desalination brine discharge modelling coupling of ... · prof., director, institute for...

MWWD 2008

Desalination brine discharge modelling - Coupling of Hydrodynamic Models for Brine Discharge Anal.doc - page 1 of 18

Desalination brine discharge modelling

Coupling of Hydrodynamic Models

for Brine Discharge Analysis

Anne Niepelt (1), Tobias Bleninger (2) and Gerhard Jirka (3)

Keywords: CORMIX, Delft3D, near-field, far-field, dense plumes, negatively buoyant Abstract The waste stream from sea water desalination plants, called brine or concentrate, is commonly discharged into coastal waters associated with several potential negative impacts on the marine environment. The few existing environmental regulations generally distinguish between local impacts around the discharge point and regional impacts further away from the source. Different hydrodynamic models are used for the prediction of either the near-field mixing (e.g. CORMIX) and/or the transport processes in the far-field (e.g. Delft3D). An optimized approach to couple both model types for brine discharge analysis has been developed and tested for a case study. The coupling algorithm includes the transformation of the output data of the near-field model CORMIX into the input data for the far-field model Delft3D-FLOW. Calculations indicate that the far-field model alone (without coupling) can not simulate the vertical concentration distribution of the plume in contrast to the near-field model. Thus, a model coupling is unavoidable for an environmental assessment. The coupling methodology, though simple, allows for an considerably improved discharge assessment. This allows for an optimized environmental hydraulic design of the outfall structure. Introduction In the last years seawater desalination has become an efficient and essential technique to produce fresh water. Thus seawater desalination contributes positively to humanity but it may also cause negative impacts on the marine environment. The waste stream from desalination plants, called brine or concentrate, is commonly discharged into coastal waters. However, brine disposal is associated with several potential negative impacts on the marine environment carrying different harmful substances (additives, corrosion by-products) and a high salt content and/or temperature into the ocean (Lattemann & Höpner, 2003). The few existing environmental regulations generally distinguish between local impacts around the discharge point and regional impacts further away from the source. Thus, the planning of brine discharges needs to optimize the design and the siting of the outfall to ensure high initial dilutions in the near-field to reduce local impacts and good siting in regions with good flushing characteristics in the far-field to reduce regional impacts (Bleninger & Jirka, 2007, 2008; Jirka, 2008).

1 Dipl.-Ing., Research assistant, Institute for Hydromechanics, University Karlsruhe, Kaiserstr. 12, 76131 Karlsruhe, Germany, [email protected], www.ifh.uni-karlsruhe.de 2 Dr.-Ing., Research assistant, Institute for Hydromechanics, University Karlsruhe, Kaiserstr. 12, 76131 Karlsruhe, Germany, [email protected], www.ifh.uni-karlsruhe.de 3 Prof., Director, Institute for Hydromechanics, University Karlsruhe, Kaiserstr. 12, 76131 Karlsruhe, Germany, [email protected], www.ifh.uni-karlsruhe.de

mailto:[email protected]

http://www.ifh.uni-karlsruhe.de/





MWWD 2008 – IEMES 2008

Different hydrodynamic models are used for the prediction of either the near-field mixing (e.g. CORMIX) and/or the transport processes in the far-field (e.g. Delft3D) since it is not possible to calculate the mixing characteristics in one over-all model. An optimized approach to couple both model types for brine discharge water analysis (also applicable for other effluent discharges) has been developed and tested for a case study. The coupling algorithm includes the transformation of the output data of the near-field model CORMIX into the input data for the far-field model Delft3D-FLOW. The main aim of the coupling approach is to improve optimal siting and design of an outfall to minimize the negative impacts on the environment. Desalination processes and brine discharge characteristics Seawater desalination produces fresh water by separating dissolved salts from saline water. The remaining water stream is called brine or concentrate characterized by an increased salt content. Nowadays two desalination processes are prevailing: the so-called multistage flash distillation (MSF) and the reverse osmosis (RO) making up about 89% of the total capacity (Glade, 2005). Developed from the first used desalting technique by thermal evaporation MSF is used predominantly with 62% market share. In the MSF process heated seawater passes through a series of vessels called stages where the ambient pressure gets progressively lower causing the water to boil without adding more heat. The vapor is converted to freshwater by condensation (Fig. 1, left). RO makes up about 27% of the seawater desalination capacity. Based on membrane separation RO is a pressure-driven process (Fig. 1, right). Pure water is separated from the dissolved salts by flowing through a water-permeable membrane (Lattemann & Höpner, 2003).

