the Medway Tunnel at Chatham in Kent....The plan area covered by the SUTRA model at Medway is shown in figure 3. The area is 4000m by 3000m. The modelled area was selected to cover

Invited Paper

The modelling of saline intrusion during the

construction of submerged tunnels

T. Roberts', J. White*, Z. Mohammed'

"WJ Engineering Resources Ltd.

* Civil Engineering, Queen Mary and Westfield College, London

University, London, UK

^ Department of Hydraulics and Irrigation, Alexandria University,

Abstract

This paper describes the procedure that may be used to apply the USGS twodimensional numerical model SUTRA to predict groundwater flow and salinitydistribution for the boundary conditions that are likely to be encountered duringthe construction of a submerged tube tunnel. The paper discusses the method ofcalibration that may be adopted to assess relevant parameters and boundaryconditions. These include coefficients of permeability, dispersion and leakage,together with the distribution of undisturbed groundwater levels and salinity atthe boundary of the model. Case study examples are included.

Introduction

Submerged tube tunnels are suited to the crossing of rivers, estuaries or shallowwater channels where ground conditions or environmental considerations areunsuitable for conventional tunnel or bridge solutions. A number of submergedtunnels have now been constructed around the world and the technique isrelatively well understood. Culverwell/ records details of 67 immersed- tuberoad and rail tunnels world-wide. These have been constructed between 1910 and1986 with 16 constructed during the latter ten years of that period. Two morerecent submerged tunnels in the UK are the Conwy Crossing in North Wales andthe Medway Tunnel at Chatham in Kent.

A feature of the construction procedure for immersed tube tunnels is thatit often relies heavily on groundwater control. The tunnel consists ofprefabricated sections which are floated out and then sunk into position in apreformed trench across the bed of the channel. The prefabrication of the tunnelsections may be carried out in a casting basin on site. This consists of a deepexcavation which may penetrate an aquifer and, because of its proximity to openwater, be constructed below the water table. The casting basin is subsequently

Transactions on Ecology and the Environment vol 7, © 1995 WIT Press, www.witpress.com, ISSN 1743-3541

32 Water Pollution

flooded in order to float the tunnel sections out into position. Temporarydewatering of the casting basin will therefore be required. Furthermore therewill be a need to dewater and excavate for the two cut and cover approach roadsat either end of the tunnel. See figure 1.

The effect of such extensive groundwater control procedures is to drawdown the water table locally, and possibly to such an extent that there will be asignificant effect on the regional or 'far field' groundwater regime. Thusthroughout the construction period the groundwater system acts like a large welldrawing towards it groundwater over large distances. Since the site is next to alarge water body which is likely to be saline, the region of groundwater flow thatis influenced by the dewatering procedure is almost bound to contain asignificant source of saline water. This will be induced into fresh water zones ofthe aquifer and may be so extensive that it will threaten any regional sources ofgroundwater supply that may be in the area. See figure 2.

Besides the influence of the dewatering system, there are two otherprocedures which may adversely effect the quality of the groundwater. Prior tofloating out the prefabricated tunnel units a trench is excavated across thechannel in order to receive the units. During this time the dewatering system willhave reduced pore pressures in the aquifer below the trench and increased pore-pressure gradients between the channel and the aquifer. The excavation for thetrench may significantly reduce the leakage path from the channel to the aquiferthus increasing the leakage of saline water into the aquifer system.

A further adverse effect may arise as the result of flooding the castingbasin to float out the tunnel units. When this happens a relatively large surfacearea of the aquifer becomes inundated with saline water for the first time. Thisprovides the potential for a large source of recharge saline water which againmay adversely effect the quality of the regional groundwater resource.

This paper describes the procedure that may be used to model thegroundwater flow and salinity distributions for the boundary conditions that arelikely to be encountered during the construction of a submerged tube tunnel.The procedure is illustrated with examples from the Medway Tunnel Project.

Model Structure

The computer modelling work for the Medway Tunnel Project was carried outusing the finite element computer program SUTRA produced by the USGeological Survey. SUTRA simulates groundwater flow and the transport ofdissolved substances in two dimensions. The model employs a hybrid finiteelement and integrated finite difference method to approximate the governingequations that describe the interdependent processes, Voss/ The input datarequired to define the boundary conditions for the model are as follows,

drawdown potential at boundarychloride concentration at boundaryaquifer transmissivity distribution


Water Pollution 33

channel leakage coefficientschannel chloride concentrationdispersion coefficientextraction flows from supply wells in the modelled areaextraction flows from the dewatering wells.

