www.elsevier.com/locate/scitotenv

Science of the Total Environm

The Wheal Jane wetlands model for bioremediation

of acid mine drainage

P.G. Whitehead*, B.J. Cosby, H. Prior

Aquatic Environments Research Centre, Department of Geography, University of Reading, Box 227, Whiteknights, Reading, RG6 6AB, UK

Abstract

Acid mine drainage (AMD) is a widespread environmental problem associated with both working and abandoned mining

operations. As part of an overall strategy to determine a long-term treatment option for AMD, a pilot passive treatment plant

was constructed in 1994 at Wheal Jane Mine in Cornwall, UK. The plant consists of three separate systems, each containing

aerobic reed beds, anaerobic cell and rock filters, and represents the largest European experimental facility of its kind. The

systems only differ by the type of pretreatment utilised to increase the pH of the influent minewater (pH b4): lime dosed (LD),

anoxic limestone drain (ALD) and lime free (LF), which receives no form of pretreatment. Historical data (1994–1997) indicate

median Fe reduction between 55% and 92%, sulphate removal in the range of 3–38% and removal of target metals (cadmium,

copper and zinc) below detection limits, depending on pretreatment and flow rates through the system. A new model to simulate

the processes and dynamics of the wetlands systems is described, as well as the application of the model to experimental data

collected at the pilot plant. The model is process based, and utilises reaction kinetic approaches based on experimental microbial

techniques rather than an equilibrium approach to metal precipitation. The model is dynamic and utilises numerical integration

routines to solve a set of differential equations that describe the behaviour of 20 variables over the 17 pilot plant cells on a daily

basis. The model outputs at each cell boundary are evaluated and compared with the measured data, and the model is

demonstrated to provide a good representation of the complex behaviour of the wetland system for a wide range of variables.

D 2004 Elsevier B.V. All rights reserved.

Keywords: Acid mine drainage; Bioremediation; Metals; Water quality; Wetlands; Modelling; Wheal Jane Mine

1. Introduction models through dynamic lumped process models, to

There are many approaches to modelling environ-

mental systems, ranging from empirical dblack boxT

0048-9697/$ - see front matter D 2004 Elsevier B.V. All rights reserved.

doi:10.1016/j.scitotenv.2004.09.012

* Corresponding author. Tel.: +44 118 987 5123; fax: +44 118

931 4404.

E-mail address: p.g.whitehead@reading.ac.uk

(P.G. Whitehead).

distributed full chemical speciation models. The

impacts of acid mine drainage (AMD) on rivers has

been modelled by Whitehead and Jeffrey (1995) and

detailed process-based models of acidification have

been developed for upland soil systems (Cosby et al.,

1985a, b). Few models have simulated the impact of

constructed wetland treatment of acid mine drainage.

Conceptual models have predicted iron retention and

ent 338 (2005) 125–135

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135126

cycling in wetland ecosystems receiving coal mine

drainage (Mitsch et al., 1981, 1983; Fennessy and

Mitsch, 1989a,b). In addition, Flanangan et al. (1994)

developed a more comprehensive model predicting

iron, manganese, aluminium and sulphate retention in

a proposed wetland remediation site in Ohio, USA.

This model was then evaluated once the wetland had

been built (Mitsch and Wise, 1998). However, there

have been few comprehensive studies of passive

treatment systems, which have resulted in useful

model developments for highly acidic systems.

A key objective of the Wheal Jane project has been

to develop generic passive treatment system models

that could be used to represent the dominant processes

controlling the behaviour of the wetland system

(Whitehead and Prior, 2004). Once developed, this

model could be used to evaluate other systems and

could also be used to identify the most suitable

operating conditions for the design and operation of

future bioremediation schemes for the treatment of

AMD.

2. Wheal Jane Pilot Plant

The passive treatment plant consists of three

schemes (Whitehead and Prior, 2004). These differ

only in the pretreatment utilised to modify the pH of

the influent minewater, namely, lime dosing to pH 5

(lime dosed, LD), an anoxic limestone drain (ALD)

and no modification (lime free, LF). All three systems

include the following: (i) constructed aerobic reed

beds designed to remove iron and arsenic, (ii) an

Fig. 1. Iron concentrations down the Wheal Jane Pilot

anaerobic cell to encourage the reduction of sulphate

and facilitate the removal of zinc, copper, cadmium

and the remaining iron as metal sulphides, and (iii)

aerobic rock filters designed to promote the growth of

algae and facilitate the precipitation of manganese.

