Determining sources and transit times of suspended sediment in the Murrumbidgee River, New South Wales, Australia, using fallout and ZøPb

Transcription

1 WATER RESOURCES RESEARCH, VOL. 34, NO. 4, PAGES , APRIL 1998 Determining sources and transit times of suspended sediment in the Murrumbidgee River, New South Wales, Australia, using fallout and ZøPb P. J. Wallbrink, A. S. Murray, and J. M. Olley CSIRO Land and Water, Canberra, A. C. T., Australia L. J. Olive School of Geography and Oceanography, Australian Defence Force Academy, Campbell, A. C. T., Australia Abstract. Sediment budgets typically require an estimate of the proportional yield from erosion sources to sediments in transport and storage. This becomes increasingly difficult as catchments become larger, and erosion, storage, and deposition processes become more complex. We demonstrate how fallout radionuclides can be used to estimate the proportional contributions to sediment load, from a tripartite classification of erosion sources in a large catchment (the mid-murrumbidgee, 13,500 km2). The three major potential sources of sediment within this catchment are cultivated lands (-22% of the surface area), uncultivated pastoral lands (-78%), and the numerous channels and gullies found in this region. Concentrations of the fallout radionuclides 21øPbex and 137Cs in representative samples from each of these three sources are significantly different. Employing these values in a simple mixing model shows that the largest contribution of material is currently derived from subsoil channel/gully sources. Alterations to the suspended sediment 21øPbex signature by in situ labeling and decay are also considered. 210, Applying the model with different concentrations of Pb x (decayed as a function of residence time in channels) suggests that the mean residence time of fine-grained material within this system is 10 _ 5 years. However, differences in?cs concentrations observed between flood and low-flow sediments, and the presence of the short-lived, cosmogcnic?be, suggest that residence time of some of this fine-grained material may be of the order of only weeks to months. 1. Introduction tribution to sediments from various erosion sources, to be measured directly, rather than inferred by manipulation of Sediment budgets typically require estimates of (1) basin- erosion rate and sediment yield terms. The residence time of wide sediment yield, (2) erosion rates from upland, channel, this material in channels can also potentially be estimated. In and floodplain sources, and (3) changes to the volume and this paper we present a method for using fallout tracers to residence time of material in downslope storages [Trimble, deconvolve the proportionate contributions to suspended sed- 1983]. Quantifying these variables is usually more practicable iment from various erosion sources in a large catchment in small catchments, [Dietrich and Dunne, 1978; Swanson et al., (13,500 kn12). We also provide some constraints for the resi- 1982] where direct measurements of erosion rates and storage dence time of suspended fine-grained material within it, and terms can be made. However, budgets are difficult to construct combine this information with the known annual sediment in very large catchments because the delivery of eroded mate- yield and surface areas of the erosion sources, to estimate the rial from slopes and its storage in channels becomes more changes in specific yield from these sources over a 10-year complex, and its measurement subjecto greater uncertainties, period. with increasing scale. As a consequence, quantifying these processes typically requires careful measurements within the framework of long-term monitoring programs (see examples in the works by Ritter [1968], Hickey [1969], Dietrich et al. [1982], Trimble [1983] and Madej [1995]). A complementary approach involves measuring the properties of fallout radionuclides in sediments. This allows key components of the sediment budget, such as the proportional con Study Area The Murrumbidgee River drains one of Australia's largest (84,000 km 2) inland basins (Figure 1). Although average suspended sediment concentrations in the river (typically 50 mg L - ) are not high by global standards [Olive and Reiger, 1986; Foster et al., 1996], they are perceived by residents along the river to be increasing. Several communities use the river as a primary supply of water for domestic and irrigation purposes. Now at Nordic Laboratory for Luminescence Dating, Riso National Domestic consumption requires the water to be clarified using Laboratory, Roskilde, Denmark. aluminum sulphate, and so an increase in sediment concentra- Copyright 1998 by the American Geophysical Union. Paper number 97WR /98/97WR $09.00 tions in the river has both economic and health implications. By comparing sedigraphs with discharge hydrographs from several stations along the river, Olive et al. [1994] demonstrated 879

