- Email: [email protected]
The influence of stream salinity on reservoir water quality
Agricultural Water Management, 4 (1981) 255--273
255
Elsevier Scientific Publishing C o m p a n y , A m s t e r d a m -- Printed in The Netherlands
THE INFLUENCE OF STREAM SALINITY ON RESERVOIR WATER QUALITY IMBERGER Department of C i v i l Engineering, U n i v e r s i t y of Western A u s t r a l i a , Nedlands, 6009, W.A. J.
ABSTRACT Imberger, J . , 1981. quality.
The influence of stream s a l i n i t y on reservoir water
Agric. Water Manage., 1981.
The dynamic reservoir simulation model DYRESMis used to investigate the response of the Wellington Reservoir to changes in streamflow s a l i n i t y and outflow strategies.
Construction of a s a l i n i t y diversion dam could be useful in sub-
s t a n t i a l l y reducing reservoir s a l i n i t i e s , but with the penalty of reduced water yield. 1
INTRODUCTION The surface water resources of the populated south-western corner of Western
A u s t r a l i a are drawn from a series of western flowing streams which rise in the plateau regions of the Darling Range, flow through increasingly defined v a l l e y systems, p a r t i c u l a r l y near the western edge of the plateau (Darling Scarp), and out across the coastal p l a i n to the Indian Ocean. Wellington Reservoir is t y p i cal of storages in the region, being located, as shown in Fig. I , in the incised v a l l e y of the C o l l i e River some i0 km from the scarp.
However, i t s catchment
d i f f e r s from those of other reservoirs in the region, in that sections have been progressively cleared of native f o r e s t and converted to annual pastures f o r agricu l t u r a l development.
This land use change has resulted in substantial increases
in the s a l i n i t y of i n f l o w to the reservoir.
Loh and Hewer (1977) indicate t h a t ,
p r i o r to 1950, with only 5% of the catchment cleared, average annual i n f l o w sali n i t i e s were below 300 mg L - I .
Current estimates of the average annual i n f l o w
s a l i n i t y , with some 23% of the catchment cleared, are 750 mg L - I T.D.S., with a yearly increase between 1971 and 1975 of around 50 mg L- I T.D.S. I t is generally accepted that the increased s a l i n i t i e s are the r e s u l t of the reduction of t r a n s p i r a t i o n a f t e r f o r e s t clearing and the consequent increase in recharge to the groundwater system.
The increased groundwater flows upset the
e x i s t i n g s a l t balance by f l u s h i n g large q u a n t i t i e s of soluble s a l t s , previously stored in the soil p r o f i l e , to the stream system (Wood, 1924; Peck and Hurle, 1973).
As groundwater systems respond over periods of tens of years (Peck et a l . ,
1977) the f u l l
e f f e c t of recent clearing has yet to be f e l t .
F i r s t order estimates
0378-3774/81/0000--'0000/$02.50 © 1981 Elsevier Scien~fic Publishing Company
256
made by Loh and Hewer (1977) indicate that i f recent l e g i s l a t i o n l i m i t s catchment clearing to i t s current l e v e l , average annual i n f l o w s a l i n i t i e s of Ii00 mg L - I T.D.S. w i l l u l t i m a t e l y develop, with dry year s a l i n i t i e s as high as 1450 mg L- I T.D.S. and peak s a l i n i t i e s as high as i0 000 mg L - I .
S12004 ~.612035j--~ • mEGION eS
~l
..........7,S;I SCALE 1:500 000
Fig. l ( a ) - Wellington Reservoir. Map showing the reservoir and the major catchment regions. Numbers r e f e r to gauging stations located on the streams. Clearly, any improvement in the supply s a l i n i t y by means of improved reservoir management w i l l be increasingly b e n e f i c i a l as such high s a l i n i t i e s develop, e s p e c i a l l y since long term solutions to the problem must involve improve catchment management and r e h a b i l i t a t i o n which take a long time to establish. In the assessment of reservoir management strategies (Fischer et a l . , 1979) two questions arise.
F i r s t , what is the e f f e c t of the streamflow s a l i n i t y and
the phase between the s a l i n i t y and streamflow in determining the overall s a l t content of the reservoir.
Second, what is the i n t e r p l a y between the streamflow
s a l i n i t y v a r i a t i o n s and the outflow volume and v a r i a b i l i t y s a l t content of the reservoir.
in determining the
The dynamic reservoir simulation model DYRESM
was used to simulate the response of the Wellington Reservoir to changing inflow s a l i n i t i e s and outflow strategies in order to assess the answers to these two questions.
257 .q ?
11 (/
//I
i / ii I f \ ,\_
/
/ tJ /! /
HAMFLTOf
VASE %~\
t i/\ .~J \\
S
f
I,,
J
le
/ (/\I}
I p
j
/
USS~L
//
t~ 5
( \
\h
,/
%..
/
{ SCALE
Fig. l(b) - Wellington Reservoir. Enlargement showing the location of the dam wall and the Mungalup gauging station.
~
'....
~ALE
1k
Fig. 1(c) - Wellington Reservoir.
Site for the proposed salinity diversion dam.
258
2
STREAJMFLOW SALINITY VARIATIONS Much has been written about the s a l i n i t y problem, i t s causes and the mechanisms
responsible for the s a l i n i t y increases observed over the last few decades in the south-west region of Western Australia. The purpose of this paper is to synthesize the effects of these increases on the salt content of the Wellington Reservoir, to establish the s e n s i t i v i t y of the result to variations in streamflow s a l i n i t y and to explore techniques for the alleviation of the salt load. The most developed model of the groundwater system active in the south-west region appears to be that proposed by Smith and Hebbert (1980), who modelled a two aquifer system as shown in Fig. 2.
The water entering the higher areas is
conducted to the deep groundwater via the preferred paths established by new and fossil root structures.
Weak pressures are set up causing a constant upflow in
the lower areas surrounding contributing creeks and streams. The upward flows negotiate a salt laden soil profile (see Johnston et a l . , 1980) introducing saline water to the streams in the winter.
During periods of no flow this slow per-
colation continues leading to a gradual buildup, when combined with evaporation, of salt on and near the surface of the lower areas.
This salt is stored ready
for flushing into the streams during the f i r s t storms of the season.
RAr~
Fig. 2. (From Smith and Hebbert, 1980) - Schematic of the double aquifer concept. Water enters the stream d i r e c t l y via surface and subsurface flows in the perched aquifer and from below under weak artesian pressures. During periods of rain the perched aquifer (Fig. 2) builds up, causing leaching of the salt accumulated by direct water ponding during the end of the previous winter and that accumulated by the upward percolation described above.
In areas
where the perched aquifer is saturated, direct runoff w i l l occur leading to a very rapid response of the streams. The thickness of the perched aquifer obviously w i l l vary greatly so that the contributing area w i l l be a function of space as well as time.
The response of the perched aquifer i t s e l f w i l l be rapid once
saturated (a few days).
The salt stored on the surface and in the perched aquifer
259
w i l l thus be flushed out once the water has travelled a distance characterized by the width of the contributing area.
Effective flushing requires a certain
volume of continuous rain in order that the movement of water be continuous and not be interspersed by periods of strong evaporation. Deep-rooted vegetation has two influences.
First, i t prevents the buildup of
the upward flows into the contributing areas and secondly the vegetation drains much of the perched aquifer during the winter months. This conceptually plausible model explained most of the features of the observed s a l i n i t y and flow d i s t r i b u t i o n , an example of which is shown in Fig. 3.
The
early rains in May and half of June would have progressively leached more and more of the upper salt laden areas.
The increasing contribution from the rising per-
ched aquifer leads to a general rise in s a l i n i t y climaxing with the very large spike around the inflow at the end of June. From that time onward the s a l i n i t y declined, indicating a general trend towards a flushed state. naturally
Each new flood
eads to a local spike as increases in flow encompass a bigger sphere
of influence in the perched aquifer due to the growth in the contributing area.
24
i i
Z2
i! !i
.................. .........
a.o
1.8
1.e
1.4
~
V~
/
o8
o.6
o4
o.2
°
MAY
JUNE
JULY
AUGUST
o.o
IS75
Fig. 3. Distribution of the s a l i n i t y and flow in the Collie River at the Mungalup gauging station. The distribution t y p i f i e s the rise in s a l i n i t y during the early flows and the lowering of s a l i n i t y as the upper surface areas are flushed.
260
Further, the model suggested by Smith and Hebbert (1980) is able to explain the very d i f f e r e n t s a l i n i t y response recorded during the July flow and the high flow at the middle of August.
A s t r a i g h t c o r r e l a t i o n of s a l i n i t y and flow as sugges-
ted by Loh and Porter (1976) is obviously not warranted even i f a correction f o r the time of year is included. double a q u i f e r model.
A more plausible answer may be derived from the
The flow in July was preceded by perched aquifer flushing
whereas the August flow was preceded by a period of low r a i n f a l l which had allowed a new s a l t load to accumulate in the c o n t r i b u t i n g areas. Smith and Hebbert (1980) have proposed this model, but i t w i l l take a few years to v a l i d a t e such a d e t a i l e d model f o r the C o l l i e Catchment and so no attempt was made to couple t h i s model to DYRESM. Instead certain features of the model were used to change the recorded s a l i n i t i e s to simulate r e f o r e s t a t i o n and increased clearing in the f o l l o w i n g way. The essential features are that the peak s a l i n i t i e s in the stream depend on the s a l t concentration in the contributing areas, the phasing of the flow in r e l a t i o n to the volume of the contributing areas and the percentage of s a l t rich area to fresh flowing catchment.
By r e t a i n i n g the phas-
ing as in 1975 to 1978, but increasing the percentage area c o n t r i b u t i o n , increased s a l i n i t y records may thus be generated.
Further, by assuming that the nature
of the clearing carried out in the l a s t 15 years is s i m i l a r to that done e a r l i e r i t is reasonable to assume the flushing volumes contributing to the s a l i n i t y have the same time responses.
These assumptions would lead to a much increased
peak s a l i n i t y , but with about the same flushing time scale in r e l a t i o n to flow v a r i a b i l i t y and the same "flushed" s a l i n i t y .
