Professional Documents
Culture Documents
The physical characteristicsof mountain streams differ from the uniform and conceptually well-
defined open channels for which the analysis of solute transport has been oriented in the past and is
now well understood. These physical conditions significantlyinfluence solute transport behavior, as
demonstratedby a transientstoragemodelsimulationof solutetransportin a very small(0.0125m3
s-) mountainpool-and-rifflestream.The applicationis to a carefullycontrolledand intensively
monitored chloride injection experiment. The data from the experiment are not explained by the
standard convection-dispersion mechanisms alone. A transient storage model, which couples dead
zones with the one-dimensionalconvection-dispersionequation, simulatesthe general characteristics
of the solute transport behavior and a set of simulation parameters were determined that yield an
adequatefit to the data. However, considerableuncertaintyremainsin determiningphysicallyrealistic
values of these parameters. The values of the simulationparametersusedare comparedto values used
by other authors for other streams. The comparison supports, at least qualitatively, the determined
parameter values.
INTRODUCTION dent, because the short distances and small volumes of water
Extensive knowledge exists concerning the mechanisms in the system suggest that similarly small-scale physical
and simulation of solute transport in streams [Fischer et al., processes occurring in the stream may have significant
1979]. From the applications perspective, interest has fo- consequences.Thus the applicationof a one-dimensional(or
cused on larger streamslocated near population centers. As pseudo-two-dimensional)transport analysisremains empiri-
a generalization, these streams are physically low slope, cal. The model equations have previously been applied by
deeper than the roughest bed feature, and relatively uniform numerous workers, mostly to improve the shortcomingsof
(possibly due to flow regulation). Chemically, these streams the advection-dispersion model, in large streams. After
contain and transport rather high concentrations of diverse presenting the model and the simulation of the experiment,
constituents. Small mountain pool-and-riffle streams are we will discussthe apparent distinctions between previous
describedby opposingcharacteristics.Chemically, there are applications and the one presented here. This comparison
fewer significantconstituents present and these are at lower supports the conclusion that the transient storage mecha-
concentrations. These streams cannot be characterized as nism is a viable first approximationto the complexity of the
uniform either in dimension or gradient, and the bed may physical processes.
significantly influence solute transport. MODEL
Of late there has been interest in studying the physical
The typical startingpoint for the study of solutetransport
[Keller and Tally, 1979;Keller, 1975;Heede, 1972], chemical in streams centers on the one-dimensional convection-dis-
[Hubbard and Striffier, 1973; Lewis and Grant, 1979], and
persion analysis. The 'tail' of a solute tracer pulse is often
solute transport [Day, 1975, 1977] characteristicsof small more pronouncedthan can be accountedfor by this analysis.
mountain pool-and-riffle streams. Certain larger mountain (See Fischer et al. [1979] for a complete discussionof solute
streams maintain the significant pool-and-riffle characteris-
transport in real streams, the one-dimensionalanalysis, and
tic. Studies of the overall hydrochemical and transport alternatives.) A common method for simulating these long
characteristics of these larger streams are now appearing tails has been to allow for storage zones or 'dead zones'
[Janda, 1977;Kennedy and Malcolm, 1977]and will presum-
along the stream channel. These storage zones are assumed
ably become more important as society continues to seek
to be stagnantrelative to the longitudinalflow of the stream
new locations to discharge its chemical and thermal pollu-
and to obey a first-order mass transfer type of exchange
tion.
relationship. That is, the exchange of solute between the
In this study, application of a transient storagemodel to a main stream channel and a storage zone is proportional to
field experiment is presented. The model must reflect the the difference in concentration between the stream and the
large variability in the physical characteristicsof the stream
storage zone. Implementation of this model requires an
in order to simulate the field experiment. Concurrently, the
estimate of the storage zone cross-sectional area and an
model is useful in smoothing out this same variability in
empirical exchange coefficient. Under a variety of nomen-
order to make the simulations tractable.
clatures, various theoretical, laboratory flume, and field
The importance of differing physical scales becomes evi-
aspectsof dead zone models have been analyzed [Hays et
al., 1966; Thackston and Krenkel, 1967; Thackston and
Copyright 1983by the American GeophysicalUnion.
