XXI CONGRESO LATINOAMERICANO DE HIDRÁULICA SÃO PEDRO, ESTADO DE SÃO PAULO, BRASIL, OCTUBRE, 2004

Transcription

1 Proceedings, XXI Congreso Latinoamericano de Hidráulica, International Association for Hydraulic Research and Engineering (IAHR), São Pedro, Brazil, October 18-22, 2004, A.M. Genovez, Ed. IAHR AIPH XXI CONGRESO LATINOAMERICANO DE HIDRÁULICA SÃO PEDRO, ESTADO DE SÃO PAULO, BRASIL, OCTUBRE, 2004 DESIGN OF MULTIPORT DIFFUSER OUTFALLS FOR COASTAL WATER QUALITY PROTECTION Gerhard H. Jirka and Tobias Bleninger Institute for Hydromechanics, University of Karlsruhe, Karlsruhe, Germany ( ABSTRACT Increased coastal zone populations often lead to regional water quality problems due to uncontrolled municipal waste water discharges. The prediction and monitoring of bacteria and nutrient concentrations in the coastal waters is essential for pollution control, where water quality regulations with environmental quality standards in addition to emission limit values promise improvements in the quality characteristics. However, the specification of where in the water body the environmental quality standards apply and how to predict and monitor is missing in most of the regulations and planning procedures. A clear mixing zone regulation and its implementation with support of coupled hydrodynamic and water quality models is demonstrated using the example of pollution control for multiport diffuser outfalls discharging municipal wastewater into coastal waters. In terms of efficiency, secureness and cost due to low operational and maintenance requirements this disposal option is seen as one of the most attractive for coastal regions, if necessary means are undertaken in consideration to avoid negative impacts. This paper reviews simple predictive equations and advanced modeling techniques for the near- and far-field analysis of effluent discharges from multiport diffusers into the coastal zone. A new design and engineering optimization program for the internal diffuser hydraulics design and operation is also presented. KEY WORDS water quality management, mixing zone, multiport diffuser

2 2 1. INTRODUCTION: COASTAL WATER QUALITY PROBLEMS Coastal waters, including estuaries, wetlands, mangrove forests, bays represent a resource of enormous natural and economic value. These productive and diverse areas host a multitude of species important for nature and man. The coastal population is attracted by and depends strongly on these values for their livelihood, food, health, and recreation, as is seen in the number of people living in coastal areas. The area within 200 km from the shoreline coastal areas hosts more than 50 % of the world population (Hinrichsen, 1998). UNEP (2004) gives estimates on further increases of the world s coastal population. The population of Latin America and the Caribbean is even more littoral (Hinrichsen, 1998): around 610 million, that is 75% of the population, live within 200 km of a coast. The majority of the Caribbean Basin s 200 million permanent residents live on or near the seashore. The resident population is swelled every year by the influx of some 100 million tourists, nearly all of whom end up on the region s beaches. Resulting environmental pressures, in addition to the direct physical coastline alterations, strongly affect coastal water quality (UNEP, 2002). Public health suffers from polluted coastal areas where beach closures are indicative of the dangers in water contact sports. This has direct effects on working and living quality in that region. Most of the local problems are directly related to badly or non-controlled point-sources like municipal and industrial discharges (UNEP, 2004). Also non-point pollution sources like surface runoff from urban or agricultural areas often contribute significantly to coastal water quality problems. But increased investments in sewage collection systems and reduction in storm water overflows will reduce the diffuse pressures and will concentrate them on controllable point-sources. Pollution control of both source types has to be in balance with the assimilative capacity of coastal waters. 2. COASTAL WATER QUALITY MANAGEMENT Pollution control mechanisms are pollutant reduction at the source, wastewater reuse, wastewater treatment and disposal. Choosing adequate techniques for regional needs and capabilities is an important step in coastal water quality management. Source control mechanisms are needed to ban persistent pollutants (e.g. CKW or highly oxidizing cleaning agents or detergents), which are neither decomposed in conventional treatment plants, nor in coastal waters after direct discharges. Wastewater reuse is, because if its high costs, only applied in coastal regions with water shortage. The wastewater treatment and disposal option is the most effective, if regions with strongly increasing sanitation coverage are considered (UNEP, 2004, 2002). The level of treatment is defined by the nutrient needs and assimilative capacities of the receiving waters. Biological wastewater treatment (the most widely applied technology) is therefore only necessary, if receiving waters have low flushing characteristics and tend to eutrophicate. Most coastal water bodies at open coast locations instead are capable to use the nutrient inputs from low level treated wastewater discharges as any other natural nutrient input. This option allows coastal communities to reduce costs for expensive treatment equipments, its maintenance and operation as well as for the resulting excess sludge. This explains the worldwide increasing utilization of submarine outfalls. Positive examples with detailed monitoring programs of the receiving waters can be seen, for example in Australia (Philip and Pritchard, 1996) or Boston (Signell et al., 2000). Regulatory goal and requirements Ragas et al. (1997) have reviewed the advantages and disadvantages of different control mechanisms in the permitting processes of releases into surface water. Emission limit values (ELV) present a direct and effective method for the limitation of pollutant loadings by restricting the concentration for the mass flux of specific pollutants. Other ways prescribing ELVs are best available technologies (BAT), for example, for sea outfalls as some form of treatment, at least primary, or enhanced (e.g. chemically) primary, or a secondary biological treatment stage for nitrogen removal, or a tertiary stage for phosphorus removal. ELVs are preferred from an administrative perspective because they are easy to prescribe and to monitor (end-of-pipe sampling). From an ecological perspective, however, a quality control that is based on ELVs alone appears illogical and limited, since it does not consider directly the quality response of the water body itself and therefore does not hold the individual discharger responsible for the water body. To illustrate that point consider a large point source on a small water body or several sources that may all individually meet the ELVs but would accumulatively cause an excessive pollutant loading. Environmental quality standards (EQS), set as concentration values for pollutions or pollutant groups, that may not be exceeded in the water body itself have the advantage that they consider directly the physical, chemical and biological response characteristics due to the discharge and therefore they put a direct respon-

