Journal of Water and Environment Technology, Vol.13, No.3, 2015 Modeling of Water Quality in Tidal River Network in ,

Masayasu IRIE, Tomo YAMAGUCHI, Shuzo NISHIDA, Yusuke NAKATANI

Department of Civil Engineering, Division of Global Architecture, Graduate School of Engineering, Osaka University, 2-1 Yamadaoka, Suita, Osaka 565-0871, Japan

ABSTRACT An urban tidal river network is one of the most difficult targets to simulate the flow and water quality in various kinds of water areas because the model must simulate both the vertical feature of the intrusion of sea water and the horizontal feature of water quality that is affected by human activities. In this paper, we determine the applicability of a three-dimensional flow and water quality model when reproducing the process of water with high concentrations of nitrogen, phosphorus and other constituents flowing through an urban tidal river network. We applied a three-dimensional model to Neya River, its tributaries and its downstream rivers in Osaka. The model can simulate well the vertical stratification of temperature and salinity near three river mouths, the mixing process of two different sources of water originating from , and Neya River and its tributaries, the longitudinal profiles of chemical oxygen demand (COD), dissolved oxygen (DO), total nitrogen and phosphorus. Calculated water levels in the upstream area of Neya River system, however, are not coincident with the observations. A 3D water quality model, which is a potent tool in ocean and lakes, proved to be a versatile enough tool to simulate complicated and tidal urban river networks.

Keywords: Neya River, three-dimensional water quality modeling, tidal river, urban river network, water quality

INTRODUCTION Most rivers flowing through large cities have been influenced by drainage water and loadings after human water usage. When there are sewage facilities and influence of tides, the temporal and spatial change and the characteristics of water quality in the river water flowing down are complicated. The water quality is poor in some rivers that are polluted and affected by human activities. Once the river water is highly contaminated by organic or toxic substances, removal of those substances is difficult if the characteristics and makeup of the water are not clarified.

Tidal rivers with vertical stratification of salinity are one of the most difficult target areas to simulate with hydraulic models. The difficulties are caused by a large variety of hydrodynamic characteristics. Tidal rivers are located close to coastlines, mostly consist of more than one water channel and often have numerous branches that require many open boundary conditions in simulations. As open boundaries at the end of a stream are much wider than the river widths, the assessment and simulation of tidal force need an utmost care in the settings for the model. The ratio of the change of surface water level to depth is larger than that in lakes and oceans. If the upstream water discharge is small and the tidal river is long, the friction on the river bottom becomes a dominant parameter, especially when the model represents the perturbation of surface height in the upstream. While the geometry of river networks requires a two-dimensional (2D), vertically integrated, hydrodynamic model at the very least, the stratification does a two-dimensional, horizontally integrated model. Therefore, the tidal river network

Address correspondence to Masayasu Irie, Department of Civil Engineering, Graduate School of Engineering, Osaka University, Email: [email protected] Received January 7, 2014, Accepted November 17, 2014. - 231 - Journal of Water and Environment Technology, Vol.13, No.3, 2015 requires combinations of unsteady, longitudinal one-dimensional (1D) and vertical 2D models, shallow water 2D and vertical 2D models, or a three-dimensional (3D) model. There are many open-source 3D hydraulic model codes with a sub-model of water quality, which can be applied to such dynamically complicated areas. Successful applications are mostly for large estuaries such as those done by Lin and Kuo (2003) and Shen and Haas (2004) for the York River system. There are, however, only rare cases in which the reproducibility of temporal and spatial distributions of substances in flow is modeled in small tidal rivers. Huang and Spaulding (1995) simulated pollutant transport induced by combined sewer overflows (CSO) in Mt. Hope Bay, Liu et al. (2008) simulated hydrodynamics in Danshuei River of Taiwan, and Huang et al. (2011) performed 3D numerical experiments in a shallow tidal river located in Tampa Bay. Since the river networks of large cities in Japan have complicated geometry and narrow widths because they were originally used and modified as canals for waterway traffic, it is extremely difficult to simulate the flow and water quality characteristics in the entire area at one time.

In this study, current, salinity, temperature and some kinds of nitrogen and phosphorus content are simulated using 3D hydrodynamic models combined with a biochemical reaction compartment downstream of Neya River and Yodo River, which flow through Osaka City, the third largest city in Japan. The annual average discharge of Neya River and its tributaries is about 2 m3/s compared to 18 m3/s of treated sewage water. The objective of the present study is to develop a model for water quality in urbanized, tidal and polluted rivers in order to clarify the characteristics of organic matter, DO, phosphorus and nitrogen.

