Modeling the mass fluxes and transformations of nutrients in the Pearl River Delta, China

Journal of Marine Systems 78 (2009) 146–167

Contents lists available at ScienceDirect

Journal of Marine Systems j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j m a r s y s

Modeling the mass fluxes and transformations of nutrients in the Pearl River Delta, China Jiatang Hu, Shiyu Li ⁎ School of Environmental Science and Engineering, Sun-Yat Sen University, Guangzhou, 510275, China

a r t i c l e

i n f o

Article history: Received 5 November 2008 Received in revised form 9 April 2009 Accepted 4 May 2009 Available online 8 May 2009 Regional index terms: China South China Sea Pearl River Delta Keywords: Biochemical oxygen demand Nutrients Physical–biological model Fluxes Biogeochemical cycle Estuarine dynamics

a b s t r a c t Over recent years, accelerated anthropogenic nutrient discharges have exerted great pressure on the water quality management in the Pearl River Delta (PRD), China. There is a concern about the eutrophication processes and hypoxia in this region. A better understanding of the origins and transport of nutrients is required before accurate prediction of impacts of nutrients on eutrophication and hypoxia in the PRD can be anticipated. Therefore a coupled physical–biological model is developed to simulate the fluxes and transformations of nutrients in the PRD. The coupled model combines a one-dimensional model for the river network (called the RNPRD) and a three-dimensional model for the Pearl River Estuary (PRE), which are both physical–biological models. The model is calibrated and validated to different sets of field data. The model results of water surface elevation, discharges, salinity, suspended sediment and water quality variables are in reasonable agreement with the observational data, suggesting that the model is robust enough to capture the physical and biogeochemical dynamics in the PRD. Also, the fluxes and transformations of carbonaceous biochemical oxygen demand (CBOD), ammonia nitrogen (NH3), nitrate plus nitrite nitrogen (NO23) and inorganic phosphorus (IP) in July 1999 (wet season) are explored and discussed. Results show that the RNPRD act as a source for NO23, but a sink for CBOD, NH3 and IP that consumes 50%, 37% and 11% of their external loads, respectively. The riverine fluxes of nutrients exported from the RNPRD to the PRE are generally controlled by high river discharge and significantly contributed by upstream inputs. The riverine fluxes are the largest inputs of nutrients to the PRE. The PRE also behaves as a source for NO23, but a sink for CBOD, NH3 and IP that consumes 90%, 80% and 16% of their external loads, respectively. The estuarine fluxes of nutrients exported from the PRE to the South China Sea are significantly contributed by the external and internal sources of nutrients in the PRE. In the RNPRD, the transformations of CBOD, NH3 (also NO23) and IP are dominated by carbonaceous oxidation, nitrification and deposition, respectively. Regarding the PRE, carbonaceous oxidation, nitrification and phytoplankton uptake are identified as the dominant processes with respect to CBOD, NH3 (also NO23) and IP. Unlike the RNPRD, the phytoplankton dynamics and internal sources of nutrients play an important role in the nutrient budgets in the PRE. Also, seasonal variations of the nutrient budgets in the PRD are discussed. Model results indicate that the dry season and wet season have a similar feature in terms of transformations of nutrients, but show significant seasonal variations in terms of nutrient fluxes. At the same time, the PRE is compared to the Changjiang and Mississippi Rivers with regard to differences in nutrient inputs between these similar river-dominated systems. © 2009 Elsevier B.V. All rights reserved.

1. Introduction The Pearl River Delta (PRD) is a very complicated large-scale estuarine system in China (Fig. 1a). It consists of a tidal river network (called the River-network in the PRD) and an estuary (called the Pearl River Estuary). In recent years, the PRD region has become one of the most densely populated and economically developed regions in China. Consequently, the water body of the PRD receives a high load of

⁎ Corresponding author. Tel.: +86 20 84113620; fax: +86 20 84110692. E-mail addresses: [email protected] (J. Hu), [email protected] (S. Li). 0924-7963/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jmarsys.2009.05.001

anthropogenic nutrients from increased agricultural activities (Neller and Lam, 1994), fish dike farming (Ruddle and Zhong, 1988) and sewage effluents (Hills et al., 1998). This increase in nutrients is likely to result in serious environmental issues, such as eutrophication, harmful red tides and hypoxia. The water quality has been extensively examined in the Pearl River Estuary (PRE), indicating that the estuary exhibits some symptoms of eutrophication and low dissolved oxygen (Yin et al., 2001; Huang et al., 2003; Tang et al., 2003; Yin et al., 2004a, b; Dai et al., 2006; Harrison et al., 2008). Furthermore, the Rivernetwork in the PRD (RNPRD), which is mainly comprised of the Xijiang River, Beijiang River and Dongjiang River (Fig. 1a), is the largest and most complicated tidal river network system in China. A large

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

147

Fig. 1. Maps showing (a) the Pearl River Delta (PRD) coastline, bottom topography, major rivers, major cities and monitoring stations in the river network (RNPRD), and (b) longitudinal circulation in the RNPRD, the Pearl River Estuary (PRE) and the South China Sea (SCS).

amount of nutrients from these rivers and wastewater discharges in the PRD transports through multiple river channels in the RNPRD, passes to the PRE through eight river outlets (Fig. 1a), and ultimately transports to the South China Sea (SCS). The transport of nutrients is

controlled by multiple forcing mechanisms including complicated topography, river discharges, monsoon winds, tides and coastal currents (Wong et al., 2003a,b; Dong et al., 2004; Mao et al., 2004), in association with complex biogeochemical processes (Cai et al.,

148

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

2004). These processes can result in major transformations in the amount and chemical nature of nutrients (Tappin, 2002; Tappin et al., 2003). Therefore it is important to provide an improved quantitative understanding of the fluxes and transformations of nutrients in the PRD, in order to assess their potential impacts on the water quality and hence provide guidelines in settling goals of pollutant reduction to achieve water quality standards. In recent decades, numerical modeling techniques have been widely applied to estuaries such as the Humber estuarine system (e.g., Tappin et al., 2003) and the Chesapeake Bay (e.g., Cerco and Cole, 1993; Xu and Hood, 2006) to simulate both physical and biogeochemical processes and study the interactions between them. The complexity of the coupled physical–biological models ranges from one-dimensional (1-D) (e.g., Tappin et al., 2003) to fully three-dimensional (3-D) models (e.g., Cerco and Cole, 1993; Zheng et al., 2004; Xu and Hood, 2006). When such numerical models are applied to the estuaries they can provide a better estimation on the fluxes of nutrients associated with their transfers and transformations (Jay et al., 1997). Some well-established models have been applied to calculate mass balance for nutrients and suspended particles in coastal river-dominated systems (as the PRD) such as the Northern Adriatic Sea (e.g., Spillman et al., 2007), the Ringkøbing Fjord (e.g., Håkanson et al., 2007) and the Humber estuarine system (e.g., Tappin et al., 2003). Results from these studies indicate a close coupling of physical and biogeochemical processes over a range of space and time scales, which demonstrates the importance of representing estuarine processes in dynamic process-based physical–biological models. There have been many dynamic models developed to describe chemical and biological transfers and transformations in the Pearl River Delta (PRD) (e.g., Guan et al., 2001a,b; Jos et al., 2007), in addition to water and suspended sediments (e.g., Chen et al., 1999; Wong et al., 2003a,b; Hu and Li, 2008). However, previous studies have mainly focused on the hydrodynamic features and water quality variations in the PRD. Little is known about the fluxes, transformations and ultimate fate of nutrients within the PRD. Furthermore, the majority of modeling work in the past has focused on the river network (RNPRD) or the Pearl River Estuary (PRE) separately, with few studies (e.g., Hu and Li, 2008; Jos et al., 2007) integrating these two regions as an entity. Since the RNPRD is comprised of narrow and shallow river channels and primarily dominated by river discharges (Fig. 1b), the physical and biogeochemical processes in the RNPRD are simulated by a cross-sectinoally integrated 1-D model. Regarding the PRE, it is a complex system where several different circulation regimes coexist and various types of fronts form between the circulation regimes such as the river plume front and coastal temperature front (Wong et al., 2003a). The value of using 3-D models for the PRE has been clearly demonstrated previously (Wong et al., 2003a,b; Hu and Li, 2008), in which the 3-D circulation in the PRE with respect to the combined forcing of multiple mechanisms is extensively explored and discussed. As the RNPRD and the PRE are affecting each other and closely interrelated (Fig. 1), it is essential to integrate them as an entity. This necessitates application of a 1-D and 3-D coupled model, which has the advantage over previous studies that only included the PRE or the RNPRD and neglected the dynamic variations of material exchange between the RNPRD and the PRE through the eight river outlets (e.g. Guan et al., 2001a,b; Wong et al., 2003a, b). In addition, a 1-D and 3-D coupled model has the advantage over a fully 3-D model for the whole area in terms of computational efficiency and model grid generation. Overall, the major objectives of this study are: • to propose a 1-D (for the RNPRD) and 3-D (for the PRE) coupled model, which includes physical and biogeochemical processes; observations are used to evaluate the model performance; • to simulate the fluxes of nutrients (nitrogen, phosphorus and CBOD) passing through the RNPRD, the PRE, and the SCS, respectively; • to construct the nutrient budgets for the entire study area, and quantify the contributions from external and internal sources of

nutrients to their budgets, and characterize the key biogeochemical processes in the transfers and transformations of nutrients; the internal sources of nutrients are defined as internal nutrient inputs from primary production, benthic releases, conversion from ammonia nitrogen via nitrification, and generation of inorganic nutrients from organic matter via mineralization or bacterial decomposition. 2. Materials and methods 2.1. Model description 2.1.1. Physical and suspended sediment transport model A 1-D model (called Riv1D) has been developed for the river network (RNPRD) (Hu and Li, 2008). The model has been expanded and refined in recent years. To date, it can be used to simulate the hydrodynamics, salinity distribution, suspended sediment (SS) transport and water quality processes in well-mixed rivers and shallow estuaries. The hydrodynamic module of Riv1D is based on the solution of Saint Venant equations of mass and momentum conservation. These equations are solved through a Preissmann implicit scheme and an iterative approach. For water quality aspects, Riv1D solves the governing advection–diffusion equation, with additional source and sink terms to account for external loads and transformations of water quality constituents. A control volume scheme has been applied to solve the advection–diffusion equation. Modules for salinity and SS dynamics are also incorporated within Riv1D, using the same model structure and computational framework as the water quality module. The governing hydrodynamic equations (Cunge et al., 1980) are shown as follows: 1 AQ AZ + = qL B Ax At 2