2 880 WALLBRINK ET AL.: SOURCES AND TRANSIT TIMES OF SUSPENDED SEDIMENT,/ Griffith L ' "x, Gundagai /./'"" ' "'"' ---'""'"" X' Narrander.... '_ -- Hay / ' -ii?:i::?:ii?:i::?:ii?: :"3( ::i::i::i::?:i::iic'- "'" """ :" ' - Burrinjuck o '--.., Io ring... i i Resei oir Figure 1. Murrumbidgee River, New South Wales, Australia, showing location of mid-murrumbidgee tributary catchments as shaded area. that the tributary catchments of the mid-murrumbidgee region been identified; the drainage density of channels and gullies up were the major source of this suspended sediment. to 6 m deep within this region has been measured at m The entire basin comprises three distinct geomorphic re- ha - [Lucas, 1996]. For catchments <10 km 2 in the upper gions: the upper, middle, and lower Murrumbidgee. The upper Murrumbidgee, Sebire [1991] measured the sediment yield region ( -20,500 km 2) is mountainous and hilly, and is sepa- from channeled landscapes as at least an order of magnitude rated from the middle region by two large storage reservoirs, higher than that from unchannelled catchments. This is con- Burrinjuck and Blowering. These dams trap most of the sedi- sistent with measurements of up to 80% contribution from ment derived from the upper region [Wasson et al., 1987], and gully sources within selected basins in the United States effectively isolate this area from the lower river. Downstream [Glymph, 1957; Bradford and Piest, 1980]. of these dams the river enters the middle section (the mid Use of Fallout Tracers 137Cs and 2]øpb Murrumbidgee, 13,500 km 2) which is characterized by rolling terrain dissected by numerous gully networks. The major trib- The source of suspended sediments can be determined by utaries of the Murrumbidgee join the river in this region; their using conservative tracers [OldfieM et al., 1979; Pearl and Wallaverage catchment size is km 2. The lower reaches of ing, 1986; Olley et al., 1993]. Burchet al. [1988] suggested that these tributaries, and the main channel of the river itself, are examination of variation in the concentrations of fallout 37Cs characterized by deposits of coarse grain to fine sand beds. In and 7Be in sediments, arising from differences in their initial steeper, headwater regions the channel grade is typically con- soil distributions, could reveal information about the origin of trolled by bed rock, separated by intermittent deposits of an- that sediment. Evidence confirming the usefulness of 37Cs, gular gravels, small cobbles and sands. Deposits of fine-grained 7Be, and 2 øpbex to describe the depth of sediment sources, and material in this system can be found in low gradient channels associated erosion processes, was presented by Wallbrink et al. contained either in stable in-stream bars, where it is trapped by [1991] and Wallbrink and Murray [1993] at Black Mountain and native plants such as the densely growing Typha orientalis and Whiteheads Creek, Australia; He and Owens [1995] used mea- Phragmites australis, or small transient bars in the channel bed. surements of 37Cs, 2 øpb, and 226Ra in a numerical mixing Downstream of Wagga Wagga the river enters the lower model to describe the contribution from cultivated lands, chanregion (50,000 km 2) where it becomes highly sinuous as it nel banks, and uncultivated lands to sediments in the River crosses the Riverine Plains to its confluence with the Murray Culm, United Kingdom. Fallout 37Cs and 2 øpb have also River, some 1500 km from its headwaters. Both flow and sed- been used as tracers of soils and sediments in a variety of other iment load of the Murrumbidgee River are highly variable. geomorphic settings [Pearl and Walling, 1986; Loughran et al., Olive et al. [1996] calculated that the sediment transported