Modeling the Fate and Transport of Nickel in the Mersey Estuary

7/26/2019 Modeling the Fate and Transport of Nickel in the Mersey Estuary

1/23

Journal of Environmental Science and Health Part A, 41:825847, 2006

Copyright C Taylor & Francis Group, LLCISSN: 1093-4529 (Print); 1532-4117 (Online)

DOI: 10.1080/10934520600614454

Modelling the Fateand Transport of Nickelin the Mersey Estuary

Michael Hartnett,1 Binliang Lin,2 Peter D. Jones,3 andAlan Berry4

1MarCon Computations International Ltd. and Department of Civil Engineering,National University of Ireland, Galway, Ireland2School of Engineering, Cardiff University, Cardiff, Ireland3U.K. Environment Agency, Warrington, England4MarCon Computations International Ltd., Galway, Ireland

Modelling heavy metals in estuarine environments is extremely complex for various

reasons; one of the primary complicating factors is that metals exist in two phases,

dissolved and particulate bound. Dynamic changes in water chemistry, and in par-

ticular salinity, affect the partitioning of metals between the two phases and hencemake it difficult to determine the relative fractions of each phase. A relatively simple

approach was developed to relate variations in partition coefficient for Ni to salinity

fluctuations in the Mersey Estuary. The functional relationship developed between

partition coefficient and salinity departs from the traditional exponential type curve,

providing a more realistic relationship.

A numerical model was then developed for predicting the transport and dis-

tribution of Ni about the Mersey Estuary. The model couples transport of metals

throughout the water along with incorporating the chemical processes controlling

how nickel is fractioned between dissolved and particulate phases through the newly

developed partition coefficient relationship. Model predictions of dissolved Ni along the

longitudinal axis of the estuary were compared with measurements of Ni for two events;very good correlation was obtained between the model results and the data.

Key Words: Partition coefficient; Heavy metal modeling; Nickel modeling; Sediment

modeling; Mersey Estuary.

INTRODUCTION

Heavy metals such as Pb, Zn, Cd, Cu and Ni are toxic to many fish species

when present in water in quite low concentrations and are also very dangerous

Address correspondence to Michael Hartnett, Department of Civil Engineering, Na-tional University of Ireland, Galway, Ireland; E-mail: [email protected]

825


2/23

826 Hartnett et al.

in drinking water sources. The EC Dangerous Substances Directive[1 ] is

implemented to control the release of dangerous substances into water. In

this directive Ni has been assigned an environmental quality standard (EQS)

of 30 g/L for coastal and estuarine waters. In implementing EU Directives

such as the Dangerous Substances Directive,[1 ] the Shellfish Directive[2 ] and

the Habitats Directive[3 ] numerical models are used to make predictions of

various water quality parameters, including heavy metals. The development of

a numerical model to accurately predict the transport of heavy metals in water

is a complex task including dynamically coupling between hydrodynamics,

sediments and the metals. The task becomes more complex when modelling

the distribution of heavy metals in large estuaries due to salinity fluctuations.One of the legacies bequeathed by the Industrial Revolution to the Mersey

Estuary is a that a large buildup of industrial discharges has left considerable

deposits of heavy metals bound to the fine sediments of the estuary. As a

result of the neglect of waste management, the estuary earned itself the

ignominious title of the most polluted estuary in Europe. During the period

18501950 increased industrial output resulted in many industries requiring

large volumes of water for cooling and processes such as bleaching, tanning and

metalworking. It is likely that during the 1960s the water quality status of the

Mersey Estuary was at its lowest being severely contaminated by toxic organic

and inorganic compounds from the local industries. At the beginning of the1970s pollution effects from heavy metals caused serious concerns. Then, UK

Government laboratories reported that fish caught in Liverpool Bay contained

elevated levels of many heavy metals. These raised levels were attributed to

the contaminated waters of the Mersey Estuary due to the wanton discharging

of industrial wastes.[4 ]

Because of the serious nature of metal contamination of the Mersey

Estuary, many other studies have been undertaken to define and better

understand the nature of the problem and, in particular, nickel reactivity in

the Mersey Estuary. Campbell et al.[5 ] present and discuss the distribution

and behaviour of both dissolved and particulate bound nickel in the Mersey

Estuary. One of the main conclusions from their work is that the distribution

of nickel throughout the Mersey Estuary is complex and clearly deviates from

simple end-member mixing behaviour.