Schnelle, 1970; Pedersen, 1977; Valentine and Wood, 1977;
Paper number 3W0373. Sabol and Nordin, 1978; Valentine and Wood, 1979a, b; Tsai
0043-1397/83/003W-0373505.00 and Holley, 1979; Nordin and Troutman, 1980].
718
BENCALA AND WALTERS: MOUNTAIN STREAM SOLUTE TRANSPORT MODEL 719
0m ......
, .............................
"
::.. ..:ii: In mountain streams, one does, nevertheless, observe
significanttails from a solute pulse. Solute mass is removed
during the rising phase of a solute pulse, temporarily re-
tained in storageuntil the pulse passes,and then returned to
281
m _ N the stream. From this observation
fact a mechanismthat presents itself as transient storage of
we infer that there is in
100
meters 3 m solute mass along the length of the stream. Hence we do not
believe that a strict dead zone model is physically descrip-
tive of the processes occurring in mountain streams, but
Fig. 1. Experimental reach of Uvas Creek (Santa Clara County, rather that the observed 'transient storage' can be empirical-
California). Injection point and five monitoringlocationsare indicat-
ed. ly simulated using the identical equations.
The distinction between describing physical dead zones
and simulating observed transient storage is crucial to the
The model equations are determination and interpretation of the model parameters.
OC
= AD
A Ox A Ox TRANSVERSE DISTANCE (m)
1 2 3
00 448 m
+ q-b--
A
(Cr- C)+ a(Cs- C) (1)
0.75
dCs A 1 2 3 4
- a (Cs - C) (2) 0o
dt As
0.25
where
A cross-sectional
areaof the channel,m2'
D dispersioncoefficient,m2
0.5
qr lateralvolumetricinflowrate(perlength),m3s- m-1.
c, soluteconcentration in lateralinflow,mg 1-' 0 1 2 3
4o
Cs soluteconcentration in the storagezone,mg1-l'
As cross-sectional
areaof the storagezone,m2' 0.25
streamstorageexchangecoefficient,
time, s,
x distance, m. POOL
00 1 2 3 4
Equation (2) and the coupling term a(Cs - C) in (1) are :)RIFFLE
deceptivelysimple,for they embodyseveralphysicalprinci-
plesand constraints.They ideally describea systemwith the
following characteristics. 0.75
Creek is crucial to the interpretation of both the experimen- Fig. 4. Measured(1973)elevationof water surfacein Uvas Creek.
tal data and the simulations. Approximately one year after
the 1972 field experiment a detailed mapping of the study
reach was made in which stream elevations and over 100 The transient storage model will be used to simulate the
channel transects were taken. Figures 2, 3, and 4 illustrate transport of chloride in Uvas Creek resulting from the
several important physical characteristics. The dominant experimental steady injection. The model equations were
feature is the spatial variability. solvedby standardmethods, usingboth finite differenceand
Figure 2 shows a detailed top view of a lower reach of finite element (linear bases), spatial discretizations, and a
Uvas Creek and six representative channel transects. The Crank-Nicolson (time-centered) time integration. There
channel is highly irregular. It is composed of alternating were no essential differences between the finite difference
pools and riffles. The transectsshow the pools to be relative- and finite element simulations; results from the former are
ly deep and uniform compared to the riffles. However, even presented herein. The model parameters were selectedby
in the pools, the bottom profile is irregular and small zones visually determiningthe set of parameterswhich yielded the
of water are separatedfrom the main channel. In the riffles, 'best fit' to the concentration data. This approach is, of
the depth goes to 0.0 m at several locations across the course, highly subjective. However, the spatial variability
transectsin places where cobble and vegetation protrude that exists in the stream makes it highly doubtful that a more
through the flow. Figure 3 quantifies the image given by complex techniquewould yield more meaningfulresults.
