospheric deposition and stream are not included in the estimation of time delay and chemistry in Appalachian forests using cross-correlation. transmission factor. Atmos. Environ., 146, 206–214, https://doi.org/10.1016/ This study ascertains the river discharge and sediment j.atmosenv.2016.09.015. relationships between upstream and downstream sta- Fang, Z., H. Bogena, S. Kollet, J. Koch, and H. Vereecken, 2015: tions, especially the lag time and attenuation thereof Spatio-temporal validation of long-term 3D hydrological between stations, which provides an important contri- simulations of a forested catchment using empirical orthogo- nal functions and wavelet coherence analysis. J. Hydrol., 529, bution to the river management and restoration and im- 1754–1767, https://doi.org/10.1016/j.jhydrol.2015.08.011. provement of pollutant control strategies in the middle Gao, B., D. Yang, and H. Yang, 2013: Impact of the Three Gorges and the lower Yangtze River basin. However, any as- Dam on flow regime in the middle and lower Yangtze River. sessment of the full impacts (including short- and long- Quat. Int., 304, 43–50, https://doi.org/10.1016/j.quaint.2012.11.023. term influence) of the dam should be based on large Gao, J. H., and Coauthors, 2014: Changes in water and sediment ex- amounts of long-term data (Zhang et al. 2012), while the change between the Changjiang River and Poyang Lake under natural and anthropogenic conditions, China. Sci. Total Environ., Three Gorges Dam has only been in operation for a short 481, 542–553, https://doi.org/10.1016/j.scitotenv.2014.02.087. time. Because of the scarcity of historical data, there is Grinsted, A., J. C. Moore, and S. Jevrejeva, 2004: Application of some inevitable uncertainty in the present study. In ad- the cross wavelet transform and wavelet coherence to dition, the dynamic balance of erosion and deposition in geophysical time series. Nonlinear Processes Geophys., 11, the river channel is the result of a complex superposition 561–566, https://doi.org/10.5194/npg-11-561-2004. Guo, H., Q. Hu, Q. Zhang, and S. Feng, 2012: Effects of the Three of many factors. Further analyses based on more exten- Gorges Dam on Yangtze River flow and river interaction with sive postdam data are needed to investigate the dam’s Poyang Lake, China: 2003–2008. J. Hydrol., 416–417, 19–27, impact on the middle and lower Yangtze River. https://doi.org/10.1016/j.jhydrol.2011.11.027. Hocking, M., and B. F. J. Kelly, 2016: Groundwater recharge and time lag measurement through Vertosols using impulse re- Acknowledgments. This research was funded by the sponse functions. J. Hydrol., 535, 22–35, https://doi.org/ Fund for Innovative Research Group of the National 10.1016/j.jhydrol.2016.01.042.