Turner et al.[6 ] undertook a detailed study of seven trace metals dis-

tributed throughout the Mersey Estuary including nickel. In their paper they

describe the estuary as a highly contaminated, organic-rich estuary and they

observed an increase in the sediment-water distribution coefficient, KD, with

increasing salinity for all metals considered except Cd. These observations

were inconsistent with inorganic speciation calculations and empirical mod-

elling studies in other estuaries, predicting an inverse relationship between

KD and salinity. The primary focus of their paper was to investigate the

factors controlling the salinity distributions of metal data based on defining the


3/23

Modelling the Fate and Transport of Nickel 827

chemical processes that control the KD salinity relationship. They developed

an empirical model based on salting out of neutral organic chemicals that

reflected the salinity distribution of the data.

This is a different approach to the one adopted in this paper; however,

the model developed by Turner et al. has promise for incorporation into

numerical model simulations also. Ng et al.[7 ] carried out a detailed contam-

inant modelling study of the Humber estuary. They outline the difficulties

of making accurate predictions of heavy metals in the Humber due to many

controlling variables including: salinity, pH and availability of complexing

species. Within the model, the partition coefficient was allowed vary as a

function of salinity in either of two ways: using a relationship developed byMillward and Turner[8 ] or using explicit tabulations of measured values of KDas a function of salinity. Millward and Turners relationship is discussed in

more detail in a later section of this paper. The paper by Ng et al. illustrates

how a geochemical module based on empirically derived partition coefficients,

coupled to a two-dimensional hydrodynamic model, was developed to form the

basis of a geochemical contaminant transport model. This basic approach of Ng

et al. is the one followed in the research undertaken by the author; however,

the authors have extended this approach with regards to the formulation of

the partition coefficient for nickel.

Martino et al.[9 ] reported upon high resolution, seasonal distributions of sixdissolved trace metals in the Mersey Estuary. The purpose of this study was

to obtain a more detailed comprehension of the mechanisms controlling metal

reactivity in the estuary. An empirical sorption was used and reproduced, with

reasonable success, axial distributions of the metals that exhibited the largest

peaks, namely Co and Pb. One of the drawbacks of this research is that the

partition coefficients used were single-valued, unrelated to salinity. Further,

the approach adopted in the paper of Martino et al. does not base predictions

of metals on the physics of the dynamical interactions between hydrodynamics

and the mechanics of sediment transport.

Comber et al.[10] undertook research into the partitioning of trace metals

between the dissolved phase and suspended solids on both the Humber

Estuary and the Mersey Estuary. One of the aims of this research was to

develop partition coefficients for a number of metals. A single-valued partition

coefficient for nickel in the Mersey was estimated at ca. 1 104 and thecorresponding value in the Humber was ca. 5 103. The reason for the higher

value of KDin the Mersey was attributed to higher loadings of sewage into the

Mersey along with differences in suspended particulate composition.

In this paper the authors describe a model that was developed to predict

distributions of Ni throughout the Mersey Estuary. An extensive water quality

monitoring programme was developed for the Mersey Estuary from the mid

1970s. Water samples are collected from up to 23 mid-stream locations and

analysed for various constituents including sanitary parameters and trace


4/23

828 Hartnett et al.

contaminants. This programme has continued to date providing the UK

Environment Agency with an extensive data set for use in many aspects of

water quality management and planning of the Mersey Estuary. It is noted

that this monitoring programme is relatively expensive, in particular the

determinations of trace metals requires more sophisticated instrumentation

and skilled personnel than is required for the determination of sanitary

parameters. Because of the high costs of trace contaminant analyses the

development of accurate numerical models to aid in the development of

monitoring programmes and in predicting distributions of trace metals is

desirable.

The development of the Mersey Estuary transport model for Ni is detailedbelow; particular emphasis is placed on how the model computes the partition

coefficient for the metal. Advantage is taken of the extensive database devel-

oped by the Environment Agency in model development and model calibration

and validation.

METHODOLOGY AND RESULTS

Overview of Approach Adopted

When metals are introduced into water containing suspended particulatematter (SPM), a portion of the metal is dissolved in the water and the

remainder is absorbed onto surface of the SPM. The distribution of the metal

fractions between these two phases can be described by a partition coefficient,

KD, defined as:

KD=P

C (1)

where P is the concentration of heavy metal absorbed on suspended sediments

(g/L) and C is the concentration of heavy metal dissolved in the water column

(g/L). The coefficient KD is function of spatially varied water chemistry and

is notoriously difficult to quantify in marine waters. It is well documented

and generally accepted that there is a strong functional relationship between

KD and salinity. In many studies only the influence of salinity are considered

and other direct water chemistry effects are ignored; this approach is adopted

in this research also. Thus in order to model, hindcast and predict heavy

metals, with a reasonable degree of accuracy, the following staged approach

was adopted.