Figure 2 and shows the measured channel cross-sectional Several simulations were run with only the convection-
areas. (The actual values of A are not germane, as there was dispersionmechanismsoperating, i.e., no exchangewith the
a higher flow rate in 1973.) The variability is so pronounced storagezone (a = 0.0 S-I). In thesesimulations,
lateral
as to appear as scatter. Figure 4 shows the water surface inflow qL was neglected because simulation experiments
elevation decrease over the study reach. The overall slope is showed that physically plausible values of qL did not alter
steep(0.03m m-) asexpectedfor a mountainstream.On a the overall agreement of the simulations to the data. The
small distance scale of tens of meters there are significant model then has two parameters,A and D, with Q fixed by the
differences in slope ranging from virtually flat to several determination of the chloride injection plateau at the moni-
times the overall value. toring station at 38 m. (Locations of monitoring stationsare
These measurementssupport the picture of the complex shown in Figure 1.) By using spatially uniform valuesof A
water movement one observes in the stream. The water is and D, values for these parameters were determined by
alternately shooting through riffles with the cross section searchingfor the best fit to the leading edgesof the pulsesat
expanding as the water 'dumps' into the pools and then two monitoring stations. The simulations based on the
passingthrough the pools with the cross sectioncontracting leading edges observed at 38 m and 619 m are shown in
as the water falls into the next riffle. Particularly in riffle Figures 5a and 5b, respectively. The agreementwith data is
sections, the water is in contact with a gravel and cobble bed not good, but more importantly, the characteristicsof the
and can enter easily accessible void spaces. data and the simulationare different. The significantfeatures
of the concentration data are (1) the decrease in the maxi-
mum concentration at the downstream locations, (2) the
2.0 i i I i i I i i i I i I i i i
12.0 ! i ! i i i i i i i
DISCUSSION
b,,, _.
o o o,o o o indicate
whythevalueofAs = 0.7m2isnotunrealistic.
Both
3.0 As and the exchangecoefiqcienta will then be discussedas
o they relate to parametervaluesused by other authorsfor
applicationsto other streams.
o 0.0 That the effective As for Uvas Creek is greater than the A
0800 12'00 ' ' 16'00' ' '20'00' 2400
TIME OF DAY (h) is a somewhat curious result of the simulation exercise.
12.0
There are severalpossiblemechanismsoperatingin Uvas
Creek that allow this to be plausible.Five suchmechanisms
E Observations are listed below. The order is from the most general mecha-
281
m nismsthat we would expect to exist in almost any streamto
z
o 9.o
0 38m
the mostspecificmechanisms likely to be mostimportantin
smallmountainstreams.Solutemay be temporallystoredby
z
'" 6.0 (1) turbulenteddiesgeneratedby large-scalebottom irregu-
o
larities, (2) largebut slowly movingrecirculatingzonesalong
the sidesof pools, particularlylocated immediatelydown-
3.0 stream of the entrance to a pool from a riffle section, (3)
o small but very rapidly mixing recirculatingzones located
behind flow obstructions,particularly located in riffle sec-
o 0.0 i i i
tions where cobble, smallboulders,and vegetationcommon-
0800 12'00' ' ' 16'00
' ' ' 2100
' ' ' 2400
TIME OF DAY (h) ly protrude through the flow, (4) side pockets of water
Fig. 5. Observed chloride concentrations and simulations at effectivelyactingas dead endsfor solutetransport,and (5)
three locationsfor steady injection at Uvas Creek. Injection started flow into, out of, and through a coarse gravel and cobble
at 08:30 and endedat 11:30.Data are presentedfrom this period until
23:30 when rain began to fall. (See Table 1 for parameter values). (a)
12.0
Simulation of convection-dispersionmechanisms; best fit to the
leading edge of the pulse at 38 m. (b) Simulation of convection- //"--'\ Observations
dispersionmechanisms;best fit to the leading edge of the pulse at ,.--,,\ 281m
619 m. /"
9.0
6.0
1 '//T simulations
without
storage
significant features of the stream concentration data listed
above. In particular, Figure 6b illustrates that the storage
zone acts as the neededtemporary sink for the apparentloss 3.0
_ '..