3 3 sibility on the discharger. But a water quality practice that would be based solely on EQSs could lead to a situation in which a discharger would fully utilize the assimilative capacity of water body up to the concentration values provided by the EQSs. Furthermore, the water quality authorities would be faced with additional burdens because of a more difficult monitoring where in the water body and how often should be measured? in the case of existing discharges or due to the increased need for a prediction modeling in case of new discharges. Nowadays most discharge regulations (e.g. US: EPA, 1994; Europe: EC-Water framework directive, 2000; Brazil: CONAMA, 2000) follow a combined approach of these control mechanisms. For most countries this policy means a considerable deviation from current water quality management practice by which the releases of pollutants has been controlled by either one of these two control mechanisms, but usually not their combination. Although the new strategies appear to be logical in their intention the actual implementation is vague and incomplete. In particular, the fact that most of the regulations do not state where precisely in the water body the EQS-values shall be applied will lead to arbitrary and contradictory interpretations on part of water authorities. Possible interpretations that the EQS-values apply either directly at the discharge point or after initial mixing are illogical and contradict the intentions (Jirka et. al, 2004). A clear definition of a regulatory mixing zone is needed. Mixing zone regulations The mixing processes due to discharges into water bodies occur according to physical principles, shown later, and lead to a spatial and temporal configuration of the mass plume and the associated concentration distribution. To what degree do the water quality control measures correspond to these physical facts? From discussions with personnel from various water authorities the authors know of two extreme interpretations regarding this omission in the WFD: 1) The EQS-value shall be applied as near as possible to the discharge point in order to obtain a good quality status in an area as large as possible. This highly restrictive interpretation negates the fact that the physical mixing process cannot be reduced to extremely small areas (in the limit this approaches an end-of-pipe demand for EQS!), but requires a certain space in particular for imposed high ELV/EQS ratios. It undermines the balanced objectives of the combined approach. 2) The EQS-value is supposed to apply after the completion of initial mixing or at the beach or at the water surface. All such qualitative statements have specific deficiencies, that make them either unenforceable or overly generous and likely to create sacrificial areas with high concentration levels not meeting a good quality status. Future regulations should contain the following approximate wording: The environmental quality standards apply in the case of point sources outside and at the edge of the mixing zone. The mixing zone is a spatially restricted region around the point source whose dimensions shall be specified either according to water body type and use or on an ad-hoc basis. The mixing zone defined in the above statement is a regulatory formulation with the following general attributes: 1) The term mixing zone signifies explicitly that mixing processes require a certain space. 2) The term spatially restricted should guarantee that the mixing zone shall be minimized by the regulatory authority for the purpose of attaining the environmental quality goals. 3) While the mixing zone includes a portion - namely the initial one - of the actual physical mixing processes, these processes will continue beyond the mixing zone where they lead to further concentration dropoffs in the pollutant plume below the EQS-values. 4) The definition is restricted to point sources since diffuse sources usually do not contain clearly distinct mixing processes. Once the principle of a mixing zone has been adopted and defined it is also necessary that national water authorities provide clear guidance for the actual specification of mixing zone dimensions. Two major possibilities exist here: a) Specification of numeric mixing zone dimensions according to water body type and biological characteristics: It seems advisable to constrain the mixing zone to a limited region around the outfall in which the initial buoyant jet mixing is dominant. In that fashion the EQS-values can be achieved within short distances. Thus the following specification appears effective: The mixing zone is a volume with vertical boundaries in the coastal water body that is limited in its horizontal extent to a distance equal to N multiples of the average water depth H ave at the outfall location and measured in any direction from the outfall structure. Thus, for a single port outfall this would be a cylindrical volume with the port in its center (Fig. 1a). For a multiport diffuser outfall with many ports arranged along a straight diffuser line it