A hydrodynamic model of the Mersey Estuary was developed, calibrated

and validated. This model allowed current speeds and directions be pre-

dicted; these parameters were necessary for subsequent modelling of various

transport processes. A solute transport model was developed, calibrated and

validated and used to predict salinity distributions throughout the estuary


5/23


at various stages of the tide using the results from the hydrodynamic model.

From field data an expression was developed relating the partition coefficient

to salinity, this is a novel component of this research. From the model

salinity predictions and the KD versus salinity relationship, the model has

the capability of computing spatially varied and temporally varied partition

coefficients as the model calculations progress forward in time. A sediment

transport model of the Mersey Estuary was then developed, calibrated and

validated and used to predict estimates of cohesive sediments. Finally, having

predicted cohesive sediment concentrations and values of KDat each grid point

of the model for each computational timestep, the model computes the fraction

of the heavy metal which is dissolved and that fraction which is absorbed ontothe cohesive sediments.

Hydrodynamic Model of Mersey Estuary

The Mersey Estuary is one of the largest estuaries in the UK, having a

catchment of some 5,000 km2 that includes the major cities of Liverpool and

Manchester. The estuary is a macro-tidal estuary with tidal ranges recorded at

Gladstone Dock of between 10.5 m on extreme spring tide to 3.5 m on extreme

neap tides. Freshwater inputs from the River Mersey vary between approxi-

mately 10 m3/s and 500 m3/s at the extremes; typical flows are in the range

2060 m3/s. A numerical model was developed of the estuary extending from

New Brighton at the seaward end to the tidal limit at Howley Weir, Figure 1

shows the extent of the model domain. The two-dimensional model DIVAST

was used during this research. The estuary was represented in the model by

a regular grid defined on two mutually orthogonal axes on a horizontal plane.

The grid spacing was set at 100 m resulting in a model size of 215 308grid squares, or 66,220 computational grid points. At each model grid point

the bathymetry of the estuary was defined from a bathymetric survey of the

estuary carried out for the Environment Agency in 2002. Figure 2 shows a plotof the Mersey Estuary bathymetry; this figure shows the complex geometry

and bathymetry of the study area. The Upper Estuary is a meandering channel

of approximately 15 km in length; below Runcorn the estuary opens into a wide

shallow basin to form the Inner Estuary with extensive mudflats and a large

salt marsh on the southern bank. Downstream further the estuary converges

to form the Narrows, a straight narrow channel of up to 30 m deep.

The hydrodynamic model is based on the following depth integrated

formulation of the Navier Stokes equations, for details of the derivation, see

Falconer.[11] The continuity equation and x, y momentum equations can be

respectively written as:

t+ Qx

x + Qy

y = 0 (2)


6/23

830 Hartnett et al.

Figure 1: Mersey Estuary Study Domain.

Qx

t + UQx

x + VQx

y = f Qy gH

x+ xw

xb

+ H

2U

x2+

2U

y2

(3)

Qy

t + UQy

x + VQy

y = f Qx gH

y+ yw

yb

+ H

2V

x2+

2V

y2

(4)

where = water elevation above (or below) datum; U, V= depth averagedvelocity components in x, y directions; Qx = UH, Qy + VH= unit widthdischarge components in x, y directions; H= + h= total water depth,


7/23


Figure 2: Mersey Estuary Bathymetry 2002.

h= water depth below datum; = momentum correction factor; f= Coriolisparameter;xw, yw= surface wind shear stress components in x,y directions;xb,yb = bed shear stress components inx,y directions; and = depth averageeddy viscosity.

Falconer[5 ] also details how Equations 24 are numerically discretised

and incorporated into the DIVAST computer code. DIVAST is used to find a

numerical solution to Equations 24, resulting in the predictions of current

speeds and directions and water surface elevations at each computationalgrid point for prescribed boundary conditions. In this research the boundary

conditions specified are the tidal dynamics at the seaward boundary and the

Mersey River discharges at Howley Weir. The model underwent extensive

calibration and validation to provide confidence in the model results. Model

predictions were compared against field measurements of currents and water

elevations at a number of different locations throughout the estuary. Figure 3

presents a comparison between model results and field data for current speeds.

The correlation between model predictions and data is considered to be good

and is typical of the correlation achieved during the calibration/validation

aspect of the research. The hydrodynamics model is now considered to be

acceptable for use in predicting the transport of nickel about the Mersey

Estuary.


8/23

832 Hartnett et al.

Figure 3: Comparisons between model predictions and measured current velocities at (a)Pluckingham and (b) Eastham.

