Integrated approach of hydrological and water quality dynamic simulation for anthropogenic disturbance assessment in the Huai River Basin, China

Science of the Total Environment 598 (2017) 749–764

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Integrated approach of hydrological and water quality dynamic simulation for anthropogenic disturbance assessment in the Huai River Basin, China Xiaoyan Zhai a,b, Jun Xia c, Yongyong Zhang c,⁎ a b c

State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, China Research Center on Flood and Drought Disaster Reduction of the Ministry of Water Resources, Beijing 100038, China Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Science and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Integrated hydrological and water quality dynamic were simulated in regulated basin. • Impacts of three anthropogenic activities were assessed for water qualityquantity. • Dam regulation deteriorated downstream water quality with current pollutants.

a r t i c l e

i n f o

Article history: Received 3 December 2016 Received in revised form 4 April 2017 Accepted 12 April 2017 Available online xxxx Editor: Jay Gan Keywords: Rainfall-runoff-nutrient relationship Dynamic processes Water quality and quantity High regulation Huai River Basin

⁎ Corresponding author. E-mail address: [email protected] (Y. Zhang).

http://dx.doi.org/10.1016/j.scitotenv.2017.04.092 0048-9697/© 2017 Elsevier B.V. All rights reserved.

a b s t r a c t Detailed depiction of hydrological process and its associated pollution processes plays a critical role in environment improvement and management at basin scale. It also provides a useful tool to assess impact of potential factors on hydrological and water quality conditions. However, it was still difficult to well capture some typical characteristics of these complicated processes including built-in nonlinearity and time-variation, water infrastructure regulations, particularly for highly regulated basins. In this study, an integrated approach of hydrological and water quality dynamic simulation was proposed to solve these difficulties and assess the impacts of several anthropogenic disturbances. The Huai River Basin which was highly disturbed and seriously polluted, was selected as the study area. The main anthropogenic activities considered were point source pollution emissions, diffuse pollutant losses and dam regulations. Results showed that the integrated simulation could well capture the variations in water level, water discharge, concentrations of permanganate index (CODMn) and ammonia nitrogen (NH4-N) in high (2007), normal (2008) and low (2004) flow years at 15 stations in the upper and middle streams of Huai River Basin. The regulation rules of downstream sluices played negative roles on water quality improvement if keeping current pollution sources, while those of middle stream sluices played positive roles on water quality improvement. However, the water quality deterioration was mainly attributed to emission of point source pollution (12%–43%), followed by diffuse pollutant loss (0–23%) and water quantity-oriented dam regulation (−29%–20%). The study was expected to provide technical supports for the implementation of water pollution control and sustainable water resources management in the Huai River Basin, and give a reference of integrated hydrological and hydrodynamic simulation. © 2017 Elsevier B.V. All rights reserved.

750

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

1. Introduction Pollution has become an increasingly serious water issue in global river basins for decades, which has threatened drinking water safety and restricted global socio-economic development, especially in China. Huai River Basin is the most densely populated, highly polluted and regulated river basin in China. In 2014, 53.8% of river lengths were polluted (Huai River Water Resources Bulletin, 2014). The water shortage issue induced by water pollution has become increasingly serious, and socalled “cancer villages” have emerged along the riparian zone since the 1970s. Currently, most of the rivers are regulated by dams or sluices (Zhang et al., 2010), which has resulted in serious river fragmentation, aquatic ecosystem degradation, lakes and wetlands shrinkage and serious water quality deterioration (Zhang et al., 2012; Xia et al., 2015). Waste water emissions together with intensive dam regulations have dramatically aggravated water pollution, especially in the middle and lower reaches of the Huai River Basin (Zhang et al., 2010). It is urgent and critical to control water pollution and regulate the dams or sluices reasonably in the Huai River Basin. The integrated simulation approach of water quality and quantity processes can provide insights into water pollution control and regulation of dams or sluices, and can facilitate the implementation of sustainable water resources management. Currently, two categories of models exist, i.e. watershed model and hydrodynamic model. The watershed model shows great superiority in describing the rainfall-runoff processes at basin scale, such as SWAT (Arnold et al., 1998), DTVGM (Xia, 1991, 2002a), HBV-N (Arheimer and Brandt, 1998) and HEQM (Zhang et al., 2016). However, model performance should still be improved for flow routing, transport and transformation of water quality, and dam regulation in river network. For example, the dynamic river flow routing can be accurately reproduced by a full Saint-Venant equation rather than a simplified water balance model commonly incorporated in a watershed model (Lian et al., 2007; Remo and Pinter, 2007; Paz et al., 2010; O'Sullivan et al., 2012; Paiva et al., 2013). The hydrodynamic model provides a widely-used approach because it is possible to derive detailed water quality and quantity processes with high spatial and temporal resolutions (Wu et al., 2004; Zhang et al., 2008; Cardona et al., 2011), such as EFDC (Hamrick, 1992), MIKE (Warren and Bach, 1992) and HEC-RAS (HEC, 1997). Whereas, enormous challenges still remain for their generalized applications, such as ambiguous land water and nutrient movement mechanisms (Yan and Kahawita, 2000; Gitis et al., 2007), flow regime and accompanying water quality processes disturbances by water infrastructure construction and regulation (Zhang et al., 2013). The yield and transport processes of land water and nutrient are essentially nonlinear issues (Xia, 1991; Gomi et al., 2008; Tarquis et al., 2011; Tuset et al., 2016). Numerous existing studies have been conducted ranging from empirical to physical approaches (Rinaldo et al., 2005; Kim et al., 2012; Hallema and Moussa, 2014; Maxwell et al., 2014). Yan and Kahawita (2000) developed a physically based numerical model to simulate pollutant dissolution and transport in overland flow, while the proper validation of the complete model required further extensive experimentation. Cea et al. (2014) analyzed the performance of a two dimensional overland flow model using empirical laboratory data, but the application is limited to DEM and mesh resolution. The physical approaches try to demonstrate the role of underlying storage on runoff and nutrient processes based on hydrodynamics, while the elaborate depiction is still confined by observation availability, elaborate parameterization, computational efficiency and understanding of physical processes (Khu et al., 2008; Kim et al., 2012). Alternatively, systematic approach, which has gained popularity for decades, is highly efficient and applicable to depict the interactions between water and nutrients, and exhibits the inherent nonlinearity and time-variant nature in above processes. Specifically, functional series, which is unconditional to priori hypothesis and is suitable for complex systems with obscured physical mechanisms (Xia, 2002a), has been introduced into the nonlinear hydrological fields since the 1960s. Xia (1991) proposed a time

variant gain model (TVGM) based on a nonlinear rainfall-runoff relationship with few parameters and limited inputs, and has been widely applied in many river basins of China. Thus, it will be an efficient way for water quantity and quality simulation to incorporate the nonlinear systematic theory into the river hydrodynamic model to efficiently and accurately reproduce the disturbed runoff and water quality processes at basin scale. Furthermore, the impact of water infrastructure regulation should be considered, as flow regulation is a global phenomenon with rivers fragmented by water infrastructures, and significantly alters natural hydrological cycle (Zhang et al., 2013). The integrated water quality and quantity simulation considering water infrastructure regulations shows great superiority in representing unsteady flow and pollutants transportation, especially in dammed water systems. Zhang et al. (2011a) developed a dynamic numerical model through a laboratory experiment, and determined that water quantity and quality variations downstream of a sluice have nonlinear relationships with upstream flow, pollutant discharge and sluice regulation. Lopes et al. (2004) simulated the hydrodynamics and water quality processes integrating the ISIS FLOW and QUALITY modules and operational conditions of the Touvedo dam in the Lima river, North Portugal. Mateo et al. (2014) combined a water resources model, a river routing model and a reservoir operation module to simulate the floodplain inundation in the Chao Phraya River Basin, Thailand. In the Huai River Basin, many studies on the integrated water quality and quantity modeling have been conducted to support sustainable water resources management (Zhang et al., 2013; Yang et al., 2016). Whereas, the challenges still existed on efficiently simulating the land movements of water and nutrient, accurately depicting the disturbed flow regime and accompanying water quality processes, and identifying impacts of various anthropogenic activities. The objectives of this study are to: (1) integrate nonlinear response module of rainfall-runoffnutrient and specific operational rules of dams with river hydrodynamic and water quality modules to describe regulated runoff and water quality variables in the Huai River Basin; (2) assess the impacts of anthropogenic activities (dam regulation, point source pollutants and diffuse pollutant losses) on hydrodynamic and water quality variations and identify critical impacted areas. 2. Materials and methods 2.1. Study area Located between Yangtze River Basin and Yellow River Basin, Huai River Basin (111°55′–121°25′E, 30°55′–36°36′N, Fig. 1), is the sixth largest river basin with high density of population in China. The drainage area is 274,657 km2. Its climate belongs to the north-south climate transition straps, with northern warm temperate zone and southern subtropical zone. The annual average precipitation and runoff depths were 883 mm and 230 mm. Both precipitation and runoff gradually decreased from south to north, and were unevenly distributed throughout the year. Most of the precipitation fell between June and September (the flood season), and runoff in the flood season accounted for 55%–82% of total annual runoff. The main land uses were dry farmland (68%) and rice (15.4%). This region confronted water disasters frequently, including droughts, floods, pollution and ecosystem degradation (Xia et al., 2011). Intensive anthropogenic activities significantly altered natural flow regimes and riverine environment (Zhang et al., 2011b; Zhai et al., 2014), such as industrial and municipal wastewater emissions, nutrient losses, excessive dam and sluice construction. The chemical oxygen demand (COD) and ammonium nitrogen (NH4-N) loads discharged into rivers in 2014 were 1.08 and 1.39 times of those targeted emissions (Huai River Basin Comprehensive Planning 2012–2030, 2013). Diffuse pollution contributed 30% for water quality deterioration in 2000. Over 10 thousand dams and sluices were