in a 1982; Wasson et al., 1987; Froelich et al., 1993; Walling and 2-week period during a one-in-12-year flood (July-August Woodward, 1995; Hutchinson, 1995]. A brief description of the 1991) accounted for -40% of the annual average sediment flux properties of these nuclides is given below. of 580,000 t at Wagga Wagga. Fallout 2 øpb (half-life 22 years; also known as 2 øpb excess, The three major land uses within these mid-murrumbidgee 2 øpbex ) is generated from the decay of 222Rn in the atmotributary catchments are cultivated land for wheat and cereals sphere. It is continually precipitated on the soil surface by (22%), pasture (59%), and forest (18%). About 20% of the rainfall and is usually defined as the excess of 2 øpb activity catchment has been assessed having a potential for sheet/rill over its parent 226Ra. Maximum concentrations of 2 øpbcx in erosion of up to 5 t ha yr-, on the basis of aerial photography soils are usually found at the surface; concentrations decrease analysis and universal soil loss equation modeling [Edwards et approximately exponentially with depth, reaching typically unal., 1989; Lucas, 1996]. Significant in-channel sources have also detectable levels at depths of -100 mm [Fisenne, 1968; Mat-

3 WALLBRINK ET AL.: SOURCES AND TRANSIT TIMES OF SUSPENDED SEDIMENT 881 thews and Potipin, 1985; Wallbrink et al., 1991]. Anthropogenic terest. The samples from these five locations (total transect 37Cs(half-life 30.2 years) is the product of aboveground nuclear weapons testing during the 1950s through the 1970s length, -50 km; n - 25) were then mixed together. The entire process was then repeated for another five transects within the [Longmore et al., 1983]. It is also distributed approximately same land use type (total n - 150). The total area of each exponentially with depth, although often with a maximum slightly below the soil surface [Walling and Bradley, 1988; Basher et al., 1995; Owens et al., 1996]. Studies in undisturbed Australian soils have found that the majority of this nuclide (>90%) is retained within the top 100 mm of the soil [Wallbrink and Murray, 1993]. Total soil inventories of 37Cs(Bq m-2) in Australi are about an order of magnitude lower than those in the northern hemisphere [Longmor et al., 1983]. The fallout of 137Cs in Australia effectively ceased by the midland use class represented using this sampling procedure was km 2. It is known that fine clays are the main particle size fraction within the suspended sediments [Olive et al., 1994], and so the <2/am fraction was separated from the mixed soil samples by a sequential process of wet sieving and particle settling, and then analyzed. Subsoil material generated from processesuch as slumping, tunneling, fluting, toppling, and scouring in channels and gullies is not exposed to direct radioactive fallout before erosion 1970s, and about half of the 2 øpbex fallout currently detectable occurs, and so contains no 2 øpbex or 137Cs. However, the in soils has occurred in the last 20 years. Consequently, the different landforms of the tributary catchments would be expected to have concentrations of these nuclides that are dischannels and gullies of the Murrumbidgee tributaries typically have near-vertical sides. As these retreat, a small amount of topsoil, labeled with 137Cs and 2 øpbex, enters the channel, tinct from one another, as a result of their differing land uses which results in measurable levels of these nuclides on the and exposure histories. detached sediments. The net radionuclide depth-averaged signature of this material (in Bq kg - ) was determined by dividing 2. Materials and Methods the tracer concentrations in the fine clays of the topsoil (defined as the region directly labeled by 2 øpbex and 137Cs, i.e., 2.1. Calculating the Contributions From the Different the top 200 mm) by the average depth of the channel bank or Land Use Types gully wall beneath it (assuming a constant clay content). The tracer concentrations of the topsoil were calculated from the For simplicity we have divided the sediment sources of the clay fraction of a series of mixed cores taken to a depth of mid-murrumbidgee into three broad categories: uncultivated -200 mm at the point locations used for characterizing the lands (78% of catchment area), cultivated lands (22% of catchland use types (n = 60). The average