An extensive record of biological, chemical and

physical data exists following the completion of the

first phase of research at the site during 1994–1997

(Hamilton et al., 1999). The second research phase

started in 1998 and included weekly water quality

monitoring of key points across the site (A1–3, A8–9,

A13, B1, B11, B15–16, B21–24, C1, C6–7 and C11;

see Whitehead and Prior, 2004).

Samples were analysed for a broad range of

determinants, including pH, dissolved oxygen

(DO), redox potential, alkalinity, dissolved metals

and anions. All analyses were completed within 1

week of collection at the site, in accordance with

UKAS and BS5750 quality assurance controls.

Comparisons between Adit water and Carnon River

water are provided, for comparison, by Neal et al.

(2004).

3. Biochemical behaviour of the Wheal Jane

wetland system

To develop a model for AMD bioremediation at the

site, historical and current water chemistry data were

analysed to provide an insight into chemical processes

occurring and set limits for the model.

Interim results of the water chemistry appear to

support the original research carried out at the site

Passive Treatment Plant (June–November 1996).

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135 127

from 1994 to 1998 (Fig. 1). All systems appear to

remove soluble Fe from the AMD, with removal

primarily taking place in the first two aerobic cells.

The ALD configuration has the highest removal rate,

with a median of 99.5%, with the LD and LF

systems removing a median of 96.5% and 94.8%,

respectively.

When comparing the removal rates of other metals

from the three systems, it is noticeable that there is a

broad difference in the capacity of each system to

remove metals in addition to a wide variance between

elements.

The relationship between potential and pH at

sites across the passive treatment plant has been

mapped to allow the interpretation of element

speciation within the three treatment systems using

adapted Pourbaix diagrams (Pourbaix, 1974, White-

head et al., 2004).

These theoretical diagrams depict the thermody-

namically most stable form of an element, as a

function of potential and pH, and show the relative

stability of an element and its predominant form

under a range of environmental conditions. This

simple analysis has allowed the theoretical determi-

nation of which chemical equilibrium equations

could be included in the model for each element

considered. However, recent work by Johnson and

Hallberg (2004), Hall and Puhlmann (2004) and

Hall et al. (2004) indicate that most of the transition

Fig. 2. Spatial variations in dissolved oxygen at Wheal Jane Pilot Passiv

aerobic cells and the rock filters, but low DO in the anaerobic zones (rea

between metal speciation forms is controlled by

microbial behaviour. Thus, using the equilibrium

approach is not the best strategy for determining

metal concentrations. Hall and Puhlmann (2004)

conducted a series of laboratory and field experi-

ments to determine the reaction rate controls and

has found that the processes are generally first-order

microbiologically controlled reactions with measur-

able reaction rates. Thus, it is possible to utilise this

knowledge when modelling the processes at Wheal

Jane and other bioremediation systems.

Spatial variations in pH, dissolved oxygen, con-

ductivity, redox and BOD at the pilot passive treat-

ment plant were determined. The system variations of

these parameters were important, and they allowed the

environmental conditions within each section or reach

of the three systems to be analysed. The maximum,

minimum and interquartile ranges for each reach were

calculated for all parameters for the period June to

November 1996, to allow limits to be established and

the determination of important processes to aid model

development. Dissolved oxygen plots for this period

(Fig. 2) show that, although heavily depleted in the

anaerobic cell, the final stages of the treatment process

in the rock filter are sufficient to return the DO

concentration to levels recorded at the end of the final

aerobic cell. This indicates that DO reaeration is an

important process in this section of the treatment

plant.

e Treatment Plant showing low initial DO, rising DO through the

ches 5, 4 and 3 in the three systems).

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135128

4. The Wheal Jane model

The model for the Wheal Jane passive treatment

system is based on a set of ordinary differential

equations that are solved using a fourth-order Runge

Kutta Integration routine with a Merson variable step

length procedure. The advantage of this system is that

it is fast and the model results are generated rapidly

(Whitehead et al., 1979, 1998a,b). This allows the

user to simulate the system quickly and repeat runs to

evaluate changing behaviour patterns. The model

consists of 22 equations for each of 17 cells. The 17

cells consist of the limestone treatment tank, the ALD,

the five aerobic cells, the anaerobic cell and the nine

rock filters, as shown in Fig. 3. The 22 equations

consist of the flow, metals, anions, cations and a

tracer.