Whilst some of this information is available prior to construction frompre-contract pumping tests, Water Authority records and existinghydrogeological mapping, the most detailed information arises shortly after thecommencement of dewatering operations when the most significant changes inthe groundwater regime and salinity distributions will be recorded.

The plan area covered by the SUTRA model at Medway is shown infigure 3. The area is 4000m by 3000m. The modelled area was selected tocover the area of influence of the tunnel dewatering works based on data fromvarious remote monitoring points. The computer mesh used by the SUTRAprogramme is shown in figure 4. The density of the mesh has been increased inthe area of the tunnel works to improve the accuracy of the simulation of thedewatering wells, trench works and casting basin flooding.

Boundary Conditions

a) Drawdown PotentialsThe boundary potentials used for the model runs are shown in figure 4. Theestuary and river were taken as OmOD with the potential rising to +2.5mODsouth of the river and to +lmOD to the west. This distribution is consistent withthe hydrogeological maps of the area and also with limited data available beforecommencement of pumping. The model takes no account of tidal fluctuations.

b) TransmissivityAt the Medway Tunnel site the aquifer is confined consisting of approximately300m of soft, fine grained, fissured chalk overlain with a narrow band of gravelwhich in turn is overlain by clay. It is thought that the groundwater flow takesplace mainly in an upper fissured region of the chalk no more than 50m in depth.The transmissivity distribution used for the model runs is shown in figure 5. Thisis based on three factors.

(i) An analysis of early drawdown data from the dewatering system usinga simple Theis analysis showed an average permeability of 1.8 x 10 m /sec.

(ii) The hydrogeological map of the area indicates that the chalk becomesconfined just to the south of the tunnel site and that a spring line is present to thewest. Historic leakage/springs at the confining boundary may well have lead to anarrow zone of high transmissivity.

(iii) The transmissivity distribution was modified following repeatedmodel runs and comparison with the steady-state drawdown pattern obtainedafter the dewatering system had been operating from about 9 months. This


34 Water Pollution

particular drawdown pattern, together with the associated chloride data wasused throughout to calibrate the model.

c) River/Estuary LeakageInitially it was thought that leakage from the river and estuary would not be ofgreat significance because of a sealing layer of alluvial silt. However, it wasfound that incorporating a leakage algorithm into the SUTRA model gave animproved fit to the calibration data. Such an algorithm was also necessary toprovide the facility to simulate leakage from the cross-channel trenching and thecasting basin during flooding.

The leakage algorithm simulates inflows or outflows at relevant nodes onthe finite-element mesh. The flows are proportional to the difference betweenthe computed potential and a reference potential. The constant ofproportionality is termed the leakage factor and is the ratio of the permeability tothe thickness of the river bed silt. This suggests a range for leakage factor valuesof 10 sec to 10 sec . Higher values might occur as the result of regulardredging for shipping whereas lower values will arise when the seal between theriver and the aquifer is improved by the presence of the overlying clay.

The reference potential applied to the river/estuary nodes was OmOD. Inthe undisturbed state, where the potential in the chalk aquifer is above thisleakage will occur from the aquifer into the river which acts as a sink. However,where the potential in the chalk aquifer is reduced due to the dewatering worksleakage will occur out of the river/estuary which then becomes a source whichmay contain a high chloride content. The distribution of the river/estuary leakagefactor used for the Medway model runs is shown in figure 6. Initially a uniformleakage factor was assumed but this was modified following repeated model runsand comparison with the calibration data.

d) Chloride concentrationThe chloride distribution for the Medway model boundary and for theriver/estuary leakage is shown in figure 7. Leakage from the estuary was takenas 14,000 mg/1 chloride, and leakage from the river as 10,000 mg/1 falling to 3,000mg/1 from north to south. Fresh water entering across the south/east boundaryand the west boundary was taken as 60 mg/1.

The effect of molecular diffusion and dispersion is small in comparisonwith advection. The Medway model does not take account of diffusion but adispersion coefficient of 70m has been used.

e) Extraction FlowsAt Medway there were a total of 47 wells providing a total extraction flow of400 litres/second. These were grouped into 26 nodes each with an extractionflow in the range 10 to 25 litres/second. Where necessary the nodes were movedfrom the rectangular grid array to give a better representation of the welldistribution.