OUTLINE OF TARGET AREA AND METHODS Target area: tidal rivers of Neya River and its tributaries and downstream Target rivers in this study are shown in Fig. 1. Neya River and its tributaries have a low-lying basin with an area of 267.6 km2 surrounded by Ikoma Mountains on the east, Yodo River on the north, on the south and Uemachidaichi Hill on the west. Neya River has many tributaries such as Furu River, Diani-Neya River, Onchi River and Hirano River (hereafter referred to Neya River and its tributaries as Neya River system which does not contain other water system except surface runoff). There is only one downstream end, Kyobashiguchi, which is located at the north end of Uemachidaichi Hill and the water flows into Okawa River which is one of the former main streams of Yodo River. Yodo River was cut and the present main stream was constructed as a wide floodway connected with directly after two massive floods in 1885 and 1891. Thus, Okawa River is controlled by a weir at the upstream end where it connects to the main stream of Yodo River. At the downstream end of Okawa River the flow divides into Higashi-yokohori canal and two rivers (Dojima River and Tosabori River) which meet and divide again into three river mouths. The river mouths are located in Port of Osaka. The upstream and the downstream side of Kyobashiguchi are hereafter referred to as Neya River system and Osaka City Rivers (OCR), respectively. Calculations were done between the port and the upstream end of Okawa River and the upstream weirs and dams of Neya River system which is still influenced by tides.

- 232 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

Fig. 1 - Neya River system and its downstream.

Legally-imposed field surveys in this area have been carried out by local governments every season at 12 stations in OCR among public water areas defined by the Water Pollution Prevention Act. Living Environment Items such as BOD (biochemical oxygen demand), CODMn (chemical oxygen demand), SS (suspended solids) and DO (dissolved oxygen), and Health Items such as cadmium and total mercury have been measured in these public surveys (Osaka Prefectural Government, 2013). In both rivers BOD shows a gradual decrease after 1990s, but remains stubbornly high and is sometimes higher than the environmental standard.

A field survey was conducted to clarify the distribution of nitrogen and phosphorus in OCR at the four tidal hours on April 25, 2006. Some of the results were reported by Irie et al. (2008). In this survey, water at the surface and bottom at 25 points was sampled and inorganic nitrogen, total nitrogen (TN), phosphate, and total phosphorus (TP) were measured. Water temperature, salinity, turbidity, and DO were also measured in situ by a CTD (conductivity, temperature and depth) sensor and DO meter. According to this survey, the water flowing from Neya River system flows down and back again, have to take longer time than semidiurnal tidal cycle to go down through the port. Consumption of DO occurs in the bottom water during the flow into OCR. Saline water comes into the streams up to 8 – 10 km away from the river mouths. A halocline is formed throughout this distance.

Flow and biochemical model: EFDC The river network in OCR consists of narrow canals that were previously used for transportation by barge and they spread widely with many confluences and branches. A horizontal 2D model is necessary to model the process of the flow. As there are clear vertical differences of salinity, temperature and water quality in the downstream due to the halocline, a vertical 2D model is needed to represent such vertical characteristics. Thus, a 3D flow model was applied directly in this study.

- 233 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

The EFDC (Environmental Fluid Dynamics Code) was originally developed by Hamrick (1992), and has been supported and developed by United States Environmental Protection Agency (U.S.EPA) (Tetra Tec Inc., 2002) and provides flow, sediment transport, toxic substances and water quality information for models. The EFDC has wide applicability in the simulation of a variety of 1D – 3D water dynamics. In 3D dynamic solutions, sigma terrain-following vertical coordinate, hydrostatic, free-surface, turbulence-averaged equations are solved with the Boussinesq approximation like other 3D physical models based on Blumberg and Mellor (1987). The finite difference method and staggered C grid are used for discretization. It adopts a mode-splitting scheme, which means the model calculates vertically averaged velocity field and free surface in “external mode” and 3D velocity field and substance transport such as salinity, water temperature, and other water quality state variables in “internal mode”. Although some options are implemented for the vertical viscosity and diffusion in this code, but two turbulence parameter model “Mellor-Yamada level 2.5 turbulence closure scheme” (Mellor and Yamada, 1982) as modified by Galperin et al. (1988) was used in this study. For the horizontal viscosity and diffusion, Smagorinsky-type subgrid formulation was used (Smagorinsky, 1963).

The water quality sub-model in EFDC was originally based on WASP5 EUTRO model developed by U.S.EPA (Ambrose et al., 1993) and implements two kinds of this simple sub-model and another full-version model. The full-version model used in this paper is functionally equivalent to CE-QUAL-ICM developed by Hydroqual Inc. for U.S. Army Corps of Engineers (Di Toro and Fitzpatrick, 1993) and Chesapeake Bay Water Quality model (Cerco and Cole, 1993). The water quality model originally has 21 state variables in five groups:

1) cyanobacteria, diatom algae, green algae, stationary algae; 2) refractory particulate organic carbon, phosphorus, and nitrogen(RPOC, RPOP, and RPON, respectively); labile particulate organic carbon, phosphorus, and nitrogen (LPOC, LPOP and LPON, respectively); dissolved organic carbon, phosphorus, and nitrogen (DOC, DOP and DON, respectively); 3) total phosphate(PO4), ammonia(NH4), and nitrate which represents the sum of nitrate and nitrite (NO2 + NO3); 4) particulate biogenic silica, dissolved available silica, and total active metal; 5) chemical oxygen demand and dissolved oxygen.

A simple schematic of the model is shown in Fig. 2.

Fig. 2 - Schematic of the water quality sub-model of EFDC.