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

751

Fig. 1. Location of study area, rainfall stations, water quantity and quality monitoring stations and dams.

constructed in the whole basin, and every 25 km2 was segmented by one dam or sluice. The total storage capacity of dams and sluices occupied 51% of the total annual runoff. The upper and middle streams of Huai River Basin were selected as the study area, because they were intensively interrupted by anthropogenic activities. 2.2. Data The collected database included meteorological, hydrological and water quality series in 2004, 2007 and 2008, and channel morphology data (Table 1). Due to data scarcity in the highly polluted and regulated Huai River Basin, flow regimes and water quality observations in different level years were selected for model calibration and validation, including the high (2007), normal (2008) and low (2004) flow years with percentage of water discharge anomaly being 28%, − 8% and −30%, respectively (GB/T 22482-2008, 2008). Daily precipitation series were obtained from 172 metrological stations, and were monitored following the industry standard of precipitation observation (SL 21-2006, 2006). The series of water discharge and level were collected from the Huai River Commission, and were monitored synchronously in the whole year following the national observation standard of water discharge and level (GB 50179-93, 1993; GB/T 50138-2010, 2010). Specifically, hourly and daily data were collected during the flood season (June–September) and non-flood season (October–May), respectively. Both water quality series and sewage discharge data were collected from the monitoring center of Water Resources Protection Bureau of Huai River, and were sampled following the national standard of water quality testing and sewage pollution monitoring (GB

11892-1989, 1989; HJ/T 373-2007, 2007). Two water quality indices were selected, i.e. CODMn and NH4-N, because they were the most important indices for water-pollution control in the Huai River Basin. The annual sewage discharge records consisted of industrial and urban domestic sewage discharge data at county scale, and the temporal variations of CODMn and NH4-N loads discharged were shown in Fig. 2. Channel morphology data at main monitoring stations were obtained from the Huai River Commission to derive geomorphologic distributions of river channels, and were monitored following the industry standard of waterway survey (SL 257-2000, 2000).

2.3. Model integration approach The integrated water quality and quantity model included the nonlinear response module of rainfall-runoff-nutrient, hydrodynamic module, water quality module and dam regulation module (Fig. 3). The critical equations of each module were presented as follows, and other equations were given in the Appendix. The major interactions between the modules were described as follows: (1) the nonlinear response module of rainfall-runoff-nutrient determined the dynamic runoff and corresponding nutrients for in-stream routing and transformation in other modules; (2) the description of in-stream unsteady flow calculated an essential hydrological boundary for the simulation of water quality transport and transformation; (3) the hydrodynamic and water quality modules provided water quantity and quality inputs for dam regulations, and dam regulation module simulated the dam-induced fluctuations in runoff and water quality.

Table 1 Basic information of water quantity and water quality monitoring stations in the Huai River Basin. Data type

Scale

Data description

Source

Standard

Meteorology

172 stations (2004, 2007 and 2008)

Daily precipitation

SL 21-2006 (2006)

Hydrology

Water level: 15 stations (2004, 2007 and 2008), water discharge: 10 stations (2004, 2007 and 2008) 11 stations (2004, 2007 and 2008)

Daily data during the non-flood season, hourly data during the flood season CODMn and NH4-N concentrations

10 stations (2004, 2007 and 2008)

Annual industrial and urban domestic sewage discharge data

The Huaihe River Commission The Huaihe River Commission The Huai River Water Resources Protection Bureau The Huai River Water Resources Protection Bureau The Huaihe River Commission

Water quality

Sewage discharge

Channel 10 stations (2004, 2007 and 2008) morphology

Cross-sections of river channels

GB 50179-93 (1993); GB/T 50138-2010 (2010) GB 11892-1989 (1989); HJ/T 373-2007 (2007) GB 11892-1989 (1989); HJ/T 373-2007 (2007) SL 257-2000 (2000)

Note: the above database in 2004 was not available at five stations (Zhoukou, Huaidian, Jieshou, Fuyang and Yingshang) in Shaying River due to data deficiency.

752

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

Fig. 2. Temporal variation of CODMn, NH4-N concentrations and corresponding point source pollution loads at four stations in 2004, 2007 and 2008. Boxes define the 25th, mean and 75th percentile values, and the vertical bars (whiskers) define the 1th and 99th percentile values. Median values, minimum and maximum values are defined by the black solid dot and dash symbols, respectively.

2.3.1. Nonlinear response simulation of rainfall-runoff-nutrient The module described the responses of direct runoff (surface and subsurface runoffs) and nutrient to storms, and acted as critical linkages for the movements of unsteady flow and nutrients in river channels in the hydrodynamic and water quality modules. Distributed Time Variant Gain Model (DTVGM) was adopted to calculate runoff generation and routing, which was equivalent to a special form of the second order of nonlinear Volterra functional series (Xia, 1991, 2002a). The module assumed that surface runoff coefficient was time variant, and the nonlinearity was mainly attributed to the differences in soil moisture (Xia, 1991, 2002a, 2002b). The detailed mathematical descriptions of DTVGM were given in Appendix A. In a given land cell, surface flow (Qs) was estimated with a convolution function (Singh, 1988), and subsurface flow (Qg) was calculated through a linear storage-outflow relationship (Zoch, 1934). The nonlinear relationship between rainfall and nutrient load rate was described as follows. n

t

t

Lx ðt Þ ¼ ∑ ∫ 0 ⋯∫ 0 U m¼1

0

x;n ðτ 1 ; ⋯; τ m Þ

m
 ∏ h t−τ j dτ1 ⋯dτm j¼1

ð1Þ

where Lx was load rate of nutrient x in the direct flow (g/s); U′x was nutrient x load rate of instantaneous unit hydrograph (mg/l), U′x = UCx, U was flow rate of instantaneous unit hydrograph (IUH), Cx was nutrient x concentration in the direct flow (mg/l); h was excess runoff (mm); x was a kind of nutrient, including nitrogen, phosphorus and others. In our study, the time variant gain factor of surface flow was linked with the load rate of instantaneous unit hydrograph. Thus, the yield and transport processes of nutrients in the direct flow were described as a second order nonlinear Volterra functional series. t

Lx;s ðt; n; k; kx Þ ¼ g 1 ∫ 0 U ðτ ÞC x;s ðτÞX ðt−τ Þdτ t t−τ þ g2 ∫ 0 ∫ 0 U ðτÞC x;s ðτ ÞU 0 ðt−τ−σ ÞX ðσ ÞX ðt−τÞdσdt

ð2Þ

where Lx,s was load rate of nutrient x in the surface flow (g/s); Cx,s was nutrient x concentration in the surface flow (mg/l); g1 and g2 were two parameters linked the time variant runoff coefficient with watershed wetness (Xia, 2002b); g1 was a linear parameter related with time variant gain factor for surface runoff Gs (Eq. (A1)), g2 was a nonlinear parameter related with time variant gain factor for surface runoff Gs; X was precipitation (mm); U 0 ðσÞ ¼ eð−σ=K e Þ  K e −1 .

Fig. 3. The model structure and the interactions among the modules.

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

The nutrients in top soil layer were considered to gradually release into the surface flow during a storm event, and this process was described by the 1st order adsorption dynamic equation (Novotny and Chesters, 1981).
 ∂C x;s ¼ kr;x  C E;x;s −C x;s ∂t

ð3Þ

where kr,x was nutrient x transfer coefficient at the interphase of soil and surface runoff (s−1), which accounted for effects of interphase specific surfaces (Zingales et al., 1984) and varied for parent soils (Novotny and Chesters, 1981); CE,x,s was equilibrium concentration of nutrient x in the surface flow (mg/l), which was seasonally affected by underlying land cover, and temperature and so on. Besides, nutrients in the subsurface flow were considered to be constantly and uniformly distributed. Thus, the nutrient followed the direct flow movement and was discharged into river channels. S1;x ðt Þ ¼ Ls;x ðt Þ þ Lg;x ðt Þ

ð4Þ

where S1,x was load rate of nutrient x from land to river channels (g/s); Lg,x was load rate of nutrient x in the subsurface flow (g/s). 2.3.2. Hydrodynamic simulation The hydrodynamic model was numerical solution of Saint-Venant equation for one dimensional unsteady flow in rivers (de SaintVenant, 1871). Rivers were discretized into minimum calculation units, where the direct flow was uniformly discharged. The module used the unconditional stable Preissmann four-point implicit difference scheme with weight coefficient no b0.50 (Preissmann, 1961), which has been widely used in open channel flow routing and nutrient transport (Han, 2011; Domeneghetti et al., 2014; Yi et al., 2014). Thus, the equation was discretized into linear equations for solution with boundary and initial conditions. Furthermore, consistent equations, including a water balance equation and an energy conservation equation, were established at the junction point of river channels. Thus, the river channels were connected, and the established hydrodynamic model could reproduce the hydrodynamic situation in the highly disturbed river basins, and provide flow fields to simulate the pollutant dynamics (Appendix B). 2.3.3. Water quality simulation The migration and transformation of nutrients in river channels were simulated using the one dimensional water quality module (Luo and Song, 2000). CODMn and NH4-N concentrations were treated as separate-state variables. The decay of water quality was estimated by empirical approaches. The comprehensive degradation coefficient (K) was calculated as a function of water temperature, hydraulic slope and flow regimes (Luo and Song, 2000). River longitudinal dispersion was expressed as a function of hydraulic and geomorphic features (Kashefipour and Falconer, 2002). The upwind implicit scheme (Warming and Hyett, 1974) was adopted to discretize the advection-diffusion equation. Correspondingly, a mass balance equation was established at the junction points of river networks (Appendix C). 2.3.4. Dam regulation simulation Dam regulation, as river hydraulic singularity, violated the hydraulic continuity assumption in the Saint-Venant equation and strengthened nonlinearities in water cycle. Thus, dam regulation should be considered in the integrated water quality and quantity model, so as to provide regulated boundaries for the dynamic flow routing and pollutants migration under various dam operation rules (Table 2) according to the water conservancy industry standard (SL 20-92, 1992). The dam regulation module in this study applied actual dam regulation rules including the number and openings of the gates, and the discharge coefficient,