I f this low state is taken as 500
mg L- I , then a possible recipe f o r simulating an increased s a l i n i t y would be : Se S = 5OO " Se
(1)
where Se is the e x i s t i n g s a l i n i t y .
This leads to a s a l i n i t y record with an
average s a l i n i t y of about 1700 mg L- I and a new peak s a l i n i t y of i i 500 mg L- I corresponding to expectations, outlined in the introduction.
Further work is
n a t u r a l l y required to v e r i f y this simple hypothesis expressed by equation ( i ) , but f o r the purposes of the reservoir simulation i t serves as a good f i r s t model. Reduced s a l i n i t i e s may again be expressed using the e x i s t i n g record by: S = 0.5 (Se-500) + 350 S = 350
S > 500 e S < 500 e
(2)
This w i l l be used to simulate conditions before the recent clearing took e f f e c t in 1960.
261
3
MODEL DESCRIPTION The dynamic reservoir simulation model, DYRESM, is a one-dimensional numerical
simulation model for the prediction of temperature and s a l i n i t y in small to medium lakes.
I t is based on the assumption that in such lakes the thermal structure
plays a dominant role and the isotherms are mostly horizontal planes.
Deviations
from this state of "rest" are allowed, but i t is assumed that such deviations are sporadic or weak and may be treated as perturbations to the evolution of the onedimensional structure. These assumptions place certain restrictions on the a p p l i c a b i l i t y of the model. In particular the Wedderburn Number (see Thompson and Imberger, 1980) should be greater than one for the majority of the wind events, the inflow and outflow internal Froude numbers (see Imberger et a l . , 1976) should be less than one and the lake should be small or narrow enough for the influence of the earth's rotation to be negligible (see Fischer et a l . , 1979). DYRESMwas developed over the last 6 years in order to predict the s a l i n i t y variations in the Wellington Reservoir, but in the meantime the model has found wide a p p l i c a b i l i t y as a base for more general water quality modelling.
The devel-
opment of the model is continuing and currently version 5 is the operational model. A f u l l and detailed description of this version may be found in Imberger and Patterson (1980) and only a very brief summary w i l l be given here. The lake is divided into uniform horizontal slabs which form the computational building blocks of the model. The construction of the slabs is Lagrangian and they advect with the vertical velocity induced by the inflow and outflow.
At
each time step the model ensures that heat, salt, mass and energy are conserved for each slab and thus the reservoir as a whole.
The vertical momentum equation
reduces to the hydrostatic pressure assumption, since the assumption of onedimensionality eliminates all motion except the slow vertical adjust required to accommodate in and outflows. Once this slab structure has been established the meteorological inputs at the surface of the lake are calculated using bulk aerodynamic formulae.
As shown in
Fig. 4, these together with the daily inflow and outflow form the daily input data The basic time step of the model has been set at one day since only daily data is most commonly available.
Certain assumptions are made regarding the distribu-
tion of these inputs over the 24-hour period and time steps down to ~ hour are used where the processes to be simulated vary rapidly during the day i t s e l f . I f more frequent data is available then this should be used. The daily surface heat inputs are used by the model to calculate the temperature and s a l i n i t y changes in the slab structure.
The updated slab structure is then
262
adjusted for mixed layer deepening and possible changes of the thermocline thickness due to the formation of shear i n s t a b i l i t i e s .
The calculation of the mixed
layer deepening incorporates both deepening due to surface turbulence and turbulence generated by the shear i n s t a b i l i t i e s at the base of the mixed layer.
Data Input
]
Initialisation
I
1 day
1
¼ hour
Da,ydet. i° t
!
Surface heat exchanges
]
Mixed layer deepening
- 12 hours
Fig. 4.
Billowing
Turbulent diffusion
]
Inflow
]
Outflow
]
End
J
Schematic of the programmed model DYRESM.
Once the new thermocline depth and thickness has been computed the model then calculates the net heat and salt transport from the bottom through the hypolimnion into the epilimnion.
The basic algorithm by which this is carried out is
an eddy diffusion parameterization.
However, i t was recognized very early in
the development of DYRESM (see Imberger et a l . , 1978) that a constant diffusion coefficient was not applicable.
Field experiments showed that the vertical trans-
fer of heat and mass was higher during periods of high wind and inflow even though both of these processes did not d i r e c t l y contribute to the turbulence in the hypolimnion.
I t was postulated that both these disturbances induce basin
scale internal oscillations which are damped at the boundary. The dissipation at the boundary in turn produces small scale turbulence mixing.
capable of causing
The active mixing in the boundary layer quickly leads to an adjustment
of densities in the boundary layer and so induces horizontal intrusions which in turn cause an adjustment of the centre of gravity of the overall lake structure.
263
This concept has since then been postulated for the ocean (Armi, 1978) and recent experiments by Ivey (1980) have established rates of transfer in an idealized laboratory experiment.
While the concept appears attractive, not enough is under-
stood about the details of the individual processes to allow a process model to be constructed.
Instead, a simple energy argument is used (Imberger et a l . , 1978).
I t is assumed that a small, but constant fraction of the wind and inflow energies is Used to generate mixing in the hypolimnion.
This leads to a vertical diffus-
ion coefficient :
H2
(3)
~z = 0.048 ST--M '
where TM is a time scale for mixing(equal to the potential energy of the density s t r a t i f i c a t i o n of the lake divided by the power input of the wind and streams); and S is a normalized water column s t a b i l i t y (=H/&p dp/dz, where H is the depth of the lake; Ap is the density difference between the bottom and the surface; and p is the density of the water at an elevation z). This parameterization is successful (see Imberger and Patterson, 1980) provided the s t a b i l i t y S is quite large.
I t obviously breaks down in the l i m i t of a homo-
geneous water mass with an energy input.
In this l i m i t both S and TM approach
zero and equation (3) predicts an i n f i n i t e diffusion coefficient. In order to prevent this, an arbitrary cut off to equation ( 3 ) o f ~z = 10-4m2s-1 is provided in the model. Further work is required to rationalize this cut off especially since its value has quite a strong influence on the predicted s a l i n i t y distribution.
The mixing processes in the hypolimnion seem to require a two
parameter model, equation (3) for strongly s t r a t i f i e d water and an analogous formulation for a homogeneous hypolimnion where the mixing may be expected to be more evenly distributed throughout the water mass. At the end of the diffusion routine, which is carried on the same time step as the mixed layer dynamics calculation, a new structure for a particular day has been calculated.
This density structure is then used to route the inflowing
water from the various contributing streams into the reservoir. allows for turbulent entrainment and subsurface intrusions.
The subroutine
Similarly, the out-
flow is calculated by the model using the structure l e f t after the inflow has been added. This routine is repeated for each day of the simulation. 4
MODEL PERFORMANCE DYRESM has been tested on a number of different lakes, but i t s major evaluation
264
and development has taken place with data from the Wellington Reservoir. The seasonal v a r i a b i l i t y of the various inputs to the reservoir over the period from the Julian day 133 in 1975 to day 365 in 1977 are shown in Fig.5.
Depicted
are the wind speeds, the short-wave solar r a d i a t i o n as computed from cloud cover records, the s a l i n i t y and temperature of the inflowing water and the total rate of inflow from the C o l l i e River which contributes approximately 85% of the t o t a l inflow.
The remaining i n f l o w is included in the simulations, but is not shown in
Fig. 5.
6 WI
26
c~
3ODD
4
2DOO z 1000 NRCL "
V-
'
'
'
-
~24 22o
w12 ~
~ 8
z _IN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D d , f ,M,R,M,J,J,I~,5,O,N,D, J , F ' M , R , M , J ' J , A , g , O , N ' I ~,J ,F ,~I,A 'M ' J ' J 'A 'S ' 0 N ' { ] ,
1975
1976
197"7
Fig. 5. Seasonal v a r i a b i l i t y of wind speed, short-wave radiation, inflow s a l i n i t y of the Collie River, inflow temperature and the inflow volume. Figs. 6(a) and 6(b) show the f i e l d temperature and s a l i n i t y structures length averaged along the C o l l i e River Valley as a function of time over the period January 1975 to August 1978.
The s a l i n i t y data gathered between October 1977 and
June 1978 is regarded as u n r e l i a b l e and is not shown.
The broken l i n e s in Fig.
6(a) indicate that no data was taken in the period covered and the thermal structure is interpreted from the structure before and a f t e r the period of i n t e r r u p t i o n The y e a r l y cycle is c l e a r l y evident in Figs. 5 and 6:
the cold s a l t y inflows
lodge in the base of the homogeneous reservoir in the months of June, July and August; summer s t r a t i f i c a t i o n
builds up u n t i l December, when surface winds begin
to mix the surface layers and a thermocline forms, protecting the waters below. In early w i n t e r , the a i r temperature f a l l s and the winds increase, with the resu l t that the reservoir is completely mixed before the f o l l o w i n g inflows a r r i v e .
265
The marked difference in the thermocline structure between 1976 and 1977 was caused by a change in the withdrawal p o l i c y .
In 1976 a l l the water was with-
drawn from the offtakes at 15 m height, whereas in 1977 a large quantity of salty water was scoured through the offtake at the bottom of the dam w a ll.
In addition
about two thirds of the water f o r i r r i g a t i o n was taken from the bottom o f f t a k e .
WELLINGTON THERMALSTRUCTURE
Fig. 6(a). Measured average r e s e r v o i r temperature as a function of time f o r the years 1975 to 1978. (From Imberger and Patterson, 1980).
WELLINGTON NACL STRUCTURE
:is ,FIELD TRIP / ~: I!,
~t 1) h L,
Fig. 6(b). Measured average reservoir s a l i n i t y (mg L-1NaCl) as a function of time f o r the years 1975 to 1978. (From Imberger and Patterson, 1980). I t is also clear from Fig. 6(a) that the temperature regime of the reservoir is determined by the inflows and the surface heating and cooling.
The bottom
temperature of the r e s e r v o i r f o r most of the year is determined by the temperature of the coldest inflows, whereas the surface temperature is determined by the meteorological forcing. There are seven constants which must be specified by the user before applying DYRESM. Of these only one is t r u l y adjustable - the others are related to well i d e n t i f i e d physical processes and are determined from experimental or f i e l d data.