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC 638 JOURNAL OF HYDROMETEOROLOGY VOLUME 19

Hudgins, L., C. A. Friehe, and M. E. Mayer, 1993: Wavelet Xia, Z., X. Li, and X. Meng, 2016: High resolution time-delay esti- transforms and atmospheric turbulence. Phys. Rev. Lett., 71, mation of underwater target geometric scattering. Appl. Acoust., 3279–3282, https://doi.org/10.1103/PhysRevLett.71.3279. 114, 111–117, https://doi.org/10.1016/j.apacoust.2016.07.016. Kumar, S., K. Singh, and R. Saxena, 2011: Analysis of Dirichlet and Xu, K., and J. D. Milliman, 2009: Seasonal variations of sediment Generalized ‘‘Hamming’’ window functions in the fractional discharge from the Yangtze River before and after im- Fourier transform domains. Signal Process., 91, 600–606, poundment of the Three Gorges Dam. Geomorphology, 104, https://doi.org/10.1016/j.sigpro.2010.04.011. 276–283, https://doi.org/10.1016/j.geomorph.2008.09.004. Li, Q., M. Yu, G. Lu, T. Cai, X. Bai, and Z. Xia, 2011: Impacts of the Yang, L., X. J. Cai, H. Zhang, and S. Hamori, 2016: In- Gezhouba and Three Gorges reservoirs on the sediment re- terdependence of foreign exchange markets: A wavelet co- gime in the Yangtze River, China. J. Hydrol., 403, 224–233, herence analysis. Econ. Modell., 55, 6–14, https://doi.org/ https://doi.org/10.1016/j.jhydrol.2011.03.043. 10.1016/j.econmod.2016.01.022. Liao, P., M. Cai, Y. Shi, and Z. Fan, 2013: Compressed air leak Yang, S. L., J. Zhang, and X. J. Xu, 2007: Influence of the Three detection based on time delay estimation using a portable Gorges Dam on downstream delivery of sediment and its multi-sensor ultrasonic detector. Meas. Sci. Technol., 24, environmental implications, Yangtze River. Geophys. Res. 055102, https://doi.org/10.1088/0957-0233/24/5/055102. Lett., 34, L10401, https://doi.org/10.1029/2007GL029472. Ling, Y., Q. Tian-Shuang, and L. Shengyang, 2015: Fractional time ——, J. D. Milliman, K. H. Xu, B. Deng, X. Y. Zhang, and X. X. delay estimation algorithm based on the maximum corren- Luo, 2014: Downstream sedimentary and geomorphic impacts tropy criterion and the Lagrange FDF. Signal Process., 111, of the Three Gorges Dam on the Yangtze River. Earth Sci. Rev., 222–229, https://doi.org/10.1016/j.sigpro.2014.12.018. 138, 469–486, https://doi.org/10.1016/j.earscirev.2014.07.006. Luan, Z., and Q. Jin, 2016: Research on water-sediment numerical ——, K. H. Xu, J. D. Milliman, H. F. Yang, and C. S. Wu, 2015: simulation of middle and lower reaches of the Yangtze River Decline of Yangtze River water and sediment discharge: Im- and estuary. Procedia Eng., 154, 582–588, https://doi.org/ pact from natural and anthropogenic changes. Sci. Rep., 5, 10.1016/j.proeng.2016.07.555. 12581, https://doi.org/10.1038/srep12581. Maheswaran, R., and R. Khosa, 2012: Wavelet–Volterra coupled Yu, H.-L., and Y.-C. Lin, 2015: Analysis of space–time non- model for monthly stream flow forecasting. J. Hydrol., 450– stationary patterns of rainfall–groundwater interactions by 451, 320–335, https://doi.org/10.1016/j.jhydrol.2012.04.017. integrating empirical orthogonal function and cross wavelet Maraun, D., and J. Kurths, 2004: Cross wavelet analysis: Signifi- transform methods. J. Hydrol., 525, 585–597, https://doi.org/ cance testing and pitfalls. Nonlinear Processes Geophys., 11, 10.1016/j.jhydrol.2015.03.057. 505–514, https://doi.org/10.5194/npg-11-505-2004. Yuan, W., B. Pang, J. Bo, and X. Qian, 2014: Fiber optic line-based Mei, X., Z. Dai, P. H. A. J. M. van Gelder, and J. Gao, 2015: sensor employing time delay estimation for disturbance de- Linking Three Gorges Dam and downstream hydrological tection and location. J. Lightwave Technol., 32, 1032–1037, regimes along the Yangtze River, China. Earth Space Sci., 2, https://doi.org/10.1109/JLT.2013.2296617. 