753

width and elevation of each weir. These regulation rules depended on the actual operation of each dam rather than the simplified rule curves of each dam. The outflow of dam was calculated as   Q iþ1 ¼ α 0 þ β0 Z 1 iþ1 −Z d

ð5Þ

where Q was dam outflow (m3/s); Z1 was upstream water level (m); Zd was elevation of weir (m); α0 and β0 were coefficients, which were functions of upstream and downstream water levels, weir properties and regulation rules, or derived from the rule curves of each dam, such as the relationships of water level and storage capacity, water discharge and storage capacity, etc. The coefficients α 0 and β 0 for free flow and submerged flow were calculated as: 

α 0 ¼ λ1 β0 ¼ 0

8 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi < α 0 ¼ λ2 Z 1 −Z 2 Z 1 −Z d or λ2 : β0 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Z 1 −Z 2

ð6Þ

where Z2 was downstream water level (m); λ 1 was coefficient for free flow calculation, and was calculated through number and openings of the gates, and discharge coefficient, width and elevation of weir, and upstream water level; λ 2 was coefficient for submerged flow calculation, and was calculated through above regulation rules of gates and properties of weir, and downstream water level. Besides, water quality variation at dam node obeyed the mass balance equation, and was described as follows: ∑Q in C in −∑QC þ S1 þ S2 ¼

ΔVC Δt

ð7Þ

where Qin and Cin were water discharge (m3/s) and corresponding water quality concentration (mg/l) flowing in dams; V was water storage volume (m3); S2 was point source pollution emission (g/s), including industrial and urban domestic sewage emissions. Based on the well calibrated and integrated model, anthropogenic disturbances in runoff and water quality were quantitatively assessed through scenario-based operation rules. Generally, an empirical relationship existed between water level and water discharge processes during the flood season with dam gates fully open, which could be identified using a linear regression model with the form as: Q ¼ γ0 þ γ1 Z 1 þ ε;

  ε∈ 0; δ2

ð8Þ

where γ0 and γ1 were regression coefficients of independent variable Z1; ε was random error. Furthermore, the relationship between water level and water discharge was identified by a statistical significance test with the hypothesis assuming that γ1 = 0. The statistical value followed an F distribution with the freedom degree being p-2. And the fitness of the empirical relationship was evaluated using the determination coefficient.  2 p ∑i¼1 Q s;i −Q R2 ¼ 2 p
∑i¼1 Q i −Q s;i

ð9Þ

where R2 was determination coefficient; Qi, Qs , i and Q were observed water discharge (m3/s), fitted water discharge using the regression function (m3/s) and average of observed water discharge (m3/s); p was length of data series. The identified relationship which passed the F test was used to drive the well calibrated and integrated model to derive the unregulated runoff and water quality processes setting dam gates fully open at each dam during the whole simulation period. Specifically, the scenario did not

754

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

Table 2 Empirical estimation of dams' outflow. Flow condition

Equation

Flow condition

Equation

Free weir flow Free orifice flow

Q = C1Bd(Z1 − Zd)1.5 Q = M1Bde(Z1 − Zd)0.5

Submerged weir flow Submerged orifice flow

Q = C2Bd(Z2 − Zd)(Z1 − Z2)0.5 Q = M2Bde(Z1 − Z2)0.5

Note: C1 and C2 are free and submerged weir flux coefficients, respectively; Bd is width of dams; Z1 is upstream water level; Z2 is downstream water level; Zd is elevation of sluice gate; M1 and M2 are free and submerged orifice flux coefficient, respectively; e is sluice opening.

change the boundary and initial conditions of river basin, and was performed under the same climatic conditions.  2 p ∑i¼1 Q s;i −Q F¼ 2 ðp−2Þ  F ð1; p−2Þ p
∑i¼1 Q i −Q s;i

ð10Þ

Parameter calibration was performed from upstream to downstream, from runoffs to water quality concentrations. The selected sensitivity parameters were manually adjusted within their proper ranges. Four evaluation indices were adopted in our study (Moriasi et al., 2007), including the relative error (Re), correlation coefficient (r), NashSutcliffe efficiency coefficient (NSE) and Willmott index (d). The corresponding descriptions were listed in Table 3. 2.5. Anthropogenic disturbances assessment

2.4. Model setup 2.4.1. Model initial and boundary conditions The study area was divided into 15 sub-basins for nonlinear response module of rainfall-runoff-nutrient, and the river network consisted of 18 nodes and 515 elements, including five dams (Linhuaigang and Bengbu dams in the Huai Mainstream, Huaidian, Fuyang and Yingshang dams in the Shaying River; see Fig. 1). The simulation period was from January 1st, 2004 to December 31st, 2004, and January 1st, 2007 to December 31st, 2008. The data available in 2007 was used for model calibration, while the flow and water quality conditions in 2004 (except at five stations in Shaying River due to data deficiency, i.e., Zhoukou, Huaidian, Jieshou, Fuyang and Yingshang) and 2008 were selected for model validation. Initial water quantity and quality conditions were configured by the observed water levels, water discharges, concentrations of CODMn and NH4-N at 0:00 am January 1st, 2007. The locations of upstream and downstream boundary points were shown in Fig. 1. The upstream conditions were determined using the observed water discharge, water level and water quality concentration series. The observed water discharge data at Bengbu was specified as the downstream boundary condition of the Huai Mainstream. 2.4.2. Parameter sensitivity analysis and model evaluation Parameter sensitivity analysis was performed using the simple and efficient perturbation analysis method, which repeatedly added a small change to one parameter (xj) at a time, while other parameters were kept unchanged (xi, i ≠ j, i = 1,…,m). The sensitivity degree was evaluated referring to Zhang et al. (2013). The evaluation standard was divided into Class I–IV, indicating that the evaluated parameter was insensitive, middle sensitive, high sensitive and very high sensitive, respectively. In the integrated model, eight hydrological parameters and nine water quality related parameters were selected for parameter sensitivity analysis.

Point source pollution emissions, diffuse pollutant losses and dam regulations were considered as main anthropogenic activities in the study area, which significantly disturbed natural water quantity and quality processes. Thus, the variation ratios (η) among various designed scenarios were calculated to quantify the impacts of various anthropogenic factors on runoff and water quality using the well-calibrated model, and to identify critical impacted areas. The assessment indices were water level and water discharge for runoff, CODMn and NH4-N concentrations for water quality in scenarios. In general, pollution control and treatment mainly impacted the variations of water quality, and dam regulation might play a key role in both runoff and water quality processes. Thus, the impacts of point source emissions and diffuse pollutant losses were assessed on water quality variations, and the impacts of dam regulation were assessed on hydrodynamic and water quality variations. η¼

∑ðM 0 −M1 Þ ∑M0

ð11Þ

where η is variation ratio of each index, including ηpoint, ηdiffuse, ηdam and ηpollutants for point source pollutant assessment, diffuse pollutant loss assessment, dam regulation assessment and pollutant sources assessment, respectively; M0 is assessment index (water level, water discharge, CODMn and NH4-N concentrations) under current anthropogenic disturbance scenario; M1 is assessment index under various scenarios described as follows. Scenario 0 was the baseline, in which the point source pollution emissions, diffuse pollutant losses and dam regulations were maintained the current conditions. Scenarios 1, 2 and 4 were designed for the impact assessments of pollutant sources. Scenario 1 was that the current point source pollutant was removed from our study area (ηpoint), while Scenario 2 was that the current diffuse pollutant was removed (ηdiffuse). Scenario 4 was that both of these pollutant sources were removed (ηpollutants). In general, ηpoint, ηdiffuse and ηpollutants

Table 3 The evaluation indices for model simulations. Definition

Reference

Category

Range

Optimal value

j Rej ¼ j Io −Is j Io ∑ðIo;i −I o ÞðI s;i −I s Þ ffi r ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 ∑ðIo;i −I o Þ ∑ðIs;i −I s Þ

Relative error

Error index

[0,1]

0

Correlation coefficient

Standard regression

[−1,1]

1

Nash-Sutcliffe efficiency coefficient

Dimensionless

(−∞,1]

1

Willmott index

Dimensionless

[0,1]

1

NSE ¼ 1− d ¼ 1−

∑ðIo;i −I s;i Þ2 2

∑ðI o;i −I o Þ

∑ðI s;i −Io;i Þ2 ∑ðjIs;i −I s jþjI o;i −I o jÞ

2

Note: I o is average of observed time series; Is is average of simulated time series; Io,i is observed data at i time; Is,i is simulated data at i time. Category refers to the category each evaluation indice belonging to in Moriasi et al. (2007).

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

were N 0, indicating that external pollutants aggregated water quality deterioration. Furthermore, Scenario 3 was that the gates of individual dams were set fully open during the whole simulation period in order to assess the impact of dam regulations (ηdam). A positive ηdam indicated that dam regulation promoted water level raising, water discharge increasing and water quality deteriorating with the current pollution sources, and a negative ηdam suggested that dam regulation promoted water level declining, water discharge decreasing and water quality ameliorating. While a zero ηdam represented no disturbance in above processes. 3. Results 3.1. Parameter sensitivity analysis, calibration and validation 3.1.1. Parameter sensitivity analysis All the parameters were selected for parameter sensitivity analysis of hydrology and water quality (Table 4). For hydrological simulation, the runoff generation parameters (g1 and g2) were the most sensitive parameters, both of which determined the nonlinear time variant gain runoff coefficient, and were key parameters for the nonlinear runoff yielding process. Besides, the roughness coefficient (n) was also selected, which was closely related with dynamic channel routing process. For water quality simulation, the dominantly sensitive parameters were physico-chemical parameters related with water temperature (water quality transfer coefficient β, and comprehensive degradation coefficient at 20 °C K20), followed by characteristic parameters of subbasins (equilibrium concentration in the mobile phase CE, and water quality transfer coefficient kr) or river channels (roughness coefficient n, hydraulic slope J, and rate constant for diffusion process α), and hydrological parameter (g2) which also impacted water quality concentration variations.