- 234 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

The governing mass-balance equation for the water quality state variables is expressed as:

휕(푚푥푚푦퐻퐶) 휕 휕 휕 + (푚 퐻푢퐶) + (푚 퐻푣퐶) + (푚 푚 푤퐶) 휕푡 휕푥 푦 휕푦 푥 휕푧 푥 푦 휕 푚푦퐻퐴ℎ 휕퐶 휕 푚푥퐻퐴ℎ 휕퐶 휕 퐴푣 휕퐶 = ( ) + ( ) + (푚푥푚푦 ) + 푆푐 (1) 휕푥 푚푥 휕푥 휕푦 푚푦 휕푦 휕푧 퐻 휕푧 where C is the concentration of a water quality state variable [g/m3]; u, v and w are the velocity components [m/s]; mx and my are the horizontal curvilinear coordinate scale factors; H is the water column depth [m]; Ah and Av are the horizontal and vertical 2 turbulent diffusivities [m /s], Sc is the internal and external sources and sinks per unit volume [g/m3]. The last term in the right-hand side represents the kinetic process and external loads for each water quality state variable. The present model solves equation (1) using a fractional step procedure, which calculates the source and sink term once after the procedure calculate the physical terms. Table 1 shows typical expressions of the source and sink terms for algae, nitrogen cycle and DO. For a detailed description refer to Tetra Tech Inc. (2007).

Physical settings The domain includes Port of Osaka and Neya River system in addition to OCR. There are three river mouths where fresh water flows into the port as shown in Fig. 1. This geological characteristic makes it difficult to simulate with the combination of 1D unsteady flow along each river and the 2D vertical flow model for the variety of vertical profiles. There is only one set of observed data of tidal water levels in the port. If a 1D longitudinal model is used, the port would have to be divided into 3 sections and a model would have to be developed that would allow for adjustment of the water levels in each section as the water levels at the river mouths have not been recorded and cannot be estimated easily. Besides, when the combination of two or more 1D- or vertical 2D- models is applied, settings of connective inflowing and outflowing discharges and flux of water quality variables are arduous. Therefore, a 3D model is employed directly. This leads to wide applicability in the evaluation of environmental policies that includes large changes of river and treated wastewater discharges and flux of substances – for example, release of advanced treated wastewater to a different upstream river which is one of the water environmental policies of Neya River system planned by the local government. The influence of tidal change on the water level reaches beyond Kyobashiguchi, which is the boundary between OCR and Neya River system. The domain for calculation includes part of Neya River system due to this influence and the setting of upper boundary conditions. As the river discharge flowing from upstream is calculated from the data obtained at gauging stations and weirs, the possible locations to set the upstream boundary conditions are limited. Figure 3 shows the domain and grids in the settings. The upstream boundaries are set at the points where there is no influence of tides, even at high spring tide, except in Okawa River, Hirano River and its diversion aqueduct, and Daini-Neya River. In these four rivers, upstream boundaries are set at weirs. The downstream, open-ocean boundaries are set in Port of Osaka between reclaimed islands.

- 235 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

The grids were generated under an orthogonal curvilinear coordinate system. The cross sections of rivers consist of one grid in Neya River system and 2 – 12 grids in OCR. The lengths of grids are 20 – 400 m. A water column is divided into 10 levels with σ coordinates. Nine sewage plants are also considered in the area.

Table 1 - Biochemical source and sink terms for water quality model (Tetra Tech Inc., 2007). Indices Biochemical terms 휕퐵 휕 푊퐵 푥 = (푃 − 퐵푀 − 푃푅 )퐵 + (푊푆 ∙ 퐵 ) + 푥 휕푡 푥 푥 푥 푥 휕푍 푥 푥 푉 3 -1 Bx: algal biomass of algal group x [g C/m ], Px: production rate of algal group x [day ], BMx: basal Algae -1 -1 metabolism rate of algal group x [day ], PRx: predation rate of algal group x [day ], WSx: positive settling velocity of algal group x [m/day] and WBx: external loads of algal group x [g C/day], V: control volume [m3] 휕푅푃푂푁 = ∑ (퐹푁푅 ∙ 퐵푀 + 퐹푁푅푃 ∙ 푃푅 ) ∙ 퐴푁퐶 ∙ 퐵 − 퐾 ∙ 푅푃푂푁 휕푡 푥 푥 푥 푥 푥 푥 푅푃푂푁 푥=푐,푑,푔,푚 휕 푊푅푃푂푁 + (푊푆푅푃 ∙ 푅푃푂푁) + RPON 휕푍 푉 FNRx: fraction of nitrogen metabolized by algal group x produced as refractory particulate organic nitrogen, FNRPx: fraction of nitrogen in predated algal group x converted to refractory particulate organic nitrogen, ANCx: nitrogen-to-carbon ratio in algal group x [g N/g C], KRPON: hydrolysis rate of -1 refractory particulate organic nitrogen [day ] and WRPON: external loads of refractory particulate organic nitrogen [g N/day] 휕퐷푂푁 = ∑ (퐹푁퐷 ∙ 퐵푀 + 퐹푁퐷푃 ∙ 푃푅 ) ∙ 퐴푁퐶 ∙ 퐵 + 퐾 ∙ 푅푃푂푁 + 퐾 휕푡 푥 푥 푥 푥 푥 푥 푅푃푂푁 퐿푃푂푁 푥=푐,푑,푔,푚 퐵퐹퐷푂푁 푊퐷푂푁 ∙ 퐿푃푂푁 − 퐾퐷푂푁 ∙ 퐷푂푁 + + DON ∆푍 푉 FNDx: fraction of nitrogen metabolized by algal group x produced as dissolved organic nitrogen, FNDPx: fraction of nitrogen in predated algal group x converted to dissolved organic nitrogen, KLPON: -1 hydrolysis rate of labile particulate organic nitrogen [day ], KDON : mineralization rate of dissolved organic nitrogen [day-1], BFDON: benthic flux of dissolved organic nitrogen[g C/m2/day], Z: thickness of the layer at the bottom and WDON: external loads of dissolved organic nitrogen [g N/day] 휕푁퐻4 = ∑ (퐹푁퐼 ∙ 퐵푀 + 퐹푁퐼푃 ∙ 푃푅 − 푃푁 ∙ 푃 ) ∙ 퐴푁퐶 ∙ 퐵 + 퐾 ∙ 퐷푂푁 휕푡 푥 푥 푥 푥 푥 푥 푥 푥 퐷푂푁 푥=푐,푑,푔,푚 퐵퐹푁퐻4 푊푁퐻4 − 퐾푁푖푡 ∙ 푁퐻4 + + NH4 ∆푍 푉 FNIx : fraction of nitrogen metabolized by algal group x produced as inorganic nitrogen, FNIPx: fraction of nitrogen in predated algal group x converted to inorganic nitrogen, PNx : preference for -1 ammonium uptake by algal group x [0