of mass. In addition to the transient storagemechanism,the
decreased maximum concentration at 281 m and 619 m could
also be attributed to a chemical loss mechanism or increased 0.0
0800 ' ' '2'00' ' ' 1'00' ' '20'00' ' ' 24oo
discharge.The pH of the streamwater was approximately8; TIME OF DAY (h)
at this pH level the authors are unaware of a chemical loss
12.0
mechanism for chloride. To simply invoke increased dis-
charge does decrease the maximum concentration but will 281 m __
not account for the clipping of the shouldersof the leading
9.0 /, Simulations
of
edges and the extent of the tails observed in the data.
Reach-by-reachvariation in the physical stream parame-
/ min-stream
_.619 concentratio
ters was used in the simulation in Figure 6. The variation is
substantial (on a relative basis). As intensive, concurrent
monitoringat severallocationsis often physicallyor riscally
difiqcult,we testedthe model's responseto settingspatially
uJ
Cross-Sectional
Concentration
Areas
at Injection Dispersion Exchange
Location Reach, Flow Q, Coefficient Stream Storage Coefficient
Simulation Figure CiNJ,mg1- m m3 s- D, m2 s-l A, m2 A, m2 a, s-
Convection-dispersion 5a 11.4 0-619 0.0125 0.12 0.30 0.00
5b 11.4 0-619 0.0125 0.48 0.45 0.00
Storage 6a, 6b 11.4 0-38 0.0125 0.12 0.30 0.00 0.00
38-105 0.0125 0.15 0.42 0.00 0.00
105-281 0.0133 0.24 0.36 0.36 0.30 X 10-4
281-433 0.0136 0.31 0.41 0.41 0.10 X 10-4
433-619 0.0140 0.40 0.52 1.56 0.45 X 10-4
Convection-dispersion 6a * * * 0.00
Storage:
variable parameters
uniform parameters 11.4 0-619 0.013 0.1 0.4 0.7 0.4 X 10-4
bed. All these mechanismsprobably operate in Uvas Creek, for surfaceroughnessalso appearswith a measureof storage
andit isplausible
thatactingtogether
theyaccount
for0.7m2 area.
of storage cross-sectional area even with a channel cross- Figure 8b is plot of dimensionlessgroupings relating a
sectionalareaof only0.4 m2. masstransfer grouping,the Nusselt number, to groupingsof
We are unaware of any a priori physical arguments that fluid, channel, and storagezone characteristics.The group-
would indicate whether or not the value for the exchange ings are defined as follows.
coefficient is plausible. The wide variety of spatial and Nbl Nusselt number, equal to k4d/ev;
temporal scales present in the above mechanisms require Re Reynolds number; equal to U4d/v;
that an effective exchange coefficient will reflect a lumping Sc Schmidt number; equal to rlev;
of many values. d/B channel aspect ratio;
Several other authors have used storage models for a wide
ds/B storage zone aspect ratio.
variety of stream conditions. Figures 8a and 8b show model
parameters determined in three other papers for nearly 40 where
applications.These figuresare revisions of figuresappearing
in the work of Pedersen [1977] and include additional data k masstransfercoefficient,
equalto ad, m s-1'
B channel and storage zone breadth, m;
points. The individual stream parameters were measured by
d channel depth, equal to A/B, m;
many different workers, reported in different formats, and
attempt to represent variable parameters by only one value.
ds storagezone depth, equal to As/B, m;
Thus there is considerable uncertainty associated with any
ev vertical dispersioncoefficient, equal to 0.0067 dU,,
m2 s-l.
individual point on these plots. Trends, not absolutevalues,
are significant.
v kinematicviscosity,m2 s-l.