755

in both the calibration and validation periods except 0.66 at Zhoukou. In general, the water level simulations fitted the observations well at all the 12 stations except at Zhoukou, which may be due to water infrastructure reconstruction in Zhoukou City during the validation period (He et al., 2011), as the construction formworks blocked water currents, and altered water level-water discharge relationship at Zhoukou. As for water discharge simulation, the Re values were within ±10% at all the stations, except at Runheji in both the calibration and validation periods, which may be due to the joint regulation of flood detention areas in emergency along Runheji river segment. The values of r, d and NSE were over 0.97, 0.97 and 0.90 at all the stations during the calibration period, and were over 0.85, 0.95 and 0.75 during the validation period, respectively. Thus, water discharge simulations were quite satisfactory considering the dam operations. 3.1.3. Water quality simulations As shown in Fig. 4 and Table 6, the variations of CODMn and NH4-N concentration series were well captured in the upper and middle streams. For the CODMn concentration simulation, the Re values varied between −8.95% (Bengbu) and 12.36% (Jieshou) in the calibration period, and varied between −11.63% (Yingshang) and 6.67% (Huaidian) in the validation period. In both the calibration and validation periods, the r values were N0.59 and 0.45 (except at Jieshou and Fuyang), and the d values were no b 0.84 and 0.75, respectively. Comparatively, the simulated NH4-N concentrations were better correlated with the observed series (r ≥ 0.65) in the whole simulation period. The Re values varied between −25.36% (Huainan) and 19.37% (Fengtai), and between −9.44% (Huaidian) and 16.35% (Bengbu) in the calibration and validation periods, respectively. Moreover, the d values were no b0.63 and 0.55 in the calibration and validation periods, respectively except at Huaidian and Jieshou. 3.2. Impact of pollution sources on water quality

3.1.2. Water level and water discharge simulations All the parameters were calibrated considering their physical characteristics, actual situation of study area and parameter sensitivity analysis. The simulation performances of water levels during the whole simulation period were shown in Fig. 4, and the evaluation results were shown in Table 5. The Re values at all the 12 stations were within ±1.80% in both the calibration and validation periods. The r values were N0.88 at all stations except that of Zhoukou (0.46) during the validation period. The NSE values were N 0.88 during the calibration period and were no b 0.64 during the validation period, despite a negative value at Zhoukou during the validation period. The d values were no b0.93

Anthropogenic disturbances were only assessed in 2007 and 2008 due to data deficiency in Shaying River in 2004. The impacts of pollutant sources on water quality were analyzed by comparing the differences of water quality concentrations between Scenario 1 or 2 and Scenario 0. The relative variation ratios of Scenarios 1 and 2 were shown in Fig. 7. The point source pollution emission was still the most critical pollutant source. The corresponding variation ratios for CODMn concentration reached 43%, 30% and 47% at Fuyang, and 29%, 30% and 34% at Bengbu in the whole year, the flood and non-flood seasons, respectively. For NH4-N concentration, the corresponding variation ratios reached 45%,

Table 4 Results of parameter sensitivity analysis for water quality and quantity simulation. Class

Parameter

Definition

Unit

Range

Water quantity

g1 g2 g3 N k KKG J n krCOD krN CECOD CEN KCOD, 20 KN, 20 βCOD βN α

Linear parameter related with time variant gain factor for surface runoff Nonlinear parameter related with time variant gain factor for surface runoff Subsurface runoff generation parameter Number of hypothesis linear reservoir Storage coefficient Parameter of groundwater linear reservoir Hydraulic slope Roughness coefficient of channel Mass transfer coefficient for CODMn Mass transfer coefficient for NH4-N Equilibrium CODMn concentration in the mobile phase Equilibrium NH4-N concentration in the mobile phase Comprehensive degradation coefficient of CODMn at 20 °C Comprehensive degradation coefficient of NH4-N at 20 °C Temperature coefficient of CODMn in water Temperature coefficient of NH4-N in water Rate constant for diffusion process

– – – – h – – – s−1 s−1 mg/l mg/l s−1 s−1 – – –

[−1,0] [0,1] [0,1] (0,+∞) (0,+∞) (0,1) [0, +∞] [0.02, 0.20] [0,+∞] [0, +∞] (0,+∞) (0,+∞) [0, +∞] [0, +∞] [0,1.5] [0,1.5] [0, +∞]

Water quality

Water discharge

NH4-N concentration

CODMn concentration

Value

Class

Value

Class

Value

Class

−0.11 0.09 0.00 0.00 0.00 0.00 −0.01 0.08 – – – – – – – – –

II II I I I I I II – – – – – – – – –

0.02 −0.04 −0.02 0.00 0.00 0.00 −0.09 −0.18 – 0.00 – 0.01 – −0.27 – 2.36 −0.11

I I I I I I II II – I – I – III – IV II

0.02 −0.06 0.00 0.03 0.03 0.00 −0.07 −0.30 0.05 – 0.11 – −0.25 – 1.06 – −0.11

I II I I I I II III II – II – III – IV – II

Note: texts in column “Value” and “Class” represent the sensitivity indices and sensitivity degrees of the sensitive parameters, respectively.

756 X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764 Fig. 4. Simulated and observed series in the calibration (2007) and validation (2004 and 2008) periods at four stations. Note: (a) water discharge series, (b) water level series, (c) and (d) CODMn and NH4-N concentration series. Water discharge process is not monitored at Fengtai station; hydrological and water quality series are not available at Jieshou station in 2004 due to data deficiency.

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

757

Table 5 Model evaluation indices for water level processes at 12 stations and water discharge processes at seven stations during the calibration (2007) and validation (2004 and 2008) periods. Process

Stations

Water level

Water discharge

Calibration

Wangjiaba Chencun Linhuaigang Zhengyangguan Lutaizi Fengtai Huainan Zhoukou Huaidian Jiehsou Yingshang Mengcheng Average Runheji Lutaizi Huaidian Jieshou Fuyang Yingshang Bengbu Average

Validation

Re (%)

r

NSE

d

Re (%)

r

NSE

d

−1.09 −0.59 0.53 −0.90 −0.82 −0.23 0.27 −0.16 0.20 −0.05 −1.73 −0.90 – 29.36 1.25 −3.12 9.93 −0.55 2.82 −3.27 –

0.99 0.98 0.99 0.99 0.99 0.99 0.99 0.88 0.98 0.98 0.99 0.95 0.98 0.99 0.99 0.98 0.99 0.99 0.99 0.99 0.99

0.98 0.93 0.98 0.97 0.98 0.99 0.99 0.94 0.98 0.98 0.96 0.88 0.97 0.93 0.98 0.96 0.97 0.98 0.97 0.98 0.97

0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.93 0.99 0.99 0.99 0.97 0.98 0.99 0.99 0.99 0.99 0.99 0.97 0.99 0.99

−1.45 −0.08 −0.83 −0.55 −0.12 0.36 0.68 −0.80 0.18 −0.10 −1.30 −0.56 – 12.69 8.35 −4.33 −4.57 −1.37 −1.24 6.37 –

0.99 0.97 0.99 0.99 0.99 0.99 0.99 0.46 0.88 0.92 0.99 0.95 0.93 0.99 0.99 0.88 0.91 0.96 0.96 0.99 0.95

0.94 0.93 0.84 0.97 0.98 0.97 0.80 −0.50 0.70 0.64 0.93 0.82 0.75 0.96 0.97 0.76 0.83 0.93 0.92 0.96 0.90

0.98 0.99 0.99 0.99 0.99 0.99 0.97 0.66 0.93 0.96 0.98 0.93 0.95 0.97 0.99 0.94 0.96 0.98 0.98 0.99 0.97

41% and 47% at Fuyang, and 29%, 30% and 34% at Bengbu in the whole year, the flood and non-flood seasons, respectively. Moreover, the contributions of diffuse pollutant losses in the flood season outweighed those in the whole year and the non-flood season for both CODMn and NH4-N concentrations. The contributions of diffuse pollutant losses even approached those of point source pollution emission for CODMn concentration variations at Yingshang and Bengbu. Thus, efforts should be taken to decrease diffuse pollutant losses in the upstream of these stations. The variations of CODMn and NH4-N concentrations showed similar spatial patterns in both Scenario 1 and Scenario 2 in Shaying River and Huai Mainstream. Specifically, the variation ratios increased at Fuyang and decreased at Yingshang along Shaying River, and they increased from Linhuaigang to Bengbu along Huai Mainstream. 3.3. Impact of dam regulations on runoff and water quality The impacts of dam regulations on runoff and water quality were analyzed by comparing the differences of water level, water discharge and water quality concentrations between Scenarios 3 and 0. The identified coefficients of the empirical Q ~ Z relationship (Eq. (8)) were listed in Table 7. The regression functions were well fitted with relatively high

Table 6 Model evaluation indices for CODMn and NH4-N concentration processes during the calibration (2007) and validation (2004 and 2008) periods at eight stations. Stations

Lutaizi Fengtai Huainan Huaidian Jieshou Fuyang Yingshang Bengbu

Period

Calibration Validation Calibration Validation Calibration Validation Calibration Validation Calibration Validation Calibration Validation Calibration Validation Calibration Validation

CODMn concentration

NH4-N concentration

Re (%)

r

d

Re (%)

r

d

−2.79 −7.11 3.20 4.52 −8.64 −6.43 3.96 6.67 12.36 −2.82 10.85 4.53 −3.42 −11.63 −8.95 0.24