height of channel banks ment area), and subsoil material from channels/gullies. It is and gullies within the tributaries was estimated to be -3 m anticipated that concentrations of 137Cs and 2 øpbex on the [Lucas, 1996]. suspended sediments derived from these categories will be Sampling points for suspended sediments included the outdifferent [see Wallbrink and Murray, 1993; He and Owens, lets of all the major tributary catchments (Jugiong, Tumut, 1995]. These differences can be incorporated into a simple Kyeamba, Tarcutta, Adelong, Billabung, and Hillas) and from model to determine their relative contributions to suspended four points along the Murrumbidgee River channel from the sediment flux in the Murrumbidgee river. If Cu, Cc, Ct, and tributary confluences to Wagga Wagga. Sampling was under- Pu, Pc, Pt, representhe 137Cs and 2 øpbex concentrations taken during both low-flow and flood conditions over a period from uncultivated, cultivated, and channel bank/gully sources of 3 years ( ). Samples were obtained using a continand Cs and P s representhe respective total concentrations of uous flow centrifuge (Alfa Laval, model MAB103B). Water 137Cs and 2 øpbex on suspended sediments, then was supplied to the centrifuge at a rate of -5 L min - for ACu + BCc + CCb = Cs (1) periods up to 2-3 hours, and masses between 5 and 300 g were APu + BPc + CPb = Ps (2) recovered, depending on suspended sediment concentrations. The mean particle size of the suspended sediment was <2/am,4 + + c: (3) measured using a Sedigraph 5100 automated particle size analyzer. where A, B, and C represent the unknown relative contributions from uncultivated lands, cultivated lands, and channel banks, respectively. The approach used to determine the radionuclide concentrations for parameters C,, Cc, Ct,, Cs, Pc, P,, Pt,, and Ps is described below. All samples were oven-dried, ashed at 450øC, and analyzed by gamma spectrometry for 137Cs, 226Ra, and 2 øpb as described by Murray et al. [1987]. Small mass samples (-1-10 g) were analyzed in a high-purity germanium well detector and counted for a minimum of 170 ks. Larger mass samples (up to 250 g) were counted on HpGe planar or "n" type coaxial 2.2. Radionuclide Signatures of Cultivated Lands, detectors for a minimum of 84 ks. Unless otherwise noted, all Uncultivated Lands, and Channel Banks mean values are reported with the associated uncertainties Equations (1), (2), and (3) require the average radionuclide equivalent to one standard error. signature on material from contemporary sheet and shallow rill erosion from uncultivated and cultivated surfaces in the mid- Murrumbidgee. The material used to characterize these lands was obtained from a number of point samples which were taken at different locations and then mixed together. For example, at a given location, five separate samples (each of surface area cm 2, depth -5 mm, mass -75 g) were taken randomly over an area of -200 m 2. The process was then repeated at four additional locations, -10 km apart, along an approximately linear transect across the land use class of in- 3. Results 3.1. Radionuclide Concentrations From Different Land Use and Landform Types The average radionuclide concentrations of particulates from the three defined sources, and the suspended sediments, are given in Table 1. The highest concentrations occur in soils from uncultivated lands, with values for 37Cs(C ) of 29 _ 3 Bq kg - (range Bq kg - ) and 2 øpbex (P ) of

4 882 WALLBRINK ET AL.: SOURCES AND TRANSIT TIMES OF SUSPENDED SEDIMENT Table 1. Radionuclide Concentrations in <2 txm Material From Various Land Use and Landform Types in the Tributaries of the Mid-Murrumbidgee Catchments (channel banks), respectively. In the above calculations we have assumed that concentrations of 137Cs and 2 øpbex do not change after initial mobilization from the soil. However, it is possible that the radionuclide concentrations used to describe Concentration, Location Radionuclide Code Bq kg- n the various land uses are being modified by processes operating within the fluvial system. The potential influence of these is Uncultivated lands 137Cs C u examined in the following discussion. Cultivated fields Cc Channel/gully banks Ct, Suspended sediments Cs 2.9o Uncultivated lands 2 øpbex Pu o Discussion Cultivated fields Pc Channel/gully banks P, 2.7o Assumptions