Solute Transport Model of Mersey Estuary

Salinity is transported into the Mersey Estuary from the Irish Sea during

flood tides; the near constant concentration of salinity in flood waters is

35 psu. The Mersey River and the 13 other tributaries to the estuary, along

with diffuse runoff from the catchment bordering the estuary, discharge

considerable amounts of freshwater water into the estuary. Thus significant


9/23


gradients of salinity prevail from Howley Weir to the seaward end of the

estuary; these gradients vary with time due to tidal dynamics and the volumes

of freshwaters entering the estuary. The salinity from the seawater mixes

with the freshwater water to produce complex salinity patterns throughout

the estuary. The depth-averaged, two-dimensional advective-diffusion equa-

tion for the salt transport processes can be obtained by integrating three-

dimensional advective-diffusion equation over the water depth, Falconer[11];

the depth-averaged governing equation for the salt transport can be

written as:

H

t +UH

x +VH

y

xHDx

x

yHDy

y = 0 (5)

where = depth-averaged salinity; Dx, Dy = depth-averaged diffusioncoefficients in x, y directions, respectively, which can be calculated as

Preston[12]:

Dx=(klU

2 + ktV2)HgU2 + V2C

+DW, Dy=(klV

2 + ktU2)HgU2 + V2C

+DW

in whichkl= longitudinal dispersion constant; kt= lateral diffusion constant;DW =wind induced dispersion coefficient;C= Chezy bed roughness coefficient;and g= gravitational acceleration. The values of constant kl and kt are 5.93and 0.15, respectively in the literature.[13,14] However, in practical studies

these values tend to be rather low,[15] with measured value for kland ktranging

from 8.6 to 7500 and 0.42 to 1.61, respectively.

Equation 5 describes how a constituent, q, is transported about a water

body based on currents and turbulent diffusion. The UK EA have a consider-

able data base of measured values of salinity throughout the Mersey Estuary

for various environmental conditions. In developing the solute transport model

a total of 44 freshwater discharges were specified from rivers and other

point sources. The model was initiated by firstly specifying an initial zero

salinity concentration at all points throughout the estuary. The model was

then executed to simulate how salinity would be transported into the model

domain and mix with the freshwater. The model was run for 20 tidal cycles

until steady state conditions were attained.

Model salinity predictions were compared against measured data at many

locations along the axis of the estuary. Figure 4 shows the comparison of

salinity predictions against measured salinity at two points along the estuary

during model calibration; the points referred to in Figure 4 are located at

Princes Pier and Runcorn. Also, Figure 5 shows the comparison of salinity

predictions against measured salinity at two points along the estuary during

model validation. From Figure 4 it is seen that there is quite good agreement

between the predicted and measured salinity values. The correlation shown


10/23

834 Hartnett et al.

Figure 4: Comparisons between model predictions and measured salinities at (a) PrincesPier and (b) Runcorn.

is typical of the correlation achieved for many comparisons between predicted

and measured salinity values.

The accurate prediction of salinity was very important during the develop-

ment of the Mersey Estuary model. First, by being able to accurately predict

salinity provided the confidence that the model would be able to accurately

predict the transport and dispersion throughout the estuary of material


11/23


Figure 5: Comparisons between model predictions and measured salinities at (a) Easthamand (b) Fiddlers Ferry.


12/23

836 Hartnett et al.

discharged into the estuary through rivers, outfalls and diffuse discharges.

Secondly, and equally important during this research, accurate salinity values

were required to compute the partitioning coefficient in Equation 1. As can be

seen from Figures 4 and 5, salinity varies significantly both temporally and

spatially with consequential changes in the partitioning coefficient, emphasiz-

ing the importance of accurate salinity predictions.

Development of Partitioning Coefficient-Salinity Relationship

The UK EA routinely collect water samples along the axes of the Mersey

Estuary at 23 locations; these samples are usually collected on a monthlybasis. The water samples are subsequently analysed for many water quality

parameters, including salinity, sediments, dissolved heavy metals and partic-

ulate bound heavy metals. A suite of heavy metals are analysed for, including

Ni. From this data a relatively simple approach was adopted to evaluate the

partitioning coefficient.

From the database of survey results for heavy metals the values for

dissolved and particulate-bound Ni were extracted, along with the associated

salinity values for each sample. Using the dissolved and particulate concen-

trations of Ni, a value for the partitioning coefficient was computed for each

sample. These partitioning coefficients were then plotted against the salinityvalue associated with each sample; Figure 6 presents the plot of KD versus

salinity. A functional relationship was developed between KDand salinity from

the data presented in Figure 6. A general relationship was postulated of the

form:

KD(s) = A+ B.S+ C.S2 +D.S3 +E.S4 + F.S5 (6)

where S is salinity.