AQ A Q + At Ax A

! + gA

ð1Þ AZ Q jQ j +g =0 Ax ACS2 R

ð2Þ

where Q is the cross-section averaged discharge; B is the water surface width; Z is the water surface elevation; qL is the lateral inflow; A is the cross-section area; g is the gravity acceleration; CS is the Chezy resistance coefficient; and R is the hydraulic radius. The 1-D SS transport model is based on the principle of mass conservation. The governing equation can be expressed by   Að AC1 Þ AðQC1 Þ A AC + − AEx 1 − Sc − Wc = 0 At Ax Ax Ax

ð3Þ

where C1 is the SS concentration (SSC); Ex is the coefficient of longitudinal dispersion; Sc is the internal sources and sinks; and Wc is the external inputs from point sources, non-point sources, fall-line loads and atmospheric input. Fall-line loads are classified as those introduced from rivers, representing the amount of nutrients entering the study region from upland areas. For SS aspects, Sc represents sediment deposition and resuspension processes. A fully 3-D estuarine and coastal ocean model coupled with a sediment transport module (Blumberg, 2002), namely, ECOMSED, is used to simulate the physical and cohesive sediment dynamics in the Pearl River Estuary (PRE) and its adjacent coastal waters. The hydrodynamic module of ECOMSED solves the Navier–Stokes equations for a water body with a free surface, under Boussinesq and hydrostatic approximations. It incorporates the Mellor and Yamada's (1982) level 2.5 turbulent closure sub-model, which was modified by Galperin et al. (1988). The prognostic variables in the hydrodynamic module are the free surface elevation, three components of velocity, turbulence kinetic energy, turbulence macroscale, temperature and salinity. The transport and fate of cohesive sediments can be simulated with the sediment transport module of ECOMSED. Cohesive sediment

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

149

Fig. 2. Schematic of water quality model used for the Pearl River Delta (PRD), a revised version of Fig. 3 in Zheng et al. (2004).

dynamics inherent in the sediment module include sediment transport, deposition and resuspension. The mechanisms of resuspension and deposition depend upon the bottom shear stress induced at the sediment–water interface. A detailed description of the hydrodynamic and sediment modules can be found in literature (Blumberg, 2002; Chen and Wang, 2008). 2.1.2. Water quality model The 3-D water quality model is based on the Row-Column AESOP (Advanced Ecological Systems Modeling Program) developed by HydroQual (Fitzpartick, 2004). The Row-Column AESOP (RCA) constitutes a complex of five interacting systems: phytoplankton dynamics, nitrogen cycle, phosphorus cycle, carbon cycle, and dissolved oxygen balance. RCA originally incorporates twenty-six state variables. The governing equation for each state variable can be described as follows:     AC AC AC AC A AC A AC +u +v +w = Ah + Ah At Ax Ay Az Ax Ax Ay Ay   A AC + Kh + S + W0 Az Az

ð4Þ

where C is the concentration of a water quality state variable; u, v, w are the water velocity components in the x, y and z directions, respectively; Ah and Kh are the horizontal and vertical diffusion coefficients; S is the internal sources and sinks of the water quality state variable; W0 is the external loads from point sources, non-point sources, fall-line loads and atmospheric input of the water quality state variable. Multiple forms of nutrients and organic carbon are modeled as a state variable in RCA (Fitzpartick, 2004). It is difficult to apply the original RCA to the Pearl River Delta because of the problem of data availability in China. Therefore modifications are made to RCA in view of data limitations in this study. First, the modified model includes eight state variables rather than twenty-six. New variables are phytoplankton carbon (PHYT), organic phosphorus (OP), inorganic phosphorus (IP),

and organic nitrogen (ON), ammonia nitrogen (NH3), nitrate plus nitrite nitrogen (NO23), carbonaceous biochemical oxygen demand (CBOD) and dissolved oxygen (DO). Two available groups of phytoplankton, Green and Diatoms, are simulated in the model, representing riverine and marine algae, respectively. Two phases of IP are considered, including ortho-phosphate (OPO4) and particle-sorbed phosphate (PO4SS). The kinetic processes in the modified model generally follow water quality analysis simulation program (called WASP5; Ambrose et al., 1993; Zheng et al., 2004), as shown in Fig. 2. Second, the effect of SSC on the light intensity for phytoplankton growth is considered because high sediment loads in estuaries can lead to rapid light attenuation in estuarine waters, which limits primary production. Meanwhile, the model includes the process of adsorption–desorption between SS and NH3, in addition to the sorption between SS and OPO4. A linear isotherm is adopted to model the distribution over dissolved and sorbed phases for NH3, while the Langmuir equilibrium isotherm is selected for IP. The formulation of equilibrium adsorption content regarding OPO4 and NH3 is shown in Table 1. The kinetic processes and water quality state variables in the water quality module of Riv1D are the same as those in the modified RCA model, i.e., five systems and eight state variables are included in Riv1D. The governing equation of each water quality state variable can be written as Eq. (3), with different sources or sink terms to account for pollutant loads and specific decay and transformations of nutrients. The equations for the water quality model are summarized in Table 1. A brief description of the kinetic processes is given below. 2.1.2.1. Phytoplankton kinetics. The growth rate of phytoplankton is a complicated function of nutrients availability, solar radiation and ambient water temperature. The specific growth rate is represented as the multiplication of maximum phytoplankton growth rate and limitation functions of each factor. Mechanisms that contribute to the phytoplankton loss mainly include endogenous respiration, grazing by zooplankton and settling to bottom sediment.

150

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

2.1.2.2. Phosphorus cycle. In the phosphorus system kinetics, the main source of OP is recycled from the phytoplankton biomass pool in the water column through endogenous respiration and predatory grazing. OP is converted to IP at a temperature-dependent mineralization rate, and its particulate fraction will sink and deposit on the bottom sediment. Available IP is utilized by phytoplankton for growth. Three internal sources for IP are (1) recycled from the phytoplankton biomass pool in the water column, (2) converted from OP via mineralization or bacterial decomposition, and (3) released from the bottom sediment. In addition, OPO4 interacts with PO4SS via an adsorption–desorption process. PO4SS will sink and deposit on the benthic sediment. 2.1.2.3. Nitrogen cycle. The nitrogen kinetics are basically similar to the phosphorus kinetics. Nitrogen is recycled from the phytoplankton biomass pool in the water column to the dissolved and particulate ON pools and to the NH3 pool. Particulate ON will settle and deposit on the benthic sediment. ON is converted to NH3 at a temperaturedependent mineralization rate, and NH3 is then nitrified at a temperature- and oxygen-dependent rate. Under low DO condition, NO23 can be denitrified at a temperature-dependent rate. NH3 and NO23 are consumed by phytoplankton uptake. NH3 is the preferred form of inorganic nitrogen for phytoplankton growth. Also, the benthic nutrient fluxes are internal sources for NH3 and NO23. In addition, particulate NH3 adsorbed on SS will sink and deposit on the bottom sediment. 2.1.2.4. CBOD kinetics. The internal source of CBOD is recycled from phytoplankton respiration and death. The internal sinks for CBOD include carbonaceous oxidation and settling of particulate carbonaceous material to the benthic sediment. The denitrifcation reaction also provides a sink for CBOD under low DO condition. 2.1.2.5. DO balance. DO is one of the most important water quality indicators (Zheng et al., 2004) as the availability of oxygen controls all life (in particular fish) in the water. In addition, DO is closely related to eutrophication and hypoxia which are increasing threat to coastal systems. The sources of DO considered in the model are reaeration and phytoplankton photosynthesis. The sinks of DO include phytoplankton respiration, nitrification, oxidation of CBOD, and sediment oxygen demand (SOD). 2.2. Dynamic integration of the 1-D and 3-D model The 1-D model (Riv1D) and 3-D model (ECOMSED-RCA) are dynamically coupled to create a single model in order to perform an overall simulation on the dynamics of the Pearl River Delta (PRD). In the 1-D and 3-D coupled model, quantities of the hydrodynamic and water quality state variables are exchanged across coupling crosssections (locations of the eight river outlets, see Fig. 1a) at each time step. For the hydrodynamic modeling, the 1-D model sends its simulated discharge at the coupling cross-sections to the 3-D model as its upstream boundary condition. As a feedback, the 3-D model sends its simulated water surface elevation at the coupling cross-sections to the 1-D model as its downstream boundary condition. For the water quality modeling, simulated mass fluxes (volume flux multiplied by concentrations) of the state variables are exchanged across

the coupling cross-sections at each time step. The exchanges are determined by the instantaneous direction of flow through the coupling cross-sections. The simulated mass fluxes from the 3-D model are transferred to the 1-D model when the flow is directed landwards (i.e. from the Pear River Estuary to the river network), whereas the simulated mass fluxes from the 1-D model are transferred to the 3-D model when the flow is directed seawards. The approach for coupling the 1-D and 3-D model is illustrated in Fig. 3. The major challenge of coupling the 1-D and 3-D model is to reproduce the realistic physical and biogeochemical processes in the 1-D domain and 3-D domain, in addition to depicting the material exchange between these two regions. The coupling approach shown in Fig. 3 has been applied successfully to the PRD in previous study (Hu and Li, 2008).

2.3. Model set-up 2.3.1. Model design Fig. 4a shows the model domain for the river network (RNPRD) and model grids for the Pearl River Estuary (PRE) and its adjacent shelf. The RNPRD is discretized into 299 reaches and 1726 crosssections, with five upstream boundaries (Fig. 1a) including Gaoyao at the Xijiang River, Shijiao at the Beijiang River, Boluo at the Dongjiang River, Laoyagang at the Liuxi River and Shizui at the Tanjiang River. The eight river outlets, including Humen, Jiaomen, Hongqili, Hengmen, Modaomen, Jitimen, Hutiaomen and Yamen, are specified as the exchange cross-sections in the 1-D and 3-D coupled model. The first four outlets are defined as the eastern four river outlets (Fig. 4b), while the last four are defined as the western four river outlets. The 3-D part of the coupled model is configured to cover the PRE and its adjacent coastal waters. An orthogonal curvilinear grid is designed to resolve the complex coastline, with a total of 183 by 186 points. There are 6 equidistant sigma layers in the vertical to represent the irregular bottom topography. The time steps are 40 s for the hydrodynamic and SS models, and 120 s for the water quality model. The 1-D model and the 3-D model have the same time steps. For the hydrodynamic modeling, the real-time freshwater discharges with zero salinity are introduced for the upstream boundaries in the RNPRD. Some small waterways in the RNPRD are not included in the 1-D model because their topography data is not available. Water from these waterways is treated as lateral inflows. However, their contributions to the RNPRD are ignored since there are no observations of water discharges for them. The 3-D part of the coupled model is driven by four dominant tidal constituents (M2, S2, K1 and O1) at the open sea boundaries, with a uniform salinity boundary condition (Jos et al., 2007). Winds are specified in the model using hourly observed data from the Taipa Grande Station (Fig. 1a) provided by the Macao Meteorological and Geophysical Bureau. In the model domain, southwesterly winds dominate during summer and northeasterly winds dominate during winter. For the SS and water quality modeling, measured data of the monitoring stations located at the upstream boundaries or nearest to the upstream boundaries have been used to assign boundary conditions in the 1-D part of the coupled model. In addition, the boundary conditions at the open sea boundaries are derived from data collected in the 1990s at the open sea stations (Jos et al., 2007). Initial conditions are derived by spin-up simulations which run for 30 days. The initial conditions are replaced by results at

[−139.34411+1.575701105T−1−6.642308107T−2+1.2438001010T−3\kern-34.35pc\lower25pt{{−8.6219491011T−4−Sa(1.767410−2−10.754T−1+2140.7T−2) ] Notes to Table 1 a Ambrose et al. (1993). b Steele (1962). c Fitzpartick (2004). d Di Toro (1978). e Chao et al. (2007).