Required for This Analysis Suspended sediments Ps It is implicit in the above analysis that when sediment enters Subscripts are standard errors on the mean. a river channel it brings with it radionuclide concentrations that have been characterized by the source signature sampling method. It is further assumed that both cesium and lead be- Bq kg - (range Bq kg- ). These are comparable with have conservatively during transport and that the delivery of similar grain size material from grazed land at Whiteheads material from the sources to the tributary confluences is rapid. Creek, New South Wales, 30 _+ 2 Bq kg- for 37Cs and 300 _+ These latter points are considered here. 30 Bq kg - for 2 øpbex [Wallbrink and Murray, 1993], but are slightly higher than those from an undisturbed surface at Black Mountain, A. C. T., 18 _+ 1 Bq kg- and 240 _+ 12 Bq kg - It is required that both cesium and lead remain associated with the particles to which they were initially sorbed at the time of fallout and that no major physical or chemical alteration ( 37Cs and 2 øpbex, respectively (P. J. Wallbrink et al., Quan- occurs to them during transport. In this respect, the desorption tifying the depth sources of suspended sediment using fallout of 137Cs from sediments has been observed in saline environnuclides, submitted to Soil Science Society of America Journal, ments such as estuaries [Santschi et al., 1983] or where con- 1997)). The concentrations of 37Cs and 21øPbex on clays from the cultivated lands, C c and Pc (Table 1), are lower than from the uncultivated soils (significant at P < 0.01 for 2 øpbex and P < centrations of 137Cs have been sufficient to saturate available adsorption sites on clay lattices [Stanners and Aston, 1981]. However, the concentrations of 137Cs on soils and sediments of the Murrumbidgee are low and would be easily accommodated 0.1 for 37Cs). However, they are higher than might be ex- by adsorption sites within clay lattices [Coman and Hockley, pected from surfaces that are continuously ploughed (and thus vertically mixed). This is because cultivation in this region generally involves tillage by disc harrows; these have a series of discs perpendicular to one another which break the soil to a depth of mm and move it from side to side over a 1982; Sawhney, 1970]. The waters of the Murrumbidgee also have very low salinities, and at these low concentrations the 137Cshould remain strongly adsorbed onto the sediments [Evans et al., 1983; Singh and Gilkes, 1990]. Lead 210 has been shown to be more particle reactive than 137Cs [Santschi et al., horizontal distance of mm. There appears to be relatively 1983] and so should remain tightly bound under the conditions little vertical mixing. Analysis of sectioned cores from these areas shows that the initial profile shape of 37Cs is still paroutlined here. Thus both 137Cs and 2 øpbex are likely to behave conservatively during transport and remain associated with the tially maintained (P. J. Wallbrink and A. S. Murray, unpub- particles to which they were initially sorbed. lished data, 1996), confirming that vertical mixing is not com- However, if the particles have been resident within the tribplete. In contrast, the concentrations on conventionally utaries for a length of time that is comparable with, or longer ploughed lands tend to be uniform to the bottom of the plough than, the 2 øpb half-life of 20.2 years, then the storage of layer [Walling and Bradley, 1988]. The tracer concentrations in sediment derived from channel and gully sources (Ct,, Pt,) are lower than those of cultivated and uncultivated lands (Table 1) and also lower than the suspended sediment values. The latter concentrations range from _ 1.4 to 6.6 _+ 0.9 Bq kg- for material may be important. This is because the 2 øpbex sediment label may have been altered by both direct atmospheric fallout and radioactive decay. Conversely, the 137Cs label will be unaffected by this process (for transitimes of <30 years), because most was deposited >35 years ago. Concentration 137Cs, and -4 _+ 5 to 109 _+ 22 Bq kg - for 2 øpbex, and change (through radioactive decay) is thus simultaneous and at presumably reflect changes in contribution amounts occurring the same rate in both soils and sediments. As stated above, from the different sources within the catchment at different most fine-grained