4 4 would be a rectangular prismatic volume with attached semicircular cylinders at the diffuser ends located along the diffuser line (Fig. 1b). For diffusers with a curved diffuser line or piecewise linear sections the volume would follow the diffuser line. The value N would typically be in the range of at least 1 to about 10 and set by the regulatory authority according to local water use and ecological sensitivity. For highly sensitive waters the minimum of 1 should be set. Common values for most coastal waters might be N = 2 to 3. c) Specification of mixing zone dimensions in an ad-hoc manner: After prior ecological evaluations or predictions the discharger can request the authority for a mixing zone with a certain dimension with the claim that this would guarantee an integrated water quality protection. Based on its own examinations the authority can agree with that proposal or else demand further restrictions. Increased use of modeling techniques for mixing processes In order to demonstrate compliance with the EQS-values it appears that both dischargers as well as water authorities must increase the application of quantitative predictions of substance distributions in water bodies (water quality parameters in general, mixing processes in particular). This holds for both existing discharges (diagnosis) as well as planned future discharges (prediction). There are several diagnostic and predictive methodologies for examining the mixing from point sources and showing compliance with EQS-values (see Fig. 2): 1) Field measurements or tracer tests can be used for existing discharges in order to verify whether EQSvalues are indeed met. Field measurements are costly, often difficult to perform, and usually limited to certain ambient conditions. Frequently, they must be supported through mathematical model predictions, on one hand, to establish a clear linkage to the considered discharge (especially if more than one discharge exists), and on the other hand, to synthesize conditions allowing for variabilities in the hydrological or oceanographic conditions or in the effluent rates. Fig. 1: Example of regulatory mixing zone specification for offshore submerged coastal discharges: The horizontal extent is defined by some multiple N of the average water depth H ave at the diffuser. 2) Hydraulic model studies replicate the mixing process at small scale in the laboratory. They are supported by similarity laws and are quite reliable if certain conditions on minimum scales are met as has been demonstrated in the past. But just like field tests, they are also costly to perform and inefficient for examining a range of possible ambient/discharge interaction conditions. 3) Simple analytical equations or nomograms are often satisfactory to predict reliably the mixing behavior of a pollutant plume. Simple equations or analytical expressions can be applied for the initial buoyant jet/plume mixing of single or multiport submerged discharges (e.g. Jirka and Lee, 1994). They have been validated through numerous data comparisons from laboratory or field measurements as shown in the existing literature. Of course, they are limited to simple discharge/ambient geometries. 4) Mixing zone models (near-field modeling) are reduced versions of more general water quality models. They describe with good resolution the details of physical mixing processes (mass advection and diffusion),

5 5 but are limited to relatively simple pollutant kinetics by assuming either conservative substances or linear decay kinetics. This is acceptable for most applications, since residence times in the spatial limited mixing zones (see previously mentioned specifications) are typically short so that chemical or biological mass transformations are usually unimportant. Ragas (2000) provides a comparison of different mixing zone models. The mixing zone model CORMIX (Doneker and Jirka, 1991; Jirka et al, 1996), in particular, is characterized by its wide applicability to many water body types (rivers, lakes, estuaries, coastal waters) and has been successfully used for water quality management under different regulatory frameworks. 5) General water quality models may be required in more complex situations. In simple water bodies, such as coastal regions or estuaries with well defined uni-directional current regimes or with simple reversals, and with moderate pollutant loadings, the use of mixing zone models alone may suffice to arrive at, or to evaluate, a design of a point source discharge that meets regulations. However, regions with multiple current regimes (inertial, tidal, wind- or buoyancy driven) and with large pollutant loadings, especially where several sources may interact and additional diffuse sources may exist, mixing zone models must be supplemented by larger-scale (far-field) transport and water quality models. The latter are capable of prediction over greater distances in the water body the concentration distributions for different pollutants, but also for nutrients and other bio-chemical parameters with due consideration of mass transformation and exchange processes. They do not, however, have the high spatial resolution that is required to predict mixing processes and the compliance with EQS-values in a limited mixing zone. An effective and sensible approach for strongly loaded water body reaches is therefore the prior application of general water quality models in order to predict background concentrations in the vicinity of the considered point source. Superimposed on these background concentrations are then the additional concentrations within the pollutant plume as predicted by the mixing zone model. Different methods for this far-field modeling exist, ranging from water quality models in estuary-type flows (e.g. model QUAL-2 of the U.S. EPA), to Eulerian coastal circulation and transport models (e.g. Delft3d of Delft Hydraulics) to Lagrangian particle tracking models (e.g. Roberts, 1999). The general issues in coastal water quality management and the linking of mixing zone models (near-field) and general water quality models (far-field) are summarized in Fig. 2. Complete outfall design procedures and algorithms are described in general guidelines for submarine outfalls (Grace, 1978; Williams, 1985; Water Research Centre, 1990; Wood et. al, 1993; UNEP, 1996). These are based on descriptions of the occurring physical mixing processes shown in Fischer et. al (1979), Telford (1989), Roberts et. al (1989a, b, c), National Research Council (1993), Jirka and Lee (1994), Akar and Jirka (1995), and Jirka (2004). Fig. 2: Methodology for coastal water quality management