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

151

Table 1 Summary of equations for the water quality model. Variables

Description/formulation

PHYT OP IP ON NH3 NO23 CBOD DO H I0 T Sa Pchl

Phytoplankton carbon (mg C l− 1) Organic phosphorus (mg P l− 1) Inorganic phosphorus (mg P l− 1) Organic nitrogen (mg N l− 1) Ammonia nitrogen (mg N l− 1) Nitrate plus nitrite nitrogen (mg N l− 1) Carbonaceous biochemical oxygen demand (mg O2 l− 1) Dissolved oxygen (mg O2 l− 1) Water column depth or thickness of the water segment (m) Incident light intensity at the segment surface (ly day− 1) Water temperature (°C) Salinity (ppt) Phytoplankton chlorophyll a concentration (µg l− 1, computed based on PHYT and the ratio of chlorophyll a to carbon) Suspend sediment concentration (mg SS l− 1) Initial IP concentration in the solution (mg P l− 1)

SSC C0 Phytoplankton kinetics Kinetic term Growth rate

Sp = (Gp − Dp) PHYT Gp = Gpmax · fN(N) · fI(I) · fT(T)   OPO4 3 + NO23 fN ðN Þ = min KmN NH + NH3 + NO23 ; KmP + OPO4 h    i I0 − Ke H fI ðI Þ = 2:718 − exp − II0s Ke H exp − Is e   8   < exp −Kb1 Topt − T 2 when TVTopt   fT ð T Þ =   : exp −K T − T 2 when T N Topt b2 opt

Nutrient limitation functiona Light limitation functiona, b Temperature dependencyc Light attenuation coefficienta, d

Ke=Kebase + 0.0088Pchl + 0.054P0.67 chl + 0.052SSC

Loss rate

Dp = KPR ΘPR

Phosphorus cycle Kinetic term of OP

SOP = apc Dp fOP PHYT − Km1 Θm1

ðT − 20Þ

VsP H

T − 20Þ + Kgrz Θðgrz

ðT − 20Þ





OP − VsOP ð1 H− fDOP Þ OP   PHYT ðT − 20Þ SIP = apc Dp ð1 − fOP ÞPHYT − apc Gp PHYT + Km1 Θm1 OP KmPc + PHYT

Kinetic term of IP

− Equilibrium adsorption content for phosphorus

+

e

Concentration of phosphorus in dissolved phase

Ne = Cd =

VsIP ð1 − fDIP Þ BOPO4 IP + H H

Kadp Cd 1 + Kadp Cd

e

1 2

" 

C0 −

Nem 1 Kadp

r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  2 1 em − SSC  Nem + − SSC  Nem + 4SSCN C0 + Kadp Kadp

Fraction of dissolved IPe

fDIP =

Nitrogen cycle Kinetic term of ON

SON = anc Dp fON PHYT − Km2 Θm2

Kinetic term of NH3

Kinetic term of NO23

PHYT KmPc + PHYT

Cd C0

ðT − 20Þ





ON − VsON ð1 H− fDON Þ ON   PHYT ðT − 20Þ SNH3 = anc Dp ð1 − fON ÞPHYT − anc Gp PNH4 PHYT + Km2 Θm2 ON KmPc + PHYT   DO VsNH ð1 − fDNH4 Þ BNH3 ðT − 20Þ NH3 + NH3 − −Kni Θni H Knitr + DO H  DO ðT − 20Þ SNO23 = − anc Gp ð1 − PNH4 ÞPHYT + Kni Θni NH3 Knitr + DO   KNO3 BNO23 ðT − 20Þ − Kdn Θdn NO23 + KNO3 + DO H NH3 NO23 ðKmN + NH3 ÞðKmN + NO23Þ 1 1 + Kadn SSC

PHYT KmPc + PHYT

NH3 KmN ðNH3 + NO23ÞðKmN + NO23Þ

Ammonium preference factor

PNH4 =

Distribution over dissolved and sorbed phases for NH3c

fDNH4 =

CBOD kinetics

   ðT − 20Þ ðT − 20Þ T − 20Þ + Kgrz Θðgrz SCBOD = aoc KPR ΘPR PHYT − Kdc Θdc

Kinetic term of CBOD

DO balance

#

+

 DO CBOD KBOD + DO   VsCBOD ð1 − fDCBOD Þ 5 32 K ðT − 20Þ NO3 CBOD −  K Θ NO23 − H 4 14 dn dn KNO3 + DO

32 48 + anc ð1 − PNH4 Þ PHYT 12 14   32 DO ðT − 20Þ ðT − 20Þ PHYT − Kdc Θdc − KPR ΘPR CBOD 12 KBOD + DO   64 DO SOD ðT − 20Þ ðT − 20Þ Θ NH3 − − Kni Θni 14 Knitr + DO H SOD

SDO = Ka ΘðaT

DO saturation concentrationc

Cs = exp −139:34411 + 1:575701  105  T − 1 − 6:642308  107  T − 2 + 1:243800  1010  T − 3   − 8:621949  1011  T − 4 −Sa 1:7674  10 − 2 − 10:754  T − 1 + 2140:7  T − 2

½

− 20Þ



Kinetic term of DO

ðCs − DOÞ + Gp



152

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

(1) Fig. 3. Approach for coupling the 1-D and 3-D model. Q(1) and F(1) represent simulated discharge, water surface elevation and mass flux (volume flux multiplied by i , Zi i (3) concentration of state variables) at the ith coupling cross-section from 1-D model, respectively, while Q(3) and F(3) are those from 3-D model. i , Zi i

Day 30 and the 1-D and 3-D coupled model runs again. These processes are repeated for 4 times when the difference between the final results and previous results of nutrient budgets is within 1%. 2.3.2. Pollutant loads in the water quality model In addition to the inflows from upstream boundaries, there are two types of pollutant loads included in the water quality model: pointsource loads and non-point-source loads. Both the point and nonpoint-source loads from the Pearl River Delta (PRD) region were estimated by the South China Institute of Environmental Sciences (Jos et al., 2007). Parameters included in the pollution loads are BOD5, NH3, Total Kjeldahl Nitrogen (TKN) and total phosphorus (TP). The pollutant loads from Macao and Hong Kong (Fig. 1a) are provided by the Environment Council of Macao Special Administrative Region and the Environment Protection Department of Hong Kong Special Administrative Region, respectively. The pollution loads for the modeled water quality state variables are converted from BOD5, NH3, TKN (ON + NH3) and TP (OP + IP) based on the stoichiometric ratios in organic waster materials (San Deigo-McGlone et al., 2000). The external loads for BOD5, NH3, total nitrogen (TN) and TP in the wet season and dry season are summarized in Fig. 5, showing a significant seasonal fluctuation. The wet (rainy) season is defined as April– September and the dry season as October–March (as Cai et al., 2004). The external loads in the wet season are much larger than those in the dry season, which arises from high inflows from the upstream boundaries and non-point-source loads in the wet season. 2.3.3. Parameters used in the water quality model The parameters used in the water quality model are specified based on observations in the PRE and scientific literature (Ambrose et al., 1993; Guan et al., 2001a; Fitzpartick, 2004; Zheng et al., 2004; Lu et al., 2005; Shi et al., 2005; Zhan et al., 2005; Chao et al., 2007; Jos et al., 2007). They are fine-tuned within ranges recommended by literature to produce the best fit between simulated values and field data. The required parameters and their descriptions in the water quality model are listed in Table 2. Although the water quality model incorporates a sediment diagenesis process model (Fitzpartick, 2004) to simulate processes in the benthic sediment, the sediment process model is not used because no data is available to support this activation. Instead, constant values are specified for benthic nutrient fluxes and SOD based on previous measurements (Guan et al., 2001a; Zhan et al., 2005). Future study is needed to represent these processes more accurately. 2.4. Field data used for model calibration and validation Three simulation periods are selected for model calibration and validation purpose, including January 1999 (dry season), July 1999 (wet season) and February 2001 (dry season). In all cases the simulation period is 30 days, which means that two spring tide/neap tide periods are covered. However, there is only 10 days or less of field data available in each simulation period due to data limitation. Two surveys were conducted by the Pearl River Water Resources

Commission on July 16–24, 1999 and February 7–16, 2001. These surveys produced sets of continuous and synchronous data in the river network (RNPRD). The observational data include hourly water surface elevation, discharge and SSC at fifty-nine stations (five stations located at the upstream boundaries). Salinity is available for February 7–16, 2001. Locations of the time-series stations in the RNPRD are shown in Fig. 1a. In the Pearl River Estuary (PRE), an investigation was carried out by Pearl River Estuary Pollution Project on July 17–27, 1999 (Wong et al., 2003a), yielding sets of hydrodynamic and water quality parameters including current speed and direction, salinity, DO, chlorophyll a, nitrate, nitrite, NH3, OPO4, SSC and chemical oxygen demand. Fig. 4b shows the locations of the survey and time-series stations in this summer cruise. Also, water surface elevation obtained from five tidal gauge stations (Fig. 4b) is used for the calibration and validation of the hydrodynamic model. Water quality observations in the wet season and dry season for the years 1998, 1999 and 2000 are used for the calibration and validation of the water quality model. These observations are available for seventy-one locations in the RNPRD (Fig. 4a). They are actually averaged from several samples taken at unknown time during the specific season. Detailed datasets for the model calibration and validation are shown in Table 3. 3. Results and discussion 3.1. Model calibration and validation A set of calibration and validation statistics are used to quantitatively assess the performance of the coupled model. Wherever possible, we suggest potential reasons for deficiencies and possible means for further improvement of the model performance. The correlation coefficients (r2) of the model results and observations, the mean errors and the relative errors of the model results are summarized in Table 4. 3.1.1. Model calibration 3.1.1.1. Physical and suspended sediment transport model. The coupled model was run for February 2001 for calibration of the hydrodynamic and SS models. Result shows that the simulated water surface elevation, discharge, salinity and SSC are in reasonable agreement with the observational values (Table 4). The correlation coefficients for water surface elevation and discharge are greater than 0.90, and those for salinity and SSC are 0.82 and 0.56, respectively. The larger error for SSC, relative to the hydrodynamic variables, may be mainly attributed to the simplified parameterization of processes controlling SS dynamics, e.g., a constant critical shear stress and resuspension rate are specified in the model. Also, these processes are too complex to be fully understood based on current studies. As the eight river outlets are the exchange cross-sections in the coupled model, they are the important locations to check if the model is capable of capturing the physical–biogeochemical dynamics in the entire domain and depicting the material exchange between the river network (RNPRD) and the Pearl River Estuary (PRE). Results of two representative river outlets including Humen and Modaomen (Fig. 1a)