material stored in this system is contained times and conditions. In particular, the lowest concentrations either in stable in-stream bars, where it is trapped by native (of both 137Cs and 21øPbex ) were generally associated with plants, or within small transient bars in the channel bed. Analflood sediments (discussed later), implying that the contribu- ysis of aerial photographs with time reveals that the vegetated tion from in-channel sources increases under these conditions. in-stream bars are sites of net aggradation; therefore production of material from them is limited and likely to represent a 3.2. Initial Calculation of Relative Contributions of the small proportion of the total sediment yield. Consequently, Three Potential Sources although the residence time of this material is unknown, it is The concentrations of 137Cs and 2mPbex on material from the cultivated and uncultivated lands and channel/gully subsoils (Cc, Pc, Ca, Pu, Ct,, and Pt,) are different, both from one another and from the suspended sediments (Cs, Ps; see Table 1). Equations (1), (2), and (3) provide estimates for A, B, and C of 2% (uncultivated), 13% (cultivated), and 85% expected to be long, and hence their influence is not considered further. Alternatively, the small transient bars have a residence approximately equal to the recurrence interval of large discharg events occurring in the channels themselves, that is, 2-5 years (R. J. Wasson, personal communication, 1996). The effects on 2 øpbex concentrations sediments con-

5 WALLBRINK ET AL.: SOURCES AND TRANSIT TIMES OF SUSPENDED SEDIMENT E._ _ 4: E _ ; 2.t v ' a) Area (ha x 1000) - 25O 200 '-' 150 _.e 100 o 50 o (...3 of first-order channels was 600 m (I.D. Prosser, personal communication, 1996). An estimate of the cumulative increase in channel length (kilometers) within these basins was then derived by integration of the stream net lengths (Figure 2a). The average drainage density of the mid-murrumbidgee catchments using this approach is calculated to be 2.25 m ha- ; this compares well with the total measured catchment density of 2.4 m ha -1 [Lucas, 1996]. The cumulative sediment yield for the tributary channels can be derived from Wasson's [1994] power function y = 33.40'94 relating sediment yield y (t yr- ) to area A (km2). This function is based on surveys of 131 farm dams in the southern tablelands area of New South Wales and in Figure 2a has been normalized to the known annual yield at Wagga Wagga (580,000 t yr- [Olivet al., 1994]). Both the sediment yield and cumulative channel length curves are similar in shape, partly because they are both based on power functions of catchment area. Nonetheless, the relationship between them (Figure 2b) is consistent with the suggestion that sediment yield from these catchments is effectively controlled by the channels [Sebire, 1991]. The intercept in Figure 2b is negligible, and the relationship is closely approximated by Cumulative channel length (km) Figure 2. (a) Calculated drainage density (dotted line), cumulative channel length (dashed line), and cumulative sediment yield (solid line) for the mid-murrumbidgee tributaries, New South Wales. (b) Relationship between cumulative channel length and sediment yield within mid-murrumbidgee tributaries. tained within these by (1) labeling by direct fallout and (2) radioactive decay are considered separately below. Once sediment enters the channel system, it can acquire additional 2 øpbex from fallout, but not 137Cs. However, as the cumulative amount of 2 øpbex which has fallen directly into the channel increases downstream, so too does the amount of sediment transported within the channel. In principle, the net effect of these contributions can give rise to either an increase or decrease in the 21øPbex concentrations transported sediments, depending on the relative proportions of sediment and 2 øpb x delivery. Calculating the net effect of this process becomes simpler if it can be shown that sediment yield is proportional to cumulative channel length in this catchment and that sediment loss from these deeply incised channels (by overbank deposition) is a small fraction of the total. This proposition is examined by deriving an estimate of the cumulative channelength within the catchment and comparing this to an estimate of the total sediment yield from it. In