94–106, https://doi.org/10.1002/2014EA000052. Zhang, Q., V. P. Singh, and X. Chen, 2012: Influence of Three Nalley, D., J. Adamowski, and B. Khalil, 2012: Using discrete Gorges Dam on streamflow and sediment load of the middle wavelet transforms to analyze trends in streamflow and pre- Yangtze River, China. Stochastic Environ. Res. Risk Assess., cipitation in Quebec and Ontario (1954–2008). J. Hydrol., 475, 26, 569–579, https://doi.org/10.1007/s00477-011-0466-8. 204–228, https://doi.org/10.1016/j.jhydrol.2012.09.049. ——, ——, C.-Y. Xu, and X. Chen, 2013: Abrupt behaviours of Stone, R., 2008: Three Gorges Dam: Into the unknown. Science, streamflow and sediment load variations of the Yangtze River 321, 628–632, https://doi.org/10.1126/science.321.5889.628. basin, China. Hydrol. Processes, 27, 444–452, https://doi.org/ Syvitski, J. P., C. J. Vorosmarty, A. J. Kettner, and P. Green, 2005: 10.1002/hyp.9278. Impact of humans on the flux of terrestrial sediment to the Zhang, Q., X.-C. Ye, A. D. Werner, Y.-L. Li, J. Yao, X.-H. Li, and global coastal ocean. Science, 308, 376–380, https://doi.org/ C.-Y. Xu, 2014: An investigation of enhanced recessions in 10.1126/science.1109454. Poyang Lake: Comparison of Yangtze River and local catch- Torrence, C., and G. P. Compo, 1998: A practical guide to wavelet ment impacts. J. Hydrol., 517, 425–434, https://doi.org/10.1016/ analysis. Bull. Amer. Meteor. Soc., 79, 61–78, https://doi.org/ j.jhydrol.2014.05.051. 10.1175/1520-0477(1998)079,0061:APGTWA.2.0.CO;2. Zhang, X., Y. Bai, and X. Chen, 2013: An improved cross- ——, and P. J. Webster, 1999: Interdecadal changes in the ENSO– correlation method based on fractional delay estimation for monsoon system. J. Climate, 12, 2679–2690, https://doi.org/ velocity measurement of high speed targets. Proc. World 10.1175/1520-0442(1999)012,2679:ICITEM.2.0.CO;2. Congress on Engineering and Computer Science, San Wang, J., Y. Sheng, C. J. Gleason, and Y. Wada, 2013: Downstream Francisco, CA IAENG, 651–654, http://www.iaeng.org/ Yangtze River levels impacted by Three Gorges Dam. Envi- publication/WCECS2013/WCECS2013_pp651-654.pdf. ron. Res. Lett., 8, 044012, https://doi.org/10.1088/1748-9326/8/ Zhao, G., X. Mu, B. Su, P. Tian, F. Wang, J. Zhai, and M. Xiong, 4/044012. 2012: Analysis of streamflow and sediment flux changes in the Wang, Y., B. L. Rhoads, and D. Wang, 2016: Assessment of the Yangtze River basin. Water Int., 37, 537–551, https://doi.org/ flow regime alterations in the middle reach of the Yangtze 10.1080/02508060.2012.681442. River associated with dam construction: Potential ecological Zhou, T., H. Li, J. Zhu, and C. Xu, 2014: Subsample time delay implications. Hydrol. Processes, 30, 3949–3966, https://doi.org/ estimation of chirp signals using FrFT. Signal Process., 96, 10.1002/hyp.10921. 110–117, https://doi.org/10.1016/j.sigpro.2013.06.004. White, M. A., J. C. Schmidt, and D. J. Topping, 2005: Application Zhou, Y., E. Jeppesen, J. Li, Y. Zhang, X. Zhang, and X. Li, 2016: of wavelet analysis for monitoring the hydrologic effects of Impacts of Three Gorges Reservoir on the sedimentation re- dam operation: Glen Canyon Dam and the Colorado River gimes in the downstream-linked two largest Chinese fresh- at Lees Ferry, Arizona. River Res. Appl., 21, 551–565, https:// water lakes. Sci. Rep., 6, 35396, https://doi.org/10.1038/ doi.org/10.1002/rra.827. srep35396.

Unauthenticated | Downloaded 10/01/21 07:51 PM UTC