Water Pollution 35

f) Trench and casting basin leakageModelling of the trench excavation is achieved by increasing the leakage factor atthe trench excavation nodes across the channel. The trench leakage factor usedat Medway was 2.2x10* sec" which gave a trench leakage flow of 100litres/second. This was considered to be a likely upper value for the leakage inpractice. Modelling of the flooding of the casting basin is achieved byintroducing leakage at appropriate nodes. For the Medway model a leakagefactor of 7.5x10 sec was used. For both the trench excavation and castingbasin the leakage water chloride concentration was taken as 9,000 mg/1.

Some results and observations

The steady state solution computed by SUTRA giving both drawdown andchloride distribution for the situation shortly after the commencement ofdewatering is shown in figures 8 and 9. The comparison between observed dataand spot heights computed by SUTRA is shown in figures 10 and 11, where itcan be seen that it is possible to obtain a good match. The possibility of obtainingan even better match by taking into account vertical anisotropy and densityeffects using a three dimensional model SWICHA has been investigated and theimprovement is also shown on figures 10 and 11. In the case of the threedimensional model the match with observed data was obtained with very muchreduced leakage coefficients. This indicates that the introduction of the leakagecomponent in the two dimensional model has to some extent compensated forthe two dimensional approximation which does not model vertical anisotropy ordensity effects. This matter is now the subject of further study by the authors.

The increased accuracy of a three dimensional model, whilst desirable, maynot be justified in view of the attendant increased time and cost. Despite theapproximations of the two dimensional model its application to the MedwayTunnel Project has made a useful contribution to the tunnel dewateringstrategy.

References1. Culverwell. Immersed tunnel techniques. Proc. of conf. Institution of

Civil Engineers Manchester 11-13 April 1989.2. Voss, C.I. A finite-element simulation model for saturated-unsaturated,

fluid-density-dependent groundwaterflow with energy transport orchemically-reactive single-species solute transport. US GeologicalSurvey Water-Resources Investigations Report 84-4369, 1984.

AcknowledgmentThe authors are grateful for the cooperation and assistance of Kent CountyCouncil (Client for the Medway Tunnel Project), Travers Morgan (theClient's Engineer), Tarmac-HBM Joint Venture (Design and BuildContractor) and Mott MacDonald (Medway Tunnel designers).


Casting basin

.--V

A. Construction of tunnel units in casting basin.

B. Float out: Casting basin flooded. Trench excavated.

C. Tunnel units installed with cut and cover approaches.

Fig. 1 Different stages for installation.

Fig. 3 Area Covered by the Model.

Deep well ground water control

Fig. 2 Flow lines illustrating seepage induced fromfar field and local saline sources.

'idr-a 2000 2400 2*00 3200 3600

1200 .1600 2000

*_aw

wP-»

Fig. 4 Boundary conditions - Potentials.


Water Pollution 37

1200 1600 2000 2400 2800 3200 3600

Fig. 5 Boundary conditions - transmissivity

2000 2400 2*» 3200 3600

Fig. 6 Boundary conditions - leakage coefficients

1800-

1400--

1000

600 "i

(

k 4-i.il• 7

S /'\ '

• • • J-^" I400 800

i |

1200 1600 2000 2400 2.we

L

•fsWWwtf̂

'•-̂ ^ C*14OOOs/\/1 r= ;r>oofl ;gggggj C= 9OOO P

.'••'• 1 c*sooo P__: 1 c = 3ooo p

3200 361

P/Af ^VA//Af

/A//A/ J

10 *

1000

200

mo

Fig. 7 Boundary conditions - chlorides levels.


38 Water Pollution

4000

1500 —

1000 —

3000

— 2500

- 2000

- 1500

- 1000

0 500 1000 1500 2000 2500 3000 3500 4000

Fig. 8 Drawdown Distribution.

3000

2500 -

2000 -

500 1000 1500 2000 2500 3000 3500 4000

1000 —

1500

500 1000 1500 2000 2500 3000 3500

Fig. 9 Chloride Distribution.


1

-2-

-10

-14

^

-14

3-D2-D

X

-10

Water Pollution 39

-6 -2Observed Head(m)

Fig. 10 Potential Head Results for 3-D and 2-D Model.

ex,

§

8

1c,O1

Ia

8000

4000 *

*

3-D2-D

*

**

4000 8000 12000Observed Chloride Concentration ppm

16000

Fig. 11 Chloride Concentration Results for 3-D and 2-D Model.


Documents

the Medway Tunnel at Chatham in Kent....The plan area covered by the SUTRA model at Medway is shown in figure 3. The area is 4000m by 3000m. The modelled area was selected to cover