6 6 3. NEAR-FIELD MODELING: INITIAL MIXING MODELS FOR MULTIPORT DIFFUSER OUTFALLS Large flows of waste effluents are commonly discharged through multiport diffuser outfalls. Such an outfall consists of two components (see Fig. 1b): a feeder pipe that transports the effluent from the head works at the coast to the offshore area and the actual multiport diffuser. The latter is a linear structure with a (usually large) number of ports mounted on, or connected through risers to, a submerged pipe laid on the ocean floor. The nozzles can be oriented in various ways, both in plan and in section, to effect particular types of mixing patterns and to take advantage of the direction of the prevailing ambient current. A multiport diffuser is an efficient mixing device, capable of rapidly diluting the effluent within a short distance. The hydrodynamics of an effluent continuously discharging into a receiving water body can be conceptualized as a mixing process occurring in two separate regions. In the first region, the initial jet characteristics of momentum flux, buoyancy flux (due to density differences), and outfall geometry influence the effluent trajectory and degree of mixing. This region, the "near-field", encompasses the buoyant jet flow and any surface, bottom or terminal layer interaction. In this near-field region, outfall designers can usually affect the initial mixing characteristics through appropriate manipulation of design variables. As the turbulent plume travels further away from the source, the source characteristics become less important and the far-field is attained. In this region ambient environmental conditions will control trajectory and dilution of the turbulent plume through buoyant spreading motions, passive diffusion due to ambient turbulence, and advection by the ambient, often time-varying, velocity field. The dilution in the near-field of a sea outfall is primarily governed by the buoyancy flux per unit length of diffuser and the depth of discharge. A review of these processes has been given by Fischer et al. (1979), Wood et al. (1993), or Jirka and Lee (1994). Initially, the effluent leaves the diffuser ports as a series of round plumes (Fig. 3). After travelling some distance, adjacent plumes merge with each other to form a rising curtain, and the flow is essentially two-dimensional. Under typical discharge conditions, it can be shown (Wallis, 1977; Brooks, 1982) that for sufficiently large depths, initial dilution predictions for the same buoyancy flux per unit diffuser length agree to within 15%, regardless of the details of the individual ports. 2-D z z Merging level 2-D Zone 3-D l a 0 Fig. 3: Cross-section and side view of a multiport diffuser. A series of round plumes merges to a 2D line plume. One of the key equations for dilution is the equation for a line plume in a stagnant unstratified ocean (Rouse et al., 1952): 1/3 j H Sc = 0.38 (1) q in which j ρ o = gq is the buoyancy flux per unit length, given by ρ o = initial density difference, ρ o ρ a = ambient density, g = gravitational acceleration and q = Q/L D is the discharge per unit length, given by Q = total discharge flow and L D = diffuser length, H is the water depth and S c presents the minimum (centreline) dilution at the surface location. For a given flow Q, the unit discharge q and unit buoyancy flux j are inversely proportional to the diffuser length L D, and Eq. (1) suggests that a higher dilution is obtained by increasing the length of the diffuser. For a line plume, the minimum dilution can be multiplied by a factor of 2 to give the average dilution.

7 7 Flow features such as the buoyant jet motion and any surface, bottom or terminal layer interaction also take place. In the near-field region, outfall designers can usually affect the initial mixing characteristics through appropriate manipulation of design variables. For example, Fig. 4 shows a laboratory demonstration of a discharge jet/plume motion that is trapped by the ambient density profile below the water surface leading to a plume spreading at a terminal level. This is a frequently employed design strategy for sewage discharge during summer stratification in coastal waters. Recent field data (Carvalho et al., 2002) provide further evidence of such plume phenomena. It has been demonstrated both theoretically and experimentally (Fischer et al., 1979) that maximum mixing can be achieved with closely spaced ports that allow some interference of adjacent jets. In relatively shallow coastal waters of typical depth 5 15 m, however, it is often the case that, given practical considerations (e.g. in order to maintain a minimum jet velocity and minimum diameter), multiport diffusers are designed to minimize interference of adjacent plumes. In such cases, the required spacing is about H/3. Fig. 4: Laboratory investigation of near-field mixing of submerged discharge into a stratified water body. The initial buoyant jet motion is followed by trapping and internal spreading at the terminal level. In case of a linearly stratified ambient with a density gradient dρ a /dz the maximum height of rise z max to the terminal level (see Fig. 4) and corresponding dilution S c are given by g dρ z = 2.84j max 1/2 1/3 a ρa dz (2) 1/3 j zmax Sc = 0.31 (3) q In a linearly stratified ambient, the spreading layer is found to occupy about 40 50% of the rise height. For computing bulk dilutions, one must allow for the thickness of the wastewater field. Simple models to account for blocking in the presence of an ambient current can be found in Fischer et al. (1979). Roberts (1979, 1980) studied the mixing of a line source of buoyancy in an ambient current, and found that 3 the shape of the flow field and the dilution is determined by the ambient Froude number F= u / j. F measures the ratio of the ambient current velocity to the buoyancy-induced velocity. For F < 0.1, the minimum surface dilution Sm is little affected by the current and is given by: 1/3 j H Sm = 0.27 (4) q Compared with Eq. (1), the smaller dilution coefficient reflects the effect of blocking of the surface layer. For higher crossflow, F > 0.1, however, the entrainment is dominated by the crossflow, and the alignment angle γ between the diffuser line and the current direction is important. Higher dilution results for a perpendicular alignment, γ = 90, in which the maximum amount of flow is intercepted while the parallel alignment, γ = 0, gives the lowest dilution. For F 100, the perpendicular alignment results in a dilution uh a Sm = 0.6 (5) q that is proportional to volumetric mixing between ambient (velocity u a ) and discharge flow, but with a reduced coefficient 0.6. For parallel alignment the dilution is lower by a factor of about four. Experiments by a