266
The constants are described below, together with experimentally determined values (i)
CD is the drag coefficient for inflowing streams.
CD was determined in-
dependently of DYRESMin a f i e l d study described by Hebbert et al. (1979).
The
value determined in that study, CD = 0.015, is used here. (ii)
nI is an extinction coefficient for short-wave solar radiation penetrat-
ing the water.
I t relates
the solar radiation received at the water surface, to
that penetrating to a depth z.
A single exponential decay formula was used as
only limited f i e l d measurements were available.
An average value of nI = 0.35
is taken, based on the fact that the Wellington is f a i r l y clear in the summer months when surface heating is an important effect. (iii)
~l is a constant occurring in the expression for the d i f f u s i v i t y calcula.
ted for the deep hypolimnetic mixing.
I t basically represents the efficiency
with which the power input from the surface wind and river inflows is converted to a gain in potential energy of the lake water due to vertical mixing.
A 9.6%
efficiency ~ l = 0.048) determined in earlier calibrations of DYRESMover the lO0-day period from day 133 to 233 has proven satisfactory throughout.
The cut-
off value of lO-4m2sec-1 was set by noting that this corresponds to the maximum value measured by a number of investigators (see Fischer et a l . , 1979). (iv)
CK is the coefficient that describes the s t i r r i n g efficiency of convec-
tive overturn.
Experimental results summarized by Fischer et al. (1979) suggest
an average value of CK = 0.125. (v)
n, in combination with CK as CKq3, is a coefficient measuring the s t i r -
ring efficiency of the wind.
The value given by Wu (1973) is adopted here, since
i t was shown by Spigel (1978) that in his experiments s t i r r i n g effects dominated the entrainment, so that shear production, temporal effects and internal wave radiation losses were negligible.
Wu's deepening law is dh/dt = 0.23 u,/Ri*
giving CKn3 = 0.23, and thus n = 1.23. (vi)
CS is the coefficient that describes the efficiency of shear production
for entrainment. (1978).
Values ranging from 0.2 to 1 are reported by Sherman et al.
CS = 0.2 was chosen as representing a good estimate for most experimen-
tal results.
A value of 0.5 was used before the energy released by billowing
was separately included as is done in version 5 of DYRESM. (vii)
CT is associated with temporal, or unsteady, non-equilibrium effects
due to changes in surface wind stress or surface cooling.
CT is constructed by
the requirement that for a turbulent front entraining into a homogeneous f l u i d , dh/dt = O.3u, (Zeman and Tennekes, 1977)giving a value of CT = CKn/0.3 = 0.510. These coefficient values were used for a 962-day simulation of the lake dynamics starting with i n i t i a l profile data on day 75133. The results from this
267
simulation are shown in Fig. 7.
~°
iI
~
~ii
-~•
< t~ '~9r'~
[ 9n6
'~977
1978
=i ~t ?
"
/
Io75
i
'.976
i
A $0 N O J 1977
~JJ
aS @N O °
1978
Fig. 7. Simulated average temperature and salinity for the period 1975 to 1978. (From Imberger and Patterson, 1980). Comparison of Figs. 6(a) and 7 shows that the thermal structure is reproduced extremely well with a l l the s t r a t i f y i n g and mixing regimes occurring at the correct times and of the correct magnitude.
The slug of 12°C water predicted
by the model in August to September 1975 does not appear in the f i e l d data.
This
slug is derived from the previous inflow and i t s presence may be due to errors in the inflow temperature data.
In any case, the predicted temperature gradient in
the bottom at this time is small, and the actual difference between f i e l d and predicted temperatures is less than I°C. The predicted and observed s a l i n i t y v a r i a t i o n s also compare well with the most s i g n i f i c a n t anomaly being the mismatch of the 600 mg L- I l i n e at the period of maximum inflow in 1976. The f i e l d data suggest somewhat more energetic surface mixing which may be due to small errors in the wind data.
A d d i t i o n a l l y , the
model does not reproduce the peak s a l i n i t i e s of the inflows.
This is due to the
construction of the model layers in terms of layer volumes - hence the depth resolution in the upper part.
The bottom three or four metres always appear as
mixed, and any s a l i n i t y slug occupying less than this w i l l be mixed with the layer above.
Thus there is no error in terms of the s a l t load, but one of dis-
play of the structure, and i t was accepted in order to save computer time.
If
greater resolution were required, this could be achieved by specifying smaller
268
minimum slab sizes. Overall, DYRESMappears f a i t h f u l l y to reproduce even very severe changes in the reservoir structure caused by such diverse forcing as large saline inflows active scouring, strong wind deepening and winter convective cooling.
Perhaps
more importantly, the model correctly simulates two independent parameters, sali n i t y and temperature to a resolution equal to that of the f i e l d programme. More accurate f i e l d data is required i f more stringent tests are to be applied. 5
INFLUENCE OF SALINITY CHANGESAND POSSIBLE ALLEVIATION STRATEGIES The simulation model DYRESMwas used to investigate the influence of increased
and decreased C o l l i e River s a l i n i t y and methods f o r a l l e v i a t i n g the strong b u i l d up of r e s e rv o i r s a l i n i t y f o r the case of increased C o l l i e River s a l i n i t y .
The
technique employed was to run the simulation f o r the three-year period using the measured i n f l o w, outflow and meteorological data.
This three-year period (1976
to 1978) represents a s t a r t of a severe drought period with the associated b u i l d up of stream s a l i n i t y and of course decreased flow.
As mentioned in Section 2
i t was f e l t that manipulating the magnitude of the s a l i n i t y of this time series would be more r a t i o n a l than generating a model set since this data set preserves the phase r e l a t i o n s h i p between the s a l i n i t y and flow v a r i a t i o n s .
The model then
simulates the subtle interactions between the s a l i n i t y structure, the outflow, mixing and inflow and allowed an assessment of the influence of any proposed changes.
Obviously, many v a r i a t i o n s could be simulated but because of the cost
involved only a set of extreme conditions w i l l be described.
I t is f e l t that
these results allow an assessment of the l i k e l y changes which may be expected. 5.1
Variations of Inflow Salinity
The simulation of the actual f i e l d data has already been discussed in Section 4.
Two further cases were run.
First, the Collie River s a l i n i t y was decreased
to that given by equation (2), the scour valve was turned o f f and flow with a s a l i n i t y greater than 1300 mg L-1NaCl was by-passed. The l a t t e r two strategies w i l l be explained later, but they were incorporated in the one run in order to save computer time.
The results are shown in Table 1 where i t is shown that
the average s a l i n i t y is lowered from 650 mg L- I to 480 mg L-1 and the dam overflows.
I t must be rememberedthat leaving the bottom scour valve closed (see
Fischer et a l . , 1979) has no effect in 1975/76 as i t was essentially closed under normal operation.
However, in 1976/77 much of the i r r i g a t i o n water (about two
thirds) was withdrawn through the scour valve and keeping the valve closed thus leaves a surplus of water in the reservoir.
However, the reduction of s a l i n i t y
is mainly due to the reduced inflow s a l i n i t y as essentially no water was by-passed
269
(salinity would have to be in excess of 2400 mg L° I ) and the scour valve salinity was always close to the average.
TABLE 1
Day o f Year Salinity of Inflow
Scour Valve Flow
Top Draw Flow
By-pass Flow
As measured
As operated
As operated
zero
. . . . y~-- F 11.4 308 12. I 450 8.1
Decreased
zero
As o p e r a t e d
Whenever S > 1300
II.2
308
12.0
14.6
413
Overflow
Increased
As operated
As o p e r a t e d
zero
11.2
308
12.0 1804 l 7.6
956
14.3
As m e a s u r e d
zero
As operated
zero
Ii.2
308
12.0 1464 114.7
584
Overflow
641
Increased
zero
As o p e r a t e d
zero
11.2
308
1320
Overflow
1648
As m e a s u r e d
As operated
As o p e r a t e d
11.2
308
12.4 !810 !14.7 i ! 12.0 450 6.3
As m e a s u r e d
zero
As o p e r a t e d
11.2
308
12.2
430
Increased
zero
As o p e r a t e d
11.2
308
9.7
371
.
S > 1300 Whenever S > 1300
~menever S > 1300
.
.
.
75133 ~Vol *Sal ~ .
Whenever
.
*
Volumes are in 107m 3
**
Salinities are in p.p.m. NaCL (average for whole vo I u m e ) .
.
.
.
.
F-.
.
.
.
.
76133 771 13 V o ~ - Sa--]--~ G Y - ~ a ~ -
354
.
.
.
.
.
77365 VJ~-S-dl
. . . . . . . 530 13.0 650 480 1345
438
3.5
480
12.9
444
Overflow
510
8.9
357
11.4
367
Second, the salinity of the Collie River inflow was increased as indicated by equation (1) and the reservoir was operated as was actually the case in the three-year period.
The increase in the average reservoir salinity was quite
dramatic reaching 1345 mg L- I at the end of the simulation period. levels would be quite unacceptable for all but a few selected crops.
Such salinity Further,
the strategy of taking irrigation water from the scour valve becomes no longer possible as the s a l i n i t i e s would rise to over 2000 mg L-1.
However, i t does
prevent an even greater increase of salinity buildup than would occur i f all the irrigation flow were to be taken from the top offtake. 5.2
Variations of the Scour Valve Flow.
The effect of the scour valve flow is quite subtle and can easily be misleading.
Table 1 shows the results from a simulation with inflow s a l i n i t i e s , for the
1975 to 1978 period, but with the bottom scour valve completely closed during the whole period.
As already mentioned this led to a strong reduction in the
water available for irrigation in 1977/78 with a consequential overflowing of the reservoir.
During the period 76133 to 77133 the scour policy had the ob-
vious effect of reducing the average salinity from 584 to 530 a small but definite benefit.
However, during the period from 77133 to 77365 scour policy actu-
ally had a detrimental influence since the water withdrawn in 1976 through the scour valve was of a lower salinity than the average salinity in this l a t t e r
270
period.
Had t h i s water been wasted then the strategy would have had an adverse
effect.