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

153

Fig. 4. Maps showing (a) model domain of the 1-D model and survey stations for water quality observations in the river network (RNPRD), and (b) model grids of the 3-D model and monitoring stations in the Pearl River Estuary (PRE). Blue empty triangles ( ) denote locations of the survey stations for water quality observations in the RNPRD. Data are available for seventy-one stations. Red triangles ( ) and blued solid diamonds ( ) are locations of survey stations and time-series stations in the PRE for the summer cruise in July 1999, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

154

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

Fig. 5. Overview of pollutant loads in the wet season (April–September) and dry season (October–March). The ‘PRD DS’ and ‘PRD PS’ terms represent non-point-source loads and point-source loads from the Pearl River Delta (PRD), respectively. The ‘HK + Macau’ term represents pollutant loads from Hong Kong and Macao combined. The ‘upstream’ term represents pollutant loads from five upstream boundaries in the river network (RNPRD), evaluated on the basis of observed nutrient concentrations and river discharges in July 1999 (wet season) and January 1999 (dry season), respectively.

are shown in Fig. 6a–b. These plots reveal that the model results of water surface elevation, discharge and SSC compare reasonably well with the observational values. 3.1.1.2. Water quality model. The coupled model was run for January 1999 for calibration of the water quality model. Fig. 6c shows the simulated and observed NH3, CBOD, TN and TP in the Guangzhou– Foshan (GF series), Zhongshan–Shunde (ZS series) and Jiangmen– Kaiping (JK series) regions (Fig. 4a). Since the observational data were collected at different times of the years, they are shown with model results of monthly-minimum, monthly-averaged and monthly-maximum concentrations. The model results of nutrient and CBOD follow the general patterns shown by the observational data, which show strong spatial gradients for the water quality components, with high concentrations in the Guangzhou–Foshan and Jiangmen–Kaiping regions and low concentrations in the Zhongshan–Shunde region. The model reproduces these gradients well. It can be noticed that only limited field data are available for TP. Insofar as field data are available, there is a reasonable agreement with the model results. However, the model tends to overestimate NH3 and underestimate CBOD in some stations in the Guangzhou–Foshan region (e.g., GF9–GF12). The local pollutant loads from non-point sources in the dry season are a potential cause for this problem, rather than point-source loads and pollutant kinetic processes, which would affect the results in the wet season. Thus, a better representation on the non-point-source loads is needed to further optimize the water quality model. In general, the model results agree reasonably with the observational data in terms of both values and distribution patterns. 3.1.2. Model validation 3.1.2.1. Physical and suspended sediment transport model. For model validation, the coupled model was run for July 1999 using the same set of parameters as used in the calibration process. The simulated water surface elevation, discharge, salinity and SSC agree well with the observational data, with correlation coefficients greater than 0.65 (Table 4). The simulated SSC in the surface layer in the Pearl River Estuary (PRE) is

compared to a satellite remote sensing image of SS (Fig. 7). The model result follows the observed spatial distribution quite closely. Fig. 8a and b shows the simulated and observed water surface elevation, discharge and SSC at Humen and Modaomen. The simulated results generally follow the observed patterns. However, there are also some significant discrepancies between the simulated SSC and observational values at Modaomen. The coupled model is able to reproduce the mean values of SS at Modaomen, but fails to capture its high variability of SS. This phenomenon is not readily explained, but may be linked to the high variability of settling of SS promoted by flocculation (not included in the 1-D model), or related to the simplified parameterization of deposition and resuspension processes. Fig. 8c shows the simulated water and SS fluxes through the eight river outlets calculated from July 16 to July 24, 1999. Good agreements are found for both water and SS fluxes. 3.1.2.2. Water quality model. It appears that the coupled model has a tendency to overestimate dissolved inorganic nitrogen (DIN) and OPO4, which corresponds to the underestimation of chlorophyll a (Table 4). Also, the model has a tendency to underestimate CBOD and overestimate DO. The correlation coefficients of the model results and observational values are highest for DO (0.68) and lowest for chlorophyll a (0.41). It is difficult to simulate the OPO4 and chlorophyll a distribution accurately because their variability is driven by many processes (see subsection 2.1.2; Table 1) and there are many uncertainties associated with these processes. Consequently, the correlation coefficients for OPO4 and chlorophyll a are relatively smaller. Satisfactory agreement can be found for the simulated and observed NH3, CBOD, TN and TP in the Guangzhou–Foshan, Zhongshan– Shunde and Jiangmen–Kaiping regions in the wet season (Fig. 9a). These plots reveal that the model is capable of reproducing reasonable patterns and acceptable magnitudes for the water quality components. Fig. 9b and c shows the comparisons between the simulated results and observations of CBOD, OPO4, NO23 and DO at the timeseries stations C1 and C2 as well as the survey stations in the Pearl River Estuary (PRE) (Fig. 4b). With some exception, the model results follow the observed patterns well in space and time. However, the model tends to overestimate or underestimate the nutrients and DO in

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167 Table 2 Parameters and constants for the water quality model.

Table 3 Calibration and validation dataset for the 1-D and 3-D coupled model.

Parameter Description

Value

Unit

Gpmax Is Topt Kb1

2.0a 250.0b 25.0a 0.004b

day− 1 ly day− 1 °C (°C)− 2

0.006b

(°C)− 2

0.01c 0.04a 0.025a 0.25a 32/12a 0.5c 0.22a 0.075a 0.1a 0.09a 0.2a 2.0b 1.5–2.5d 0.5a 0.5a 0.5a 0.5a 0.5a 0.5a 1.045a

day− 1 day− 1 Unitless Unitless Unitless m− 1 day− 1 day− 1 day− 1 day− 1 day− 1 day− 1 g O2 m−2 day−1 m day− 1 m day− 1 m day− 1 m day− 1 m day− 1 m day− 1 Unitless

1.000a

Unitless

Kb2 KPR Kgrz apc anc aoc Kebase Km1 Km2 Kni Kdn Kdc Ka SOD VsP VsOP VsIP VsON VsNH VsCBOD ΘPR Θgrz Θm1 Θm2 Θni Θdn Θdc Θa ΘSOD KmN KmP KmPc Knitr KNO3 KBOD fOP fON fDOP fDIP fDON fDNH4 fCBOD BNH3 BNO23 BOPO4 Kadp Nem Kadn a b c d e f g h i

Maximum phytoplankton growth rate Optimal light intensity Optimum temperature Temperature correction effect on growth rate below Topt Temperature correction effect on growth rate above Topt Phytoplankton respiration rate Mortality rate due to grazing Ratio of phosphorus to carbon Ratio of nitrogen to carbon Ratio of oxygen to carbon Background light attenuation coefficient OP mineralization rate ON mineralization rate Nitrification rate Denitrification rate CBOD oxidation rate Reaeration rate Sediment oxygen demand Settling velocity of PHYT Settling velocity of particulate OP Settling velocity of sorbed IP Settling velocity of particulate ON Settling velocity of sorbed NH3 Settling velocity of CBOD Temperature coefficient for phytoplankton respiration rate Temperature coefficient for zooplankton grazing rate Temperature coefficient for OP mineralization rate Temperature coefficient for ON mineralization rate Temperature coefficient for nitrification rate Temperature coefficient for denitrification rate Temperature coefficient for CBOD oxidation rate Temperature coefficient for reaeration rate Temperature coefficient for SOD Half saturation constant for nitrogen uptake Half saturation constant for phosphorus uptake Half saturation constant for phytoplankton limitation Half saturation concentration for oxygen limitation of nitrification Half saturation concentration for oxygen limitation of denitrification Half saturation concentration for oxygen limitation of CBOD oxidation Fraction of dead and respired phytoplankton recycled to the OP pool Fraction of dead and respired phytoplankton recycled to the ON pool Fraction of dissolved OP fraction of dissolved IP Fraction of dissolved ON Fraction of dissolved NH3 Fraction of dissolved CBOD Bottom NH3 flux Bottom NO23 flux Bottom OPO4 flux Ratio of adsorption and desorption rate coefficient for phosphorus Maximum concentration of phosphorus in solid phase Partition coefficient for sorbed NH3

Ambrose et al. (1993). Fitzpartick (2004). Jos et al. (2007). Guan et al. (2001a). Zheng et al. (2004). Chao et al. (2007). Zhan et al. (2005). Lu et al. (2005). Shi et al. (2005).

155

a

1.080 1.080a 1.080a 1.080a 1.047a 1.028a 1.080a 0.025a 0.001a 1.0a

Unitless Unitless Unitless Unitless Unitless Unitless Unitless mg N l− 1 mg P l− 1 mg C l− 1

1.0a

mg O2 l− 1

0.1a

mg O2 l− 1

0.5a

mg O2 l− 1

0.65e

Unitless

0.65e

Unitless

f

0.5 Table 1 0.5f Table 1 0.5a 14.4g −1.65g 0.062g 8.25h

Unitless Unitless Unitless Unitless Unitless mg m− 2 day− 1 mg m− 2 day− 1 mg m− 2 day− 1 l mg− 1 SS

0.015h

mg P mg− 1 SS

548.3i

l kg− 1 SS

Data period

Domain

Models to be calibrated/validated

Purpose

February 7–16, 2001 Dry season for the years 1998, 1999 and 2000 July 16–24, 1999 July 17–27, 1999 Wet season for the years 1998, 1999 and 2000

1-D 1-D

Hydrodynamic, SS Water quality

Calibration Calibration

1-D 3-D 1-D

Hydrodynamic, SS Hydrodynamic, SS, water quality Water quality

Validation Validation Validation

Note the wet season is defined as April–September and the dry season as October– March. SS: Suspended sediment.