Fig. 1: Basic components of a MSF plant (top) and a RO plant (bottom) (source: Buros, 2000)

Desalination brine discharge modelling - Coupling of Hydrodynamic Models for Brine Discharge

Anal.doc - page 2 of 18


Thermal desalination has a high energy consumption therefore it is mostly used in the Middle East where large energy reserves are available at low prices. MSF plants are often paired with power generation plants. Waste heat from the power plant is used to heat the seawater providing cooling for the power plant at the same time. In contrast the RO process is less energy intensive and so it is predominant in countries with small oil or gas resources like Spain. A sharp distinction in brine volume exists between the two desalination processes. RO plants have a conversion ratio from 20 to 50%. In contrast, MSF plants have lower recovery rates (10-20%) because of being additionally mixed with cooling water (Goebel, 2005). Thus, the effluent flow rate is 4-5 times higher for thermal desalination than for RO processes referring to the same amount of produced fresh water (compare Tab. 1). In case of a MSF plant coupled with a power plant the effluent rates only about 4% of the total intake (Lattemann & Höpner, 2003). Salinity and temperature are essential properties that distinguish between the two processes. That also causes differences of the effluent density since it varies with salinity and temperature - the higher the salinity the higher the density, the higher the temperature the lower the density (T > 4°C). In case of RO the salt concentration of brine is about twice as high as the concentration of seawater. No heating or phase change takes place (Buros, 2000). This results in a strongly increased effluent density. The brine of MSF plants is extremely heated (T > 100°C) and is therefore blended with cooling water to reduce temperature to about 10°C above the receiving water temperature. The increased effluent temperature minimizes the density difference arising from the heightened salt content (increased by 15%). Thus, the density deviates little or none from the intake water. Usually coupled to power generation plants the effluent produced by MSF is additionally mixed with cooling water from the power plant. As a result the effluent is lighter than the receiving water (Lattemann & Höpner, 2003). Summarized this means that in contrast to a MSF plant a RO plant rejects less water with a higher salinity and a higher density. Table 1: Comparison of properties of MSF and RO plants (following Goebel, 2005). Discharge

characteristics are assumed for a typical MSF plant (recovery: 10%, ΔT = 10°C) and a RO plant (recovery: 32%) with a fresh water production of Qfresh = 345 Ml/d = 4 m³/s according to Lattemann & Höpner (2003). "o" indicates the effluent characteristics. Ambient properties: ρ = 1023.5 kg/m³; Sal = 36.3 ppt; T = 27.7°C. Q = flow rate, T = temperature, Sal = salinity, ρ = density. (compare Fig. 2)

MSF RO

market share 62% 27% process distillation membrane separation driving force increased temperature osmotic pressure energy demand thermal (95%)

≡ 13 kWhel electrical 4-5 kWhel

recovery rate (Qfresh:Qintake) 10-20% 20-50% cooling required yes no ΔT = To - Ta 5-15°C 0°C example study: Qfresh 4 m³/s 4 m³/s Qintake 39 m³/s 12.5 m³/s Qcool 27 m³/s - Qbrine 8 m³/s 8.5 m³/s Qo = Qbrine + Qcool 35 m³/s 8.5 m³/s To 37.7°C 27.7°C Salo 40.4 ppt 65.3 ppt ρo 1022.9kg/m³ 1045.5 kg/m³




The various density differences between the brine and the receiving water represented by the buoyancy flux causes different flow characteristics of the discharge. On the one hand the dense RO effluent flow has the tendency to fall as negatively buoyant plume. On the other hand the MSF effluent is distinguished by a neutral to positive buoyant flux causing the plume to rise. Fig. 2 illustrates the typical behaviour of positively or negatively buoyant jets discharging into the receiving water through a submerged single port.

Fig. 2: Brine discharge characteristics of desalination plants (compare Tab. 1) Mixing processes and their characteristics The hydrodynamic mixing behavior of brine discharges depends on discharge characteristics and on ambient environment conditions. The hydrodynamics is generally separated into two regions called near-field and far-field in which different physical mechanisms dominate (Fig. 2). - Near-field characteristics Discharge characteristics primarily dominate the mixing behavior in the near-field region which extends from tens of meters up to a few hundred meters from the outfall location. The initial volume flux Qo, the initial momentum flux Mo, the buoyancy flux Jo and outfall configurations significantly influence the jet trajectory and the intensity of mixing of submerged brine discharges. The discharge fluxes controlling the flow structure of (negatively) buoyant single port discharges are: The initial volume flux: Qo = Uoao (1) The initial mass flux: Qco = Uocoao (2) The initial momentum flux: Mo = Qoao = Uo