I t is therefore seen that a judgement must always be made whether the
present year's poor q u a l i t y water is l i k e l y to be of better q u a l i t y than the best q u a l i t y of the f o l l o w i n g year. The reverse e f f e c t is noticed in the increased s a l i n i t y case (see Table i ) . Closing the bottom scour valve once again led to a f u l l
r e s e r v o i r , but with an
average s a l i n i t y as high as 1648 mg L -1 at the end of the simulation period. Hence, l i t t l e
b e n e f i t would be gained by reducing the i r r i g a t i o n supply in an-
t i c i p a t i o n of a better f o l l o w i n g year.
Under such circumstances i t would be
better to i r r i g a t e generously in 1976 and bank one's p r o f i t s in order to overcome the hardship of the f o l l o w i n g very s a l t y inflows. Fig. 8 shows the isohalines for t h i s very extreme strategy.
The higher s a l i n -
i t i e s are reflected in the very much i n t e n s i f i e d s t r u c t u r e , the increased peak s a l i n i t i e s and the strong v e r t i c a l s a l i n i t y (and thus density) gradients.
These
gradients prevented mixing, leading one to suspect much increased e f f i c i e n c y of the scour p o l i c y (not shown in Table i ) .
4O0
197S
800
1976
1000
1977
WATERSURFACIE
1978
Fig. 8. Simulated s a l i n i t y v a r i a t i o n s in 1975 to 1978 f o r increased streamflow s a l i n i t y , no by-pass and no scour valve operation. 5.3
The Effect of By-passin 9 High S a l i n i t X Water.
The f u l l assessment of the benefits to be gained by by-passing the very high saline slug of C o l l i e River water would involve a great many simulations. In -1 t h i s paper a cut o f f s a l i n i t y of 1300 mg L has a r b i t r a r i l y been chosen to represent an acceptable upper l i m i t .
Introducing such a strategy of by-pass in
addition to the scour practice of 1975 to 1978 reduced the average s a l i n i t y from 650 mg L - I to 480 mg L - I , but the extra 1.8xlO7m 3 water scoured would have reduced the water volume to a dangerously low level of 3.5xlO7m 3 at the end of the simulation period and at the s t a r t of the 1978 i r r i g a t i o n season.
"~.
271
For this reason the by-pass policy was combined with a major reduction in the 1976 i r r i g a t i o n volume. This was achieved by closing the bottom scour valve cutting the i r r i g a t i o n flow in 1976/77 to about one third that actually delivered. The average s a l i n i t y increased marginally, but the reservoir f i l l e d by the end of 1977. Obviously, a compromise between these two l a t t e r approaches would be most advantageous; a certain by-pass flow should be compensated for by a corresponding reduction in the i r r i g a t i o n volume. The by-pass strategy was also evaluated for the case of increased s a l i n i t i e s (see Table i ) .
In anticipation of a low volume i t was decided again to close
the scour valve.
This resulted in a dramatic reduction in s a l i n i t y from 1345
mg L-1 (1648 mg L- I corresponding no scour case) to a very modest value of 367 mg L-I at the end of the simulation period.
The price paid for this reduction
in s a l i n i t y was a decrease in i r r i g a t i o n and an only p a r t i a l l y f u l l reservoir. ~7 14
~oJ
24
l e 14
J,F,M,A,M,J,J,F I975
2O
16
26
24
210
2~
WATER SURFACE
j,RrS,@,N,D.
1976
1977
~
~JFMAMJJ 1978
14
14
1976
JAS@NOJFMAMJJ 1977
1978
~
WATER SURFACE
F M A M J ,j,AT~O N,8 [978
Fig. 9. Simulated temperature and s a l i n i t y variation in 1975 to 1978 for increased streamflow s a l i n i t y , by-pass for s a l i n i t i e s greater than 1300 mg L-1 and scour valve operating as recorded. The effectiveness of by-pass strategy depends greatly on variations of the Collie River s a l i n i t y and most certainly equation ( I ) accentuates the benefits accrued by the by-pass strategy. Fig. 9 shows the temperature and s a l i n i t y structure resulting from such
°
272
a policy.
The r e s e r v o i r s t i l l
s t r a t i f i e s due to temperature gradients, but only
weak s a l i n i t y gradients remain. 5.4
Implementation of a By-pass Policy.
I t was seen in Section 5.3 that a wise balance between reduction in i r r i g a t i o n and by-passing highly saline water could lead to very marked reduction in average reservoir s a l i n i t i e s . flexibility
The purpose of this paper was to demonstrate that such a
may e x i s t and should therefore be explored f u l l y .
However, an i n i t -
ial suggestion f o r a f e a s i b l e diversion dam able to handle the by-pass water w i l l now be made. The choice of a cut o f f s a l i n i t y of 1300 mg L- I requires quite a large quantity of water to be by-passed yet the diversion dam should be as f a r upstream as possible in order to avoid contamination of the main reservoir.
The location of the town-
ship of C o l l i e f u r t h e r requires that the upstream level should not be raised above the e x i s t i n g high water marker. I t is proposed to s i t e a secondary dam at station C85 (see Figs. l(b) and (c)) where the r i v e r v a l l e y has a deep hole.
This would require a 14 km long pipe-
l i n e l y i n g in the bed of the old r i v e r channel.
The deep hole would allow a
deep offtake making maximum use of the possible v e r t i c a l s t r a t i f i c a t i o n to in h i b i t drawdown over the offtake. The pipe diameter required to prevent overflow of the saline water into the parent r e s e rv o i r would be between 1.75 and 2 m.
Such a diameter would ensure
s u f f i c i e n t flow under a 25 m head to allow the upstream storage of about 4xlO7m3 to buffer the peak flows. I t is important to note that neither the dam nor the p i p e l i n e would be exposed to any s i g n i f i c a n t pressures as both would be submerged. Light p l a s t i c construction should s u f f i c e . 6
CONCLUSIONS An examination of the s a l i n i t y v a r i a t i o n s in the C o l l i e River revealed a struc-
ture dependent on the peak and base flow s a l i n i t y , the phase during storms and the flushing time of the whole system.
This v a r i a t i o n was exploited in order to
modify an e x i s t i n g s a l i n i t y record to simulate possible future trends.
I t was
shown that the prediction of the exact nature of the s a l i n i t y v a r i a t i o n s is not important as the mixing in the r e s e r v o i r acts as a strong f i l t e r . overall sequences of the s a l i n i t y v a r i a t i o n s are important.
However, the
The simulation res-
u l t s revealed that without drastic action the s a l i n i t y in the Wellington w i l l soon become unacceptable.
A by-pass strategy is suggested which could lead to
273
dramatic reductions in s a l i n i t y , but at the expense of a reduction in water for irrigation.
This conclusion confirms the i n i t i a l
findings detailed in Fischer
et a l . (1979). 7
REFERENCES Armi, L., 1978. Some evidence for boundary mixing in the deep ocean. J. Geophys. Res., 83: 1971-9. Fischer, H.B., L i s t , E.J., Koh, R.Y.C., Imberger, J. and Brooks, N.H., 1979. Mixing in inland and coastal waters. Academic Press, New York, 483 pp. Hebbert, B., Imberger, J . , Loh, I . and Patterson, J . , 1979. Collie River underflow into the Wellington Reservoir. J. Hydraulics Div. ASCE, 105, No. HY5, 533-45. Imberger, J . , Thompson, R. and Fandry, C., 1976. Selective withdrawal from a f i n i t e rectangular tank. J. Fluid Mech., 78: 389-512. Imberger, J . , Patterson, J . , Hebbert, B. and Loh, I . , 1978. Dynamics of reserv o i r of medium size. J. Hydraulics Div. ASCE. 104, No. HY5, 725-43. Imberger, J. and Patterson, J.C., 1980. A dynamic reservoir simulation model DYRESM 5. Proc. Symp. on Predictive A b i l i t i e s of Surface Water Flow and Transport Models, Berkeley, August, 1980, 75 pp. Ivey, G., 1980. Boundary mixing in a s t r a t i f i e d f l u i d in a rectangular tank. PhD Thesis, Univ. of C a l i f o r n i a , Dept. of C i v i l Engineering, Berkeley. Johnston, C.D., McArthur, W.M. and Peck, A.J., 1980. Distribution of soluble salts in s o i l s of the Manjimup Woodchip Licence Area, Western Australia. CSIRO Aust. Div. Land Resources Manage. Tech. Pap. No. 5, pp. 1-29. Loh, I . , and Porter, J . , 1976. Simulation of monthly flow and s a l t load inputs to Wellington Reservoir. Tech. Report No. 65. Water Resources Sec. Planning, Design and Investigation Branch, Public Works Department, Western Australia. Loh, I.C. and Hewer, R.A., 1977. S a l i n i t y and flow simulation of a catchment reservoir system. Proc. Hydrol. Symp., Instn. Engrs., Aust., Brisbane. Peck, A.J. and Hurle, D.H., 1973. Chloride balance of some farmed and forested catchments in south-west Australia. Water Resour. Res., 9: 648-57. Peck, A.J., Hewer, R.A. and Slessar, G.C., 1977. Simulation of the effects of bauxite mining and dieback disease on r i v e r s a l i n i t y . Tech. Rep. No. 5, CSIRO Aust. Div. Land Resources Manage. Sherman, F.S., Imberger, J. and Corcos, G.M., 1978. Turbulence and mixing in stably s t r a t i f i e d waters. Ann. Rev. Fluid Mech. I0: 267-288. Smith, R.E. and Hebbert, R.H.B., 1980. Manuscript in preparation. Spigel, R.H., 1978. Wind mixing in lakes. Ph.D. Thesis, Univ. of C a l i f o r n i a , Berkeley. Thompson, R.O.R.Y. and Imberger, J., 1980. Response of a numerical model of a s t r a t i f i e d lake to wind stress. Proc. 2nd. Int. Symp. on S t r a t i f i e d Flows, Trondhelm, June 1980. Wood, W.E., 1924. Increase of s a l t in soil and streams following the destruction of the native vegetation. J. Roy. Soc. West. Aust. 10, 35-47. Wu, J., 1973. Wind induced entrainment across a stable density interface. J. Fluid Mech. 61: 275-78. Zeman, O. and Tennekes, H., 1977. Parameterisation of the turbulent energy budget at the top of the daytime atmospheric boundary layer. J. Atmos. Sci. 34: 111-23.