the bottom layer (Fig. 9b). This discrepancy is probably due to the oversimplified parameterization of benthic processes in the PRE, i.e. a uniform and constant benthic nutrient fluxes and SOD are assumed in the model, which can not reflect the realistic benthic processes. A sediment diagenesis process model is certainly required for further improvement of the model performance, and more detailed information are needed to support the implementation of the sediment diagenesis model. Fig. 9c shows the simulated CBOD in the surface layer reaches to high concentrations in the upper PRE (due to the inflows directly from the eight river outlet), but declines with distance to offshore. This CBOD pattern is consistent with the observational data. A similar pattern is also observed for NO23, and again, it is well reproduced by the model. Through comparisons with observations we show that the coupled model is robust enough to generate reasonable results in both wet season and dry season, although there are also some discrepancies between the model results and observations. In general, the results are quite encouraging. 3.2. Fluxes and transformation processes of nutrients in the wet season July 1999, a representative wet season with moderately high river discharge (~22,840 m3 s− 1) and occurrence of hypoxia (Yin et al., 2004b), is selected as the period for further analysis on the fluxes and transformations of nutrients in the Pearl River Delta (PRD). The flux estimates (residual fluxes) are determined by averaging the nutrient fluxes (volume flux multiplied by nutrient concentrations) calculated at every time step over 30 days (~58 M2 tidal cycles). To clarify the nutrient fluxes introduced from or exported to different regions, we define fluxes introduced from the upstream boundaries, fluxes passing through the eight river outlets and fluxes exported from the Pearl River Estuary (PRE) to the South China Sea (SCS) as

Table 4 Calibration and validation statistics for the 1-D and 3-D coupled model. Simulation period Variables

Domain Mean error

Water surface elevation Water surface elevation Discharge Salinity SSC Water surface elevation July 1999 (wet season for Water surface elevation Discharge validation) Salinity SSC SSC Chlorophyll a DIN (NH3 + NO23) OPO4 CBOD DO

1-D 3-D 1-D 1-D 1-D 1-D 3-D 1-D 3-D 1-D 3-D 3-D 3-D 3-D 3-D 3-D

February 2001 (dry season for calibration)

− 0.03 m 0.06 m 55.9 m3 s− 1 − 0.08 PSU 1.2 mg l− 1 − 0.10 m − 0.03 m 203.8 m3 s− 1 − 2.7 PSU − 20.0 mg l− 1 − 3.4 mg l− 1 −0.0004 µg l− 1 0.13 mg l− 1 0.01 mg l− 1 − 0.04 mg l− 1 0.67 mg l− 1

Relative r2 error (%) − 7.5 9.5 10.9 − 18.6 22.7 − 7.7 − 9.3 8.6 − 26.4 − 28.1 − 38.6 − 39.3 29.3 32.5 − 19.1 25.1

0.95 0.91 0.96 0.82 0.56 0.98 0.92 0.97 0.94 0.77 0.67 0.41 0.67 0.48 0.57 0.68

156

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

Fig. 6. Comparisons of simulated and observed results for calibration: (a) water surface elevation and discharge, (b) SSC at two representative river outlets (Humen and Modaomen) in February 2001 (dry season), and (c) NH3, CBOD, TN and TP in the Guangzhou–Foshan region (GF series), Zhongshan–Shunde region (ZS series) and Jiangmen–Kaiping region (JK series) in January 1999 (dry season); observational data are available for the dry season for the years 1998 (blue circles), 1999 (pink squares) and 2000 (green triangles). Locations of the survey stations in the river network (RNPRD) are shown in Fig. 4a. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

the upstream fluxes, the riverine fluxes and the estuarine fluxes, respectively. The exchange cross-section between the PRE and the SCS is shown in Fig. 1a. At the same time, the pollutant loads from point sources and non-point sources are defined as waste loads in this section in order to distinguish them from the upstream fluxes. There are thirty-one kinetic terms related to the transformations of nutrients (Table 1). The effective reaction term for each water quality component is determined by averaging each kinetic term calculated at every time step over 30 days. The nutrient budgets are constructed based on the estimated fluxes and effective reaction

terms. In this section, we will focus on the fluxes and budgets for CBOD, NH3, NO23 and IP for the river network (RNPRD) and the PRE combined in July 1999 (wet season), which are shown in Fig. 10, together with information on the relative importance of different sources and sinks. It should be noted that the results do not distinguish between terrestrial inputs and primary productivity sources (or aquatic inputs). Besides, the budgets are not completely conservative because the simulation considers only one month and the final conditions can not return to the initial conditions over this short period.

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

157

Fig. 7. Comparison of (a) the simulated SSC and (b) that from remote sensing image in the wet season.

3.2.1. Fluxes and transformation processes of CBOD In the RNPRD, the external loads of CBOD are estimated as 4256 t d− 1 (Fig. 10a). The upstream flux dominates the external loads, with 69% of the total. For individual rivers, CBOD loads from the Xijiang River (Fig. 1a), which are primarily contributed by waste loads and intense erosion processes in the upstream drainage area of the Xijiang River (Gao et al., 2002), account for 93% of the upstream flux. Waste loads in the RNPRD, significantly influenced by the point sources, are also important sources for CBOD. Among the waste loads, the non-pointsource loads contribute 24% to the total, which reflects the significance of non-point-sources pollution in the wet season compared to that in the dry season (Fig. 5). Fig. 10a shows that the RNPRD acts as a sink for CBOD that consumes 50% of the external loads. The loss of CBOD from the simulated system is dominated by carbonaceous oxidation which is also the most important sink for DO. The oxidation of CBOD removes 36% of the external loads and contributes 71% to its total loss. Another major loss of CBOD is via deposition on the benthic sediment, at 578 t d− 1 (~14% of the external loads). As such, the denitrification process also provides a sink for CBOD, but this reaction is insignificant in comparison with the oxidation and deposition processes. In relation to the cumulative CBOD in the benthic sediment, its ultimate fate associated with benthic diagenesis processes remains unclear. In addition, the cumulative CBOD is available for resuspension when the current is strong enough to erode the sediment during flood periods or storm events. More CBOD will be produced and the subsequent decomposition can cause water quality deterioration and hypoxia in the RNRPD. A further research on the diagenesis processes and resuspension of CBOD is required to address these important questions. Regarding the internal primary source for CBOD, it comes from the recycle of detrital phytoplankton carbon due to endogenous respiration and zooplankton predation. This contribution, however, is quite trivial because primary productivity in the RNPRD is strongly inhibited due to high turbidity associated with high sediment loads in the wet season (Fig. 8c). The turbidity leads to reduced light penetration, low phytoplankton biomass and low primary production (Yin et al., 2000). Model results show that the chlorophyll a concentrations in the RNPRD are generally less than 0.1 µg l− 1, and the net primary productivity is less than 0.1 g C m− 2 d− 1 in July 1999. This low primary productivity (b0.1 g C m− 2 d− 1) was also reported by Yin et al. (2000). The simulated riverine flux of CBOD shows the river network (RNPRD) to be a net exporter of CBOD to the Pearl River Estuary (PRE). The riverine flux has a relatively high value, indicating the residence time of CBOD in the RNPRD is short in the wet season. Based on Yin et al. (2000), the water residence time in the PRE is about 2.08–4.91 days in

the wet season. CBOD in the water column can not complete its transformation processes due to such short residence time. From Fig.10a it can be observed that the riverine flux of CBOD appears to follow the distribution pattern of the water flux (Fig. 8c) quite closely. This phenomenon suggests that the riverine flux is strongly controlled by high river discharge and significantly contributed by the upstream flux which is the most important external source in the RNPRD as discussed above. For individual river outlets, the flux through Modaomen dominates the riverine flux, with 28% of the total. Jiaomen, Humen and Hengmen are the other major receiving river outlets for CBOD. The proportions of the fluxes through Jiaomen, Humen and Hengmen to the riverine flux are 21%, 17% and 14%, respectively. As for the eastern four river outlets, its flux is much larger than that through the western four river outlets. This is mainly attributed to the higher river discharge through the eastern four river outlets (~61% of the total river discharge) relative to that through the western four river outlets. In the PRE, the external loads of CBOD, at 2839 t d− 1, are contributed by the riverine flux and waste loads discharged in the PRE. The riverine flux is the largest input of CBOD to the PRE, accounting for 81% of the external loads in the PRE. In contrary to the RNPRD, the dynamics of phytoplankton play an important role in the PRE. The internal source for CBOD recycled from algal death in the PRE, at 1243 t d− 1, is simulated to be at least two orders of magnitude higher than in the RNPRD. It is approximately equivalent to 44% of the external loads. Fig. 11 shows the simulated surface SSC and chlorophyll a distributions in the PRE. It can be observed that the concentrations of chlorophyll a at the inner part of the PRE (close to the RNPRD) are low because of the turbidity induced by high sediment loads from the eight river outlets. At the outer part of the PRE, water transparency increases due to the substantial decreases in SSC, and as a result, the concentrations of chlorophyll a increase. These phenomena are explored and discussed in more detail in literature (Yin et al., 2000; Yin et al., 2001; Yin et al., 2004a,b; Harrison et al., 2008). The amount of CBOD recycled from algal death at the outer part of the PRE, at 1173 t d− 1, is simulated to be approximately 15–16-fold greater than that at the inner part of the PRE. However, the internal source for CBOD is exceeded by large consumptions via oxidation, denitrification and deposition processes. The PRE thus behaves as a sink for CBOD that consumes 90% of the external loads. The oxidation process dominates the simulated CBOD behavior, resulting in a loss at 2608 t d− 1 (~ 94% of the external loads). The loss of CBOD via denitrification is simulated to be roughly equivalent to the loss via deposition, at 575 t d− 1 (~ 20% of the external loads). As most of CBOD is depleted in the water column, only a small amount of CBOD, at 288 t d− 1, is exported from the PRE to

158

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

Fig. 8. Comparisons of model results and observational data in July 1999 (for validation). Data include (a) water surface elevation and discharge, (b) SSC at Humen and Modaomen, and (c) water and SS fluxes through the eight river outlets during July 16 and July 24, 1999. G1–G8 represents the eight river outlets (Fig. 1a): Humen, Jiaomen, Hongqili, Hengmen, Modaomen, Jitimen, Hutiaomen and Yamen, respectively.

the South China Sea (SCS). The estuarine flux of CBOD may be primarily marine original, as most of the terrestrial CBOD is depleted inside the PRE and the contribution from algae death is very significant. This feature is supported by the studies of Jia and Peng (2003) and Hu et al. (2006). They both stated that the terrestrial derived organic matter contributed significantly to the bottom sediment at the inner part of the PRE, while organic matter in the sediment at the outer part of the PRE and adjacent shelf were mainly marine original. 3.2.2. Fluxes and transformation processes of ammonia nitrogen The fluxes and budgets of NH3 for the river network (RNPRD) and the Pearl River Estuary (PRE) combined are summarized in Fig. 10b. The external loads of NH3 in the RNPRD are estimated as 712 t d− 1. The upstream flux provides 57% to the external loads, with the Xijiang River being the most important source (~90% of the upstream flux). Percentage contribution of the waster loads of NH3 to the external loads has increased compared to that of CBOD. Source apportionment

of the waster loads is found as follows: 71% for the point sources and 29% for the non-point sources. In the RNPRD, nitrification is identified as the dominant process in the transformations of NH3, leading to a loss at 276 t d− 1 (~39% of the external loads) which accounts for 99% of its removal. This process also contributes significantly to the high NO23 concentrations and low DO saturations in the water column both simulated and observed in the wet season. Deposition of particlesorbed NH3 also provides a sink for NH3, but this process is insignificant in comparison with the nitrification process. Although NH3 is the preferred form of inorganic nitrogen for phytoplankton growth, the loss of riverine NH3 via phytoplankton uptake is trivial since the primary productivity in the RNPRD is strongly inhibited due to high turbidity. The internal sources for NH3 are quite small relative to the loss via nitrification. Conclusively, the RNPRD behaves as a sink for NH3 that consumes 37% of the external loads. The proportion of the simulated riverine flux of NH3 to the external loads in the RNPRD (~63%) is higher than that of CBOD (~ 54%), which