²ao (3) The initial buoyancy flux: Jo = go’Qo = go’Uoao (4) where Uo is the discharge velocity, ao the port area, ρo the discharge density, ρa the ambient density and go’ = g(ρa-ρo)/ρo defines the initial buoyant acceleration. The effluent discharging into the receiving water generates an intense shear flow due to velocity discontinuity between the effluent and the ambient flow causing turbulent mixing. The velocity discontinuity may arise from an initial momentum flux (pure jet), from a buoyancy flux leading to vertical acceleration (pure plume) or generally from a combination of these two fluxes (buoyant jet or forced plume). The entrainment of ambient fluid dilutes the effluent and hence decreases differences in concentration or fluid properties (density, salinity, temperature (concerning MSF)) between the effluent and the ambient water. The momentum flux and the brine concentration become diffused into the ambient water field. This mixing process is referred to as submerged buoyant jet mixing. The ambient conditions play a minor role but do also influence the mixing process in the near-field. Ambient currents deflect the jet trajectory into the current direction inducing higher dilution. Density stratification has a negative effect on dilution since it inhibits vertical acceleration leading the plume to be trapped at a terminal level. Depending on the dynamic and geometric characteristics of the discharge flow, boundary interaction processes can occur at vertical ambient boundaries such as the water surface, the sea bed, or pycnoclines (internal boundaries). Boundary interaction processes define the transition




from near-field to far-field mixing. For example a wake attachment can be forced by the a crossflow if the effluent is discharged near the bottom. To determine the plume behavior and the influence of certain hydrodynamic mechanisms (cross-flow, stratification) on effluent mixing appropriate length scales are formed by dimensional analysis based on the discharge fluxes Qo, Mo and Jo (Nash, 1995, Doneker & Jirka, 2007). Five important length scales for steady-state submerged jets are defined as: Jet/plume transition length scale: the distance at which transition from jet to plume takes place

LM = Mo

3/4

Jo1/2 (5)

Jet-to-crossflow length scale: the distance beyond which the jet is strongly deflected by the crossflow

Lm = Mo

1/2

ua (6)

Plume-to-crossflow length scale: the distance beyond which the plume is strongly deflected by the crossflow

Lb = Jo

ua3 (7)

Jet-to-crossflow length scale: the distance beyond which the jet is strongly affected by the stratification

Lm‘ = Mo

1/4

ε1/2 (8)

Plume-to-crossflow length scale: the distance beyond which the plume is strongly affected by the stratification

Lb‘ = Jo

1/4

ε3/8 (9)

where ua defines the ambient velocity and ε defines the buoyancy gradient. Tidal currents are characterized by flows which reverse direction. During the reversal period, the so-called slack tide the ambient water may be momentarily stagnant. When slack tide is approached meaning ua ! 0 the steady state length scale Lm becomes unbounded and thus an unsatisfactory measure for the jet behavior (Nash, 1995). A relationship between the ambient acceleration |dua/dt| and the discharge momentum flux Mo gives a measure for describing the unsteady trajectory leading to following scales: Jet-to-unsteady crossflow length scale, a measure of the distance of the forward propagation into the ambient flow of a discharge during the reversal episode.

Lu = ⎝⎜⎜⎛

⎠⎟⎟⎞Mo

dua/dt

1/2 (10)

Jet-to-unsteady crossflow time scale, a measure of the duration over which an effluent may be considered as discharging into stagnant water while the velocity field is reversing.

Tu = ⎝⎜⎜⎛

⎠⎟⎟⎞Mo

dua/dt 1/4

1/6 (11)

Jirka et al. (1981) showed that buoyant jet deflection is primarily influenced by discharge momentum and not by buoyancy thus, scales for the interaction of the buoyancy flux Jo and dua/dt are not considered to be dominant.