255
Elsevier Scientific Publishing C o m p a n y , A m s t e r d a m -- Printed in The Netherlands
THE INFLUENCE OF STREAM SALINITY ON RESERVOIR WATER QUALITY IMBERGER Department of C i v i l Engineering, U n i v e r s i t y of Western A u s t r a l i a , Nedlands, 6009, W.A. J.
ABSTRACT Imberger, J . , 1981. quality.
The influence of stream s a l i n i t y on reservoir water
Agric. Water Manage., 1981.
The dynamic reservoir simulation model DYRESMis used to investigate the response of the Wellington Reservoir to changes in streamflow s a l i n i t y and outflow strategies.
Construction of a s a l i n i t y diversion dam could be useful in sub-
s t a n t i a l l y reducing reservoir s a l i n i t i e s , but with the penalty of reduced water yield. 1
INTRODUCTION The surface water resources of the populated south-western corner of Western
A u s t r a l i a are drawn from a series of western flowing streams which rise in the plateau regions of the Darling Range, flow through increasingly defined v a l l e y systems, p a r t i c u l a r l y near the western edge of the plateau (Darling Scarp), and out across the coastal p l a i n to the Indian Ocean. Wellington Reservoir is t y p i cal of storages in the region, being located, as shown in Fig. I , in the incised v a l l e y of the C o l l i e River some i0 km from the scarp.
However, i t s catchment
d i f f e r s from those of other reservoirs in the region, in that sections have been progressively cleared of native f o r e s t and converted to annual pastures f o r agricu l t u r a l development.
This land use change has resulted in substantial increases
in the s a l i n i t y of i n f l o w to the reservoir.
Loh and Hewer (1977) indicate t h a t ,
p r i o r to 1950, with only 5% of the catchment cleared, average annual i n f l o w sali n i t i e s were below 300 mg L - I .
Current estimates of the average annual i n f l o w
s a l i n i t y , with some 23% of the catchment cleared, are 750 mg L - I T.D.S., with a yearly increase between 1971 and 1975 of around 50 mg L- I T.D.S. I t is generally accepted that the increased s a l i n i t i e s are the r e s u l t of the reduction of t r a n s p i r a t i o n a f t e r f o r e s t clearing and the consequent increase in recharge to the groundwater system.
The increased groundwater flows upset the
e x i s t i n g s a l t balance by f l u s h i n g large q u a n t i t i e s of soluble s a l t s , previously stored in the soil p r o f i l e , to the stream system (Wood, 1924; Peck and Hurle, 1973).
As groundwater systems respond over periods of tens of years (Peck et a l . ,
1977) the f u l l
e f f e c t of recent clearing has yet to be f e l t .
F i r s t order estimates
0378-3774/81/0000--'0000/$02.50 © 1981 Elsevier Scien~fic Publishing Company
256
made by Loh and Hewer (1977) indicate that i f recent l e g i s l a t i o n l i m i t s catchment clearing to i t s current l e v e l , average annual i n f l o w s a l i n i t i e s of Ii00 mg L - I T.D.S. w i l l u l t i m a t e l y develop, with dry year s a l i n i t i e s as high as 1450 mg L- I T.D.S. and peak s a l i n i t i e s as high as i0 000 mg L - I .
S12004 ~.612035j--~ • mEGION eS
~l
..........7,S;I SCALE 1:500 000
Fig. l ( a ) - Wellington Reservoir. Map showing the reservoir and the major catchment regions. Numbers r e f e r to gauging stations located on the streams. Clearly, any improvement in the supply s a l i n i t y by means of improved reservoir management w i l l be increasingly b e n e f i c i a l as such high s a l i n i t i e s develop, e s p e c i a l l y since long term solutions to the problem must involve improve catchment management and r e h a b i l i t a t i o n which take a long time to establish. In the assessment of reservoir management strategies (Fischer et a l . , 1979) two questions arise.
F i r s t , what is the e f f e c t of the streamflow s a l i n i t y and
the phase between the s a l i n i t y and streamflow in determining the overall s a l t content of the reservoir.
Second, what is the i n t e r p l a y between the streamflow
s a l i n i t y v a r i a t i o n s and the outflow volume and v a r i a b i l i t y s a l t content of the reservoir.
in determining the
The dynamic reservoir simulation model DYRESM
was used to simulate the response of the Wellington Reservoir to changing inflow s a l i n i t i e s and outflow strategies in order to assess the answers to these two questions.
257 .q ?
11 (/
//I
i / ii I f \ ,\_
/
/ tJ /! /
HAMFLTOf
VASE %~\
t i/\ .~J \\
S
f
I,,
J
le
/ (/\I}
I p
j
/
USS~L
//
t~ 5
( \
\h
,/
%..
/
{ SCALE
Fig. l(b) - Wellington Reservoir. Enlargement showing the location of the dam wall and the Mungalup gauging station.
~
'....
~ALE
1k
Fig. 1(c) - Wellington Reservoir.
Site for the proposed salinity diversion dam.
258
2
STREAJMFLOW SALINITY VARIATIONS Much has been written about the s a l i n i t y problem, i t s causes and the mechanisms
responsible for the s a l i n i t y increases observed over the last few decades in the south-west region of Western Australia. The purpose of this paper is to synthesize the effects of these increases on the salt content of the Wellington Reservoir, to establish the s e n s i t i v i t y of the result to variations in streamflow s a l i n i t y and to explore techniques for the alleviation of the salt load. The most developed model of the groundwater system active in the south-west region appears to be that proposed by Smith and Hebbert (1980), who modelled a two aquifer system as shown in Fig. 2.
The water entering the higher areas is
conducted to the deep groundwater via the preferred paths established by new and fossil root structures.
Weak pressures are set up causing a constant upflow in
the lower areas surrounding contributing creeks and streams. The upward flows negotiate a salt laden soil profile (see Johnston et a l . , 1980) introducing saline water to the streams in the winter.
During periods of no flow this slow per-
colation continues leading to a gradual buildup, when combined with evaporation, of salt on and near the surface of the lower areas.
This salt is stored ready
for flushing into the streams during the f i r s t storms of the season.
RAr~
Fig. 2. (From Smith and Hebbert, 1980) - Schematic of the double aquifer concept. Water enters the stream d i r e c t l y via surface and subsurface flows in the perched aquifer and from below under weak artesian pressures. During periods of rain the perched aquifer (Fig. 2) builds up, causing leaching of the salt accumulated by direct water ponding during the end of the previous winter and that accumulated by the upward percolation described above.
In areas
where the perched aquifer is saturated, direct runoff w i l l occur leading to a very rapid response of the streams. The thickness of the perched aquifer obviously w i l l vary greatly so that the contributing area w i l l be a function of space as well as time.
The response of the perched aquifer i t s e l f w i l l be rapid once
saturated (a few days).
The salt stored on the surface and in the perched aquifer
259
w i l l thus be flushed out once the water has travelled a distance characterized by the width of the contributing area.
Effective flushing requires a certain
volume of continuous rain in order that the movement of water be continuous and not be interspersed by periods of strong evaporation. Deep-rooted vegetation has two influences.
First, i t prevents the buildup of
the upward flows into the contributing areas and secondly the vegetation drains much of the perched aquifer during the winter months. This conceptually plausible model explained most of the features of the observed s a l i n i t y and flow d i s t r i b u t i o n , an example of which is shown in Fig. 3.
The
early rains in May and half of June would have progressively leached more and more of the upper salt laden areas.
The increasing contribution from the rising per-
ched aquifer leads to a general rise in s a l i n i t y climaxing with the very large spike around the inflow at the end of June. From that time onward the s a l i n i t y declined, indicating a general trend towards a flushed state. naturally
Each new flood
eads to a local spike as increases in flow encompass a bigger sphere
of influence in the perched aquifer due to the growth in the contributing area.
24
i i
Z2
i! !i
.................. .........
a.o
1.8
1.e
1.4
~
V~
/
o8
o.6
o4
o.2
°
MAY
JUNE
JULY
AUGUST
o.o
IS75
Fig. 3. Distribution of the s a l i n i t y and flow in the Collie River at the Mungalup gauging station. The distribution t y p i f i e s the rise in s a l i n i t y during the early flows and the lowering of s a l i n i t y as the upper surface areas are flushed.
260
Further, the model suggested by Smith and Hebbert (1980) is able to explain the very d i f f e r e n t s a l i n i t y response recorded during the July flow and the high flow at the middle of August.
A s t r a i g h t c o r r e l a t i o n of s a l i n i t y and flow as sugges-
ted by Loh and Porter (1976) is obviously not warranted even i f a correction f o r the time of year is included. double a q u i f e r model.
A more plausible answer may be derived from the
The flow in July was preceded by perched aquifer flushing
whereas the August flow was preceded by a period of low r a i n f a l l which had allowed a new s a l t load to accumulate in the c o n t r i b u t i n g areas. Smith and Hebbert (1980) have proposed this model, but i t w i l l take a few years to v a l i d a t e such a d e t a i l e d model f o r the C o l l i e Catchment and so no attempt was made to couple t h i s model to DYRESM. Instead certain features of the model were used to change the recorded s a l i n i t i e s to simulate r e f o r e s t a t i o n and increased clearing in the f o l l o w i n g way. The essential features are that the peak s a l i n i t i e s in the stream depend on the s a l t concentration in the contributing areas, the phasing of the flow in r e l a t i o n to the volume of the contributing areas and the percentage of s a l t rich area to fresh flowing catchment.
By r e t a i n i n g the phas-
ing as in 1975 to 1978, but increasing the percentage area c o n t r i b u t i o n , increased s a l i n i t y records may thus be generated.
Further, by assuming that the nature
of the clearing carried out in the l a s t 15 years is s i m i l a r to that done e a r l i e r i t is reasonable to assume the flushing volumes contributing to the s a l i n i t y have the same time responses.
These assumptions would lead to a much increased
peak s a l i n i t y , but with about the same flushing time scale in r e l a t i o n to flow v a r i a b i l i t y and the same "flushed" s a l i n i t y .