0.72 0.64 0.59 0.71 0.66 0.64 0.69 0.63 0.66 −0.09 0.71 0.15 0.76 0.49 0.67 0.51

0.98 0.95 0.97 0.98 0.97 0.96 0.86 0.79 0.88 0.75 0.87 0.82 0.84 0.84 0.98 0.97

4.05 4.07 19.37 2.08 −25.36 −9.44 −2.61 −5.44 −9.48 −4.82 −13.56 6.36 −9.31 4.00 −10.01 16.35

0.86 0.83 0.98 0.90 0.94 0.78 0.74 0.65 0.68 0.85 0.85 0.80 0.87 0.82 0.84 0.65

0.92 0.85 0.98 0.90 0.93 0.71 0.56 0.43 0.63 0.39 0.64 0.55 0.66 0.59 0.87 0.72

determination coefficients (R2 N 0.60). The variations of water discharge and water level in Scenarios 3 and 0 were shown in Fig. 5, and several hydrologic metrics were shown to describe the variations of average (average and median values), low (minimum value, 75th and 99th percentile values) and high (maximum value, 1th and 25th percentile values) flow events according to Zhang et al. (2012). In Scenario 3, average, low and high water discharges increased at Huaidian and Yingshang in Shaying River, and at Linhuaigang and Bengbu in Huai Mainstream in both the non-flood season and the whole year, while they decreased at Fuyang in Shaying River. The annual variations were not remarkable (within ± 5%) except at Fuyang (ηdam = − 19.61%) and Yingshang (ηdam = 23.83%), while the disturbances were quite extraordinary in the non-flood season with the variation ratios over ±7% except at Huaidian. Besides, the alterations of water discharge were considered to be negligible in the flood season when dam gates were almost fully open in Scenario 0. Correspondingly, average, low and high water levels had opposite variation patterns with those of water discharges except at Yingshang (ηdam = 12.51%, 8.77% and 14.34% in the whole year, the flood and non-flood seasons, respectively), where water level increased synchronously with water discharge in Scenario 3. Water level increased about 1.8 m at Fuyang and Yingshang in Shaying River, decreased about 2.4 m at Huaidian in Shaying River, and decreased about 1.3 m at Linhuaigang and Bengbu in Huai Mainstream, and the variations were the greatest in the non-flood season. Thus, the current water dispatch rules raised water level at Huaidian, Linhuaigang and Bengbu, and increased water discharge at Fuyang. While water level tended to ascend at Fuyang and Yingshang, and water discharge tended to increase at Huaidian, Yingshang, Linhuaigang and Bengbu by setting dam gates fully open.

Table 7 Empirical coefficients of fitted water level-water discharge relationship at main dams. River

Huai Mainstream Shaying River

Dam

Linhuaigang Bengbu Huaidian Fuyang Yingshang

R2

Fitted coefficient α0

α1

−13,593.00 −14,754.00 −14,357.00 −6581.55 −5393.00

702.09 1060.50 420.56 278.81 262.43

Note: all the estimated coefficients pass the F test at 90% confidence level.

0.91 0.90 0.79 0.73 0.63

758

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

The variations of CODMn and NH4-N concentrations were shown in Fig. 6 in Scenarios 3 and 0. Water quality deterioration was alleviated at Yingshang and Bengbu in Scenario 3, while it was aggravated at Huaidian, Fuyang and Linhuaigang in the whole year, the flood and non-flood seasons. Specifically, the average variations of CODMn and NH4-N concentrations were the most remarkable in the whole year and the non-flood season at Yingshang (18% b ηdam b 21%), and were the most remarkable in the flood season at Bengbu (21% b ηdam b 27%). While the disturbances in CODMn and NH4-N concentrations were negligible at Fuyang (−2% b ηdam b 0%) except for NH4-N concentration in the flood season (ηdam = 7%), and were slight at Huaidian (− 6.5% b ηdam b −2%). Whereas, the variations of CODMn and NH4-N concentrations were the most remarkable in the flood season (−33% b ηdam b 36%). Thus, the current water dispatch rules alleviated water quality deterioration at Huaidian, Fuyang and Linhuaigang, while they aggregated water pollution at Yingshang and Bengbu with the current pollution sources. 3.4. Integrated impact of pollution sources and dam regulations on water quality The impacts of pollution sources and dam regulations on water quality variations were analyzed in Scenarios 3 and 4. Water quality deterioration was alleviated through dam regulations at Huaidian, Fuyang and Linhuaigang (−36% ≤ ηdam ≤ 0) in the whole year, the flood and nonflood seasons, while it was aggravated through dam regulations at Yingshang (15% ≤ ηdam ≤ 25%) and Bengbu (6% ≤ ηdam ≤ 25%) with current pollution sources (Fig. 7). Moreover, the contributions of pollution sources (ηpollutants) varied between 20% and 56% for the variations of both CODMn and NH4-N concentrations in the whole year, the flood and non-flood seasons. The contribution was over 50% at Fuyang and Bengbu for both CODMn and NH4-N concentrations in the whole year, the flood and non-flood seasons. On the whole, water quality variation was mainly impacted by point source pollution emissions (12%–43%), diffuse pollutant losses (0–23%) and dam regulations (−29%–20%) except at Yingshang and Linhuaigang. Specifically, the contributions of pollution sources (point source pollution emissions and diffuse pollutant losses) outweighed those of dam regulations. The impacts of dam regulation on CODMn or NH4-N variations at Yingshang (15%–20%) outweighed those of point source pollution emissions (7%–32%) or diffuse pollutants losses (1%–14%), respectively. Dam regulations contributed the most for water quality variation at Linhuaigang (25%–36%) except in the flood season. Thus, specific water quality managements should be adopted and strengthened to ensure regional water environment treatment. For example, pollution sources reduction and wastewater pipeline construction should be given priority to water pollution control at Huaidian, Fuyang and Bengbu. Besides, attention should also be paid to diffuse source pollution prevention in the flood season at Bengbu (23%–31%). As for Yingshang, various water quality managements should be adopted, such as scientific dam regulations, pollution sources reduction, wastewater pipeline construction, nonpoint source pollution prevention. 4. Discussion 4.1. Integrated simulation approach for anthropogenic disturbance assessment The integrated simulation provides an effective approach to assess water quantity variation and water quality deterioration by considering the complicated hydrological-hydrodynamic interactions and various

759

anthropogenic interferences at basin scale. On the one hand, the integrated simulation approach takes full advantage of hydrological system and dynamic theories. The systematic response function method is a simple but efficient alternative to describing the complex rainfallrunoff-nutrient interactions at basin scale (Rinaldo et al., 2005), and it is flexible to be integrated with the existing dynamic illustrations in rivers. The disturbed river flow routing and pollutants transportation can be accurately depicted by incorporating dam regulation rules into the hydrodynamic and water quality modules. On the other hand, the assessment of anthropogenic interferences can excavate the critical pollution sources in regional water quality problems, and quantify the impacts of various anthropogenic factors. In comparison with watershed model and hydrodynamic model, the integrated model shows great significances in solving actual hydrological and water pollution issues, especially in large scale river basins or in river basins which are lack of detailed watershed topographic data. Comparatively, the integrated model provided competitive simulation results for hydrological and water quality dynamics in the Huai River Basin (Fig. 4; Tables 5 and 6). The existing studies using DTVGM (F. Ma et al., 2014; ZK. Ma et al., 2014), Xinanjiang model and SWAT (Shi et al., 2011) in the Huai River Basin, showed that Re, r and NSE were b15%, 0.85 and 0.70 for daily water discharge processes without considering dam regulations. As more detailed dam regulation rules were incorporated in the model, water quality processes were better simulated than other studies, such as improved SWAT (Zhang et al., 2013) for CODMn and NH4-N loads at unregulated stations (Re: ± 8% and ± 13%; r: 0.50 and 0.42) and at sluices (Re: ± 6% and ± 13%; r: 0.42 and 0.22). However, further improvements are still needed, such as different kinds of nutrients were treated as separate-state variables, and their migration and transportation processes were comprehensively simulated with individual parameters. The more detailed transformation processes among various kinds of nutrients should be incorporated in further studies. The rainfall-runoff-nutrient relationship can be quantitatively determined through field or laboratory experiments, and can be incorporated into the integrated model so as to strengthen the mathematical descriptions of water and nutrient interactions in soil layers. Although the simulation accuracy was acceptable in describing the hydrological and water quality variations in low flow, normal flow and high flow years, the model performance should be further validated with more data series. Model uncertainty analysis should also be performed, and relative approaches should be adopted to alleviate uncertainties, such as acquiring multi-source data, enhancing data quality, and improving model structures. 4.2. Impacts of dam regulation Various disturbance patterns were discerned on runoff and water quality processes in Huai Mainstream and Shaying River. Huaidian, Yingshang, Linhuaigang and Bengbu dams could be coarsely classified as storage dams in the non-flood season and peak flow regulation dams in the flood season (Poff and Hart, 2002; McManamy, 2014; Alrajoula et al., 2016), which were characterized as having large storage capacities, long residence times and high control over water-release rates in the non-flood season, and attenuating peak flow in the flood season, respectively. The hydrological responses to these dam regulations were consistent with those reported in other studies (McManamy, 2014; ZY. Zhang et al., 2015). Water discharge tended to decrease and water level tended to increase in the non-flood season as the dams stored water volume and raised water level for water supply, and attenuated flood peak for flood prevention in the flood season. In addition, a synchronously increasing pattern existed in water discharge

Fig. 5. Water discharge and water level variations at five dams from 2007 to 2008 in Scenario 3 and Scenario 0. Boxes define the 25th, mean and 75th percentile values, and the vertical bars (whiskers) define the 1th and 99th percentile values. Median values, minimum and maximum values are defined by the black solid dot and dash symbols, respectively. Note: Scenario 0 represents baseline with current pollution sources and dam regulation rules. Scenario 3 represents dam gates setting fully open.