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

159

Fig. 9. Comparisons of simulated and observed results for validation: (a) NH3, CBOD, TN and TP in the Guangzhou–Foshan region, Zhongshan–Shunde region and Jiangmen–Kaiping region in July 1999 (wet season); observational data are available for the wet season for the years 1998 (blue circles), 1999 (pink squares) and 2000 (green triangles), (b) CBOD, OPO4, NO23 and DO for different layers at the time-series stations C1 and C2, and (c) CBOD, DO, NO23 and OPO4 for the surface layer at the survey stations in the Pearl River Estuary (PRE) in July 1999. The number in bottom panel denotes the survey station number (Fig. 4b). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

is mainly attributed to the higher percentage contribution of the waste loads of NH3 (~43%) relative to that of CBOD (~ 31%). The riverine flux of NH3 behaves in a similar manner to that of CBOD, except the NH3 flux through Humen exceeds Jiaomen. This difference is related to the

fact that most of the waste loads discharged from Dongguan, Foshan and Guangzhou (the city with the largest waste discharge, see Fig. 1a) are direct to the upstream area of Humen, and hence lead to a remarkable increase in the NH3 flux through Humen. The proportion

160

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

of the flux through Humen to the riverine flux has increased to 21%. Like CBOD, the eastern four river outlets are the major receiving river outlets for NH3, transferring 61% of the riverine flux to the PRE. The external loads of NH3 in the PRE are estimated as 577 t d− 1, of which 78% are contributed by the riverine flux. The interactions

between multiple mechanisms for NH3 appear to be more complicated in the PRE than in the RNPRD. The simulated NH3 behavior in the PRE is controlled under the combined effects of phytoplankton kinetics, nitrification, ON mineralization and benthic processes. The active dynamics of phytoplankton play an important role in the

Fig. 10. Monthly-average fluxes and budgets for (a) CBOD, (b) NH3, (c) NO23, and (d) IP for the river network (RNPRD) and the Pearl River Estuary (PRE) combined in July 1999 (wet season). Data in square brackets are ratios of waste loads, effective reaction terms and fluxes of nutrients to the external loads in each system. Positive values represent inputs and negative values represent losses. Regarding the riverine fluxes, data in square brackets are ratios with respect to the RNPRD (the former) and the PRE (the later). Data in round brackets are proportions of the nutrient fluxes through the eastern four river outlets (G1–G4) to the riverine fluxes.

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

161

Fig. 10 (continued).

transformations of NH3 in the water column, resulting in a net loss equivalent to 15% of the external loads in the PRE. During phytoplankton growth, NH3 is utilized by phytoplankton for growth, at 132 t d− 1 (~23% of the external loads). The large consumption of NH3 via phytoplankton uptake corresponds to the increased primary production in the PRE which also leads to a significant loss of NO23 at

107 t d− 1. During phytoplankton death, NH3 is recycled from the phytoplankton biomass pool in the water column, at 47 t d− 1 (~ 8% of the external loads). The benthic NH3 flux and conversion from ON via mineralization also provide significant contributions to the NH3 budget, together at 340 t d− 1 (~59% of the external loads). However, these sources for NH3 are exceeded by large consumptions mainly via

162

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

Fig. 11. Simulated surface (a) SSC and (b) chlorophyll a distributions in the Pearl River Estuary (PRE) in July 1999 (wet season, monthly-average).

nitrification and phytoplankton uptake. Thus the PRE behaves as a sink for NH3 that consumes 80% of the external loads. The loss of NH3 via deposition is trivial, indicating that the sorption process between SS and NH3 is so insignificant that it can be ignored in the transformations of NH3. The major loss of NH3 is via nitrification which is simulated to be larger than the external loads in the PRE (Fig. 10b). Ultimately, only 121 t d− 1 of NH3 are exported from the PRE to the SCS due to the high depletion of NH3 in the PRE. The proportion of the simulated estuarine flux of NH3 (~21%) is higher than that of CBOD

(~10%), presumably due to the significant contributions from the internal sources of NH3. 3.2.3. Fluxes and transformation processes of nitrate plus nitrite nitrogen From Fig. 10c it can be observed that the upstream flux of NO23 dominates the external loads of this constituent in the river network (RNPRD) which are estimated as 1975 t d− 1. NO23 loads from the Xijiang River provide 92% to the upstream flux. With respect to the internal sources for NO23, the conversion from NH3 via nitrification is

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

the leading input for NO23, which is simulated to be equivalent to 14% of the external loads in the RNPRD. The major loss of NO23 is via denitrification, at 14 t d− 1 (~ 1% of the external loads); although this process plays a minor role in terms of NO23 budget. The internal sinks of NO23 such as phytoplankton uptake and denitrification are trivial relative to its generation from NH3 nitrification. As a consequence, the RNPRD acts as a source for NO23 and all of the NO23 loads added to the RNPRD are lost to the PRE. It can be concluded that NO23 behaves quasi-conservatively in the water column of the RNPRD. The simulated riverine flux of NO23 is particularly high under the combined contribution from nitrification and external loads of NO23. It shows a similar distribution pattern to that of NH3. Approximately 68% of the riverine flux of NO23 is exported from three major river outlets including Modaomen, Humen and Jiaomen. The proportions to the riverine flux are 25% for Modaomen, 23% for Humen and 20% for Jiaomen. The high NO23 flux through Humen is significantly contributed by the conversion from NH3 via nitrification, which implies an intense nitrification rate within this region (Harrison et al., 2008). The eastern four river outlets remain the major receiving outlets in terms of nutrient fluxes. In addition, the riverine flux of NO23 accounts for 84% of the riverine flux of DIN (~2713 t d− 1), indicating that NO23 is the major phase of DIN. The riverine flux with the waste loads adds up to a total external input of 2268 t d− 1 of NO23 to the Pearl River Estuary (PRE) which is dominated by the former. Much alike the RNPRD, the PRE acts as a source for NO23 due to the significant internal compensation from NH3 nitrification which is simulated to be equivalent to 32% of the external loads in the PRE. Regarding the internal sinks for NO23, about 5% of the external loads (~107 t d− 1) are utilized via phytoplankton uptake, while another 9% of the external loads (~201 t d− 1) is consumed via denitrification. These two processes contribute 95% of the removal of NO23. In general, the internal sinks of NO23 appear to be relatively small, which indicates the importance of nitrification in the maintenance of NO23. As a result, a particularly high flux of NO23 (~2709 t d− 1) is exported to the South China Sea (SCS). Also, the proportion of the simulated estuarine flux of NO23 to the estuarine flux of DIN (~2830 t d− 1) is over 96%, a value much higher than that for the riverine flux of NO23. All of the NO23 loads added to the PRE are lost to the SCS. This suggests the quasiconservative behavior of NO23 in the water column of the PRE and the accumulation of NO23 in the SCS. 3.2.4. Fluxes and transformation processes of inorganic phosphorus In the river network (RNPRD), the external loads of IP are estimated as 103 t d− 1 (Fig. 10d). Source apportionment of the external loads is found as follows: 79% for the upstream flux (of which 94% is contributed by the Xijiang River), and 21% for the waste loads (of which 70% is contributed by the point sources). From Fig. 10d it can be observed that the RNPRD acts as a sink for IP that consumes 11% of the external loads. Deposition of PO4SS is the dominant process in the transformations of IP, and this process leads to a loss of IP at 12 t d− 1 which accounts for more than 99% of its removal. This indicates that the behavior of IP is greatly influenced by SS which exhibits an intense adsorption on OPO4. Similar to DIN in the RNPRD, the loss of riverine IP via phytoplankton uptake is trivial. It suggests that phytoplankton production in the RNPRD is mostly limited by light due to high turbidity, rather than nutrients availability. This phenomenon can also be reflected by the simulated SSC and chlorophyll a distributions at the inner part of the PRE (close to the RNPRD), as shown in Fig. 11. The simulated riverine flux of IP is particularly high because of the low consumption in the RNRPD. It shows a similar distribution pattern to those for CBOD, NH3 and NO23, with Modaomen, Jiaomen, Humen and Hengmen being the major receiving river outlets. One difference is that the IP flux through Humen exceeds Modaomen, and it becomes the largest one out of the eight river outlets. This is because Modaomen as well as Jiaomen carries a higher SS load relative to Humen (Fig. 8c), and thus in the regions of Modaomen and Jiaomen, IP experiences a significant loss via deposition ascribed to the intense adsorption of OPO4