- Far-field processes The further away from the source the less important the discharge characteristics. In the far-field extending from hundreds of meters to tens of kilometers the ambient conditions are dominating the mixing processes. The established plume is transported through passive advection by a generally unsteady ambient current. Large scale motions as buoyant spreading processes and passive diffusion control the slow mixing and the trajectory of the plume. Buoyant spreading motions only occur for positively or negatively buoyant discharges (Doneker & Jirka, 2007). Buoyant forces caused by density differences spread the mixed effluent flow over large distances in lateral direction. A plume of substantial thickness can thereby decrease essentially to a thin but wide layer. The transverse spreading flow is a density current like motion with rather small mixing processes due to entraining ambient fluid at the frontal head of the current. Passive ambient diffusion is a far-field mixing process which arises due to existing ambient turbulence. The established plume increases in width and thickness until it interacts with boundaries (bottom or banks). The strength of passive diffusion depends mainly on ambient flow characteristics and the degree of stratification. In case of open coastal areas the plume size affects the diffusivities leading to accelerative plume growth described e.g. by the Richardson's "4/3 law" of diffusion. In a stable ambient stratification buoyancy in general strongly damps the vertical diffusive mixing processes. The effluent flow and the effluent mixing are described by the continuity equation and the Navier-Stokes equation stating conservation of mass and conservation of momentum and forces. The implementation of simplifying assumption as hydrostatic pressure, incompressibility, the Boussinesq approximation, and Reynolds decomposition reduces the governing equations mentioned above into the so-called U-RANS (unsteady Reynolds averaged Navier-Stokes) equations used for far-field modeling e.g. in Delft3D (after Rodi, 1993). Coupled hydrodynamic modelling of brine discharges The mixing processes of brine discharges have a wide range of length and time scales. Since it is not possible to simulate them with one overall model separate models are used in the near-field and far-field and then linked together. The developed coupling interface is expands on an existing approach for the linking of the near-field model CORMIX with the far-field model Delft3D (Bleninger, 2006). - Hydrodynamic models CORMIX (Cornell Mixing Zone Expert System) is a well proved near-field model consisting of software systems for the analysis, prediction and design of discharges into diverse water bodies. The main emphasis is on the geometry and dilution characteristics of the initial mixing zone. Boundary interaction, buoyant spreading and passive ambient diffusion are also considered for far-field predictions. Although it is in principle a steady-state model unsteady mixing in tidal environments can also be analyzed. Rule-bases provide essential information on the expected dynamic discharge behavior. CORMIX uses a flow classification system based on length scales to determine the discharge/environment interaction and the flow processes that control initial near-field mixing. A sequence of appropriate simulation modules based on buoyant jet similarity theory, buoyant jet integral models, ambient diffusion theory, stratified flow theory, and on simple dimensional analysis are composed for every flow class to predict the trajectory and dilution characteristics. CorTime is a new tool already implemented in CORMIX which allows the calculation of time series. Varying ambient conditions (e.g. currrent, water level) or discharge characteristics during certain period can thereby be considered. Delft3D developed by Deltares (former Delft Hydraulics, Netherlands) is a common software package for the far-field modeling of flow, waves, water quality, ecology, sediment transport and bottom morphology and the interactions between those processes. It consists of several modules which are capable to interact with each other. Delft3D-FLOW is the hydrodynamic module to simulate two-dimensional or three-dimensional unsteady flow and transport phenomena resulting




from tidal and meteorological forcing. It also includes the effect of density differences due to a non-uniform temperature and salinity distribution. Delft3D-FLOW solves the unsteady shallow water equations in two or three dimensions for incompressible free surface flow where pressure distribution in the vertical is assumed to be hydrostatic. Source and sink terms are included in the equations to model effluent discharges. Finite difference method is applied to solve these equations in combination with initial and boundary conditions on a rectilinear or curvilinear boundary fitted grid. - Coupling algorithm The coupling algorithm is described for the hydrodynamic models CORMIX and Delft3D, but with few modifications it is adaptable to other programs with similar capabilities. The required input files and transformations are generated by the use of routines. The pre- and post-processing routines and the model linking routines are coded within the commercial software MatLab® Version 7.4 R2007a from the company MathWorks®. The so-called m-files are ASCII files thus, they may simply be recoded for other languages as well if MatLab® is not available. In addition, macros are created for the completion of the generated input files. They are coded within the text editor UltraEdit®. The coupling algorithm is usually run in the following sequence:

1. Pre-processing: preparation of field measurements as time series input for the near-field model

2. CorTime: time series near-field modeling with CORMIX based on measured data 3. Post-processing: analysis and presentation of the results of CorTime 4. Model linking: preparation of near-field results as time series input for the far-field model 5. Delft3D-FLOW: hydrodynamic modeling based on measured field data and the output data

of the near-field model A short overview of the single steps is given in the following. A more detailed description is to be found in Niepelt (2007). Coupling step 1: Pre-processing CorTime is the time series analysis tool of CORMIX. For a mixing zone analysis with CorTime two documents are required: a CORMIX input file (*.cmx) and a CorTime time series input file (CorTimeInput.txt). The CORMIX input file contains all conditions and configurations of the ambient, the effluent and the discharge. This is the same input file which is required for a single CORMIX simulation (explained in detail in the CORMIX user manual). The CorTime file CorTimeInput.txt includes the varying input data (e.g. ambient velocity, current direction, effluent flow rate) for each time step. A routine was developed to generate CorTimeInput.txt which has a definite structure given by CorTime. This routine converts input data (measured field data and discharge conditions) of a defined range of time steps into the necessary parameters. For every single parameter a separate vector is created. These vectors have all the same amount of components equal to the considered amount of time steps. Finally a single matrix is build by adding the vectors in the required order and then is stored in a text file. Furthermore, the input parameters can optionally be displayed in several figures e.g. as a function of time or statistical analysis. Coupling step 2: CorTime The CORMIX input file mentioned above has to be generated within CORMIX. The data input occurs in the forms-based graphical user interface (GUI) with a series of tabs subdivided into data groups (effluent, ambient, discharge, etc.). A detailed description is given in the CORMIX user manual (Doneker & Jirka, 2007). CorTime runs through following sequence:




• Loading the *.cmx-file and clicking on the 'Validate&Run' button to ensure that the defined configuration runs successfully.