I f this low state is taken as 500
mg L- I , then a possible recipe f o r simulating an increased s a l i n i t y would be : Se S = 5OO " Se
(1)
where Se is the e x i s t i n g s a l i n i t y .
This leads to a s a l i n i t y record with an
average s a l i n i t y of about 1700 mg L- I and a new peak s a l i n i t y of i i 500 mg L- I corresponding to expectations, outlined in the introduction.
Further work is
n a t u r a l l y required to v e r i f y this simple hypothesis expressed by equation ( i ) , but f o r the purposes of the reservoir simulation i t serves as a good f i r s t model. Reduced s a l i n i t i e s may again be expressed using the e x i s t i n g record by: S = 0.5 (Se-500) + 350 S = 350
S > 500 e S < 500 e
(2)
This w i l l be used to simulate conditions before the recent clearing took e f f e c t in 1960.
261
3
MODEL DESCRIPTION The dynamic reservoir simulation model, DYRESM, is a one-dimensional numerical
simulation model for the prediction of temperature and s a l i n i t y in small to medium lakes.
I t is based on the assumption that in such lakes the thermal structure
plays a dominant role and the isotherms are mostly horizontal planes.
Deviations
from this state of "rest" are allowed, but i t is assumed that such deviations are sporadic or weak and may be treated as perturbations to the evolution of the onedimensional structure. These assumptions place certain restrictions on the a p p l i c a b i l i t y of the model. In particular the Wedderburn Number (see Thompson and Imberger, 1980) should be greater than one for the majority of the wind events, the inflow and outflow internal Froude numbers (see Imberger et a l . , 1976) should be less than one and the lake should be small or narrow enough for the influence of the earth's rotation to be negligible (see Fischer et a l . , 1979). DYRESMwas developed over the last 6 years in order to predict the s a l i n i t y variations in the Wellington Reservoir, but in the meantime the model has found wide a p p l i c a b i l i t y as a base for more general water quality modelling.
The devel-
opment of the model is continuing and currently version 5 is the operational model. A f u l l and detailed description of this version may be found in Imberger and Patterson (1980) and only a very brief summary w i l l be given here. The lake is divided into uniform horizontal slabs which form the computational building blocks of the model. The construction of the slabs is Lagrangian and they advect with the vertical velocity induced by the inflow and outflow.
At
each time step the model ensures that heat, salt, mass and energy are conserved for each slab and thus the reservoir as a whole.
The vertical momentum equation
reduces to the hydrostatic pressure assumption, since the assumption of onedimensionality eliminates all motion except the slow vertical adjust required to accommodate in and outflows. Once this slab structure has been established the meteorological inputs at the surface of the lake are calculated using bulk aerodynamic formulae.
As shown in
Fig. 4, these together with the daily inflow and outflow form the daily input data The basic time step of the model has been set at one day since only daily data is most commonly available.
Certain assumptions are made regarding the distribu-
tion of these inputs over the 24-hour period and time steps down to ~ hour are used where the processes to be simulated vary rapidly during the day i t s e l f . I f more frequent data is available then this should be used. The daily surface heat inputs are used by the model to calculate the temperature and s a l i n i t y changes in the slab structure.
The updated slab structure is then
262
adjusted for mixed layer deepening and possible changes of the thermocline thickness due to the formation of shear i n s t a b i l i t i e s .
The calculation of the mixed
layer deepening incorporates both deepening due to surface turbulence and turbulence generated by the shear i n s t a b i l i t i e s at the base of the mixed layer.
Data Input
]
Initialisation
I
1 day
1
¼ hour
Da,ydet. i° t
!
Surface heat exchanges
]
Mixed layer deepening
- 12 hours
Fig. 4.
Billowing
Turbulent diffusion
]
Inflow
]
Outflow
]
End
J
Schematic of the programmed model DYRESM.
Once the new thermocline depth and thickness has been computed the model then calculates the net heat and salt transport from the bottom through the hypolimnion into the epilimnion.
The basic algorithm by which this is carried out is
an eddy diffusion parameterization.
However, i t was recognized very early in
the development of DYRESM (see Imberger et a l . , 1978) that a constant diffusion coefficient was not applicable.
Field experiments showed that the vertical trans-
fer of heat and mass was higher during periods of high wind and inflow even though both of these processes did not d i r e c t l y contribute to the turbulence in the hypolimnion.
I t was postulated that both these disturbances induce basin
scale internal oscillations which are damped at the boundary. The dissipation at the boundary in turn produces small scale turbulence mixing.
capable of causing
The active mixing in the boundary layer quickly leads to an adjustment
of densities in the boundary layer and so induces horizontal intrusions which in turn cause an adjustment of the centre of gravity of the overall lake structure.
263
This concept has since then been postulated for the ocean (Armi, 1978) and recent experiments by Ivey (1980) have established rates of transfer in an idealized laboratory experiment.
While the concept appears attractive, not enough is under-
stood about the details of the individual processes to allow a process model to be constructed.
Instead, a simple energy argument is used (Imberger et a l . , 1978).
I t is assumed that a small, but constant fraction of the wind and inflow energies is Used to generate mixing in the hypolimnion.
This leads to a vertical diffus-
ion coefficient :
H2
(3)
~z = 0.048 ST--M '
where TM is a time scale for mixing(equal to the potential energy of the density s t r a t i f i c a t i o n of the lake divided by the power input of the wind and streams); and S is a normalized water column s t a b i l i t y (=H/&p dp/dz, where H is the depth of the lake; Ap is the density difference between the bottom and the surface; and p is the density of the water at an elevation z). This parameterization is successful (see Imberger and Patterson, 1980) provided the s t a b i l i t y S is quite large.
I t obviously breaks down in the l i m i t of a homo-
geneous water mass with an energy input.
In this l i m i t both S and TM approach
zero and equation (3) predicts an i n f i n i t e diffusion coefficient. In order to prevent this, an arbitrary cut off to equation ( 3 ) o f ~z = 10-4m2s-1 is provided in the model. Further work is required to rationalize this cut off especially since its value has quite a strong influence on the predicted s a l i n i t y distribution.
The mixing processes in the hypolimnion seem to require a two
parameter model, equation (3) for strongly s t r a t i f i e d water and an analogous formulation for a homogeneous hypolimnion where the mixing may be expected to be more evenly distributed throughout the water mass. At the end of the diffusion routine, which is carried on the same time step as the mixed layer dynamics calculation, a new structure for a particular day has been calculated.
This density structure is then used to route the inflowing
water from the various contributing streams into the reservoir. allows for turbulent entrainment and subsurface intrusions.
The subroutine
Similarly, the out-
flow is calculated by the model using the structure l e f t after the inflow has been added. This routine is repeated for each day of the simulation. 4
MODEL PERFORMANCE DYRESM has been tested on a number of different lakes, but i t s major evaluation
264
and development has taken place with data from the Wellington Reservoir. The seasonal v a r i a b i l i t y of the various inputs to the reservoir over the period from the Julian day 133 in 1975 to day 365 in 1977 are shown in Fig.5.
Depicted
are the wind speeds, the short-wave solar r a d i a t i o n as computed from cloud cover records, the s a l i n i t y and temperature of the inflowing water and the total rate of inflow from the C o l l i e River which contributes approximately 85% of the t o t a l inflow.
The remaining i n f l o w is included in the simulations, but is not shown in
Fig. 5.
6 WI
26
c~
3ODD
4
2DOO z 1000 NRCL "
V-
'
'
'
-
~24 22o
w12 ~
~ 8
z _IN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D d , f ,M,R,M,J,J,I~,5,O,N,D, J , F ' M , R , M , J ' J , A , g , O , N ' I ~,J ,F ,~I,A 'M ' J ' J 'A 'S ' 0 N ' { ] ,
1975
1976
197"7
Fig. 5. Seasonal v a r i a b i l i t y of wind speed, short-wave radiation, inflow s a l i n i t y of the Collie River, inflow temperature and the inflow volume. Figs. 6(a) and 6(b) show the f i e l d temperature and s a l i n i t y structures length averaged along the C o l l i e River Valley as a function of time over the period January 1975 to August 1978.
The s a l i n i t y data gathered between October 1977 and
June 1978 is regarded as u n r e l i a b l e and is not shown.
The broken l i n e s in Fig.
6(a) indicate that no data was taken in the period covered and the thermal structure is interpreted from the structure before and a f t e r the period of i n t e r r u p t i o n The y e a r l y cycle is c l e a r l y evident in Figs. 5 and 6:
the cold s a l t y inflows
lodge in the base of the homogeneous reservoir in the months of June, July and August; summer s t r a t i f i c a t i o n
builds up u n t i l December, when surface winds begin
to mix the surface layers and a thermocline forms, protecting the waters below. In early w i n t e r , the a i r temperature f a l l s and the winds increase, with the resu l t that the reservoir is completely mixed before the f o l l o w i n g inflows a r r i v e .
265
The marked difference in the thermocline structure between 1976 and 1977 was caused by a change in the withdrawal p o l i c y .
In 1976 a l l the water was with-
drawn from the offtakes at 15 m height, whereas in 1977 a large quantity of salty water was scoured through the offtake at the bottom of the dam w a ll.
In addition
about two thirds of the water f o r i r r i g a t i o n was taken from the bottom o f f t a k e .
WELLINGTON THERMALSTRUCTURE
Fig. 6(a). Measured average r e s e r v o i r temperature as a function of time f o r the years 1975 to 1978. (From Imberger and Patterson, 1980).
WELLINGTON NACL STRUCTURE
:is ,FIELD TRIP / ~: I!,
~t 1) h L,
Fig. 6(b). Measured average reservoir s a l i n i t y (mg L-1NaCl) as a function of time f o r the years 1975 to 1978. (From Imberger and Patterson, 1980). I t is also clear from Fig. 6(a) that the temperature regime of the reservoir is determined by the inflows and the surface heating and cooling.
The bottom
temperature of the r e s e r v o i r f o r most of the year is determined by the temperature of the coldest inflows, whereas the surface temperature is determined by the meteorological forcing. There are seven constants which must be specified by the user before applying DYRESM. Of these only one is t r u l y adjustable - the others are related to well i d e n t i f i e d physical processes and are determined from experimental or f i e l d data.
266
The constants are described below, together with experimentally determined values (i)
CD is the drag coefficient for inflowing streams.