8 8 Mendez-Diaz and Jirka (1996) have examined the different plume trajectories for various crossflow strengths. Experiments to study the dilution and wastefield formation of a line plume in a steady linearly densitystratified current have been reported by Roberts (1989). The results show that mixing behaviour is determined by the crossflow parameter F, buoyancy flux j, and the ambient density gradient. For all diffuser alignments γ, an increase in current speed results in increase of dilution, and decrease of rise height and wastefield thickness. As in unstratified current, maximum mixing is achieved with the perpendicular alignment. However, compared with the unstratified case, the dilution is less sensitive to diffuser orientation. The simple dilution equations given in the foregoing are useful for initial design screening of alternatives. They are limited to simplified ambient conditions. For final design evaluations and for more general and complex ambient oceanographic conditions more comprehensive numerical models must be employed. Clearly, the most general and flexible analysis and design system for the initial mixing of outfall effluents is represented by the CORMIX system (Doneker and Jirka, 1990; Jirka and Akar, 1991). CORMIX addresses the full range of discharge geometries and ambient conditions, and predicts flow configurations ranging from internally trapped plumes, buoyant plumes in uniform density layers with or without shallow water instabilities, and sinking (negatively-buoyant) plumes. Boundary interaction, upstream intrusion, buoyant spreading and passive diffusion in the far field are also considered. In total, CORMIX considers 32 generic flow classes describing different types of diffuser plume configurations, ranging from stagnant water to weak ambient crossflow, uniform or density-stratified environments, different types of alignments and port orientations, weak or strong discharge momentum and buoyancy characteristics, and other detailed features. It uses a system of prediction modules to assemble an overall prediction of the near-field mixing characteristics, in tenure of plume geometry and concentration levels. At the heart of CORMIX is an integral jet model COR- JET which provides detailed buoyant jet analysis under complex ambient conditions (Jirka, 2004). CORMIX also contains several pre- and post processing options of the assembly and consistency checking of input data, for the rapid evaluation (batch sequence) of a range of ambient conditions (e.g. variable ambient current spreads), and for the graphical 3-D display of the prediction results. Fig. 5 shows an example of a CORMIX prediction of a diffuser discharge into a stratified flowing environment with an oblique alignment angle (γ = 45 ) leading to an internally trapped plume. CORMIX has been validated with a wide range of fundamental laboratory data sources. The amount of comprehensive and reliable field data for actually operating diffusers that can be used for model validation is still limited at present. The field survey of Carvalho et a. (2992) for the Ipanena outfall in Rio de Janeiro has provided a highly satisfactory validation for the CORMIX model; as regards its predictive ability and accuracy not only for the immediate near-field but also the transition to the far-field in form of the buoyant spreading of the internally or surface-trapped plume (see Jirka and Doneker, 2003). Other available plume models (e.g. plumes or RSB) are clearly limited in that regard.

9 9 Fig. 5: Example of a CORMIX prediction of a diffuser discharge into a stratified flowing environment with an oblique alignment angle (γ = 45 ) seen in the plan view, leading to an internally trapped plume, seen in the side view, resulting in a concentration profile along the plume centerline of an effluent concentration of 100%. 4. FAR-FIELD MODELING: COASTAL CIRCULATION AND WATER QUALITY MODELS As the turbulent plume travels further away into the far-field, the source characteristics become less important. Conditions existing in the ambient environment will control trajectory and dilution of the turbulent plume through buoyant spreading motions, passive diffusion due to ambient turbulence, and advection by the ambient, usually time-varying velocity field. The required detail of a far field prediction can vary considerably from case to case. It depends on the complexity of the coastal ocean environment, the availability of data, and finally, the severity of the pollution problem. For an open coastal environment with a prevailing current structure a simple plume model may suffice if a far field calculation is needed at all. For estuaries or semi-enclosed bays, a flushing analysis may be required to ascertain the net flow-through and potential long-term accumulation of pollutants. Finally, in coastal environments or tidal networks of complex topography and current structure, numerical transport models have to be employed for solution of the convective diffusion equation in order to predict the far-field pollutant distribution. The velocity field in these transport models may have to be calculated from a separate circulation model or be obtained from detailed field measurements. Beyond the initial mixing zone governed by buoyant jet mixing, a pool of wastewater is formed either at the surface or at the level of submergence under stratification conditions. The wastewater could be brought shoreward either by lateral turbulent diffusion in a predominantly alongshore current or directly by windgenerated onshore currents. A common water quality objective is to prevent high concentrations of pollutants from reaching coastal recreational waters or shell fisheries. For a line diffuser normal to a current, the initial width of the wastewater field can be taken to be equal to the diffuser length; the depth of the field in moving stratified water can be estimated on the basis of continuity considerations and accounting for partial blocking above the base of the field (Brooks, 1972): h = z QS/ u L z / 1 + QS/ u L z (6) {( ) ( )} max a D max a D max where S, the average dilution, and z max are obtained from the near-field solution (Eqs. 2 and 3). Eq. 5 is expected to hold for SQ/(u a L D z max ) 2. These dimensions are needed as initial conditions for transport calculations in the far field. Further examples of the coupling of near field solutions to far field plume models are contained in Adams et al. (1975) and Jirka et al. (1976). In still water, the surface field with its residual buoyancy spreads in an unsteady fashion. Much research is devoted to the dynamics of the intermediate field in which buoyancy plays a significant role (e.g. Koh, 1983; Akar and Jirka, 1994a, b).