- 236 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

Osaka City Neya River Rivers(OCR) system

Furu R.

Neya R.

Okawa R. W1 W2 Dojima R. Neya R.

Daini-Neya R. Tosabori R. W6 W5 W3 Onchi R. Aji R.

W9 Hirano R. Diversion Hirano R. Shirinashi R. W4 W7 Daini-Neya R. W8 Port of Osaka Kizu R.

Fig. 3 - Domain and grids and the location of sewage plants for physical settings.

To reproduce the hydrodynamics on April 25, 2006 when the field survey was conducted by us, the model was run from April 1. Air temperature, solar radiation and wind were taken from hourly observations at Osaka District Meteorological Observatory which is located in the area and used to calculate heat flux at water surface in the hydrodynamic model. Water temperature and solar radiation were also used to represent the dependency of algal growth. Since the rivers are very narrow and enclosed by high sheet pile walls and vertical concrete walls, and the wind at water surface was not observed, wind stress was not considered. Precipitation was used to set rainfall flux on water surface, but not done to recalculate inflowing discharges from sewage plants. Initial values of water temperature and salinity in OCR were determined by reference to the observations of April 25. Initial water temperature and salinity in Okawa River were set at 13.3°C and 0.01, respectively, and those in Neya River system were set at 14.0°C and 0.01, respectively. Discharge, temperature and salinity of inflow from the upper boundaries were determined by the observations taken by the Ministry of Land, Infrastructure, Transport and Tourism (MLIT) and the database of environment in public waters deployed by (Osaka Prefectural Government, 2013). The hourly discharge and water temperature were set at the upper boundary of Okawa River. The discharge and water temperature at the upper boundaries in Neya River system were set by reference to those observed on another day in April, 2006.

On the lower boundaries located at the entrance of Port of Osaka, water level is given hourly by the tidal level observed in the port (JMA, 2013). Salinity and water temperature were set by reference to the observations on that month and our field survey on April 25. The salinity and water temperature were respectively 31 and 13°C at the surface and 35 and 11°C at the bottom. Values in the middle layer were set by linear interpolation.

- 237 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

There are many tributaries, conduits and drains in the target area. Sixteen water inflows in Neya River system add their discharge to that of the main streams and were therefore considered in this simulation. They were calculated from the difference achieved by subtracting the average discharge of the main streams from the discharge at Kyobashiguchi, the lower end of Neya River system. The inflowing discharges from the upstream boundaries in Neya River system were 0.1 – 0.45 m3/s and water temperature of the inflows were set at 15.1 – 16.2°C. Lateral inflows with the discharge of 0.24 – 1.9 m3/s and temperature of 15.1 – 16.8°C were also considered in Neya River system. Okawa River has a large discharge of 70 – 120 m3/s. Discharge amounts and temperature of treated sewage from nine plants were taken from sewage statistics compiled by Osaka prefectural and city government. Those discharges changed from 0.52 m3/s (Plant W7) to 3.4 m3/s (Plant W1). The water temperature of the treated sewage was set at a constant value of 20.0°C in this calculation.