• Selection of the appropriate time series input file CorTimeInput.txt (Post-Processing/Advanced menu option, sub-menu option CorTime). By loading the file the simulation begins instantly.

• All time steps will be run thru by CorTime. CorTime will parse input for each step, load the GUI with each steps values, validate and run.

• Three different output reports are created and saved for each step: a CORMIX input file, a Prediction File and a Session Report. At the end of running all steps a CorTime-Status Report will be created, saved as *.txt-file.

The Prediction File lists all simulation input data and the predicted plume properties. It gives details of the simulation flow modules executed for the given flow classification. The Session Report summarizes all discharge input data and global plume features including compliance with mixing zone regulations. The CorTime-Status Report.txt lists all plume characteristics at the end of the near-field region (NFR) and of a specified regulatory mixing zone (RMZ) for each time step indicating which steps were run successfully and which did not. Out of this file the required input parameters for the coupled far-field modeling are determined. Together with some data of CorTimeInput.txt the predicted plume characteristics will be transformed into appropriate input data as explained in coupling step 4. Coupling step 3: Post-processing The predicted mixing zone processes and the plume behavior of every single time step can be analyzed within CORMIX by means of the Session Report, the Prediction File, the visualization tool CorVue or further output features. Time series analysis can be carried out of the CorTime-StatusReport.txt. After conversion of the data by the use of MatLab® routines a statistical analysis of the predicted plume parameters can be executed. Furthermore, the plume predictions can be plotted for example as a function of time or statistically as relative frequency and cumulative distribution. Coupling step 4: Model linking The main step of the model coupling is the transformation of the output data of the near-field model into input data for the far-field model. In Delft3D-FLOW outfalls are considered as localized discharges of water and dissolved substances. The source is added at the center of a cell and then immediately distributed over the entire cell volume. The discharge locations and the respective flow rates can be read from so-called attribute files. The application of coupling implies that in the far-field model the discharge is located at the predicted endpoints of the near-field. Fig. 3 shows the transformation of the location of the end of the near-field (NFRX, NFRY), the plume width pw, and the plume thickness pt calculated in CORMIX into the corresponding grid cells of Delft3D. If the plume expands over more than one grid cell (MNK) the flow rate Qo has to be apportioned in i parts (i = number of cells) while Qo = ∑Qi to follow mass conservation. Qi depends on the cell size. The concentration C, the salinity Sal, and the temperature T remain the same since they are properties of Q: Ci = Co, Sali = Salo, Ti = To.




Fig. 3: Transformation of CORMIX plume coordinates (NFRX, NFRY, NFRZ) into corresponding grid cells (MNK) of Delft3D via global coordinates (x,y,z). Division of the flow rate Qo in i parts (i = number of cells): Qo = ∑Qi, Qi depends on the cell size. Concentration C, salinity Sal, temperature T remain the same since they are properties of Q: Ci = Co, Sali = Salo, Ti = To.

Several routines were developed to transform the coordinates of the predicted plume location into the corresponding grid cells and then to generate the required input files such as the source location file and the discharge flow rate file. The input file for the source location (*.src) lists all discharge locations identified by the name of the location, the type of interpolation, and its grid coordinates (M N K) while the flow rate file (*.dis) contains the discharge rate and its properties (concentration, salinity, temperature) as a function of time for each discharge location. The generation of these two input files is divided into two steps. In the first step the transformation of the horizontal coordinates is accomplished. In the second step the two files are created as stated above considering the vertical distribution and the allocation of the flow rate. The coordinate transformation is shown schematically in Fig. 4. Coupling step 5: Delft3D-FLOW To execute a flow simulation with Delft3D-FLOW various kinds of information are required. These include the extent of the model area, location of boundaries and its conditions, the bathymetry, geometrical details of the area such as breakwaters, structures, discharges and the definition of which and where results of the simulation need to be stored. Finally, a numerical grid needs to be generated on which all location related parameters are being defined. The data are stored in separate so-called attribute files such as the flow rate file mentioned above. These files are produced by the use of external programs or other Delft3D Tools, manually offline and/or in the Delft3D-FLOW input processor. The main input file for the hydrodynamic simulation program is the so-called Master Definition Flow-file (MDF-file). In the MDF-file further




information (e.g. output options) are defined and a reference is made to the attribute files instead of including all data. Delft3D-FLOW usually runs through following sequence: • Starting the FLOW input processor of Delft3D and loading the selected MDF-file. All data saved

in the attribute files are loaded automatically into the GUI. • The input parameters can be modified in the corresponding data groups. If that is the case the

data have to be saved anew. • Specifying of which computational results are to be stored, then saving the MDF-file. • Execution of the computation after verifying of the MDF-file. The results of a flow simulation

are stored in four types of output files: history file, map file, drogue file, communication file. The history file contains all quantities as a function of time in the specified monitoring points and/or cross-sections. The map file contains results of all quantities in all grid points at a user-specified time interval. The drogue file contains the (x,y)-position of all drogues at each computational time step in the time interval between release and recovery time. The communication file includes all data which are required for other modules of Delft3D such as the hydrodynamic results for a water quality simulation. The results of all quantities in all grid points are stored. The results can be visualized by the use of the post-processing tool Delft3D-QUICKPLOT. Depending on the type of result file a large variety of graphs can be displayed. The graphs can be saved in the figure file format of MatLab® for further analysis.