A regression analysis was carried out to find the coefficients of Equation

6 that would provide the best fit with data. From this analysis, the followingcoefficients were derived relating partitioning coefficient to salinity for Ni:

A= 7310B = 3003C = 457D = 23.4E = 0.4

F = 0The curve in Figure 6 shows the graphical representation of Equation 6

when the preceding coefficient values are used. The following points are


13/23


Figure 6: Plot of partition coefficient for Ni versus salinity.

worth observing from the approach adopted and the relationship that hasbeen developed. First, there is a considerable spread between the data and

Equation 6. The data shows quite a degree of variation and by inspection we

can see that, due to the complex chemical relationships governing partitioning

coefficients, a simple relationship based purely on salinity will never be able

to fully capture the fluctuations in the partitioning coefficients. Second, it is

very interesting to notice that there is not a unique relationship between

partitioning coefficient and salinity; for example, low salinity values and

salinities in the range of 1525 psu give approximately the same value

of KD.

From Figure 6 it is clear that the values of KD are higher for low and

high salinity values than for intermediate values; Equation 6 reflects this

characteristic of the KD versus salinity relationship. Most other approaches

to computing KD are indeed based on unique relationships between KDand salinity, obviously neglecting the important property illustrated above.

Further work undertaken by the authors show that other heavy metals in

both the Mersey Estuary and the Ribble Estuary exhibit salinity relationships

similar to that shown in Figure 6.

Sediment Transport Model of Mersey EstuaryThe manner in which heavy metals are distributed and transported

about an estuary is significantly influenced by bed sediments and suspended


14/23

838 Hartnett et al.

particulate matter; this is certainly the case in the Mersey Estuary. Bed

sediments throughout the Mersey Estuary have been heavily contaminated

with heavy metals, mineral oils and other toxic contaminants resulting from

industrial and municipal discharges and accidental spills. Due to spring tidal

ranges of in excess of 10 m, tidal dynamics throughout the estuary are often

quite strong resulting in the resuspension of bed sediments during periods of

strong tidal induced bed shear stresses. During periods of low hydrodynamic

activity, typically around times of low and high water, suspended particulate

matter is allowed settled to the bed again. Thus concentrations of SPM vary

quite considerably throughout the tidal cycle with relatively large amounts of

sediments being shunted up and down the estuary by tidal action.The dynamics of the SPM in the Mersey Estuary is important for two

main reasons. Firstly, during certain periods of the tidal cycle the bed becomes

a source of heavy metals into the water column and at other times the bed

becomes a sink for heavy metals from the water column. Obviously, when

sediments are being resuspended we must know the mass of the heavy metal

that is bound to these sediments to quantify the source terms. Hence in order

to accurately model heavy metals it is necessary to know the distribution of

sediments throughout an estuary and also the mass of particulate bound heavy

metals. Second, as the amount of SPM changes in the water column more or

less metal is exchanged from the dissolved stage to the particulate bound stage.Thus the accurate prediction of heavy metals relies on the accurate predictions

of sediments in the water column.

The two-dimensional equation governing the transport and dispersion of

cohesive sediments throughout the estuary is given as:

SH

t + SUH

x + SVH

y

x

HDx

S

x

y

HDy

S

y

= qero + qdep (7)

where S= depth-averaged cohesive sediment concentration; qdep, qero= thedeposition and erosion rates, respectively.

For the non-cohesive sediment transport we have the following two-

dimensional depth-averaged equation for the non-cohesive sediment transport:

SH

t + SUH

x + SVH

y

x

HDx

S

x

y

HDy

S

y

= Eae+wsS (8)

where Eae= erosion constant (kg m2 s1) and ws= settling velocity. Furtherdetails of the derivation of Equations 7 and 8 can be found in Wu and

Falconer.[9 ]

The sediment model is an integral part of the DIVAST numerical wa-

ter quality model. The sediment transport model uses the same computa-

tional/bathymetric grid as the hydrodynamic model and sediment concentra-

tions are computed at the same computational timestep. In Equations 7 and


15/23


Table 1: Non-cohesive sediments grading.

Grain size finer than (%) Effective grain size (m)

D16 10D50 50D84 175D90 205

8 the U and V components of velocity at each mode grid point and each

computational timestep are obtained from the hydrodynamic model.The bed of the Mersey Estuary contains both cohesive and non-cohesive

sediments. By and large, the non-cohesive sediments lie on the estuary bed

between Howley Weir and Runcorn as shown in Figure 1, from Runcorn

seawards the bed sediments are predominantly cohesive. These observations

are made from assessment of bed sediment data collected by the UK EA along

the axis of the estuary. The non-cohesive sediments grading distribution is

given in Table 1. The grading shown in Table 1 is based on an averaging of 15

sediment samples taken along the axis of the bed. Cohesive sediments are of

more importance in this research as heavy metals become adsorbed onto the

surfaces of these sediments; the cohesive floc size of the cohesive sediments,C50, is 64 m.