Biochemical settings This simulation considered 19 state variables excluding stationary algae and total active metal considered in the original sub-model. The initial distributions of these concentrations were set at one value without spatial change at the start time on April 1. The values were given by the data gathered by local governments in May, 2006 (Osaka Prefectural Government, 2013) because the detailed data was not observed in April. In this application the hydrodynamics assumed more important role than biochemical reactions in the reproducibility of the distributions of water quality parameters. It is because the time required for river water to flow down to the river mouth was very short and might be less than two days. Appropriate setting of the boundary conditions of water quality state variables are more important than the tuning of the kinetic parameters. The default values (Cerco et al., 2000) and empirical values are used for the kinetic parameters in the present water quality model. The empirical values were set based on our numerical simulation in Osaka Bay (Irie et al., 2004). The boundary conditions were also taken from the same observations. The simulation was run from April 1 to reproduce a similar spatial distribution as observed on May 16, 2006. Table 2 briefly shows the initial and boundary conditions of the water quality state variables. The concentrations of the water quality state variables were based on the data of treated sewage compiled by Osaka prefectural and city government and verified to reproduce their spatial distributions. The rate of oxygen consumption by the river bottom were set to be 1.5 g/m2/day (upper section) to 7.0 g/m2/day (port) by reference to a field survey (Irie et al., 2010).

- 238 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

Table 2 - Concentrations of water quality parameters for initial and boundary conditions. Parameters [mg/L] Conditions RPOC LPOC DOC RPOP LPOP DOP PO4 RPON LPON DON NH4 NO2+NO3 COD DO 0.300 0.300 3.300 0.002 0.002 0.015 0.050 Initial conditions 0.250 0.240 0.710 0.030 0.680 2.000 11.000 Upstream 0.400 0.400 2.000 0.004 0.005 0.010 0.050 Okawa River boundary 0.060 0.060 0.100 0.030 0.680 2.000 9.400 conditions 0.500 0.500 10.000 0.080 0.080 0.600 0.730 Hirano River 0.500 0.500 1.000 2.300 1.900 8.000 3.300 0.200 0.200 7.800 0.030 0.030 0.060 0.320 Furu River 0.100 0.100 0.200 1.100 3.600 8.000 4.500 0.400 0.400 5.000 0.005 0.005 0.010 0.090 Neya River 0.110 0.110 0.300 0.020 1.360 2.000 11.000 0.500 0.500 7.000 0.030 0.030 0.100 0.220 Daini-Neya River 0.200 0.200 1.000 0.490 2.200 2.000 7.900 1.000 1.000 10.000 0.060 0.060 0.200 0.400 Onchi River 0.900 0.800 2.000 3.500 0.200 4.000 5.800 Lateral Tributaries of 0.400 0.400 4.400 0.010 0.010 0.020 0.270 boundary Neya River 0.010 0.010 0.080 0.200 1.500 4.000 9.300 conditions Tributaries of 0.500 0.500 7.000 0.060 0.060 0.120 0.240 Onchi River 0.200 0.200 0.700 1.800 1.600 4.000 3.900 Sewage 1.000 1.000 10.000 0.010 0.010 0.010 0.360 W1 plants 0.400 0.500 1.000 0.760 12.300 6.000 10.000 (compen 1.000 1.000 10.000 0.500 0.500 0.120 0.100 W2 -dium) 0.200 0.200 1.150 6.150 4.800 6.000 10.000 0.600 0.600 7.000 0.010 0.010 0.010 0.680 W3 0.500 0.500 1.000 0.180 6.300 8.000 10.000 0.600 0.600 6.900 0.010 0.010 0.010 0.360 W4 0.410 0.410 1.000 6.450 4.230 6.000 10.000 0.500 0.500 7.450 0.100 0.100 0.100 1.200 W6 east 0.020 0.020 0.050 1.700 8.360 2.000 10.000 1.000 1.000 9.500 0.100 0.100 0.100 0.600 W6 west 0.150 0.150 0.360 10.350 4.490 2.000 10.000 1.500 1.500 11.000 0.010 0.010 0.010 0.360 W8 0.300 0.300 1.050 0.350 13.500 2.000 10.000 0.400 0.400 6.050 0.010 0.010 0.010 0.400 W9 0.800 0.800 1.000 3.200 1.740 2.000 10.000

RESULTS AND DISCUSSION Water levels, salinity and temperature In Fig. 1, gauge stations for Osaka Prefecture to measure water levels in the area are denoted by triangles and names initialized by “G”. Figure 4 shows the calculated water levels and the observations for the last 10 days of April. The calculated peak times induced by tidal change are coincident with those in the observations at most stations even though a spatial difference in the time was observed. At Sta. G5 and G6, located upstream, there was a small difference in the phase. Calculated water level has good correspondence with the observations, but the water level upstream in Neya River system at low tide is 50 cm lower than the observations at maximum. Even though the models have sufficient reproducibility of the water levels when used on oceans and lakes, this implies that it is difficult for this kind of 3D model to reproduce the water levels when there is a wide spatial variety of rivers and the river discharge is extremely small. The reproducibility in OCR, however, is highly credible due to Neya River system acting as a buffer zone at the upper boundaries. It should be mentioned especially, that the change of the calculated water level at Sta. G3 is coincident with the

- 239 - Journal of Water and Environment Technology, Vol.13, No.3, 2015 observations, even though there are three river mouths on the downstream side of OCR where sea water intrudes along the bottom.