order to calculate channel lengths within the catchment we first derived a drainage density (Figure 2a) for the averagesized catchments of the mid-murrumbidgee ( km2). This is approximated by a power function of the form d = aa '; where d is drainage density (m ha- ); A is stream order basin area (hectares), and a and b are constants (8.0 and -0.11, respectively). This relationship was based on combining Horton's [1945] laws of basin areas and stream segments (given by Strahler [1969]) with our measurements of basin area and length of first-order channels. It was assumed that the area ratio was 4, the bifurcation ratio was 3.2, and the mean length 25O M = ax (4) where M is the sediment yield (t yr- ), x is the channel length (kilometers), and a is a constant. From this we assume that sediment yield in the tributaries increases approximately linearly with channel length and that long-term sediment loss from these deeply incised channels is considered to be a small fraction of the total. Since the 2 øpbcx fallout rate (Bq m -2 yr - ) can be assumed spatially constant, the total amount of 2 øpbex that falls within the channels, P (Bq yr- ), must be a linear function of their length; thus we use P = bwx (5) where P is the cumulative 2 øpbex fallout (Bq yr - ) within the channel system, b is a constant, w is the channel width (4 m, assumed constant), and x is the channel distance (kilometers). Then the 2 øpbex concentration from direct fallout at any point in the channel system is P/M = bw/a (6) which is independent of time and distance. Therefore under the assumptions given here, the concentration of 2 øpbcx added directly from fallout to the channel sediments, as a function of distance, constant. The annual amount of fallout 2 øpbex in this region, b, is estimated to be 64 _+ 8 Bq m-2 yr- based on inventories of 2 øpbcx in cores taken in undisturbed soils from the tributary catchments (2100_+ 73 Bq m -2, n = 55; see Moore and Poet [1976] for a description of the methodology involved). An estimate of a, in (6), the average amount of sediment moving within the channel network per unit length, is -190 kg m- yr - based on the cumulative channel length of Figure 2a and the annual sediment load at Wagga Wagga. If we then assume the system to be in steady state, that is, the annual flux of sediment entering the channel system equals the amount leaving, then the addition of 2 øpbcx by direct deposition is 4b/a 1.35 Bq kg -. This is independent of position in the channel and residence time. However, it does not allow for radioactive decay of this added component, which will tend to reduce the concentration below this value, depending on residence time. As stated above, the derivation of this value

6 WALLBRINK ET AL.: SOURCES AND TRANSIT TIMES OF SUSPENDED SEDIMENT o E 60 o : 40.m o 2o Channel/gully sources Cultivated land 80 J... of Maximum mean length Uncultivated land residence time Residence time in tributary channels (yrs) Figure 3. Relative contribution amounts from subsoils (channel banks/gullies) and cultivated and uncultivated land sources estimated as a function of increasing residence time. Solid lines indicate the mean contributions from the different sources; dotted lines represent one standard error on the mean. Dashed line with dots represents practical limits to estimated source contributions. assumes a constant channel width of 4 m; channels of widths 1co that, on average, the contribution of suspended sediment from subsoils increases during flood events, and more importantly that this effect is observable within the time period of the flood itself, that is, within about a 4-week period. Furthermore, concentrations (up to 330 +_ 40 Bq kg - ) of the cosmogenic 7Be were measurable on five of the flood sediment samples. This nuclide is also an indicator of surface soils [Burch et al., 1988], but its half-life is only 53 days. Typical concentrations of this nuclide on fines in Australian surface soils range from 150 to 350 Bq kg - [Wallbrink and Murray, 1993, 1996]. For these flood sediments to contain the high concentrations observed, their transpor time must be of the order of only a few weeks to months. This suggests that on occasion the flood sediments must also contain a significant surface soil contribution, even though, on average, the subsoil contribution increases during