Apart from the resuspension of bed sediments the other significant sedi-

ment load was from the River Mersey. The UK EA has extensive measurements

of flow rates and associated sediment loads discharging from the Mersey River

into the estuary. From these data it was decided to calibrate and validate

the model using a 10-day average river flow from the Mersey River with its

associated sediment load.

The suspended sediment transport model was calibrated by comparing

sediment concentrations model predictions against field measurements. Com-

parisons were made against data collected on 18th September 1989 at six

locations along the estuary. During the field survey a spring tide of 9.36 m

was recorded at Gladstone Docks and the 10-day average Mersey River flow

was estimated to be 1,050 103m3/day.The model was calibrated so that model predictions were as accurate as

possible. The model empirical coefficients were tuned to improved accuracy,

the best fit between predictions and data were obtained when the coefficients

had the following values:

Empirical erosion constant 0.00004 kg/N/sCritical shear stress for erosion 1.5 N/m2

Critical shear stress for deposition 0.25 N/m2


16/23

840 Hartnett et al.

Figure 7: Comparisons between model predictions and measured concentrations of SPM at(a) Princes Pier and (b) Fiddlers Ferry.

Figure 7 shows comparisons between model predictions and data at two

locations, Princess Pier and Fiddlers Ferry; similar correlations were achieved

at other locations along the estuary. From Figure 7 it is seen that the model can

reasonably accurately predict SPM and the model reflects the peaks in SPM at

various stages in the tide when bed shear stress are high. Figure 7 is typical

of SPM model calibration in general.


17/23


Heavy Metal Model of Mersey Estuary

As can be seen, modelling heavy metals in an estuary is dependent on

accurately modelling hydrodynamics, salinity and SPM as well developing a

good understanding of the relationship between KD and salinity. This section

of the paper details the numerical predictions of dissolved Ni throughout the

Mersey Estuary. Details of the formulation of the heavy metal differential

equations and the numerical model can be found in Wu and Falconer. [16]

The main sources of heavy metals in the Mersey Estuary are from: bed

sediments; river discharges; domestic discharges; and seawater.

Table 2 presents the average flow rates and concentrations of Ni in the

flows from the river (total Ni) and domestic/industrial (dissolved Ni) wastewa-ter discharges. The Ni load derived from flooding seawater was prescribed by

specifying a concentration of 1.0 g/L of dissolved Ni in the concentration of

the seawater. Within the bed sediments the Ni concentration, mass of Ni per

unit mass of sediment, was specified at 60 g/g of cohesive sediment. At each

computational timestep the model tracks the mass of cohesive sediment being

deposited and resuspended and then computes the mass of Ni from this source.

When Ni enters from the water column above sources, the model computes

the salinity at each location, it then computes partitioning coefficients and

fractions the Ni between dissolved and sediment bound based on the available

cohesive sediments at each model grid point. All the sources of Ni defined in

the model were derived from field data collected by the UK EA.

Model calibration was necessary to ensure high correlation between model

predictions and actual concentrations of dissolved Ni in the estuary. In partic-

ular, calibration was necessary to evaluate the empirically derived KD-salinity

formulation presented in Equation 6 above. Model calibration was carried out

by comparing model predictions of dissolved Ni against measurements of Ni

taken by the UK EA on 9th February 2001. Figure 8 shows the correlation

Table 2: Sources of Ni discharging into Mersey Estuary.

Source Flow rates (m3/s) Ni concentration (g/l)

Wastewater TreatmentPlants

Sandon 3.385 25Warrington 0.941 6Croda Colloids 0.045 6

River DischargesWeaver 5 7Mersey 30100 7

Sankey 2.62 7

Ditton 1 7MSC@Easham 7 8

Dissolved and particulate Niotherwise values relate to dissolved Ni.


18/23

842 Hartnett et al.

Figure 8: Metal Calibrationcomparison between model predictions and measureddissolved Ni concentrations along axis of Mersey Estuary on 9th September 2001.

between the model predictions and measured data for Ni along the axis of the

estuary downstream from Howley Weir.

Further, model validation was subsequently carried out against data

collected on 28th March 2002. During validation the model parameters remain

unchanged, the environmental conditions prevailing on the survey dates arespecified to the model and the model is executed. Figure 9 presents comparison

between model predictions and measured values of dissolved concentrations of

Ni.