The reproducibility of the longitudinal, sectional distributions of salinity and temperature is the most important factor for model users and therefore receives the most attention. Notably, as the stream branches off and rejoins in the zone where saline sea water intrudes at the bottom, it is more difficult to represent. There are no observations of the longitudinal distributions of salinity and temperature, except in the data measured by Irie et al. (2008). Here, the reproducibility on the survey date at the four tidal hours is evaluated. The observation stations are denoted by filled circles and names initialized by “A” in Fig. 1. A small boat was used in the observation to cruise through the points at four tidal hours: flood, high, ebb and low tides. That means one sectional distribution has small temporal variety spatially. Figure 5 shows the distributions of calculated and observed salinity in Aji River. In the calculation, the saline seawater intrudes into the upstream side of Sta. A4 along the river bottom. The vertical stratification of salinity keeps the strength even though the water depth is less than 5 m. The simulation has better representations at high and ebb tides than at low and flooding tides. Especially at the flooding tides, the distance from the river mouth to the upstream end which the saline water intrudes to is different between the computed results and observations. It is due to the temporal difference in the distributions of observed values. The survey boats could not keep up with the reciprocation of the intruding sea water. The distributions of temperature are well reproduced (figures are not shown due to space constraints). The distribution of the observed temperature has the highest value in the middle of the fresh water mass, which might have originated from a sewage plant every tidal hour. The simulation represents the highest water mass except at ebb tide. The water mass with the highest temperature disappears because the mass drifted further towards the downstream side than the observations. As two ends of the calculated reciprocation correspond to the observed ends, there is not much difference between the values in the observation and calculation. The total flux of calculated passive substances including salinity, temperature and other water quality parameters for one tidal cycle can be used to discuss the dynamics of nitrogen and phosphorus.

- 240 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

Fig. 4 - Temporal change of water levels in OCR and Neya River system.

Fig. 5 - Longitudinal sectional distributions of calculated and observed salinity in Aji River and the observed values shown by Irie et al. (2008). (See the location in Fig.1.)

- 241 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

Nitrogen and phosphorus First, the calculated concentrations of the water quality parameters are compared with the observations measured by local governments. Figure 6 shows the location of observation stations. The results of municipal surveys for public waters are used because the data for various kinds of water quality parameters measured at the same time in the entire area of OCR and Neya River system are limited. Irie et al. (2008) showed the characteristics of how the water mass with a high concentration of nitrogen and phosphorus flows through OCR. The water coming from Okawa River contains a lower concentration of nitrogen and phosphorus because this water originally comes from Yodo River from which the river discharge is much larger than that of Neya River system. The water mass supplied from Neya River system contains much larger quantities of nutrients and organic matter than that from Yodo River. Figure 7 shows the comparison of the measured COD, DO, TN and TP at survey points for the public water area on May 16, 2006 and the calculated values on May 16, 2006 at 15:00.

The observed concentrations of COD in Neya River system are higher than those in OCR, showing almost 8 mg/L or more except in the uppermost part of Neya River. Those in Onchi River and Hirano River especially show the highest concentrations. Those in OCR show lower concentrations of less than 6 mg/L, because the water coming from Okawa River dilutes the concentration. The computed results agree well with the observations except in Onchi River and Kyobashi (Sta. 5) where there are differences of the concentrations of all four variables. The cause of the difference in Onchi River is not clear. The concentration at Kyobashi is variable due to tides. The time when the local government measured at the same point is not same as that at which the calculated values are shown here.

Fig. 6 - Location of observation stations for the comparison of water quality parameters.

- 242 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

Fig. 7 - Longitudinal profiles of COD, DO, TN and TP. (Red denotes the observations, blue the calculated results.)

The observed DO was higher in the upper zone of OCR and the upper zones of Neya River and Daini-Neya River above confluences with other rivers in which DO is lower. DO was less than 4 mg/L in the lower zone of OCR where it is located near the port and has deep depth. The model concentrations of DO agree better where they are high and worse where they are low. This discrepancy might be caused by the fact that some biochemical reactions, such as the consumption by organic matter in a river bed and/or the first reaction of treated sewage water when it flows into a river are not sufficiently modeled or verified due to the lack of fundamental data.

The concentrations of TN and TP are higher in Neya River system, especially in the lower zones which are influenced by the sewage plants than those in OCR. While the model TN and TP agree well in the entire areas, the small differences can be seen at some points. This is most likely caused by the uncertainty of the inflow flux of TN and TP from sewage plants.

Table 3 shows the results of sensitivity analyses to evaluate the impact of the biochemical reaction on the surface distributions of TN and DO. The values are the average of the difference by subtracting the original results from the new results in three areas of the river system. The impact is very small because the water goes through to the port quickly.

Effect due to the Introduction of 3D Model The effect caused by the introduction of the 3D model appears to be the simultaneous reproduction of the vertical, longitudinal-sectional distributions (as shown in Fig. 7) and the horizontal distributions of substances with widely varied concentrations. Figure 8 shows the horizontal distributions of observations and the calculated results of TN in the upper zone of OCR. The water with a high concentration of TN flowing from Neya River system joins with the water of low concentration flowing from Okawa River just

- 243 - Journal of Water and Environment Technology, Vol.13, No.3, 2015 after passing through Kyobashiguchi (the point indicated by the X mark in Fig. 1). However, these two waters are not well mixed in the confluent zone, and then they flow into two divided streams, Dojima and Tosabori rivers. In Tosabori River, TN is higher than that in Dojima River because Tosabori River is more influenced by Neya River system. This different characteristic of Tosabori and Dojima rivers has different influence on the water quality in Aji and Kizu rivers which are located downstream. Water in that flows southward is more influenced by Neya River system and more contaminated with TN than Aji River. The model simulates well the lack of mixing in the confluent zone and that the high TN water of Neya River has more influence on Tosabori and Kizu rivers.