760

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

761

Fig. 7. Variation ratios of point source pollution emissions, diffuse pollutant losses and dam regulations at five dams in different scenarios. Note: Scenario 1 represents current point source pollutants removed, Scenario 2 represents current diffuse pollutant losses removed, Scenario 3 represents dam gates setting fully open, and Scenario 4 represents both pollutants removed.

and water level series at Yingshang in gates fully open scenario. Currently, low discharge was maintained in the non-flood season at Yingshang, which was occasionally closed for water storage. Thus, Yingshang was in both water storage and water discharge conditions for water supply and flood prevention throughout the whole year. Fuyang dam could be placed into run-of-river dam category, which typically had little control over water releasing. The regulation impacts of Fuyang dam were consistent with results determined in other studies (Fitzhugh and Vogel, 2011; YY. Zhang et al., 2015). Without dam's regulation, water discharge decreased and water level ascended at Fuyang, which was projected to discharge floods throughout the year for flood prevention and shipping. Furthermore, the fate and transport processes of pollutants were significantly impacted by the regulation rules, and the disturbance patterns were consistent with some existing results (Zhang et al., 2010; Suwarnoa et al., 2014; Morling et al., 2017). Huaidian and

Linhuaigang dams, as the storage dams in the non-flood season and peak flow regulation dams in the flood season, acted positively on water quality improvement with a relatively high water environment capacity during water storage period (Zhang et al., 2013). The upstream polluted inflows were gradually released into downstream rivers after a long residence time and decay process from the impoundments segmented by dams. The environmental impacts of dam regulation were slightly positive at the run-of-river dam (Fuyang) with preferable self-purification functions of rivers. Whereas, Yingshang and Bengbu dams, as the storage dams, aggravated water quality deterioration with the current regulation rules and pollution sources (Xia et al., 2008). Despite the above positive hydrologic impacts on water environment, water quality was probably deteriorated due to upstream highly polluted inflows, point source pollutants and diffuse pollutant losses (accounting for 26%– 43% of water quality deterioration) continuously discharging into

Fig. 6. CODMn and NH4-N concentration variations at five dams from 2007 to 2008 in Scenario 3 and Scenario 0. Boxes define the 25th, mean and 75th percentile values, and the vertical bars (whiskers) define the 1th and 99th percentile values. Median values, minimum and maximum values are defined by the black solid dot and dash symbols, respectively. Note: Scenario 0 represents baseline with current pollution sources and dam regulation rules. Scenario 3 represents dam gates setting fully open.

762

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

the impoundments with a low flow discharge. The fluvial mobility would be strengthened with dam gates fully open, and might promote water quality improvement. 4.3. Integrated impacts of pollutant sources and dam regulation Spatial heterogeneity existed in the regional water quality problems, and various anthropogenic activities acted diversely on water environment in the Huai River Basin. In general, the identified impacts of pollutants and dam regulations on water quality were in accordance with other studies (Zhang et al., 2010). Specifically, point source pollution emission was still found to be the most primary pollutant source for water quality deterioration (Zhang et al., 2010; Zhai et al., 2014). Diffuse pollutant losses also affected CODMn and NH4-N concentration variations as a significant external pollutant source. Specifically, Fuyang and Yingshang sub-basins were identified as the critical intensive regions because of livestock breeding and rural sewage discharge, which was in accordance with other studies (Hao, 2014). Fuyang City has the largest rural population in Anhui province in 2007. Moreover, livestock breeding was much more prosperous in Fuyang City than that in Luan, Huainan and Bengbu Cities (Anhui Statistical Yearbook, 2008). Although the impacts of diffuse pollutant losses were relatively slight on regional water quality variations, water quality concentrations decreased by approximately 30% and 40% in the whole year and the flood season, respectively in the whole basin if all the diffuse pollutants were removed. Thus, diffuse pollutant loss has become a great threat to water quality conditions in the whole basin, although it was not an extraordinary regional water quality problem in 2007 and 2008. Moreover, the regulated hydrological regimes played basic and crucial roles in water quality variations through acting on pollutant sources (Zhai et al., 2014). Water quality might be improved in the non-flood season with low flow events, especially in winter with relatively low water temperature and decreased diffuse pollutant losses, which was in accordance with other studies (Zhai et al., 2014). Dam gates were closed for water storage, which cut rivers off and prevented upstream highly polluted inflow from pouring into downstream. Saturated and dissolved oxygen contents increased with the decreasing of water temperature, which would promote the oxidation reaction of pollutants. Besides, diffuse pollutants being washed away by streams were significantly decreased in the non-flood season with occasional storm events. 5. Conclusions We simulated the regulated water quality and quantity processes under various regulation patterns, and then discerned the impacts of several anthropogenic disturbances using integrated hydrologicalhydrodynamic simulation approach. The results showed that: (1) The simulation performance was satisfactory as for describing the regulated runoff and water quality variables in Huai River Basin, China. The relative error was within ± 1.80%, ± 30%, ±14%, ±32% for water level, water discharge, CODMn and NH4N concentrations, respectively. The average correlation coefficient was 0.95, 0.97, 0.61 and 0.81. The average Willmott index was 0.97, 0.98, 0.91 and 0.71. The average Nash-Sutcliffe efficiency coefficient was 0.87 and 0.93 for water level and water discharge, respectively. The integrated model was suitable for anthropogenic disturbance assessment in Huai River Basin. (2) Dam gates fully open resulted in the increasing of water discharges at Huaidian, Yingshang, Linhuaigang and Bengbu, and raising of water levels at Fuyang and Yingshang. Water quality tended to be improved by dam removal in the downstream of Shaying River and Huai Mainstream with current pollutant sources.

(3) Water quality variation was mainly impacted by point source pollution emissions (12%–43%), diffuse pollutant losses (0– 23%) and dam regulations (− 29%–20%) except at Yingshang and Linhuaigang. The impacts of dam regulation on water quality at Yingshang (15%–20%) outweighed those of point source pollution emissions (7%–32%) or diffuse pollutant losses (1%–14%), respectively. Dam regulation contributed the most for water quality variation at Linhuaigang (25%–36%) except in the flood season. Specific water quality managements should be adopted and strengthened to ensure regional water environment treatment. Conflict of interest We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled. Acknowledgements This study was supported by the Natural Science Foundation of China (No. 41671024), the China Youth Innovation Promotion Association CAS (No. 2014041) and the Program for “Bingwei” Excellent Talents in IGSNRR CAS (No. 2015RC201). Appendix A. Nonlinear response module of rainfall-runoff-nutrient The nonlinear description of rainfall-runoff process was described as 8 Rs ðt Þ ¼ Gs ðt Þ  X ðt Þ > > > > < Rg ðt Þ ¼ Gg ðt Þ  APIZðt Þ t > APIðt Þ  X ðt−σ Þdσ > Gs ðt Þ ¼ g 1 þ g 2  > > 0 : Gg ðt Þ ¼ g3

ðA1Þ

where X was precipitation (mm); Rs and Rg were surface and subsurface runoffs, respectively (mm); API(t) = ∫t0U0(σ) ⋅ X(t − σ)dσwas antecedent precipitation index function (mm), U 0 ðσÞ ¼ eð−σ=K e Þ  K e −1 ; Ke was recession rate of soil moisture; Gs and Gg were time variant gain factors for surface and subsurface runoffs, respectively; g3 was coefficient in subsurface runoff generation. g1, g2 and g3 were closely related with geomorphic characteristics, soil parameters and land use types (Song et al., 2016). Thus, spatial heterogeneities were considered in runoff yielding. The direct flow could be calculated as 8 Z t > > > ð t; N; k Þ ¼ U ðτ; N; kÞ  Rs ðt−τÞdτ Q > s < 0 Ao > þ Q g  ðt−Δt Þ  KKG Q g ðt Þ ¼ Rg  ð1−KKGÞ  > > 3:6t > : qðt Þ ¼ Q s ðt Þ þ Q g ðt Þ

ðA2Þ

where U(τ,N,k) was flow rate of instantaneous unit hydrograph (IUH), U τ

Ao ðτ; N; kÞ ¼ kΓðNÞ  ðτkÞN−1  e−k ; N was number of hypothesis linear reser-

voir; k was storage coefficient; Ao was area of land cell (km2); KKG ¼ kg −0:5Δt ; kg þ0:5Δt

kg was constant coefficient of linear groundwater reservoir

(s); q was direct flow (m3/s), which was uniformly discharged into river channel for water and nutrient routing in the hydrodynamic and water quality modules.