CD was determined in-
dependently of DYRESMin a f i e l d study described by Hebbert et al. (1979).
The
value determined in that study, CD = 0.015, is used here. (ii)
nI is an extinction coefficient for short-wave solar radiation penetrat-
ing the water.
I t relates
the solar radiation received at the water surface, to
that penetrating to a depth z.
A single exponential decay formula was used as
only limited f i e l d measurements were available.
An average value of nI = 0.35
is taken, based on the fact that the Wellington is f a i r l y clear in the summer months when surface heating is an important effect. (iii)
~l is a constant occurring in the expression for the d i f f u s i v i t y calcula.
ted for the deep hypolimnetic mixing.
I t basically represents the efficiency
with which the power input from the surface wind and river inflows is converted to a gain in potential energy of the lake water due to vertical mixing.
A 9.6%
efficiency ~ l = 0.048) determined in earlier calibrations of DYRESMover the lO0-day period from day 133 to 233 has proven satisfactory throughout.
The cut-
off value of lO-4m2sec-1 was set by noting that this corresponds to the maximum value measured by a number of investigators (see Fischer et a l . , 1979). (iv)
CK is the coefficient that describes the s t i r r i n g efficiency of convec-
tive overturn.
Experimental results summarized by Fischer et al. (1979) suggest
an average value of CK = 0.125. (v)
n, in combination with CK as CKq3, is a coefficient measuring the s t i r -
ring efficiency of the wind.
The value given by Wu (1973) is adopted here, since
i t was shown by Spigel (1978) that in his experiments s t i r r i n g effects dominated the entrainment, so that shear production, temporal effects and internal wave radiation losses were negligible.
Wu's deepening law is dh/dt = 0.23 u,/Ri*
giving CKn3 = 0.23, and thus n = 1.23. (vi)
CS is the coefficient that describes the efficiency of shear production
for entrainment. (1978).
Values ranging from 0.2 to 1 are reported by Sherman et al.
CS = 0.2 was chosen as representing a good estimate for most experimen-
tal results.
A value of 0.5 was used before the energy released by billowing
was separately included as is done in version 5 of DYRESM. (vii)
CT is associated with temporal, or unsteady, non-equilibrium effects
due to changes in surface wind stress or surface cooling.
CT is constructed by
the requirement that for a turbulent front entraining into a homogeneous f l u i d , dh/dt = O.3u, (Zeman and Tennekes, 1977)giving a value of CT = CKn/0.3 = 0.510. These coefficient values were used for a 962-day simulation of the lake dynamics starting with i n i t i a l profile data on day 75133. The results from this
267
simulation are shown in Fig. 7.
~°
iI
~
~ii
-~•
< t~ '~9r'~
[ 9n6
'~977
1978
=i ~t ?
"
/
Io75
i
'.976
i
A $0 N O J 1977
~JJ
aS @N O °
1978
Fig. 7. Simulated average temperature and salinity for the period 1975 to 1978. (From Imberger and Patterson, 1980). Comparison of Figs. 6(a) and 7 shows that the thermal structure is reproduced extremely well with a l l the s t r a t i f y i n g and mixing regimes occurring at the correct times and of the correct magnitude.
The slug of 12°C water predicted
by the model in August to September 1975 does not appear in the f i e l d data.
This
slug is derived from the previous inflow and i t s presence may be due to errors in the inflow temperature data.
In any case, the predicted temperature gradient in
the bottom at this time is small, and the actual difference between f i e l d and predicted temperatures is less than I°C. The predicted and observed s a l i n i t y v a r i a t i o n s also compare well with the most s i g n i f i c a n t anomaly being the mismatch of the 600 mg L- I l i n e at the period of maximum inflow in 1976. The f i e l d data suggest somewhat more energetic surface mixing which may be due to small errors in the wind data.
A d d i t i o n a l l y , the
model does not reproduce the peak s a l i n i t i e s of the inflows.
This is due to the
construction of the model layers in terms of layer volumes - hence the depth resolution in the upper part.
The bottom three or four metres always appear as
mixed, and any s a l i n i t y slug occupying less than this w i l l be mixed with the layer above.
Thus there is no error in terms of the s a l t load, but one of dis-
play of the structure, and i t was accepted in order to save computer time.
If
greater resolution were required, this could be achieved by specifying smaller
268
minimum slab sizes. Overall, DYRESMappears f a i t h f u l l y to reproduce even very severe changes in the reservoir structure caused by such diverse forcing as large saline inflows active scouring, strong wind deepening and winter convective cooling.
Perhaps
more importantly, the model correctly simulates two independent parameters, sali n i t y and temperature to a resolution equal to that of the f i e l d programme. More accurate f i e l d data is required i f more stringent tests are to be applied. 5
INFLUENCE OF SALINITY CHANGESAND POSSIBLE ALLEVIATION STRATEGIES The simulation model DYRESMwas used to investigate the influence of increased
and decreased C o l l i e River s a l i n i t y and methods f o r a l l e v i a t i n g the strong b u i l d up of r e s e rv o i r s a l i n i t y f o r the case of increased C o l l i e River s a l i n i t y .
The
technique employed was to run the simulation f o r the three-year period using the measured i n f l o w, outflow and meteorological data.
This three-year period (1976
to 1978) represents a s t a r t of a severe drought period with the associated b u i l d up of stream s a l i n i t y and of course decreased flow.
As mentioned in Section 2
i t was f e l t that manipulating the magnitude of the s a l i n i t y of this time series would be more r a t i o n a l than generating a model set since this data set preserves the phase r e l a t i o n s h i p between the s a l i n i t y and flow v a r i a t i o n s .
The model then
simulates the subtle interactions between the s a l i n i t y structure, the outflow, mixing and inflow and allowed an assessment of the influence of any proposed changes.
Obviously, many v a r i a t i o n s could be simulated but because of the cost
involved only a set of extreme conditions w i l l be described.
I t is f e l t that
these results allow an assessment of the l i k e l y changes which may be expected. 5.1
Variations of Inflow Salinity
The simulation of the actual f i e l d data has already been discussed in Section 4.
Two further cases were run.
First, the Collie River s a l i n i t y was decreased
to that given by equation (2), the scour valve was turned o f f and flow with a s a l i n i t y greater than 1300 mg L-1NaCl was by-passed. The l a t t e r two strategies w i l l be explained later, but they were incorporated in the one run in order to save computer time.
The results are shown in Table 1 where i t is shown that
the average s a l i n i t y is lowered from 650 mg L- I to 480 mg L-1 and the dam overflows.
I t must be rememberedthat leaving the bottom scour valve closed (see
Fischer et a l . , 1979) has no effect in 1975/76 as i t was essentially closed under normal operation.
However, in 1976/77 much of the i r r i g a t i o n water (about two
thirds) was withdrawn through the scour valve and keeping the valve closed thus leaves a surplus of water in the reservoir.
However, the reduction of s a l i n i t y
is mainly due to the reduced inflow s a l i n i t y as essentially no water was by-passed
269
(salinity would have to be in excess of 2400 mg L° I ) and the scour valve salinity was always close to the average.
TABLE 1
Day o f Year Salinity of Inflow
Scour Valve Flow
Top Draw Flow
By-pass Flow
As measured
As operated
As operated
zero
. . . . y~-- F 11.4 308 12. I 450 8.1
Decreased
zero
As o p e r a t e d
Whenever S > 1300
II.2
308
12.0
14.6
413
Overflow
Increased
As operated
As o p e r a t e d
zero
11.2
308
12.0 1804 l 7.6
956
14.3
As m e a s u r e d
zero
As operated
zero
Ii.2
308
12.0 1464 114.7
584
Overflow
641
Increased
zero
As o p e r a t e d
zero
11.2
308
1320
Overflow
1648
As m e a s u r e d
As operated
As o p e r a t e d
11.2
308
12.4 !810 !14.7 i ! 12.0 450 6.3
As m e a s u r e d
zero
As o p e r a t e d
11.2
308
12.2
430
Increased
zero
As o p e r a t e d
11.2
308
9.7
371
.
S > 1300 Whenever S > 1300
~menever S > 1300
.
.
.
75133 ~Vol *Sal ~ .
Whenever
.
*
Volumes are in 107m 3
**
Salinities are in p.p.m. NaCL (average for whole vo I u m e ) .
.
.
.
.
F-.
.
.
.
.
76133 771 13 V o ~ - Sa--]--~ G Y - ~ a ~ -
354
.
.
.
.
.
77365 VJ~-S-dl
. . . . . . . 530 13.0 650 480 1345
438
3.5
480
12.9
444
Overflow
510
8.9
357
11.4
367
Second, the salinity of the Collie River inflow was increased as indicated by equation (1) and the reservoir was operated as was actually the case in the three-year period.
The increase in the average reservoir salinity was quite
dramatic reaching 1345 mg L- I at the end of the simulation period. levels would be quite unacceptable for all but a few selected crops.
Such salinity Further,
the strategy of taking irrigation water from the scour valve becomes no longer possible as the s a l i n i t i e s would rise to over 2000 mg L-1.
However, i t does
prevent an even greater increase of salinity buildup than would occur i f all the irrigation flow were to be taken from the top offtake. 5.2
Variations of the Scour Valve Flow.
The effect of the scour valve flow is quite subtle and can easily be misleading.
Table 1 shows the results from a simulation with inflow s a l i n i t i e s , for the
1975 to 1978 period, but with the bottom scour valve completely closed during the whole period.
As already mentioned this led to a strong reduction in the
water available for irrigation in 1977/78 with a consequential overflowing of the reservoir.
During the period 76133 to 77133 the scour policy had the ob-
vious effect of reducing the average salinity from 584 to 530 a small but definite benefit.
However, during the period from 77133 to 77365 scour policy actu-
ally had a detrimental influence since the water withdrawn in 1976 through the scour valve was of a lower salinity than the average salinity in this l a t t e r
270
period.
Had t h i s water been wasted then the strategy would have had an adverse
effect.