10 10 Under high dilution conditions in moving water, the wastewater field is essentially advected passively in a tidal current. Vertical mixing is damped by buoyancy, so that the subsequent dilution is mainly due to horizontal mixing by the turbulent eddies. Brooks (1960) gives a widely used method of estimating the subsequent dilution of a wastewater field due to lateral mixing by oceanic turbulence. A line source is assumed with initial conditions determined by the initial mixing phase as discussed above; the horizontal transport of a pollutant layer of constant thickness in a steady uniform current normal to the source is then formulated. By assuming a 4/3 power dependence of the eddy diffusivity on the local plume width and a first order decay with rate constant k, the centerline concentration in the plume is given by: cmax 3/2 = erf 3 c i 8Eit li 1/2 e kt where t = travel time = x/u a, E i = diffusivity corresponding to the initial width of the plume l i, and c i = initial concentration at the end of the near-field. Subject to the global flushing constraints, Equation 6 is useful in giving a conservative estimate of the subsequent dilution and elucidating the relative importance of horizontal diffusion and decay processes. It is, however, limited in two respects: a) nearshore tidal currents are unsteady, and b) a wind-generated onshore surface current is usually accompanied by a compensatory offshore bottom current; Munro & Mollowney (1974) have shown that in shallow coastal waters vertical mixing in such a counter current system can lead to substantial additional reductions in concentration. This result is supported by detailed observations of bacterial distributions at several sea outfalls (Gameson, 1982; Munro, 1984). The above equations are restricted to simple oceanographic conditions. As such they have also been incorporated as an extension of near-field models (e.g. in the case for CORMIX). For more general conditions see also Fig. 3 and the discussion in Section 2 more general coastal circulation, mass transport and transformation (water quality parameters) models need to be employed. These models fall into two categories: (a) Eulerian flow and transport models, and (b) Lagrangian tracking models. a) Eulerian flow and transport models: The solution of the advective diffusion equation governing mass transport with appropriate near-field source conditions and ambient boundary conditions determines the farfield pollutant distribution. In general, this equation is transient and three-dimensional. Given typical shallow coastal environments, simplification to two horizontal dimensions, or, with channel-like geometries, even to one dimension are often used, implying some averaging over the missing dimension. Knowledge of the detailed advective velocity field is an important element for the solution of the transport equation. The velocity field may be obtained from a separate numerical calculation of the equations of motion (a so-called circulation model ) or from field surveys, or occasionally from a physical scale model. In recent years, increasingly powerful computer models have become available. The tidal circulation can be computed by a finite difference (FD) or finite element (FE) or finite volume model which solves 3-D or the vertically averaged 2-D equations of motion and continuity. Appropriate closure techniques that link the turbulent stress terms to the mean flow quantities (such as eddy viscosity concepts or k-ε models) need to be employed. The heat (temperature) and salinity conservation equations are usually solved in parallel with the equations of motion since these parameters are linked to the water density by an equation of state. Given the velocity field, the pollutant concentration field is typically obtained by solving the Eulerian convective diffusion equation in two or three dimensions. Vertical averaging is applied over the entire (shallow) water depth, or in case of deep tratified bodies over an upper layer above a pycnocline. A number of public-domain or commercial codes are available at present to aid in the prediction and engineering design of coastal effluent discharge schemes. These include the models Delft3D (Delft Hydraulics), ECOM (Hydroqual), Mike 3 (Danish Hydraulic Institute), POM (Princeton University) and Telemac-3D (HR Wallingford). Eulerian models always require a substantial amount of input data, in terms of detailed topographical information, and ambient data at the open boundary conditions (such as current speeds, temperature and salinity distributions). As an example, Fig. 5 shows the computational grid (plan view) for the Hong Kong/Pearl River water quality simulation using Delft3D. (6)

11 11 Fig. 5: Plan view of computational grid for the Hong Kong / Pearl River quality simulation using Delft3D (Delft Hydraulics, 2001) b) Lagrangian plume tracking models: Thee models treat the development of the unsteady far-field pollutant plume by superimposing individual patches ( particles ) of released pollutant masses. During a time step each pollutant patch is advected by the instantaneous advective velocity field (assumed uniform and homogeneous over the entire domain) and diffused by a random time-dependent diffusion process. In addition, reaction or decay processes can be simulated. Various realistic features (e.g. the return of previously diluted sewage over the outfall and different source conditions) can be readily simulated by superposition methods. However, these methods (as opposed to the general numerical models) are limited to a uniform current field and only allow very simple coastal boundary conditions (e.g. straight coastline). Chin and Roberts (1985) developed a Langrangian random-walk model that directly uses data from spatially distributed, continuously recording current meters to simulate far-field dispersion. The concept of visitation frequency has also been suggested and applied to estimate the probability of the wastefield reaching a particular coastal location (Csanady & Churchill, 1987; Roberts, 1999). Fig. 6 shows an example of a Lagrangian model application for an offshore diffuser outfall (Tetra Tech, 1000). Fig. 6: Two plume traces from an outfall tracked by the prevalent velocity field, Probability diagrams resulting from repeated runs of far-field Lagrangian particle tracking model. Left: Travel time probabilities show probable plume location after discharge. Right: Concentration envelopes at a given vertical elevation (Tetra Tech, 2000)