DISCUSSION

The measured dissolved Ni data taken on both the calibration and validation

dates show different conditions prevailing with respect to dissolved Ni concen-

trations in the water column; the model results reflect these variations and

exhibit a high degree of correlation with the measured data. Data shows Ni

concentrations varying in both time and spatially due to varying flow rates and

tidal dynamics. Although the estuary bed has been modeled as having a sharp


19/23


Figure 9: Metal Validationcomparison between model predictions and measureddissolved Ni concentrations along axis of Mersey Estuary on 28th March 2002.

divide between the locations of cohesive and non-cohesive sediments, model

predictions are still quite good considering how complex it is to understand

and then model heavy metals. The general equation of the form shown in

Equation 5 relating partitioning coefficient to salinity has departed from

previous approaches and has given a greater degree of flexibility in defining theform of the KD-salinity relationship. Conventionally the chemical behaviour of

heavy metals in brackish waters have been based on exponential KD-salinity

relationships, such as the following relationship developed by Turner and

Millward:[8 ]

ln KD= b[ln(S+ 1)]+ ln K0D (9)

where b is the slope of ln(KD) versus ln(S+ 1) and K0D is the partitioning infreshwater.

Turner and Millward[12] provide values of b and K0Dfor three heavy metals,

Cd, Cs and Zn; these values are presented in Table 3. Figure 10 graphical

illustrates KD for Zn using Equation 9 and the values from Table 3. From


20/23

844 Hartnett et al.

Table 3: Partitioning coefficient parameters for various metalsSource Turner and

Millward.[8 ]

Metal b K 0D

Cd 1.36 199Cs 0.688 3.5Zn 0.374 32.9

Figure 10 the exponential shape of KDis obvious, showing that KDhas a unique

relationship with salinity and is shown to decrease with increasing salinity.

Importantly, by inspection of Figure 6 we see that the exponential decrease

of KDwith salinity, that is often adopted by modellers, does not happen for the

case of Ni in the Mersey Estuary. The data shows relatively high values for

KD at low salinities, a drop in KD values at intermediate salinities and a rise

again in KD values for high salinities. There is an obvious trend to this data

and so the model must take this into account. Equation 6 as plotted in Figure

6 also exhibits this trend and hence is considered to define a more realistic

relationship between KD and salinity than an exponentially decreasing func-

tion. Equation 6 importantly allows the non-unique relationship between KD

and salinity be expressed mathematically, so that for two different salinitiesthe relationship may compute the same KD value.

Figure 10: KD-salinity plot for Zn-based on Turner and Millward.[12]


21/23


The development of Equation 6 is conceptually very simple, it is based on

sampling estuarine water and analysing the sample for dissolved and partic-

ulate metals and salinity. The methodology inherently includes variations in

the salinity of the actual estuarine waters that provides an indicator for vari-

ations in water chemistry. Based on this approach, once the numerical model

accurately predicts the spatial and temporal salinity variations partitioning

coefficients can be simply calculated.

Figure 6 shows that there is a considerable spread in the KD values

computed from data for a given salinity for Ni, however, this spread is

relatively small when we consider other metal constituents of the Mersey

Estuary such as lead. Figure 11 shows a KD-salinity plot for Hg based on mea-sured data of dissolved and particulate Hg and on measurements of salinity.

Equation 6 represents an average relationship for KD-salinity; however, from

the model results this relationship appears to work quite well in predictions

the transport and dispersion of Ni throughout the Mersey Estuary. Thus by

developing KD-salinity relationships from field measurements the extent of

fluctuations in water chemistry becomes apparent and can be incorporated into

the relationship in an averaged manner.

The authors are aware of the work done by Turner and Millward[8 ]

in considering conceptual aspects of KD-salinity relationships to develop

improved predictive bio-geochemical models. Their work was based on aseries of laboratory experiments. Issues addressed by Turner and Millward

included the dependency of solid-solution partitioning on particle size and

isotope speciation in solution, and the relative contribution and implications of

flocculant products to absorbed contaminant concentrations. However, in order

Figure 11: Plot of KD-salinity for Hg in the Mersey Estuary.


22/23

846 Hartnett et al.

to use the approaches outlined by Turner and Millward it is necessary to have

considerable more information on estuarine processes such as floc formulation.

Even if these processes were well understood, this is would require very

extensive computational time making the modelling tasks extremely difficult

to undertake.