Fig. 4: Transformation of coordinates of the plume at the end of the near-field: Conversion of the CORMIX output data (NFRX, NFRY±BH) [ ] into global coordinates (xleft/right, yleft/right) via rotation of the coordinate system [ ]. Adding the GPS data of the outfall location leads to the geographical coordinates (Xleft/right, Yleft/right). Resolution of the plume width pw in ten points to determine in which cells (M, N) of the numerical grid the plume is located [ ]. Case Study The present coupling approach has been applied on the existing analysis of the planned waste water outfall in Cartagena, Colombia (Bleninger, 2006) with some modifications of the discharge conditions to make use of the existing environmental data but for a virtual brine discharge situation. Instead of waste water a large brine effluent of a RO plant (Qo = 7 m³/s) is discharged through a submerged single port. The ambient conditions are based on the field data for the Cartagena outfall following Bleninger (2006). The simulation time of one week was selected






including 169 hourly time steps. This period contains current directions mainly oriented to the south-west with a range of about 130° with a mean velocity Ua,mean = 0.44 m/s and a mean density ρa,mean = 1023.5 kg/m³. According to the ambient conditions the effluent characteristics result in To = 27.7°C, Salo = 66.9 ppt, ρo = 1046.8 kg/m³. Further conditions illustrated in Fig. 5 are the vertical angle of discharge θ = 30°, the port diameter Do = 1.2 m, the discharge velocity Uo = 6.19 m/s and the discharge concentration Co = 100%. More details are listed in Niepelt (2007).

Fig. 5: Sketch of the applied outfall design In the far-field model the computational domain was chosen about 20km around the outfall location. A structured curvilinear grid was used for the horizontal discretization with higher resolution in the outfall region (Fig. 6, left). In the vertical a σ-grid with 13 layers was used. The cross-sectional views through the model domain are given on the right in Fig. 6. The outfall is located in grid point M49 N51 in the layers 11 and 12.

Fig. 6: Computational domain of the far-field model. Left: Plan view of the numerical grid. Right: Cross-sectional views of the numerical grid: west ↔ east cross-section along N 51 (top), south ↔ north cross-section along M 49 (bottom)

- Results of the near-field modeling Over the whole simulation time only two different flow characteristics were predicted. Two time steps (I and II) were selected which are representative for these two plume types visualized in Fig. 7. Depending on constant parameters (Uo, Qo) the momentum flux Mo is identical for all time steps. The ambient density difference varies only slightly over time so that the buoyancy flux Jo is




considered equal. As a result the jet/plume transition length scales LM are the same since LM depends on Mo and Jo. These parameters result in: • momentum flux: Mo = 43.33 m4/s2 • buoyancy flux: Jo = -1.56 m4/s3 (go

’ = -0.223 m/s²) • jet/plume transition: LM = 13.53 m The ambient velocity differs by 0.47 m/s. This leads to different plume-to-crossflow and jet-to-crossflow length scales: time step I: time step II: • jet-to-crossflow: Lm = 10.88 m jet-to-crossflow: Lm = 47.35 m • plume-to-crossflow: Lb = 7.04 m plume-to-crossflow: Lb = 580.37 m ⇒ strong current: LM/Lm > 1 ⇒ weak current: LM/Lm < 1 As shown in Fig. 7 the predicted flow class at time step I (occurring 34% of all calculated time steps) describes a plume which rises to a maximum height and then descends toward the bottom. Initially dominated by the effluent momentum the plume gets strongly deflected by the predominant ambient current (LM/Lm > 1) after a short distance. At time step II (occurring 66% of all time steps) the discharge strength dominates the flow. After some distance the plume gets deflected by the weak ambient current (LM/Lm < 1) and is vertically mixed over the full water depth.