163

by SS. The proportions to the riverine flux are 25% for Humen, 24% for Modaomen, 20% for Jiaomen and 12% for Hengmen. The proportion of the flux through the eastern four river outlets (~64%) is relatively high, which is associated with its large river discharge and further enhanced by the increased proportion of the flux through Humen. In relation to the different phases of IP in the water column, the riverine fluxes for OPO4 and PO4SS are estimated as 44 and 52 t d− 1, respectively. The latter accounts for 57% of the riverine flux of IP, indicating that a large amount of IP transports in the form of sorbed phase. It is clear that both the riverine fluxes for OPO4 and PO4SS are so important that none of them can be ignored or individually represent the riverine flux of IP. Also, since more OPO4 will be desorbed from the riverine PO4SS due to elimination of sorption capacity of SS in the Pearl River Estuary (PRE), it is inappropriate to regard the riverine flux of OPO4 as the only source for available IP that is exported from the RNPRD to the PRE. In the PRE, the riverine flux dominates the external loads of IP, with 91% of the total (~105 t d− 1). The internal sources for IP are of importance to the IP budget. The amounts of IP recycled from the algae death and converted from OP via mineralization are estimated as 6 and 18 t d− 1, respectively. Spurred by the increased primary productivity, the loss of IP is dominated by phytoplankton uptake which is estimated as 30 t d− 1 (~28% of the external loads). Another major loss of IP is via deposition on the benthic sediment, at 11 t d− 1 (~10% of the external loads). This serves to emphasize the significant influence of SS on the behavior of IP. Overall, the PRE acts as a sink for IP that consumes 16% of the external loads in the PRE. The simulated estuarine flux of IP has a relatively high value, estimated as 92 t d− 1 (~88% of the external loads in the PRE). It can be found the estuarine flux is significantly contributed by the internal sources of IP. In addition, the estuarine fluxes for OPO4 and PO4SS are estimated as 92 and 1 t d− 1, respectively. The former one accounts for 99% of the estuarine flux of IP, suggesting that OPO4 is the major phase of IP transporting in the PRE. 3.3. Seasonal variations of nutrient budgets in the Pearl River Delta As shown in Fig. 5, the external loads of nutrients exhibit significant seasonal variations in terms of upstream fluxes and non-point-source loads. Fig. 12 is meant to provide an insight into the nutrient budgets in January 1999 (dry season), and to ascertain if the river network (RNPRD) and the Pearl River Estuary (PRE) are sinks or sources for nutrients in the dry season. Model results show that the transformations of nutrients in the dry season are generally in a similar manner to those in the wet season (Fig. 10), in terms of the key biogeochemical processes for nutrients and the contribution to the nutrient budgets from phytoplankton dynamics and internal sources of nutrients. For instance, carbonaceous oxidation and nitrification are still the dominant processes in the transformations of CBOD and NH3, respectively, in the RNPRD and the PRE in the dry season. Also, the RNPRD and the PRE appear to be sinks for CBOD, NH3 and IP, and sources for NO23 in the dry season when the RNPRD and the PRE have a common feature of retaining a high proportion of nutrient inputs. This implies that in the dry season as well as the wet season, a great amount of nutrients from terrestrial inputs or primary production have been consumed through a series of intense biogeochemical processes, e.g., phytoplankton uptake, settling to the seafloor and nitrification. As a result, eutrophication and its consequences (e.g., harmful algal blooms and hypoxia in bottom waters) may occur. These impacts are an increasing threat to coastal ecosystems and are believed to be associated with high nutrient loads delivered by rivers to the PRD and its coastal waters. Regarding a potential hypoxia, model results show that consumption of DO in bottom waters is linked to the oxidation process of CBOD and ammonia nitrification, in addition to SOD fuelled by cumulative organic matter in the benthic sediment from primary production along with terrestrial inputs. By comparing with the wet season, the magnitude of internal sources for nutrients in the dry season is much smaller, but their contributions to the nutrient budgets are more significant, because the external loads in

164

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

the dry season are of less importance due to low river discharge condition (~2620 m3 s− 1). In the dry season, the external loads of nutrients are generally contributed by the waste loads in the RNPRD and the PRE since the upstream fluxes and the riverine fluxes are relatively small (Fig. 12). The proportions of the simulated riverine fluxes of nutrients to the external loads have remarkably decreased due to a

longer residence time. From Figs. 10 and 11 it is clear that the seasonal variations of nutrient fluxes are very significant. The upstream fluxes of nutrients in the wet season are 12–27-folds greater than those in the dry season, while the simulated riverine fluxes and estuarine fluxes are 5– 11-folds and 1–5-folds larger, respectively. This indicates that most of the annual nutrients occur in the wet season, which is closely linked to

Fig. 12. Monthly-average fluxes and budgets for (a) CBOD, (b) NH3, (c) NO23, and (d) IP for the river network (RNPRD) and the Pearl River Estuary (PRE) combined in January 1999 (dry season).

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

165

Fig. 12 (continued).

the water discharge volume. It is estimated that about 80% of the freshwater discharge occurs in the wet season (April–September) and only 20% in the dry season (October–March) (Cai et al., 2004). 3.4. Comparison of nutrient inputs among the Pearl River Estuary, Changjiang and Mississippi Rivers The Pearl River Estuary (PRE) and similar river-dominated systems such as the Changjiang and Mississippi Rivers have a common feature

of being highly dynamic and complicated environments with a close coupling between riverine nutrients, net productivity and hypoxia (Spillman et al., 2007; Rabouille et al., 2008), therefore comparison of nutrients inputs among these large river systems will be instructive for further studies of the PRE in terms of eutrophication processes and hypoxia. Freshwater discharge, riverine DIN and DIP inputs, and nutrient molar ratios in the PRE, the Changjiang and Mississippi Rivers are shown in Table 5. To facilitate the comparison, we estimate the annual value of nutrient inputs in the PRE by averaging the results

166

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167

Table 5 Freshwater discharge, riverine nutrient input and nutrient ratios in three river-dominant systems. System

Data period

Discharge (103 m3 s− 1)

Pearl River Estuarya

July 1999 January 1999 Annual⁎

Changjiang Riverb Mississippi Riverb

Annual Annual

23 3 13 27 16

Nutrient input (t d− 1) DIN

DIP

2713 348 1531 1306 1955

44 10 27 45 133

DIN:DIP (mol:mol) 137 77 126 65 33

a This work. Annual ⁎ is estimated by averaging monthly results obtained in July 1999 (wet season) and January 1999 (dry season). b Dagg et al. (2004).

obtained in July 1999 (wet season) and January 1999 (dry season). The annual values of riverine DIN and DIP fluxes in the PRE, estimated as 1531 and 27 t d− 1, respectively, are similar to the values of 1074 and 25 t d− 1 estimated by Cai et al. (2004). Among these rivers, the Mississippi has the largest inputs of DIN and DIP, but it has a lower DIN:DIP ratio due to its relatively high DIP input. Nitrogen limitation is more frequent in the Mississippi River due to efficient recycling of phosphorus and loss of nitrogen through denitrifcation (Rabalais et al., 2002). Riverine DIN input for the PRE and the Changjiang is similar, although the Changjiang discharge is two times larger. However, the PRE has a relatively low DIP input, resulting in a high DIN:DIP ratio over 126, a value much higher than other two river systems. The seasonal variation of the riverine DIN flux is clearly more significant than the riverine DIP flux in the PRE (Table 5), thus resulting in a significant seasonal variation in nutrient ratio. This is mainly ascribed to the influence by the upstream fluxes of DIN and DIP in association with their biogeochemical processes in the river network, e.g., loss of DIN via phytoplankton uptake and denitrification is quite trivial (see subsection 3.2 and 3.3). Response of the ecosystem in the PRE to seasonal and annual variations in eutrophication pressures is an intriguing aspect of the problems that need to be addressed in the future. The DIN:DIP ratio is particularly high in the wet season when the DIN input is very large and phosphorus is generally believed to be the main limiting nutrient for phytoplankton growth (Yin et al., 2000; 2004a; Harrison et al., 2008), although light also limits primary production (Fig. 11). This feature of phosphorus limitation in the PRE is similar to the Changjiang River where the DIN:DIP ratio is much higher than the Redfield ratio (16:1), but in contrast to the Mississippi River where nitrogen appears to be the more common limiting nutrient (Lohrenz et al., 2008). 4. Conclusions A 1-D and 3-D coupled model, which includes physical and biogeochemical processes, is used to simulate the fluxes and transformations of nutrients in the Pearl River Delta (PRD). The model has been calibrated and validated to available field data collected in February 2001, January 1999 and July 1999. It is paramount to see the model results of physical and water quality state variables are in reasonable agreement with the observational data, suggesting that the model is sufficiently robust to capture the physical and biogeochemical dynamics in the PRD. At the same time, the fluxes and budgets for CBOD, NH3, NO23 and IP in July 1999 (a representative wet season) are presented in this study. Results show that the river network (RNPRD) behaves as a source for NO23, but a sink for CBOD, NH3 and IP that consumes 50%, 37% and 11% of the external loads for CBOD, NH3 and IP, respectively. The simulated riverine fluxes of nutrients exported from the RNPRD to the Pearl River Estuary (PRE) have relatively high values due to the short residence time in the RNPRD. The riverine fluxes are generally controlled by high

river discharge and significantly contributed by the upstream fluxes. In addition, the riverine fluxes are the largest inputs to the PRE, with 78–100% of the external loads in the PRE. The PRE also acts as a source for NO23 and a sink for CBOD, NH3 and IP that consumes 90%, 80% and 16% of the external loads for CBOD, NH3 and IP, respectively. The simulated estuarine fluxes exported from the PRE to the SCS are significantly contributed by the internal sources of nutrients, in addition to the external loads in the PRE. Regarding the transformation processes of nutrients in the RNPRD, carbonaceous oxidation, nitrification and deposition are identified as the dominant processes with respect to CBOD, NH3 (also NO23) and IP. The phytoplankton dynamics play a minor role in the transformations of nutrients as phytoplankton growth is strongly limited due to high turbidity associated with high sediment loads in the wet season. The internal sources of nutrients are also trivial. With respect to the PRE, the interactions between multiple mechanisms for nutrients are more complicated than in the RNPRD. The transformations of CBOD, NH3 (also NO23) and IP are dominated by carbonaceous oxidation, nitrification and phytoplankton uptake, respectively. Different from the RNPRD, the phytoplankton dynamics in the PRE appear to play an important role in the transformations of nutrients due to the increased primary production. Furthermore, the nutrient budgets in the PRE are significantly contributed by the internal sources of nutrients, in association with the external loads. At the same time, the nutrient budgets in January 1999 (dry season) are discussed. Conclusively, the transformations of nutrients in the dry season are generally in a similar manner to those in the wet season, while the nutrient fluxes show significant seasonal variations. It should be noted that the nutrient budgets in this study are constructed based on the results in one month (July 1999) due to data limitations. July 1999 is a representative wet season in terms of hydrodynamic conditions. In this sense, the nutrient budgets are representative for the wet season. Also, eutrophication and occurrence of hypoxia are reported and widely studied in this period (e.g., Yin et al., 2001, 2004a,b; Harrison et al., 2008). This work would provide a basis for heuristic studies of the PRD estuarine system in terms of understanding of the severity of eutrophication, variation of primary production, and formation, occurrence and intensity of hypoxia. Long-term simulations (e.g., one year or more) are certainly required to perform a better estimation on the nutrient budgets in the PRD in the future. Acknowledgements We would like to thank the Macao Meteorological and Geophysical Bureau for providing us the wind data. We also acknowledge the contribution of information on pollutant loads from the Environment Council of Macao Special Administrative Region, Environment Protection Department of Hong Kong Special Administrative Region, and the South China Institute of Environmental Sciences. In addition, we also want to thank the Pearl River Estuary Pollution Project and the Pearl River Water Resources Commission for sharing the field data used for model calibration and validation. This research was financially supported by the 908 Project of Guangdong (GD908-02-03). References Ambrose Jr., R.B., Wool, T.A., Martin, J.L., 1993. The water quality analysis simulation program, WASP5, part a: model documentation. Technical Report. InU.S. Environmental Protection Agency, Atlanta, GA. Blumberg, A.F., 2002. A primer for ECOMSED version 1.3 user manual. Technical Report. InHydroQual, Inc., Mahwah, New Jersey. Cai, W.J., Dai, M.H., Wang, Y.C., Zhai, W.D., Huang, T., Chen, S.T., Zhang, F., Chen, Z.Z., Wang, Z.H., 2004. The biogeochemistry of inorganic carbon and nutrients in the Pearl River estuary and the adjacent North South China Sea. Continental Shelf Research 24, 1301–1319. Cerco, C.F., Cole, T., 1993. Three-dimensional eutrophication model of Chesapeake Bay. Journal of Environmental Engineering 119, 1006–1025.