The basic model structure comprises of the

following: (i) input minewater chemistry, (ii) lime

dosing section (which may be bypassed for LF

system), (iii) ALD section (which may be bypassed

for LD and LF system), (iv) aerobic cells, (v)

anaerobic cells, (vi) rock filter and (vii) output

chemistry. A total of 22 variables is being modelled

(flow, D.O., B.O.D., Al, As, Ca, Cd, Cl, Cu, Fe, Hg,

K, Mg, Mn, N, NH4, Na, Ni, P, Pb, S, Zn and tracer).

Processes modelled include the following:

! Temperature, evapotranspiration and rainfall

effects;

! Water balance, mass balance, first-order temper-

ature dependent reaction kinetics for metals and

nutrients;

! Cation–anion balance, pH calculation;

Fig. 3. The Wheal Jane Model configuration showing lime dosing,

anoxic limestone drain (ALD), aerobic cells, anaerobic cells and the

rock filters.

! Microbial mediation of sulphide reactions, macro-

phyte effects on aerobic cells, algal biomass effects

and uptake of Mn; and

! BOD decay, DO reaeration.

To model water quality, it is necessary to first

simulate water flow in all cells in the system. The

first stage of the flow model development is to

compute a mass balance over the cells. The model

for flow variation in each reach is based on a

nonlinear reservoir model, and thus, the model may

be viewed in hydrological flow routing terms as one

in which the relationship between inflow (I) from

the upstream cell, outflow (Q) from the cell, the

volume (V) of water in the cell and hydrologically

effective rainfall (HER) in each reach is represented

by the continuity equation:

dQ tð Þdt

¼ Q tð ÞV tð Þ I tð Þ þ HER tð Þ � Q tð Þ½ � ð1Þ

The hydrologically effective rainfall is determined

from the actual daily rainfall less the estimated

evapotranspiration calculated using the UK Met

Office MORECS system (MetOffice, 1981). The form

of Eq. (1) has been used and evaluated by Whitehead

et al. (1998a,b) to model a wide range of stream and

catchment systems. The Q/V term represents an

estimate of the reciprocal of the mean residence time

in the cell so that as volume in the cell increases or the

flow rate decreases, the residence time will increase.

The dynamic water quality model is based on a

similar mass balance approach but includes factors to

allow for the nonconservative nature of the water

quality variables. For example, dissolved oxygen in

the system is a balance between the various sources

(e.g., reaeration) and sinks of oxygen (e.g., oxygen

loss by biochemical decay processes). The mass

balance equations required to simulate the behaviour

of any variable can be written in differential equation

form for a reach as follows:

dX tð Þdt

¼ U tð ÞT tð Þ � X tð Þ

T tð Þ FZ tð Þ ð4Þ

where X refers to the downstream (cell output)

concentration, mg l�1; U refers to the upstream (cell

input) concentration, mg l�1; T is the cell residence

time, which will vary as a function of flow and

Fig. 4. Daily rainfall (mms) and temperature (8C) at Wheal Jane (June–December 1996).

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135 129

volume of the cell as described above; and Z refers to

additional sources or sinks affecting the cell.

The differential equations used in the model are as

follows.

Tracers

dX1 tð Þdt

¼ U1 tð ÞT tð Þ � X1 tð Þ

T tð Þ FZ1 tð Þ ð5Þ

Metals

dX2 tð Þdt

¼ U2 tð ÞT tð Þ � X2 tð Þ

T tð Þ

� k11:047h�20ð ÞX2 tð ÞFZ2 tð Þ ð6Þ

Fig. 5. Wheal Jane Mine chemistry time series results—Reach 7 observed

Zone (June–December 1996).

Nitrogen

dX3 tð Þdt

¼ U3 tð ÞT tð Þ �X3 tð Þ

T tð Þ �k21:047h�20:ð ÞX3 tð ÞFZ3 tð Þ

ð7Þ

Dissolved oxygen

dX4 tð ÞT tð Þ ¼ U4 tð Þ

T tð Þ � X4 tð ÞT tð Þ

þ k31:047h�20:ð Þ Cs � X4 tð Þð Þ

� 4:2 k21:047h�20:ð ÞX3 tð Þ

� k41:047h�20:ð ÞX5 tð ÞFZ2 tð Þ ð8Þ

(dots) and simulated anions and cations at outflow of the Aerobic

Fig. 6. DO and BOD simulated concentrations down the Wheal Jane System.