Fig. 7: CORMIX plume visualizations for the time step I (left) and II (right) Fig. 8 shows the horizontal distribution of the endpoints of the near-field with their concentrations. Here, the two different plume characteristics are visible as well. In the near distance at about 50 m – representative shown at time step I – the plume gets less diluted (S ≈ 18) and concentrations of C = 4 to 5.87% are expected. For the second flow class, higher dilution (S ≈ 35-100) is achieved so that the concentrations range from C = 1 to 4%. The near-

plume.

field has a range of 106 m to 757 m in mean flow direction and is spread over distance of ±150 m transversal to the mean flow. In the far-field model the effluent will be discharged at the endpoints with respect to the width and the thickness of the




Fig. 8: Scatter plot of the predicted endpoints of the near-field with their concentrations - Coupling configurations In order to draw a comparison three coupling configurations were investigated (Fig. 9): 1. no coupling

All results of CorTime are ignored. (Fig. 9, 1st row).

. vertical coupling

th) and near-field distance determined with CorTime are considered.

are affected. The plume is vertically distributed over the whole water depth or over layer 7 to 13. A uniform concentration distribution is assumed over both the horizontal and the vertical (Fig. 9, 3rd row).

- Results of the (coupled) far-field modeling For the analysis the same time steps (I and II) as for the near-field analysis are considered. Further, three observation points (outfall, A and B) were regarded: outfall is placed at the grid point of the outfall location, A is situated in 500 m, B in 2300 m downstream of the outfall location in the mean flow direction.

⇒ Effluent discharges at the outfall location so that two cells are affected2

The vertical plume distribution computed with CorTime is considered while the determined near-filed distance and the plume width are neglected. ⇒ Effluent discharges at the grid point of the outfall location, vertically distributed over the whole water depth or over layer 7 to 13. A uniform concentration distribution is assumed over the vertical (Fig. 9, 2nd row).

3. full coupling The plume size (thickness and wid

⇒ Effluent discharges at the calculated endpoints of the near-field. Having regard to the plume width 20 grid points




Fig. 9: Distribution of the discharge locations for the three coupling configurations.

Left: plan views, middle: south↔north cross-sections, right: east↔west cross-sections ns, ×: grid points where layer 7 to 13 are affected separately

. 11 show maps for the three coupling configurations at the time steps I and II. Fig. ution in layer 11 in which the outfall is located. Fig. 11 disp he

a larger concentration scale is used. Fig. 12 shows concentration profiles ove Regcasdefl s.

A shdire tfall location the concentration exceeds 5.0% (no coupling). Considering at least the (ve In F pling approaches are clearly visible. In the case of no coupling the effluent is discharged at the height of the outfall location. The plume

2 in the mean current direction. No vertical mixing

•: discharge locatio Fig. 10 and Fig

10 shows the concentration distriblays cross-sectional views of the concentration distribution along the grid line M49. In t

case of no couplingr the depth at the observation points outfall, A and B.

arding the plan views given in Fig. 10 at one time step the plume looks similar for all coupling es. The plume is travelling in the mean flow direction. Unsteady variations cause slight ections and stretching of the plume

arp distinction exists in the concentration of the plumes. If the effluent is discharged ctly at the oudistribution over the vertical predicted by CorTime the concentration amounts less than 0.8% rtical and full coupling).

ig. 11 and 12 the differences of the investigated cou

travels along the sea bed in layer 11 and 1occurs. The largest concentration (5.3%) arise at the outfall location. While traveling further away from the source the concentration decreases due to horizontal spreading.




Fig. 10: Plan views of the concentration distributions in layer 11 (height of outfall) the three

o coupling a

or ve me time

ion is reduced by almost half as in the

ling are higher compared to the full coupling at each

discharge location is not fixed in one point. As a result the effluent

e concentration differences between the full and vertical coupling approaches are negligible.

for ncoupling configurations at time step I (top) and II (bottom). In the case of

larger concentration scale is used.

rtical and full coupling the shape of the plume is almost identical regarding the saFsteps in the cross-sectional views. For the time step I the plume is distributed over layer 7 to 13. In time step II the plume is distributed over the whole water depth. Hence, the two different plume characteristics that were predicted with CorTime were transmitted correctly. For the vertical coupling approach the largest concentration is about 0.74% be found at the utfall location. At the observation point A the concentrato

case of no coupling. At the point B the plume concentrations are decreased to less than 0.05%. In the case of full coupling the highest concentration (0.26%) occurs at the observation point A. The concentrations at the observation point B are lower than 0.04%. Consequently, the oncentrations resulting from vertical coupc

observation point. That is because considering only the vertical plume distribution for the coupling the effluent is continuously discharged in the grid point of the outfall location. Taking account of the predicted ndpoint of the plume the e

concentration does not accumulate at the outfall location and is therefore lower in this grid point compared to the vertical coupling. Additionally the sources are distributed in the transverse direction in the case of full coupling. That is why the concentration in point A and B are of minor value as well compared to vertical coupling since A and B are located in the mean flow direction. However, at a distance of about 2300 m downstream from the outfall location th




Fig. 11: Concentration distributions in the cross-section along M49 for the three coupling

configurations at time steps I and II. In the case of no coupling a larger concentration scale is used.