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

(d−1); J was hydraulic slope; α was rate constant for the dispersion propffiffiffiffiffiffiffiffi cess; u ¼ gHJ was friction velocity of river channel (m/s); H was water depth (m). The above advection-diffusion equation was discretized as

Appendix B. Hydrodynamic module The one-dimensional Saint-Venant equation was 8 ∂Z ∂Q > > > : þ α0 u þ gA−Bu2 −u2 jz þ gn2 4=3 ¼ qV x ∂t ∂x ∂x ∂x AR

ðB1Þ

where A was cross-section area of river channel (m ); R was hydraulic radius (m); B was water surface width (m); Z was water level (m); Q was water discharge (m3/s); v was water velocity (m/s); Δt was time step (s); x was spatial interval (m); n was roughness coefficient of river channel; g was gravitational acceleration (m/s2); Vx was longitudinal water velocity of lateral inflow (m/s); α'was momentum correction coefficient. The above equation was discretized as iþ1 þ Q iþ1 þ C j Z iþ1 −Q iþ1 j jþ1 þ C j Z j jþ1 ¼ D j iþ1 iþ1 E j Q iþ1 þ G j Q iþ1 j − F jZ j jþ1 þ F j Z jþ1 ¼ Ψ j

ðB2Þ

where i and j were time step and river section; C, D, E, F, G and Ψ were coefficients which were calculated using variables at ith time; θ was Bn 1 Δx j

q jþ1 Δx j n n n 2 − 1−θ θ ðQ jþ1 −Q j Þ þ C j ðZ jþ1 þ 2Δtθ ,D j ¼ θ Δx j Δx j n n n g∣u∣ n Z j Þ ; E j ¼ 2Δtθ −ðαuÞ j þ ð 2 Þ Δx j ; F j ¼ ðgAÞ jþ1 ; G j ¼ 2Δtθ þ ðαuÞnjþ1 2θC R j 2 n Δx þð g∣u∣2 Þ Δx j ; Ψ j ¼ 2Δtθj ðQ njþ1 þ Q nj Þ− 1−θ ½ðαuQ Þnjþ1 −ðαuQ Þnj Þ− 1−θ θ θ 2θC R jþ1

weight coefficient; C j ¼

jþ 2

ðgAÞnjþ1 ðZ njþ1 −Z nj Þ. 2

The consistent equations at the junction point of river channels were 8 dV > > <∑Qj ¼ dt j > > : E j ¼ ρgZ j þ 1 ρv j 2 þ p j ¼ E 2

ðB3Þ

where Q and Z were water discharge (m3/s) and water level (m) of tributaries flowing through the junction point; V was water storage (m3); E was total mechanical energy per volume at the junction point (kg/m/ s2); ρ was water density (kg/m3); p was atmospheric pressure at water surface (kg/m/s2). Appendix C. Water quality module The dynamics of water quality indices was described as  ∂ðAC Þ ∂ðQC Þ ∂ ∂C S1 −KAC þ þ S2 þ ¼ AD x ∂t ∂x ∂x ∂x

ðC1Þ

where C was water quality concentration of CODMn or NH4-N (mg/l); D was sum of molecular diffusion coefficient (m2/s), turbulent diffusion coefficient (m2/s) and longitudinal dispersion coefficient (m2/s), which approximately equaled to longitudinal dispersion coefficient; K was comprehensive degradation coefficient (d−1). The comprehensive degradation coefficient and the river longitudinal dispersion were calculated as  v K ¼ βðT w −20Þ K 20 þ 0:197J 0:599 R D¼α

 2:1  0:7 B v Ru R u

iþ1 ω j C iþ1 þ η j C iþ1 j−1 þ ξ j C j jþ1 ¼ Ω j

ðC4Þ

where ω, ξ, η and Ω were coefficients which were calculated using var-

2

(

763

ðC2Þ

ðC3Þ

where β was empirical temperature coefficient of K, which was determined through previous studies (Luo and Song, 2000; Chen et al., 2012), βCOD = 1.047, βN = 1.017 in our study; Tw was water temperature (°C); K20 was comprehensive degradation coefficient at 20 °C

iþ1 Δt ðADÞ j−1 − Δx Q iþ1 þ iables at ith time; ω j ¼ −Δt j−1; ξ j ¼ A j Δx2 j j

Δt þ

ðADÞiþ1 j−1 Δx2j

Δt; η j ¼ −

ðADÞiþ1 j Δx2j

Δt; Ω j ¼

ðACÞij −K ij Aijþ1 C ij Δt 2

Q iþ1 j Δx j

þ

Δt þ

ðADÞiþ1 j Δx2j

Sij Δt.

The mass balance equation at the junction points of river networks was ∑ Q j C j ¼ CAs j

dZ dt

ðC5Þ

where QC was load rate of nutrient in river channels (g/s), C was water quality concentration of tributaries flowing through the junction point (mg/l); As and Z were water surface area (m2) and water level (m) at the junction point. References Alrajoula, M.T., Zayed, I.S.A., Elagib, N.A., Hamdi, M.R., 2016. Hydrological, socio-economic and reservoir alterations of Er Roseires dam in Sudan. Sci. Total Environ. 566-567, 938–948. Anhui Statistical Yearbook (2008). 2008. (Beijing: China Statistics Press, In Chinese) Arheimer, B., Brandt, M., 1998. Modelling nitrogen transport and retention in the catchments of southern Sweden. Ambio 27, 471–480. Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large-area hydrologic modeling and assessment: part I. Model development. J. Am. Water Resour. Assoc. 34, 73–89. Cardona, C.M., Martin, C., Salterain, A., Castro, A., San Martin, D., Ayesa, E., 2011. CALHIDRA 3.0 - new software application for river water quality prediction based on RWQM1. Environ. Model. Softw. 26, 973–979. Cea, L., Legout, C., Darboux, F., Esteves, M., Nord, G., 2014. Experimental validation of a 2D overland flow model using high resolution water depth and velocity data. J. Hydrol. 513, 142–153. Chen, Y.C., Liu, Z.W., Zhu, D.J., 2012. Methods and Applications of Hydrodynamic and Water Quality Simulation. Science Press, Beijing (In Chinese). de Saint-Venant, A.J.C.B., 1871. Theorie du mouvement non permanent des eaux, avec application aux crues des rivieres et a l'introduction de marees dans leurs lits. C. R. Hebd. Seances Acad. Sci. 73, 147–154. Domeneghetti, A., Tarpanelli, A., Brocca, L., Barbetta, S., Moramarco, T., Castellarin, A., Brath, A., 2014. The use of remote sensing-derived water surface data for hydraulic model calibration. Remote Sens. Environ. 149, 130–141. Fitzhugh, W., Vogel, R.M., 2011. The impact of dams on flood flows in the United States. River Res. Appl. 27, 1192–1215. GB 11892-1989, 1989. The State Environmental Protection Standards of the People's Republic of China: Water Quality-determination of Permanganate Index (In Chinese). GB 50179-93, 1993. The State Standards of the People's Republic of China: Code for Liquid Flow Measurement in Open Channels (In Chinese). GB/T 22482-2008, 2008. The State Standards of the People’s Republic of China: Standard for Hydrological Information and Hydrological Forecasting (In Chinese). GB/T 50138-2010, 2010. The State Standards of the People's Republic of China: Standards for Water Level Observation (In Chinese). Gitis, V.G., Petrova, E.N., Pirogov, S.A., Yurkov, E.F., 2007. Mathematical modeling of the pollutants overland flow and transport. Autom. Remote. Control. 68 (9), 1643–1653. Gomi, T., Sidle, R.C., Miyata, S., Kosugi, K., Onda, Y., 2008. Dynamic runoff connectivity of overland flow on steep forested hillslopes: scale effects and runoff transfer. Water Resour. Res. 44, W08411. http://dx.doi.org/10.1029/2007WR005894. Hallema, D.W., Moussa, R., 2014. A model for distributed GIUH-based flow routing on natural and anthropogenic hillslopes. Hydrol. Process. 28, 4877–4895. Hamrick, J.M., 1992. A Three-dimensional Environmental Fluid Dynamics Computer Code: Theoretical and Computational Aspects, Special Report 317. The College of William and Mary, Williamsburg, US. Han, D., 2011. A coupled 1-D and 2-D channel network mathematical model used for flow calculations in the middle reaches of the Yangtze river. J. Hydrodyn. Ser. B 23 (4), 521–526. Hao, S.N., 2014. Research on Water Function Zone Non-point Source Pollution Control of Ying River Basin. North China University of Water Resources and Electric Power, Zhengzhou (Master thesis, In Chinese). He, Y.C., Zhang, T.Y., Liu, J.J., 2011. Cause analysis of high flood level at Zhoukou in Shaying River in 2010. Henan Water Resources & South-to-North Water Diversion. 13, pp. 29–30 (In Chinese). HEC (Hydrologic Engineering Center), 1997. HEC-RAS River Analysis System: Hydraulic Reference Manual. Hydrologic Engineering Center, Davis, CA. HJ/T 373-2007, 2007. The State Environmental Protection Standards of the People's Republic of China: Technical Specifications of Quality Assurance and Quality Control for Monitoring of Stationary Pollution Source (In Chinese).