Figure 8a is a plot of the ratio of storage area to channel These groupingshave been defined to be analogousto those
area,As/A, againstf-I/2 where used in chemical engineeringoperations for mass transfer
f frictionfactor,equalto 8(U,/U)2; from a flowing gas or liquid to a flat plat [Bird et al., 1960;
McCabe and Smith, 1967; Welty et al., 1969]. In forming the
U streamvelocity,equalto Q/A, m s-;
U shearvelocity,equalto (gSd)1/2,m s-;
g accelerationof gravity,m s-2' 12.0
d channel depth, m;
Observations
S water surface slope,
,,.111\ 281m
9.O Simulation with ___
Taking U, as the shear velocity, f is then a measure of '/ i uniform
parameter
s 0Iltl
m
friction [Henderson, 1966; Fischer et al., 1979]. Thackston ,
12'00
i i
16'00 2000
i
2400
0800
flow, and to a lesser extent in open channel flow, there is TIME OF DAY (h)
empirical evidence supportinga linear relationshipbetween
Fig. 7. Observed chloride concentrations and simulations at
f-l/2 and the logarithmof the ratio of lengthscalesfor two locations. In-stream concentrationsare compared for simula-
surfaceroughnessto flow depth. Figure 8a showsconsider- tions with spatially varied and uniform parameter. (See Table I for
able scatter in the data; however, the basic trend observed parameter values).
BENCALA AND WALTERS: MOUNTAIN STREAM SOLUTE TRANSPORT MODEL 723
River of northern California, U.S. Geol. Surv. Open File Rep. 78- longitudinal mixing in prismatic channels with storage in separa-
205, 1977. tion zones, Hydraul. Eng. Ser. 35, UILU-ENG-79-2010, Univ. of
Lewis, W. M., and M. C. Grant, Relationships between stream Ill. at Urbana-Champaign, 1979.
discharge and yield of dissolved substancesfrom a Colorado Valentine, E. M., and I. R. Wood, Longitudinal dispersion with
mountain watershed, Soil Sci., 128(6), 353-363, 1979. dead zones, J. Hydraul. Div. Am. $oc. Civ. Eng., 103(HY9), 975-
McCabe, W. L., and J. C. Smith, Unit Operations of Chemical 990, 1977.
Engineering, 2nd Ed., McGraw-Hill, New York, 1967. Valentine, E. M., and I. R. Wood, Experiment in longitudinal
Nordin, C. F., Jr., and B. M. Troutman, Longitudinal dispersion in dispersionwith dead zones, J. Hydraul. Div. Am. Soc. Civ. Eng.,
rivers; the persistence of skewness in observed data, Water 105(HY8), 999-1016, 1979a.
Resour. Res., 16(1), 123-128, 1980. Valentine, E. M., and I. R. Wood, Dispersion in rough rectangular
Pedersen, F. B., Prediction of longitudinal dispersion in natural channels,J. Hydraul. Div. Am. $oc. Div. Eng., 105(HY 12), 1537-
streams, Ser. Pap. 14, Inst. of Hydrodyn. and Hydraul. Eng., 1553, 1979b.
Tech. Univ. of Denmark, Copenhagen, 1977. Welty, J. R., C. E. Wicks, and R. E. Wilson, Fundamentals of
Sabol, G. V., and C. F. Nordin, Dispersion in rivers as related to Momentum, Heat and Mass Transfer, John Wiley, New York,
storage zones, J. Hydraul. Div. Am. $oc. Civ. Eng., 104(HY5), 1969.
695-708, 1978. Zand, S. M., V. C. Kennedy, G. W. Zellweger, and R. J. Avanzino,
Thackston, E. L., and A.M. Krenkel, Longitudinal mixing in Solute transportand modelingof water quality in a small stream,
natural streams,J. Sanit. Eng. Div. Am. Soc. Civ. Eng., 93(SA5), J. Res. U. $. Geol. Surv., 4(2), 233-240, 1976.
67-90, 1967.
Thackston, E. L., and K. B. Schnelle,Jr., Predictingeffectsof dead
zones on stream mixing, J. Sanit. Eng. Div. Am. Soc. Civ. Eng., (Received April 7, 1982;
96(SA2), 319-331, 1970. revised November 18, 1982;
Tsai, Y. H., and R. R. Holley, Temporal and spatial moments for accepted December 6, 1982.)