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135130

Biochemical oxygen demand

dX5 tð Þdt

¼ U5 tð ÞT tð Þ � X5 tð Þ

T tð Þ � k41:047h�20:ð ÞX5 tð Þ

� k5X5 tð ÞFZ5 tð Þ ð9ÞHere, k1, k2, k3, k4 and k5 are first-order

temperature-dependant rate coefficients relating to

Fig. 7. Flow and chemical simulation p

metal loss, denitrification, reaeration, biochemical

decay and sedimentation (or biofloculation), respec-

tively, in units of days�1. All the rate coefficients,

with the exception of the sedimentation rate, are

temperature dependant via the term 1.047(u�20.),

where h is the cell water temperature (8C). Cs is the

saturation concentration of dissolved oxygen and is

lotted as frequency distributions.

Fig. 8. Metal concentrations simulated halfway down the aerobic cell system.

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135 131

calculated from the equation below (Whitehead et

al., 1979)

Cs ¼ 14:51233� 0:3928026h þ 0:00732326h2

� 0:00006629h3 ð10Þ

The model has been structured so that any of the

three systems at the Wheal Jane Pilot Plant can be

simulated. In addition, the model is quite generic that

any number of aerobic cells, anaerobic cells or rock

filters can be simulated. This allows for altered

wetland designs to be evaluated.

Table 1

Median removal rates of major elements at Wheal Jane Passive

Treatment Plant, 1999–2001

System Al

(%)

Cd

(%)

Cu

(%)

K

(%)

Mg

(%)

Mn

(%)

Na

(%)

S

(%)

Zn

(%)

LD 65 78 73 24 37 54 30 47 66

ALD 90 98 95 41 41 60 33 39 73

LF 35 53 42 a 2 26 45 20 37 47

LD—lime-dosed system; ALD—anoxic limestone drain system;

and LF—lime-free system.

5. Model results and discussion

The model simulates the dynamic response of the

wetland system on a daily time step. In addition to the

mine drainage driving the behaviour of the system,

environmental factors such as rainfall, temperature

and evapotranspiration affect the performance of the

wetland system.

Fig. 4 shows the rainfall and temperature patterns

for the 6-month period from June 1996 with low

rainfall and high temperature in the summer and

increasing rainfall in the winter with corresponding

cooler temperatures. Clearly, these environmental

variables will affect a whole range of processes in

the wetland system, including plant growth, evapo-

ration and, hence, flow levels and chemical and

microbial reaction rates. All these effects have to be

included in the model, and simulation results are

presented in Fig. 5 for flow, anions and cations. The

low summer rainfall and high temperatures keep the

simulated flows low, as shown in Fig. 5, and flows

increase as autumn proceeds. The chemistry of the

wetlands also responds to this changing flow pattern

and simulated, and the observed behaviour of Ca, Mg,

Na, K, Cl and SO4 all reduce in the autumn as the

flows increase. However, the simulated response

matches the observed behaviour well especially, given

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135132

that no calibration has been undertaken at this stage in

the modelling programme.

Another key aspect of the model is the ability to

simulate the DO and BOD behaviour down the

system. Fig. 6 shows the simulated patterns of DO

and BOD and indicate that the model represents the

low influent DO levels rising through the aerobic

zone, then dropping to zero in the anaerobic zone,

before recovering again in the rock filters. BOD

shows a major increase at the anaerobic filter, derived

Fig. 9. Metal simulated concentration

from the organic material present, and then a decline

in the rock filters. These changing simulated DO and

BOD conditions compare well with the observed

trends shown in Fig. 2 and create the ideal conditions

for aerobic and anaerobic bacteria to thrive and cause

metal precipitation or loss.

An alternative method of presenting the simulation

results is to compute the frequency distribution and

cumulative frequency distribution of the model out-

puts. Fig. 7 shows these results in a graphical form.

s down the Wheal Jane system.

Fig. 10. Tracer simulation at Cell 5 in the Aerobic Zone.

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135 133

This is useful in that it is possible to see how the

distributions of the data vary down the cell system and

also determine statistics such as the 95 percentile

levels. These are useful, as Environment Agency

standards are often presented is percentile or statistical

terms.

The main objectives of the modelling study are to

simulate the effects of the wetland system on the

heavy metals response and to maximise the removal

of the toxic heavy metals in the wetland ecosystem.

Fig. 8 gives a typical response of the metal behaviour

at a point midway down the aerobic cells. The

simulated metal concentrations show a significant

reduction in the first few cells of the aerobic zone, as

is observed in the actual system (see Figs. 1 and 8 and

Table 1). This loss in the model is achieved by a first-

order reaction rate decay in the model equation for

each metal. However, from a geochemical perspec-

Fig. 11. Iron concentrations shown as a profile down the wetland system u

days�1 and a low flow rate of 0.5 l�1 (x), (b) simulation using a low iro

simulation using a high loss rate of 20 day�1 (from Hall et al., this volum

tive, the process of oversaturation and then precip-

itation of metals could, in theory, be modelled more

accurately by a thermodynamic equilibrium approach.