12 12 The coupling of near- and far-field model analyses represents an important issue in computational fluid mechanics. No definite procedures exist at the present time. Suggested approaches (e.g. Blumberg et al., 1996, or Zhang and Adams, 1999) have various redundancies or shortcomings. Generally, the coupling is less complicated for the case of integral near-field model and a Lagrangian tracking far-field model since both model types have a similar structure. However, considerable conceptual and data handling difficulties exist for linking a near-field integral model to a Eulerian far-field circulation model. Despite enormous advances in computational resources the use of a single complete model that is accurate for both near- and far-field is a long way into the future for engineering practice. 5. INTERNAL DIFFUSER HYDRAULICS AND OPERATIONAL ASPECTS Ocean outfalls consist of three components (Fig. 7): the onshore headworks (e.g gravity or pumping basin); the feeder pipeline which conveys the effluent to the disposal area; and the diffuser section where a set of ports releases and disperses the effluent into the environment. Internal hydraulics affect the flow partitioning (pressure losses) in the manifold and the resulting discharge profile along the diffuser. Fig. 7: Outfall configuration showing feeder pipe and diffuser from side view and top view, defining the pipelines and port/riser configurations Typical engineering design objectives include: 1. An optimized, mostly uniform flow distribution among the orifices in order to meet dilution requirements and to prevent operational problems (e.g. intrusion of ambient water through ports with low flow), 2. Minimal costs for construction achieved with short, small diameter pipes and low operational costs with minimal pumping heads, 3. Prevention of off-design operational problems (i.e. particle deposition, salt water intrusion, plugged ports) during low flow or no-flow periods, 4. Performance tests during unsteady operation mode (purging during start-up or intermittent pumping cycles, wave induced circulations, water-hammer). Conflicting design parameters demand for compromises, which are often not solved sufficiently. Many outfalls have internal hydraulic problems: total or partial blockage, high head losses, non-uniform flow distribution resulting in poor dilution (Charlton and Neville-Jones, 1988), in most cases resulting from inadequate attention to diffuser hydraulics or wrong estimates of future flows. Existing design algorithms and diffuser programs do not comply with these demands. Diffuser programs used in prior practice appear to stem from two main sources: 1) the work of Koh (Fischer et al., 1979) implemented in the design code PLUMEHYD; and 2) the work of Wood (Wood et al., 1993) as the design code DIFF. Both codes as well as most of other in-house spreadsheets have considerable deficiencies for the computation of diffuser details and operating conditions as they can occur in actual practice. They only consider short risers with negligible friction losses and local losses. They lack the representation of long risers (like in deep-tunneled outfalls) with meaningful frictional and local losses, Y-shaped diffusers, complex port/riser configurations, changing geometry along the diffuser, external pressure variations due to density stratifications, multiple risers at one location or multiple ports on one riser, duckbill valves or other complex port losses and performance checks for off-design operational modes. Furthermore it was common practice to design diffusers only for the ultimate design flow, which caused long-term malfunction during low-flow periods. Todays practice are expanding diffusers (Avanzini, 2003), diffusers designed to meet the initial and final requirements by closing initially a certain number of ports (either with fixed closures or backpressure regulations, which open autonomous if

13 13 enough flow enters the system). Commonly used technologies are duckbill valves. Depending on the internal pressure a rubber like material valve changes its cross section. A new computer program CorHyd has been developed to allow a more detailed analysis and optimization of the internal hydraulic diffuser performance. It is planned to include this program in the mixing zone model CORMIX after its testing phase. CORHYD applies a port-to-port analysis starting at the most seaward port. It applies for steady flow conditions, where the continuity equations at each flow division and the work-energy equation along pipe segments with constant or known flowrate are solved. Necessary input data is the geometry of the discharge structure with sets of node locations x, y, z and pipe segment geometries (i.e. cross-sections, lengths and roughness). The ambient is described by its density and water level resulting in the external hydrostatic pressure at all port orifices along the diffuser pipe. The effluent is described by its fluid density and either the total flow rate Q or the total head at the headworks. Suitable expressions for local losses at orifices, bends, contractions and duckbill valves are taken from Fischer et al. (1979), Idelchik (1986), Miller (1990) and Lee et al. (1998). The Swamee and Jain (1976) friction coefficient is used for calculating the wall friction. Such a system can be solved for all port discharges. The postprocessor includes an optimization cycle and detailed graphical results and simple performance checks for off-design conditions. Preliminary results from CORHYD allow to analyze and optimize the internal hydraulic performance. Table 1 summarizes the effects for the discharge profile and the total head, if increasing the observed parameters for a base system. It is distinguished between horizontal and sloped diffuser lines where either the port elevations are at equal depth or varying along the diffuser. Increase of Parameter Resulting Effect on Total Head Homogeneity Total discharge (no slope) (with slope) or Ambient water depth (no slope) (with slope) or Density difference (no slope) - - (with slope) or Feeder length 0 Diffuser length (constant total length) Diffuser pipe diameter or Pipe roughness 0 Number of risers (constant diffuser length) 0 Riser spacing (variable diffuser length) Riser height 0 Ports per riser Port diameter Flexible valves / = moderate in- / decrease / = strong in- / decrease 0 = neutral or small changes Table 1: Sensitivity of various parameter increases on total head required and homogeneity of the discharge profile. Analyzing and optimizing the discharge profile distribution Application of dilution equations show that a 10% discharge variation along a diffuser would result in 7% dilution difference after mixing. Often these differences are not considered and so far could harm the environment or at least could lead to critical concentrations for the discharge permit. The most effective parameter for changing the discharge profile are the port diameters. Decreasing fixed port diameter for the whole diffuser or only for those sections where discharge is too high leads to a more homogeneous profile. This is even more effective if duckbills are attached. But both increase the local losses for all (fixed orifices) or some of the flowrates (duckbills). Decreasing fixed port diameters leads to increases of the losses and the total head due to the increased velocities and pressures. Increasing nominal port diameter