764

X. Zhai et al. / Science of the Total Environment 598 (2017) 749–764

Huai River Basin Comprehensive Planning 2012–2030, 2013. The Huai River Commission of the Ministry of Water Resources, People's Republic of China (In Chinese). Huai River Water Resources Bulletin, 2014. The Huai River Commission of the Ministry of Water Resources, People's Republic of China. 2014 (In Chinese). Kashefipour, S.M., Falconer, R.A., 2002. Longitudinal dispersion coefficients in natural channels. Water Res. 36 (6), 1596–1608. Khu, S., Madsen, H., Di Pierro, F., 2008. Incorporating multiple observations for distributed hydrological model calibration: an approach using a multi-objective evolutionary algorithm and clustering. Adv. Water Resour. 31 (10), 1387–1398. Kim, J., Warnock, A., Ivanov, V.Y., Katopodes, N.D., 2012. Coupled modeling of hydrologic and hydrodynamic processes including overland and channel flow. Adv. Water Resour. 37, 104–126. Lian, Y.Q., Chan, I.C., Singh, J., Demissie, M., Knapp, V., Xie, H., 2007. Coupling of hydrologic and hydraulic models for the Illinois River Basin. J. Hydrol. 344 (3), 210–222. Lopes, L.F.G., Carmo, J.S.A.D., Cortes, R.M.V., Oliveira, D., 2004. Hydrodynamics and water quality modelling in a regulated river segment: application on the instream flow definition. Ecol. Model. 173 (2), 197–218. Luo, W.S., Song, X.Y., 2000. Water Environment Analysis and Forecasting. Wuhan University Press, Wuhan (In Chinese). Ma, F., Ye, A.Z., GongW, Mao Y.N., Miao, C.Y., Di, Z.H., 2014a. An estimate of human and natural contributions to flood changes of the Huai River. Glob. Planet. Chang. 119, 39–50. Ma, Z.K., Fan, Z.W., Zhang, M., Su, Y.L., 2014b. Flood risk control of dams and dykes in middle reach of Huaihe river. Water Sci. Eng. 7 (1), 17–31. Mateo, C.M., Hanasaki, N., Komori, D., Tanaka, K., Kiguchi, M., Champathong, A., Sukhapunnaphan, T., Yamazaki, D., Oki, T., 2014. Assessing the impacts of reservoir operation to floodplain inundation by combining hydrological, reservoir management, and hydrodynamic models. Water Resour. Res. 50, 7245–7266. Maxwell, R.M., Putti, M., Meyerhoff, S., Delfs, J.O., Ferguson, I.M., Ivanov, V., Kim, J., Kolditz, O., Kollet, S.J., Kumar, M., Lopez, S., Niu, J., Paniconi, C., Park, Y.J., Phanikumar, M.S., Shen, C.P., Sudicky, E.A., Sulis, M., 2014. Surface-subsurface model intercomparison: a first set of benchmark results to diagnose integrated hydrology and feedbacks. Water Resour. Res. 50, 1531–1549. McManamy, R.A., 2014. Quantifying and generalizing hydrologic responses to dam regulation using a statistical modeling approach. J. Hydrol. 519 (Part A), 1278–1296. Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Binger, R.L., Harmel, R.D., Veith, T.L., 2007. Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans. ASABE 50 (3), 885–900. Morling, K., Herzsprung, P., Kamjunke, N., 2017. Discharge determines production of decomposition of and quality changes in dissolved organic carbon in pre-dams of drinking water reservoirs. Sci. Total Environ. 577, 329–339. Novotny, V., Chesters, G., 1981. Handbook of Nonpoint Pollution: Sources and Management. Van Nostrand Reinhold Company. O'Sullivan, J.J., Ahilan, S., Bruen, M., 2012. A modified Muskingum routing approach for floodplain flows: theory and practice. J. Hydrol. 470-471, 239–254. Paiva, R.C.D., Collischonn, W., Buarque, D.C., 2013. Validation of a full hydrodynamic model for large-scale hydrologic modelling in the Amazon. Hydrol. Process. 27, 333–346. Paz, A.R., Bravo, J.M., Allasia, D., Collischonn, W., Tucci, C.E.M., 2010. Large-scale hydrodynamic modeling of a complex river network and floodplains. J. Hydrol. Eng. 15 (2), 152–165. Poff, N.L., Hart, D.D., 2002. How dams vary and why it matters for the emerging science of dam removal. Bioscience 5, 660–668. Preissmann, A., 1961. Propagation des intumescences dans les canaux et les rivières. First Congress of the French Association for Computation, Grenoble, France, pp. 433–442. Remo, J.W.F., Pinter, N., 2007. Retro-modeling the Middle Mississippi river. J. Hydrol. 337 (3), 421–435. Rinaldo, A., Bertuzzo, E., Botter, G., 2005. Nonpoint source transport models from empiricism to coherent theoretical frameworks. Ecol. Model. 184 (1), 19–35. Singh, V.P., 1988. Hydrologic Systems, Vol 1: Rainfall-Runoff Modeling. Prentice Hall, Englewood Cliffs, New Jersy, US. Shi, P., Chen, C., Srinivasan, R., Zhang, X.S., Cai, T., Fang, X.Q., Qu, S.M., Chen, X., Li, Q.F., 2011. Evaluating the SWAT model for hydrological modeling in the Xixian watershed and a comparison with the XAJ model. Water Resour. Manag. 25, 2595–2612. SL 20-92, 1992. The Ministry of Water Resources of the People's Republic of China: Measurement of Discharge by Hydraulic Structures (In Chinese). SL 21-2006, 2006. Industrial Standard of the People's Republic of China: Standard for Observations of Precipitation (In Chinese). SL 257-2000, 2000. Industrial Standard of the People's Republic of China: Code for Waterway Survey (In Chinese). Song, J.Y., Xia, J., Zhang, L.P., Wang, Z.H., Wan, H., She, D.X., 2016. Streamflow prediction in ungauged basins by regressive regionalization: a case study in Huai River Basin, China. Hydrol. Res. 47 (5), 1053–1068.

Suwarnoa, D., Löhr, A., Kroezea, C., Widianarkob, B., Strokal, M., 2014. The effects of dams in rivers on N and P export to the coastal waters in Indonesia in the future. Sustainability Water Qual. Ecol. 3-4, 55–66. Tarquis, A.M., de Lima, J.L.M.P., Krajewski, W.F., Cheng, Q., Gaona'h, 2011. Preface “nonlinear and scaling processes in hydrology and soil science”. Nonlinear Process. Geophys. 18, 899–902. Tuset, J., Vericat, D., Batalla, R.J., 2016. Rainfall, runoff and sediment transport in a Mediterranean mountainous catchment. Sci. Total Environ. 540, 114–132. Warming, R.F., Hyett, B.J., 1974. The modified equation approach to the stability and accuracy analysis of finite-difference methods. J. Comput. Phys. 14 (2), 159–179. Warren, I.R., Bach, H.K., 1992. MIKE 21: a modelling system for estuaries, coastal waters and seas. Environ. Softw. 7 (4), 229–240. Wu, W.M., Vieira, D.A., Wang, S.S.Y., 2004. One-dimensional numerical model for nonuniform sediment transport under unsteady flows in channel networks. J. Hydraul. Eng. 130 (9), 914–923. Xia, J., Zhang, Y.Y., Zhan, C.S., Ye, A.Z., 2011. Water quality management in China: the case of the Huai River Basin. Int. J. Water Resour. Dev. 27 (1), 167–180. Xia, J., Zhang, Y.Y., Zhao, C.S., Bunn, S.E., 2015. Bioindicator assessment framework of river ecosystem health and the detection of factors influencing the health of the Huai River Basin, China. J. Hydrol. Eng.:1943–5584 http://dx.doi.org/10.1061/(ASCE)HE (0000989, 04014008). Xia, J., Zhao, C.S., Liu, M., Wang, G.S., Zhang, Y.Y., Liu, Y., 2008. Impact assessment of dams & flood gates projects of Huaihe river on river ecosystem: a case study of the Bengbu site. J. Nat. Resour. 23 (1), 48–60 (In Chinese). Xia, J., 2002b. A system approach to real-time hydrologic forecast in watersheds. Water Int. 27 (1), 87–97. Xia, J., 1991. Identification of a constrained nonlinear hydrological system described by Volterra functional series. Water Resour. Res. 9, 2415–2420. Xia, J., 2002a. The Systematic Theory and Method of Nonlinear Hydrology. Wuhan University Press, Wuhan (In Chinese). Yan, M., Kahawita, R., 2000. Modelling the fate of pollutant in overland flow. Water Res. 34 (13), 3335–3344. Yang, X.Y., Liu, Q., Fu, G.T., He, Y., Luo, X.Z., Zheng, Z., 2016. Spatiotemporal patterns and source attribution of nitrogen load in a river basin with complex pollution sources. Water Res. 94, 187–199. Yi, Y.J., Tang, C.H., Yang, Z.F., Chen, X., 2014. Influence of Manwan Reservoir on fish habitat in the middle reach of the Lancang River. Ecol. Eng. 69, 106–117. Zhai, X.Y., Xia, J., Zhang, Y.Y., 2014. Water quality variation in the highly disturbed Huai River Basin, China from 1994 to 2005 by multi-statistical analyses. Sci. Total Environ. 496, 594–606. Zhang, M.L., Shen, Y.M., Guo, Y.K., 2008. Development and application of a eutrophication water quality model for river networks. J. Hydrodyn. Ser. B 20 (6), 719–726. Zhang, Y.Y., Arthington, A.H., Bunn, S.E., Mackay, S., Xia, J., Kennard, M., 2012. Classification of flow regimes for environmental flow assessment in regulated river: the Huai River Basin, China. River Res. Appl. 28, 989–1005. Zhang, Y.Y., Shao, Q.X., Xia, J., Bunn, S.T., Zuo, Q.T., 2011b. Changes of flow regimes and precipitation in Huai River Basin in the last half century. Hydrol. Process. 25, 246–257. Zhang, Y.Y., Shao, Q.X., Ye, A.Z., Xing, H.T., Xia, J., 2016. Integrated water system simulation by considering hydrological and biogeochemical processes: model development, with parameter sensitivity and autocalibration. Hydrol. Earth Syst. Sci. 20, 529–553. Zhang, Y.Y., Xia, J., Liang, T., Shao, Q.X., 2010. Impact of water projects on river flow regimes and water quality in Huai River Basin. Water Resour. Manag. 24, 889–908. Zhang, Y.Y., Xia, J., Shao, Q.X., Zhai, X.Y., 2013. Water quantity and quality simulation by improved SWAT in highly regulated Huai River Basin of China. Stoch. Env. Res. Risk A. 27 (1), 11–27. Zhang, Y.Y., Xia, J., Shao, Q.X., Zhang, X., 2011a. Experimental and simulation studies on the impact of sluice regulation on water quantity and quality processes. J. Hydrol. Eng. 17 (4), 467–477. Zhang, Y.Y., Zhai, X.Y., Shao, Q.X., Yan, Z.Q., 2015b. Assessing temporal and spatial alterations of flow regimes in the regulated Huai River Basin, China. J. Hydrol. 529, 384–397. Zhang, Z.Y., Huang, K.L., Huang, Y.L., Hong, H.S., 2015a. Streamflow variability response to climate change and cascade dams development in a coastal China watershed. Estuar. Coast. Shelf 166 (Part B), 209–217. Zingales, F., Marani, A., Rinaldo, A., Bendoricchio, G., 1984. A conceptual model of unitmass response function for nonpoint source pollutant runoff. Ecol. Model. 26, 285–311. Zoch, R.T., 1934. On the relation between rainfall and streamflow. U.S. Dept. of Commerce, Monthly Weather Review, Part I. 62, pp. 315–322.