Table 3 - Sensitivity analyses of surface TN and DO. Difference of TN [mg/L] Difference of DO [mg/L] Parameter conditions downstream middle upstream downstream middle upstream area area area area area area Maximum growth rate (110%) -0.01 -0.01 0.00 0.01 0.01 0.02 Maximum growth rate (90%) 0.00 -0.02 0.00 0.00 0.02 0.01 Basal metabolism rate (120%) -0.01 -0.01 0.01 0.01 0.00 0.03 Basal metabolism rate (80%) -0.01 -0.02 -0.01 0.01 0.02 0.01

Fig. 8 - Horizontal distributions of TN [mg/L] in the upper zone of OCR.

- 244 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

Temporal and spatial change of water quality parameters in tidal cycles Figure 9 shows the temporal change of TN and water levels at Sta. 24 (close to the port), 22 (confluent area already mentioned above), 5 (Kyobashi) and 3 (middle of Neya River). In the areas close to the port or in Neya River, the temporal changes of the concentration of TN are smaller than those at the other two stations. At Sta. 24 the influence of Neya River system is smaller than the other river mouths and the influence appears at low tides when the water originating in Neya River system reaches here. At Sta. 3 the influence of tidal change appears at high tides because the less contaminated water flows down to this point at other tide periods, but the water containing the treated sewage water with the high concentration of TN flows upward to this point in the inverse. Since Sta. 5 is located at the downstream end of Neya River system, the water of Neya River system is dominant, but the low contaminated water originating from Okawa River flows upstream at the high tides. That is why there is a wide range of change here. The same phenomenon is simulated at Sta. 22, but there are two minimal values in a period of high tide. This implies the complicated mixing system around the station. Figure 10 shows the horizontal distribution of TP at low and flood tides. This shows that the water originating from Neya River cannot go into the port in one tidal cycle of 12.5h even though the distance between the port and Kyobashiguchi is only about 10 km. This system gives more time for organic matter to settle to the river bed and consume oxygen while the water with high contamination goes through in OCR. These results show the possible implementation of water environmental policies that improve the water quality of a small section and reduce the residence time of Neya River water by modifying the outflow discharge and time of the treated water from the sewage plants.

4.0 2.5 12.0 2.5 3.5 Sta. 24 Sta. 22 10.0 Water level [m] 2.0 Water level [m] 2.0 3.0 8.0 2.5 1.5 1.5 2.0 6.0 [mg/L] [mg/L] 1.5 1.0 1.0 4.0 TN 1.0 TN 0.5 2.0 0.5 0.5 0.0 0.0 0.0 0.0 22 22 23 23 24 24 25 25 26 26 27 22 22 23 23 24 24 25 25 26 26 27 day day 12.0 2.5 4.0 2.5 Sta. 5 3.5 Sta. 3 10.0 Water level [m] 2.0 Water level [m] 2.0 3.0 8.0 1.5 2.5 1.5 6.0 2.0 [mg/L] 1.0 1.5 1.0 4.0 TN TN [mg/L] TN 1.0 2.0 0.5 TN 0.5 0.5 Water level 0.0 0.0 0.0 0.0 22 22 23 23 24 24 25 25 26 26 27 22 22 23 23 24 24 25 25 26 26 27 day day Fig. 9 - Temporal changes of TN and water levels at four stations.

- 245 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

Fig. 10 - Horizontal distribution of TP at a low and a flood tides.

CONCLUSIONS In this paper we simulated, via a 3D flow and water quality model, the processes induced by three kinds of waters: saline sea water, less organic Okawa River water originating from Yodo River, and water that contains treated sewage water with high concentrations of nutrients originating from Neya River system. The targeted rivers have a complicated spatial structure with three river mouths, many upstream boundaries, and many branches and confluences. This is the largest reason for applying the 3D model directly and the 3D model can obtain good reproduction of the horizontal and vertical characteristics of salinity, temperature, nutrients and other water quality parameters and the process of water flowing from the sources of two rivers and sewage plants through the port. However, some challenges remain. The height and phase of water level change are well represented except in the upstream area with a small river discharge. Since the water depth is small compared to a relatively large oscillation of tides and the current is slow in that area, it might be pointed out that there is a limitation of the simulation using a 3D model with the feature of σ coordinates. Additional field surveys are necessary for the simulation to accurately represent the consumption of DO because the representation showed that the consumption rate of DO varies throughout the rivers. A 3D water quality model, which is a potent tool in ocean and lakes, proved to be a versatile enough tool to simulate small, but complicated and tidal rivers. This is especially true in the case examined in this study, and is shown by the correspondence of the change of the water levels at the three river mouths, the characteristics of the intrusion of saline water and the mixing process of the two different waters. This creates calculated water quality parameters that agree well with the observations.