However, these equilibrium approaches assume stable

conditions, and this is certainly not the case in the

wetland ecosystem. However, some combination of

equilibrium modelling and rate-dependant process

model may be required at some stage to improve the

performance of the model (Jaffe et al., 2002).

Certain metals in the wetland system, such as

cadmium and zinc, are not reduced significantly in the

aerobic cells but are removed in the anaerobic cell as

metal sulphides are formed in the extreme anoxic

environment. The reduction in the anaerobic cell is

also modelled by a temperature-dependant first-order

process, but bacterial populations also play a large

role in enhancing the rate of reaction. At Wheal Jane,

there is a major source of these reducing bacteria from

the mine itself, and these are transported into the

wetland along with the mine effluent waters (Fig. 9).

A third process operating in the system, which

affects metal behaviour, occurs in the rock filter

section. Here, manganese is removed by precipitation

as the pH rises in the rock filters. The pH increase is

driven by algal growth over the summer, which

respires and removes the carbon dioxide in the water

and this results in a pH increase. The alkaline

conditions cause the precipitation of manganese.

Finally, it is possible with the model to simulate the

movement of tracers through wetland systems, and

Fig. 10 illustrates a typical tracer simulation for the

passive treatment system.

nder three conditions: (a) simulation using a low iron loss rate of 4

n loss rate of 4 days�1 and a high flow rate of 5 l�1 (E), and (c)

e) and a high flow rate of 5 l�1 (n).

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135134

6. Modelling changed operating conditions

The advantage of a model such as the Wheal Jane

model is that the system can be simulated under a

range of differing operating conditions. For example,

Hall et al. (2004) suggest from their iron precip-

itation studies in the aerobic cells that the capacity of

the system is significantly underused. They predict

that the flows could be increased by a factor of 10 in

the aerobic zone; in other words, the wetland system

could treat 10 times the effluent currently treated. We

can test this out using the model, and the results are

presented in Fig. 11. Here, three simulations to

predict the effects of changing flow under differing

iron loss rates, using the model, are shown. The

baseline iron profile (the lower line on Fig. 11)

shows the iron concentrations under low flows and

low iron loss rate, indicating a rapid decline in

concentration, as might be expected given the very

low flow rates. The second simulation (shown as the

upper line) shows the effects of increasing the flows

by a factor of 10. As can be seen, the iron

concentrations fall, but not so quickly, and the low

iron loss rate is not low enough to remove all of the

iron. However, in the intermediate line of iron loss,

the flows are again 10 times higher, but the iron loss

rates are now high and equivalent to those estimated

by Hall et al. (this volume). Here, the iron reduces

quickly, and the total mass of iron removal is 10

times the current loss from the wetland system.

Thus, it does suggest that the passive treatment

system could be made 10 times more efficient by

increasing the flow rates by a factor of 10. It also

emphasises the importance of the microbiological

behaviour in the controlling iron precipitation and

removal down the aerobic zone.

7. Conclusion

The model development in the Wheal Jane project

has produced a prototype model for the simulation of

a wide range of variables observed in wetland

ecosystems used for bioremediation. The combination

of the field data, experimental data and modelling

provides a powerful method of evaluating the com-

plex processes operating in such systems. In such

systems, the physical, chemical, macrobiological and

microbiological processes interact to create a highly

complex set of behaviours. Creating a new model for

such a system represents a major challenge, and the

use of first-order reaction kinetics is shown to be a

good method of modelling the complex microbial

processes controlling metal precipitation and removal

along the passive treatment plant. Moreover, the

model can be used to evaluate the operational

conditions in the wetlands system and, for example,

be used to assess alternative methods of improving

the efficiency of the plant. The dynamic model is

written as a generic model and, as such, can be set up

for a range of different wetland systems. The model is

available from Prof. Paul Whitehead at the University

of Reading.

Acknowledgements

This work is produced as part of the DTI LINK

Project BTL/20/71 RC140.

References

Cosby BJ, Wright RF, Hornberger GM, Galloway JN. Modelling

the effects of acid deposition: assessment of lumped parameter

model of soil and water and stream chemistry. Water Resour Res

1985a;2(1);54–63.