14 14 for variable area orifices has minor influences on head loss or total head. General optimization constraints are: a 50 mm minimum port size for secondary- or tertiary-level treated effluent and storm water inflow to the sewage system was suggested by Wilkinson and Wareham (1996), thus avoiding the risk of blockage; furthermore a minimum port size of 70 to 100 mm for primary treatment plants (just screening and sedimentation). A design rule, often mentioned in literature (Grace 1978), recommends to keep the ratio between the cumulative port areas ΣA p and the diffuser pipe area A d smaller than one: ΣA p,i /A d <1, with the explication that "it is impossible to make a diffuser flow full if the aggregate jet area exceeds the pipe cross-section area, since that would mean that the average velocity of discharge would have to be less than the velocity of flow in the pipe" (Fischer et al. 1979). This statement is unnecessarily conservative because the average velocity is not the effective measure for manifold flows as diffuser velocities change along the diffuser (for non-tapered and especially for tapered ones) as can do the port velocities if port diameters are varied along the diffuser line. Further recommendations in literature follow that the best ratio "is usually between 1/3 and 2/3" (Fischer et al. 1979, p.419). These suggestions are helpful for large diameter ports (D > 0.75 m), but too restrictive for commonly used smaller ports. If cheaper designs propose other ratios, in particular bigger than unity, a detailed optimization is necessary. This can easily be done in CorHyd by changing the geometries or using the optimization routine. The discharge profile for diffusers with horizontal pipe axis (or equal port elevations) is invariant on total discharge changes, but changes strongly for sloped diffusers. Especially under low flow conditions the profile is weighted to the shore. Due to the constant influence of the pipe slope on the discharge profile the discharge profile asymptotically approaches the non-sloped profile for increasing total discharges. Changes in the main diffuser pipeline geometry, i.e. tapering in order to achieve higher velocities for pipe scouring, do only have minor effects on the discharge profile. Total head optimization The available total head is either the pumping head or the water level in the storage tank. The total head is necessary to drive the system, i.e. to allow a high velocity discharge at each port and to overcome the hydraulic losses. Head loss in the generally long feeder pipes is dominated by wall friction. Bends or contractions (if occurring often gradual) are of minor importance for the feeder. The head loss in the diffuser section is often dominated by local losses at flow separations, bends and contractions, while wall friction is only important for long riser systems. A comparison allows to optimize the most effective parameter, either increasing the feeder diameter or changing the material, or increasing port diameters and simplifying the diffuser. Typically feeder friction is about 60 % of the total head for head loss is in the order of cm for short 100 m and 1 2 m for long outfalls (L = 1000 m). Increasing the diffuser diameter leads to drastic reductions of the head loss and the total head due to decreasing velocities. If tapering is applied (in order to achieve higher velocities for pipe scouring) the head loss is only slightly bigger compared to a non-tapered diffuser, due to higher diffuser pipe velocities in the tapered sections. However, the pipeline diameter is usually set to achieve self-scouring discharge velocities (in excess of m/s) on a daily basis (Wood et al 1993, pp.119, Wilkinson and Wareham, 1996). Off design performance Under low-discharge conditions diffuser are confronted especially with issues of scouring and/or intrusion of seawater. Seawater intrusion is seldom to be avoided for all discharges. There are duckbill valves, but they require additional pumping or higher headwork storage buildings. Intrusion can be prevented by requiring the port densimetric Froude numbers to exceed unity (Wilkinson, 1988), resulting in a critical port velocity V crit = ( ρ/ρ a gd) 0,5, where D denotes the port diameter. For discharges, where this criterion is not met, saltwater could enter into the system and unsteady two-layer flows can occur. This has to be analyzed in detailed numerical or physical models. Unsteady conditions Gradually varying boundary condition variations like diurnal discharge variations, rainfall events or tidal influences are considered as quasi -steady and are included into the calculations described above. Very short pumping cycles instead demand for additional detailed numerical calculation and/or laboratory experiments.