I t is therefore seen that a judgement must always be made whether the
present year's poor q u a l i t y water is l i k e l y to be of better q u a l i t y than the best q u a l i t y of the f o l l o w i n g year. The reverse e f f e c t is noticed in the increased s a l i n i t y case (see Table i ) . Closing the bottom scour valve once again led to a f u l l
r e s e r v o i r , but with an
average s a l i n i t y as high as 1648 mg L -1 at the end of the simulation period. Hence, l i t t l e
b e n e f i t would be gained by reducing the i r r i g a t i o n supply in an-
t i c i p a t i o n of a better f o l l o w i n g year.
Under such circumstances i t would be
better to i r r i g a t e generously in 1976 and bank one's p r o f i t s in order to overcome the hardship of the f o l l o w i n g very s a l t y inflows. Fig. 8 shows the isohalines for t h i s very extreme strategy.
The higher s a l i n -
i t i e s are reflected in the very much i n t e n s i f i e d s t r u c t u r e , the increased peak s a l i n i t i e s and the strong v e r t i c a l s a l i n i t y (and thus density) gradients.
These
gradients prevented mixing, leading one to suspect much increased e f f i c i e n c y of the scour p o l i c y (not shown in Table i ) .
4O0
197S
800
1976
1000
1977
WATERSURFACIE
1978
Fig. 8. Simulated s a l i n i t y v a r i a t i o n s in 1975 to 1978 f o r increased streamflow s a l i n i t y , no by-pass and no scour valve operation. 5.3
The Effect of By-passin 9 High S a l i n i t X Water.
The f u l l assessment of the benefits to be gained by by-passing the very high saline slug of C o l l i e River water would involve a great many simulations. In -1 t h i s paper a cut o f f s a l i n i t y of 1300 mg L has a r b i t r a r i l y been chosen to represent an acceptable upper l i m i t .
Introducing such a strategy of by-pass in
addition to the scour practice of 1975 to 1978 reduced the average s a l i n i t y from 650 mg L - I to 480 mg L - I , but the extra 1.8xlO7m 3 water scoured would have reduced the water volume to a dangerously low level of 3.5xlO7m 3 at the end of the simulation period and at the s t a r t of the 1978 i r r i g a t i o n season.
"~.
271
For this reason the by-pass policy was combined with a major reduction in the 1976 i r r i g a t i o n volume. This was achieved by closing the bottom scour valve cutting the i r r i g a t i o n flow in 1976/77 to about one third that actually delivered. The average s a l i n i t y increased marginally, but the reservoir f i l l e d by the end of 1977. Obviously, a compromise between these two l a t t e r approaches would be most advantageous; a certain by-pass flow should be compensated for by a corresponding reduction in the i r r i g a t i o n volume. The by-pass strategy was also evaluated for the case of increased s a l i n i t i e s (see Table i ) .
In anticipation of a low volume i t was decided again to close
the scour valve.
This resulted in a dramatic reduction in s a l i n i t y from 1345
mg L-1 (1648 mg L- I corresponding no scour case) to a very modest value of 367 mg L-I at the end of the simulation period.
The price paid for this reduction
in s a l i n i t y was a decrease in i r r i g a t i o n and an only p a r t i a l l y f u l l reservoir. ~7 14
~oJ
24
l e 14
J,F,M,A,M,J,J,F I975
2O
16
26
24
210
2~
WATER SURFACE
j,RrS,@,N,D.
1976
1977
~
~JFMAMJJ 1978
14
14
1976
JAS@NOJFMAMJJ 1977
1978
~
WATER SURFACE
F M A M J ,j,AT~O N,8 [978
Fig. 9. Simulated temperature and s a l i n i t y variation in 1975 to 1978 for increased streamflow s a l i n i t y , by-pass for s a l i n i t i e s greater than 1300 mg L-1 and scour valve operating as recorded. The effectiveness of by-pass strategy depends greatly on variations of the Collie River s a l i n i t y and most certainly equation ( I ) accentuates the benefits accrued by the by-pass strategy. Fig. 9 shows the temperature and s a l i n i t y structure resulting from such
°
272
a policy.
The r e s e r v o i r s t i l l
s t r a t i f i e s due to temperature gradients, but only
weak s a l i n i t y gradients remain. 5.4
Implementation of a By-pass Policy.
I t was seen in Section 5.3 that a wise balance between reduction in i r r i g a t i o n and by-passing highly saline water could lead to very marked reduction in average reservoir s a l i n i t i e s . flexibility
The purpose of this paper was to demonstrate that such a
may e x i s t and should therefore be explored f u l l y .
However, an i n i t -
ial suggestion f o r a f e a s i b l e diversion dam able to handle the by-pass water w i l l now be made. The choice of a cut o f f s a l i n i t y of 1300 mg L- I requires quite a large quantity of water to be by-passed yet the diversion dam should be as f a r upstream as possible in order to avoid contamination of the main reservoir.
The location of the town-
ship of C o l l i e f u r t h e r requires that the upstream level should not be raised above the e x i s t i n g high water marker. I t is proposed to s i t e a secondary dam at station C85 (see Figs. l(b) and (c)) where the r i v e r v a l l e y has a deep hole.
This would require a 14 km long pipe-
l i n e l y i n g in the bed of the old r i v e r channel.
The deep hole would allow a
deep offtake making maximum use of the possible v e r t i c a l s t r a t i f i c a t i o n to in h i b i t drawdown over the offtake. The pipe diameter required to prevent overflow of the saline water into the parent r e s e rv o i r would be between 1.75 and 2 m.
Such a diameter would ensure
s u f f i c i e n t flow under a 25 m head to allow the upstream storage of about 4xlO7m3 to buffer the peak flows. I t is important to note that neither the dam nor the p i p e l i n e would be exposed to any s i g n i f i c a n t pressures as both would be submerged. Light p l a s t i c construction should s u f f i c e . 6
CONCLUSIONS An examination of the s a l i n i t y v a r i a t i o n s in the C o l l i e River revealed a struc-
ture dependent on the peak and base flow s a l i n i t y , the phase during storms and the flushing time of the whole system.
This v a r i a t i o n was exploited in order to
modify an e x i s t i n g s a l i n i t y record to simulate possible future trends.
I t was
shown that the prediction of the exact nature of the s a l i n i t y v a r i a t i o n s is not important as the mixing in the r e s e r v o i r acts as a strong f i l t e r . overall sequences of the s a l i n i t y v a r i a t i o n s are important.
However, the
The simulation res-
u l t s revealed that without drastic action the s a l i n i t y in the Wellington w i l l soon become unacceptable.
A by-pass strategy is suggested which could lead to
273
dramatic reductions in s a l i n i t y , but at the expense of a reduction in water for irrigation.
This conclusion confirms the i n i t i a l
findings detailed in Fischer
et a l . (1979). 7
REFERENCES Armi, L., 1978. Some evidence for boundary mixing in the deep ocean. J. Geophys. Res., 83: 1971-9. Fischer, H.B., L i s t , E.J., Koh, R.Y.C., Imberger, J. and Brooks, N.H., 1979. Mixing in inland and coastal waters. Academic Press, New York, 483 pp. Hebbert, B., Imberger, J . , Loh, I . and Patterson, J . , 1979. Collie River underflow into the Wellington Reservoir. J. Hydraulics Div. ASCE, 105, No. HY5, 533-45. Imberger, J . , Thompson, R. and Fandry, C., 1976. Selective withdrawal from a f i n i t e rectangular tank. J. Fluid Mech., 78: 389-512. Imberger, J . , Patterson, J . , Hebbert, B. and Loh, I . , 1978. Dynamics of reserv o i r of medium size. J. Hydraulics Div. ASCE. 104, No. HY5, 725-43. Imberger, J. and Patterson, J.C., 1980. A dynamic reservoir simulation model DYRESM 5. Proc. Symp. on Predictive A b i l i t i e s of Surface Water Flow and Transport Models, Berkeley, August, 1980, 75 pp. Ivey, G., 1980. Boundary mixing in a s t r a t i f i e d f l u i d in a rectangular tank. PhD Thesis, Univ. of C a l i f o r n i a , Dept. of C i v i l Engineering, Berkeley. Johnston, C.D., McArthur, W.M. and Peck, A.J., 1980. Distribution of soluble salts in s o i l s of the Manjimup Woodchip Licence Area, Western Australia. CSIRO Aust. Div. Land Resources Manage. Tech. Pap. No. 5, pp. 1-29. Loh, I . , and Porter, J . , 1976. Simulation of monthly flow and s a l t load inputs to Wellington Reservoir. Tech. Report No. 65. Water Resources Sec. Planning, Design and Investigation Branch, Public Works Department, Western Australia. Loh, I.C. and Hewer, R.A., 1977. S a l i n i t y and flow simulation of a catchment reservoir system. Proc. Hydrol. Symp., Instn. Engrs., Aust., Brisbane. Peck, A.J. and Hurle, D.H., 1973. Chloride balance of some farmed and forested catchments in south-west Australia. Water Resour. Res., 9: 648-57. Peck, A.J., Hewer, R.A. and Slessar, G.C., 1977. Simulation of the effects of bauxite mining and dieback disease on r i v e r s a l i n i t y . Tech. Rep. No. 5, CSIRO Aust. Div. Land Resources Manage. Sherman, F.S., Imberger, J. and Corcos, G.M., 1978. Turbulence and mixing in stably s t r a t i f i e d waters. Ann. Rev. Fluid Mech. I0: 267-288. Smith, R.E. and Hebbert, R.H.B., 1980. Manuscript in preparation. Spigel, R.H., 1978. Wind mixing in lakes. Ph.D. Thesis, Univ. of C a l i f o r n i a , Berkeley. Thompson, R.O.R.Y. and Imberger, J., 1980. Response of a numerical model of a s t r a t i f i e d lake to wind stress. Proc. 2nd. Int. Symp. on S t r a t i f i e d Flows, Trondhelm, June 1980. Wood, W.E., 1924. Increase of s a l t in soil and streams following the destruction of the native vegetation. J. Roy. Soc. West. Aust. 10, 35-47. Wu, J., 1973. Wind induced entrainment across a stable density interface. J. Fluid Mech. 61: 275-78. Zeman, O. and Tennekes, H., 1977. Parameterisation of the turbulent energy budget at the top of the daytime atmospheric boundary layer. J. Atmos. Sci. 34: 111-23.