J. Hu, S. Li / Journal of Marine Systems 78 (2009) 146–167 Chao, X.B., Jia, Y.F., Shields, F.D., Wang, S.S.Y., Cooper, C.M., 2007. Numerical modeling of water quality and sediment related processes. Ecological Modelling 20, 385–397. Chen, Y., Wai, W.H.O., Li, Y.S., Lu, Q.M., 1999. Three-dimensional numerical modeling of cohesive sediment transport by tidal current in Pearl River Estuary. International Journal of Sediment Research 142, 107–123. Chen, B., Wang, K., 2008. Suspended sediment transport in the offshore near Yangtze Estuary. Journal of Hydrodynamics 20, 373–381. Cunge, J.A., Holly Jr., F.M., Verwey, A., 1980. Practical Aspects of Computational River Hydraulics. InPitman Publishing Limited, London. 420pp. Dagg, M, Benner, R., Lohrenz, S., Lawrance, D., 2004. Transformations of dissolved and particulate materials on continental shelves influenced by large rivers: plume processes. Continental Shelf Research 24, 833–858. Dai, M., Guo, X., Zhai, W., Yuan, L., Wang, B., Wang, L., Cai, P., Tang, J., Cai, W.J., 2006. Oxygen depletion in the upper reach of the Pearl River estuary during a winter drought. Marine chemistry 102, 159–169. Di Toro, D.M., 1978. Optics of turbid estuarine waters: approximations and applications. Water Research 12, 1059–1068. Dong, L.X, Su, J.L, Wong, S.L.A., Cao, Z., Chen, J.C., 2004. Seasonal variation and dynamics of Pearl River plume. Continental Shelf Research 24, 1761–1777. Fitzpartick, J.J., 2004. A user's guide for RCA (release 3.0). Technical Report. HydroQual, Inc., Mahwah, New Jersey. Galperin, B., Kantha, L.H., Hassid, S., Rosati, A., 1988. A quasi-equilibrium turbulent energy model for geophysical flows. Journal of the Atmospheric Sciences 45, 55–62. Gao, Q.Z., Shen, C.D., Sun, Y.M., Yi, W.X., 2002. Riverine organic carbon in the Xijiang River (South China): seasonal variation in content and flux budget. Environmental Geology 41, 826–832. Guan, W.B., Wong, L.A., Xu, D.F., 2001a. Modeling nitrogen and phosphorus cycles and dissolved oxygen in the Zhujiang River Estuary, part I. model development. Acta Oceanologica Sinica 20, 493–504. Guan, W.B., Wong, L.A., Xu, D.F., 2001b. Modeling nitrogen and phosphorus cycles and dissolved oxygen in the Zhujiang River Estuary, part II. model results. Acta Oceanologica Sinica 20, 504–514. Håkanson, L., Bryhn, A.C., Eklund, J.M., 2007. Modelling phosphorus and suspended particulate matter in Ringkøbing Fjord in order to understand regime shifts. Journal of Marine Systems 68, 65–90. Harrison, P.J., Yin, K.D., Lee, J.H.W., Gan, J.P., Liu, H.B., 2008. Physical–biological coupling in the Pearl River Estuary. Continental Shelf Research 28, 1405–1415. Hills, P., Zhang, L., Liu, J., 1998. Transboundary pollution between Guangdong Province and Hong Kong: threats to water quality of the Pearl River Estuary and their implications for environmental policy planning. Journal of Environment and Planning Management 41, 375–396. Hu, J.F., Peng, P.A., Jia, G.D., Mai, B.X., Zhang, G., 2006. Distribution and sources of organic carbon, nitrogen and their isotopes in sediments of the subtropical Pearl River estuary and adjacent shelf, southern China. Marine Chemistry 98, 274–285. Hu, J.T., Li, S.Y., 2008. One-dimensional salinity and three-dimensional baroclinic coupled model for simulating the flow in Pearl Rive Delta. Journal of Hydraulic Engineering 39, 1174–1182 (in Chinese, with English abstr.). Huang, X.P., Huang, L.M., Yue, W.Z., 2003. The characteristics of nutrients and eutrophication in the Pearl River estuary, South China. Marine Pollution Bulletin 47, 30–36. Jay, D.A., Geyer, W.R., Uncles, R.J., Vallino, J., Largier, J., Boynton, W., 1997. A review of recent developments in estuarine scalar flux estimation. Estuaries 20, 262–280. Jia, G.D., Peng, P.A., 2003. Temporal and spatial variations in signatures of sedimented organic matter in Lingding Bay (Pearl estuary), southern China. Marine Chemistry 82, 47–54. Jos, V.G., Klaas-Jan, V.H., Dirk, S., Erik, D.G., Firmijn, Z., Daniel, T., Joost, I., Reinaldo, P., Karen, M., Simon, G., 2007. Pearl River Delta Water Quality Model: final study report. Technical Report. WL delft hydraulics, Netherlands. Lohrenz, S.E., Redalje, D.G., Cai, W.J., Acker, J., Dagg, M., 2008. A retrospective analysis of nutrient and phytoplankton productivity in the Mississippi River plume. Continental Shelf Research 28, 1466–1475. Lu, P.L., Huang, W.D., Li, J., 2005. Study on adsorption experiment for phosphorus pollutant with suspended sediment. Water Resources and Hydropower Engineering 36, 93–96 (in Chinese, with English Abstr.). Mao, Q., Shi, P., Yin, K., Gan, J., Qi, Y., 2004. Tides and currents in the Pearl River Estuary. Continental Shelf Research 24, 1797–1808.

167

Mellor, G.L., Yamada, T., 1982. Development of a turbulence closure model for geophysical fluid problems. Reviews of Geophysics 20, 851–875. Neller, R.J., Lam, K.C., 1994. The Environment. In: Yeung, Y.M., Chu, D.K.Y. (Eds.), Guangdong: Survey of a Province Undergoing Rapid Change. The Chinese University of Hong Kong Press, Hong Kong. Rabalais, N.N., Turner, R.E., Wiseman, W.J., 2002. Gulf of Mexico hypoxia, a.k.a. “the dead zone”. Annual Review of Ecological Systems 33, 235–263. Rabouille, C., Conley, D.J., Dai, M.H., Cai, W.J., Chen, C.T.A., Lansard, B., Green, R., Yin, K., Harrison, P.J., Dagg, M., McKee, B., 2008. Comparisons of hypoxia among four riverdominated ocean margins: the Changjiang (Yangtze), Mississippi, Pearl, and Rhône rivers. Continental Shelf Research 28, 1527–1537. Ruddle, K., Zhong, G., 1988. Integrated Agriculture–Aquaculture in South China: the Dike–Pond System of the Zhujiang Delta. InCambridge University Press, Cambridge, p. 173. San Deigo-McGlone, M.L., Smith, S.V., Nicolas, V.F., 2000. Stoichiometric interpretations of C:N:P ratios in organic waste materials. Marine Pollution Bulletin 40, 325–330. Shi, H.X., Liu, H.J., Qu, J.H., Dai, R.H., Wang, M.H., 2005. Property of adsorption about ammonia on suspended mineral matters in eutrophic water. Environmental Science 26, 72–76 (in Chinese, with English Abstr.). Spillman, C.M., Imberger, J., Hamilton, D.P., Hipsey, M.R., Romero, J.R., 2007. Modelling the effects of Po River discharge, internal nutrient cycling and hydrodynamics on biogeochemistry of the Northern Adriatic Sea. Journal of Marine Systems 68, 167–200. Steele, J.H., 1962. Environmental control of photosynthesis in the sea. Limnology and Oceanography 7, 137–150. Tang, D., Kester, D.R., Ni, L.H., Qi, Y., Kawamura, H., 2003. In situ and satellite observations of a harmful algal bloom and water condition at the Pearl River estuary in late autumn 1998. Harmful Algae 2, 89–99. Tappin, A.D., 2002. An examination of the fluxes of nitrogen and phosphorus in temperate and tropical estuaries: current estimates and uncertainties. Estuarine Coastal and Shelf Science 55, 885–901. Tappin, A.D., Harris, J.R.W., Uncles, R.J., 2003. The fluxes and transformations of suspended particles, carbon and nitrogen in the Humber estuarine system (UK) from 1994 to 1996: results from an integrated observation and modelling study. The Science of the Total Environment 314–316, 665–713. Wong, L.A., Chen, J.C., Xue, H., Dong, L.X., Su, J.L., Heinke, G., 2003a. A model study of the circulation in the Pearl River Estuary (PRE) and its adjacent coastal waters: 1. simulations and comparison with observations. Journal of Geophysical Research 108 (C5), 3156. doi:10.1029/2002JC0011451. Wong, L.A., Chen, J.C., Xue, H., Dong, L.X., Guan, W.B., Su, J.L., 2003b. A model study of the circulation in the Pearl River Estuary (PRE) and its adjacent coastal waters: 2. sensitivity experiments. Journal of Geophysical Research 108 (C5), 3156. doi:10.1029/2002JC0011452. Xu, J.T., Hood, R.R., 2006. Modeling biogeochemical cycles in Chesapeake Bay with a coupled physical–biological models. Estuarine Coastal and Shelf Science 69, 19–46. Yin, K.D., Qian, P.Y., Chen, J.C., Hsieh, D.P.H., Harrison, P.J., 2000. Dynamics of nutrients and phytoplankton biomass in the Pearl River estuary and adjacent waters of Hong Kong during summer: preliminary evidence for phosphorus and silicon limitation. Marine Ecology Progress Series 194, 295–305. Yin, K.D., Qian, P.Y., Wu, M.C.S., Chen, J.C., Huang, L.M., Song, X.Y., Jian, W.J., 2001. Shift from P to N limitation of phytoplankton growth across the Pearl River estuarine plume during summer. Marine Ecology Progress Series 221, 17–28. Yin, K.D., Song, X.X., Sun, J., Madeline, C.S.W., 2004a. Potential P limitation leads to excess N in the Pearl River estuarine coastal plume. Continental Shelf Research 24, 1895–1907. Yin, K.D., Lin, Z.F., Ke, Z.Y., 2004b. Temporal and spatial distribution of dissolved oxygen in the Pearl River Estuary and adjacent coastal waters. Continental Shelf Research 24, 1935–1948. Zhan, D.R., Chen, F.R., Yang, Y.Q., Zhang, L., Lu, Y., 2005. Benthic fluxes of nutrients in neritic zone outside Zhujiang River Estuary in summer. Journal of Tropical Oceanography 24, 53–60 (in Chinese, with English Abstr.). Zheng, L.Y., Chen, C.S., Zhang, F.Y., 2004. Development of water quality model in the Satilla River Estuary, Georgia. Ecological Modelling 178, 475–482.