Fig. 12 Concentration profiles at the observation points outfall, A and B for the three coupling

configurations no, vertical and full coupling. Time steps: I and II. Conclusions The analysis of the concentration distributions shows the importance of model coupling. The far-field model can not simulate the strong mixing processes occurring in the near-field particularly with regard to the vertical mixing. Thus, the vertical distribution of the plume needs to be calculated by the use of a near-field model and must then be transferred into the far-field model. Regarding the horizontal plume distribution of all coupling configurations the plumes were passively advected by the ambient current and slowly mixed by the ambient turbulence in the same way. A horizontal coupling is considered to be unimportant since the concentration differences between the vertical and the full coupling are negligible. In addition, the dilutions resulting from the vertical coupling are lower compared to the full coupling. Hence, considering only the vertical distribution gives a conservative estimate regarding the mean flow direction. Consequently, the calculations indicate that the far-field model alone (without coupling) can not simulate the vertical concentration distribution of the plume in contrast to the near-field model. Thus, a model coupling is unavoidable for an environmental assessment. The coupling methodology, though simple, allows for an considerably improved discharge assessment. This allows for an optimized environmental hydraulic design of the outfall structure.

Furtherminteractions of different discharges, and consequences of different pre-treatment/operational

ore, model results allow for an optimized intake location to avoid recirculation. Also




schemes can be studied. References Bleninger, T. (2006). Coupled 3D hydrodynamic models for submarine outfalls: Environmental

hydraulic design and control of multiport diffusers, Ph.D. Dissertation, Institute for Hydromechanics, University of Karlsruhe

Bleninger, T., Jirka, G. H., Weitbrecht, V. (2006). Optimal discharge configuration for brine effluents from desalination plants, Proc. DME- Congress, Deutsche Meerwasser Entsalzung, 04.-06.04.2006, Berlin

Bleninger, T., Jirka, G. H. (2007). Modeling and environmentally sound management of brine discharges from desalination plants, Proc. EDS Congress Desalination and the Environment, April 22-25, 2007, Halkidiki, Greece

Buros, O. K. (2000), The ABCs of desalting, http:\\www.idadesal.org\pdf\ABCs1.pdf, 01.10.2007 Delft3D-FLOW(2005). Simulation of multi-dimensional hydrodynamic flows and transport

phenomena, including sediments, User Manual version 3.12, WL|Delft Hydraulics, Delft Doneker, R. L., Jirka, G. H. (2007). CORMIX User Manual: A Hydrodynamic Mixing Zone Model

and Decision Support System for Pollutant Discharges into Surface Waters, MixZon Inc., Portland, OR

Doneker, R. L., Jirka, G. H., Nash, J. D., (2004). Pollutant transport and mixing zone simulation of sediment density currents, J. Hydraulic Engineering, ASCE, 130 (4)

Glade, H. (2005). Design of seawater distillation plants, DME Seminar Introduction to Seawater Desalination, June 20th, Berlin

Goebel, O. (2005). Markets and desalination technologies in brief, DME Seminar Introduction to Seawater Desalination, June 20th, Berlin

Jirka, G. H. (2004). Integral model for turbulent buoyant jets in unbounded stratified flows. Part 1: Single round buoyant jet, Environmental Fluid Mechanics, Vol. 4

Jirka, G. H. (2006). Improved discharge configurations for brine e²uents from desalination plants, J. Hydraulic Engineering, ASCE, submitted

Jirka, G. H., Adams, E., Stolzenbach, K. (1981). Properties of surface buoyant jets, J. of

Lattema f brine and chemical discharge on the marine environment, Institute for Chemistry and Biology of the Marine Environment (ICBM), University of Oldenburg

995). Buoyant discharges into reversing ambient current, Master thesis, DeFrees

Hydraulics Div., ASCE, 106 (HY11) nn S., Höpner, T. (2003). Seawater desalination: Impacts o

Nash, J. D. (1Hydraulics Laboratory, Cornell University, Ithaca NY

Niepelt, A. (2007). Development of interfaces for the coupling of hydrodynamic models for brine discharges from desalination plants, Diploma thesis, Institute for Hydromechanics, University of Karlsruhe

Roberts, P. J. W. (2004). Additional water quality modeling for the Cartagena ocean outfall, Report for the World Bank, Atlanta

Rodi, W. (1993). Turbulence models and their application in hydraulics - a state of the art review, International Association for Hydraulic Research, Delft, 3rd edition, Balkema

Shiklomanov, I. A. (2003). World water resources at the beginning of the 21st century, International Hydrology Series/UNESCO, Cambridge University Press, Cambridge

desalination brine discharge modelling coupling of ... · prof., director, institute for...

Documents