ACKNOWLEDGMENTS A part of this work was supported by JSPS KAKENHI (Grants-in-Aid for Scientific Research) Grant Number 25630208. We thank MLIT and local governments for providing data and Yosuke Ida for preparing some figures. We also thank anonymous reviewers and editors whose comments led to substantial improvements.

REFERENCES Ambrose R. B., Wool T. A. and Martin J. L. (1993) The water quality analysis and

- 246 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

simulation program, WASP5: Part A, model documentation version 5.1. U.S. Environmental Protection Agency, Environmental Research Laboratory, Athens, GA, USA. Blumberg A. F. and Mellor G. M. (1987) A description of a three-dimensional coastal ocean circulation model. In: Three-dimensional Coastal Ocean Models, Coastal and Estuarine Sciences, 4, N. S. Heaps (ed.), American Geophysical Union, Washington DC., USA, pp.1-19. Cerco C. F. and Cole T. (1993) Three-dimensional eutrophication model of Chesapeake Bay. J. Environ. Eng., 119(6), 1006-1025. Cerco C. F., Bunch B. W., Teeter A. M. and Dortch M. S. (2000) Water quality model of Florida Bay. ERDC/EL TR-00-10, U.S. Army Engineer Research and Development Center, Vicksburg, MS, USA. Di Toro D. M. and Fitzpatrick J. F. (1993) Chesapeake Bay sediment flux model. Contract Report EL-93-2, prepared for U.S.EPA Chesapeake Bay Program, U.S. Army Engineer District and U.S. Army Engineer Waterways Exp. Sta. HydroQual, Inc., Mahwah, NJ, USA. http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA267189 (accessed April 1, 2015). Galperin B., Kantha L. H., Hassis S. and Rosati A. (1988) A quasi-equilibrium turbulent energy model for geophysical flows. J. Atmos. Sci., 45(1), 55-62. Hamrick J. M. (1992) A Three-dimensional environmental fluid dynamics computer code: Theoretical and computational aspects. Special Report 317, The College of William and Mary, Virginia Institute of Marine Science, VA, USA. Huang W. and Spaulding M. (1995) Modeling of CSO-induced pollutant transport in Mt. Hope Bay. J. Environ. Eng. ASCE, 121(7), 492-498. Huang W., Liu X., Chen X. and Flannery M. S. (2011) Critical flow for water management in a shallow tidal river based on estuarine residence time. Water Resour. Manage., 25(10), 2367-2385. Irie M., Nakatsuji K. and Nishida S. (2004) Numerical Simulation of Dynamics of Anoxic Water in Osaka Bay. Annu. J. Coast. Eng., JSCE, 51, 926-930. (In Japanese) Irie M., Nishida S. and Shoji Y. (2008) Temporal and spatial characteristics of water quality and its responses to tide in urban tidal river network. Annu. J. Hydraulic Eng., JSCE, 52, 1099-1104. (in Japanese) Irie M., Teranaka K., Yamaguchi T. and Nishida S. (2010) Applicability of nutrient and sulfur benthic model to sediment in river mouth. Annu. J. Hydraulic Eng., JSCE, 54, 1639-1644. (in Japanese) JMA (2013) Chosekikansokushiryo (Tidal observation database). Japan Meteorological Agency, Tokyo, Japan. http://www.data.jma.go.jp/gmd/kaiyou/db/tide/genbo/ (accessed April 1, 2015). (in Japanese) Lin J. and Kuo A. Y. (2003) A model study of turbidity maxima in the York River Estuary,Virginia. Estuaries, 26(5), 1269-1280. Liu W., Chen W., Kuo J. and Wu C. (2008) Numerical determination of residence time and age in a partially mixed estuary using three-dimensional hydrodynamic model. Cont. Shelf Res., 28(8), 1068-1088. Mellor G. L. and Yamada T. (1982) Development of a turbulence closure model for geophysical fluid problems. Rev. Geophys. Space Phys., 20(4), 851-875. Osaka Prefectural Government (2013) Kokyoyosuiikito Kankyo Database System (Database of environment in public waters). Osaka Prefectural Government, Osaka, Japan. http://www.pref.osaka.lg.jp/kankyohozen/osaka-wan/database.html (accessed

- 247 - Journal of Water and Environment Technology, Vol.13, No.3, 2015

April 1, 2015). (in Japanese) Shen J. and Haas L. (2004) Calculating age and residence time in the tidal York River using three-dimensional model experiments. Estuarine Coastal Shelf Sci., 61(3), 449-461. Smagorinsky J. (1963) General circulation experiments with the primitive equations, part I: the basic experiment. Mon. Weather Rev., 91(3), 99-152. Tetra Tec Inc. (2002) User’s manual for environmental fluid dynamics code hydro version(EFDC-Hydro). Tetra Tec Inc., Fairfax, VA, USA. http://www.epa.gov/athens/wwqtsc/html/efdc.html (accessed April 1, 2015). Tetra Tech Inc. (2007) The environmental fluid dynamics code theory and computation. Volume 3, Water Quality Module. Tetra Tech Inc., Fairfax, VA, USA.

- 248 -