CONCLUSION

The general methodology outlined above and its application to the Mersey

Estuary is similar to many other heavy metal modelling studies. The main dif-ference in approach has been the manner in which the partitioning coefficient

has been developed; this has significance for future heavy metal modelling

studies. As in many previous studies, KD is expressed as a function of salinity

that reflects total water chemistry. The KD-salinity relationship developed for

Ni is based on direct sampling of the estuary waters and subsequent chemical

analysis. Having developed this relationship, it is quite easy to incorporate it

into a numerical model, requiring only to compute salinity before calculating

the value of KD. The metal model developed for the Mersey has been proven

to predict Ni concentrations that are quite close to measured data, proving the

approach is promising in this case.The authors are developing KD-salinity relationships for other metals

in the Mersey Estuary and elsewhere using the same approach. For some

metals the application of Equation 6 to data results in an exponential type

of KD-salinity relationship, for other metals the approach does not result in an

exponentially decreasing relationship. Thus one of the advantages of this ap-

proach is that the form of the KD-salinity relationship is not overly prescribed.

Through the application of the above approach, the authors are in the

process of developing a series of KD-salinity relationships for various metals

and in different estuaries. The completion of this work will shed more light

on how KD-salinity relationships vary between metals and between estuaries.

The authors will publish this work on its completion. This will provide other

researchers with useful initial relationships for heavy metal modelling in the

absence of data for the system of interest.

ACKNOWLEDGMENTS

This work was funded by the UK Environment Agency under Contract No.

12153: Development of a Water Quality Model for the Mersey Estuary. The

authors wish to thanks the UK Environment Agency for permission to publish

this paper.


23/23


REFERENCES

1. Council of the European Communities. Council Directive 76/464/EEC of 4 May1976 on pollution caused by certain dangerous substances discharged into the aquaticenvironment of the Community. Official Journal 1976,L129, 2329.

2. Council of the European Communities. Council Directive 79/923/EEC of 30 October1979 on the quality required of shellfish waters. Official Journal 1979,L281, 4752.

3. Council of the European Communities. Council Directive 92/43/EEC of 21 May 1992on the conservation of natural habitats and of wild fauna and flora. Official Journal1994,L206, 750.

4. Jones, P.D. The Mersey EstuaryBack from the Dead? Solving a 150-Year OldProblem. J. CIWEM. 2000,14, 124130.

5. Campbell, J.A.; Whitelaw, K.; Riley, J.P.; Head, P.C.; Jones, P.D. Contrastingbehaviour of dissolved and particulate nickel and zinc in a polluted estuary. Sci. TotalEnviron. 1988,71, 141155.

6. Turner, A.; Martino, M.; Le Roux, S.M. Trace metal distribution coefficients in theMersey Estuary, UK: Evidence for salting out of metal complexes. Environ. Sci. Technol.2002,36, 45784584.

7. Ng, B.; Turner, A.; Tyler, A.O.; Falconer, R.A.; Millward, G.E. Modelling contami-nant geochemistry in estuaries. Water Res. 1996,30, 6374.

8. Turner, A.; Millward, G.E. Partitioning of trace metals in a macrotidal estuary.Implications of contaminant transport models. Estuarine, Coastal Shelf Sci. 1994, 39,4558.

9. Martino, M.; Turner, A.; Nimmo, M.; Millward, G.E. Resuspension, reactivity andrecycling trace metals in the Mersey Estuary, UK. Mar. Chem. 2002,77, 171186.

10. Comber, S.D.W.; Gunn, A.M.; Whalley, C. Comparison of the partitioning of tracemetals in the Humber and Mersey estuaries. Mar. Poll. Bull. 1995,30, 851860.

11. Falconer, R.A. An introduction to nearly horizontal flows. In Coastal, Estuarialand Harbor Engineers Reference Book; Abbott, M.B.; Price, W.A.; Price, E.; Spoon, F.N.,Eds.; London, 1993; 2736.

12. Preston, R.W.Representation of Dispersion in Two-Dimensional Water Flow, Re-port No. TPRD/L/2783/N84, Central Electricity Research Laboratories, England, 1985;113.

13. Elder, J.W. The dispersion of marked fluid in turbulent shear flow. J. Fluid Mech.1959,5, 544560.

14. Fischer, H.B. Mixing and dispersion in estuaries. Annu. Rev. Fluid Mech.1976,8,107133.

15. Fischer, H.B.; List, E.J.; Koh, R.C.Y.; Imberger, J.; Brooks, N.H.Mixing in Inlandand Coastal Waters; Academic Press: London, 1979.

16. Wu, Y.; Falconer, R.A. A mass conservative 3-D numerical model for predictingsolute fluxes in estuarine water. Adv. Water Res. 2000,23, 531543.

Documents

Modeling the Fate and Transport of Nickel in the Mersey Estuary