Cosby JB, Wright RF, Hornberger GM, Galloway JN. Modelling

the effects of acid deposition: estimation of long term water

quality responses in a small forested catchment. Water Resour

Res 1985b;21(11);1591–601.

Fennessy MS, Mitsch WJ. Design and use of wetlands for

renovation of drainage from coal mines. In: Mitsch WJ,

Jørgensen SE, editors. Ecological engineering: an introduction

to ecotechnology. New York7 John Wiley and Sons; 1989a.

Fennessy MS, Mitsch WJ. Treating coal mine drainage with and

artificial wetland. Res J Water Pollut Control Fed 1989b;61:

1691–701.

Flanangan NE, Mitsch WJ, Beach K. Predicting metal retention in a

constructed mine drainage wetland. Ecol Eng 1994:135–59.

Hall GH, Puhlmann T. Spatial distribution of iron oxidation in the

aerobic cells of the Wheal Jane pilot passive treatment plant. Sci

Total Environ 2004;338:73–80 [this volume].

Hall G, Swash P, Kitilainen S. The importance of biological

oxidation of iron in the aerobic cells of the Wheal Jane pilot

passive treatment system. Sci Total Environ 2004;338:67–72

[this volume].

Hamilton QUI, Lamb HM, Hallett C, Proctor JA. Passive treatment

systems for the remediation of acid mine drainage at Wheal Jane,

Cornwall. J Chart Inst Water Environ Manag 1999;13:93–103.

P.G. Whitehead et al. / Science of the Total Environment 338 (2005) 125–135 135

Jaffe PR, Sookyun W, Kallin PL, Smith SL. The dynamics of

arsenic in saturated porous media: fate and transport modelling

for deep aquatic sediments, wetland sediments and groundwater

environments. Spec Publ-Geochem Soc 2002;7:379–97.

Johnson DB, Hallberg J. Biogeochemistry of the compost bioreactor

components of a composite acid mine drainage passive

remediation system. Sci Total Environ 2004;338:81–93 [this

volume].

Meteorological Office J. The MORECS system. Hydrological

Memorandum 1981;45:91.

Mitsch WJ, Wise KM. Water quality, fate of metals and predictive

model validation of a constructed wetland treating acid mine

drainage. Water Res 1998;32(6);1888–900.

Mitsch WJ, Bosserman RW, Hill Jr PL, Smith F. Models of

wetlands amid surface coal mining regions of Western

Kentucky. In: Mitsch WJ, Bosserman RW, Klopatek J,

editors. Energy and ecological modelling. Amst.7 Elsevier;

1981. p. 103–13.

Mitsch WJ, Taylor JR, Benson KB. Classification, modelling and

management of wetlands—a case study in Western Kentucky.

In: Skogerboe GV, Flug M, Lauebroth WK, editors. Analysis of

ecological systems, state of the art in ecological modelling.

Amst.7 Elsevier; 1983.

Neal C, Whitehead PG, Jeffrey H, Neal M. The water quality of the

River Carnon, West Cornwall, November 1992 to March 1994:

the impacts of Wheal Jane discharges. Sci Total Environ

2004;338:23–39 [this volume].

Pourbaix M. Atlas of electrochemical equilibira in aqueous

solutions. NACE, Cebelcor. UK7 Pergamon Press; 1974.

Whitehead PG, Jeffrey H. Heavy metals from acid mine drain-

age–impacts and modelling strategies. IAHS Publ 1995;230:

55–68.

Whitehead PG, Hall G, Neal C, Prior H. Chemical behaviour of the

Wheal Jane bioremediation system. Sci Total Environ

2004;338:41–51 [this volume].

Whitehead PG, Prior H. Bioremediation of acid mine drainage: an

introductory overview of the Wheal Jane wetlands project. Sci

Total Environ 2004;338:15–21 [this volume].

Whitehead PG, Young PC, Hornberger GE. A systems model of

flow and water quality in the Bedford Ouse River system: Part I

Streamflow modelling. Water Res 1979;13:15.

Whitehead PG, Wilson EJ, Butterfield D. A semi-distributed

integrated nitrogen model for multiple source assessment in

catchments (INCA): Part I Model structure and process

equations. Sci Total Environ 1998a;210/211:547–58.

Whitehead PG, Wilson EJ, Butterfield D, Seed K. A semi-

distributed integrated nitrogen model for multiple source

assessment in catchments (INCA): Part II Application to large

river basins in South Wales and Eastern England. Sci Total

Environ 1998b;210/211:559–83.