floods. In combination, these 137Cs and 7Be data supporthe contention that the average transit time for some clay-size material within this system is relatively short Relative Uncertainties and Model Sensitivity The relative uncertainties for A, B, and C are estimated at +_4%, 8%, and 5%, respectively, and are shown as the dotted lines in Figure 3. These have been calculated from the sums of the squares of the deviations from the original solutions of A, B, and C when each of the parameters of Ca, Cc, Ct,, Cs, Pu, etc., are altered by plus and minus one standard error. Because A + B + C = 1, the uncertainties ona, B, and C are 2-6 m are equivalent to +_ 0.7 Bq kg -z of the calculated correlated. For example, the maximum contribution from cul- 2zøPbex concentration amount. It is thereforexpected that the tivated land is estimated to be 16%, (depending on residence 2zøPbex concentration of the suspended sediments will increase time); if we add one standard error, then the estimated conby 1.35 _+ 0.7 Bq kg - on average, while in transit within the tributary channels due to direct fallout. This represents a small correction (---6%) required for the interpretation of the observed concentration of 2 øpbex in channel sediments (Ps) in tribution from this source becomes ---24%. However, the contributions from the other sources must still sum to 100%, and so their contributions decrease and give estimates from uncultivated and channel sources of -4% and 80%, respectively. Table 1. Such mixtures are not physically possible, and so set additional If sediments remain in channels for a period comparable to bounds on uncertainties. the half-life of 2zøpb, then radioactive decay of the catchment- The sensitivity of the model to changes in Ca, Cc, derived 2zøPbex signal could be important. (The actual catch- Pu, etc., has been examined by adding and subtracting up to ment 2 øpbex signatures time = 0 do not change; they are continually replenished by fallout). The 2zøPbex values for the source terms have been corrected for radioactive decay as a three standard errors on each parameter, while holding the other values constant (Figure 4). This has been undertaken using a residence time of 5 years. In each case the changes in function of residence time by decreasing their concentrations the estimates of, l, B, and C sum to zero. The greatest changes according to the half-life of 2zøpb for a postulated number of in values of,4, B, and C are associated with variations in Pu years of "residence." A constant value (1.35 Bq kg -z yr -z) was and Ps. The model is least sensitive to variations in 137Cs then also subtracted from Ps, to account for the addition of concentrations from channel banks, Ct,, and 2zøPbex concen- 2zøPbex from direct fallout, as described above. The adjusted trations from channel banks and cultivated lands, Pc and Pt,. data have been used with (1), (2), and (3) to derive Figure 3. The estimated contributions from uncultivated (A), cultivated (B), and channel banks (C) at time = 0 years is now 0%, 16%, and 84% respectively. The overwhelming contribution of material is still predicted to be from in-channel sources. It can be seen that at a residence time of 10 +_ 5 years the predicted contribution from cultivated lands reaches 0% (Figure 3); residence times beyond this (for which a <0% cultivated input is predicted) have no physical basis. Evidence that this may be an upper limit for residence times of the fine particulates comes from our knowledge of the transient behavior of the small clay-rich sediment bars, and the difference It is possible that the model predictions of source contribution could be improved by better estimating the 137Cs and 2 øpbex concentrations (Cs and Ps) on suspended sediments. However, these estimates of Cs and Ps already represent the average of many (n = 71) samples over a 3-year period. Sediment delivery within the catchment is a dynamic process. For example, the variations observed in Ps and Cs imply that the mix of sediment from the various sources must vary over periods at least as short as the sampling regime, that is, days to months. However, the mix from the catchment sources may also systematically alter from year to year, or decade to decade, in response to different catchment management practices or between 137Cs concentrations observed on suspended sedi- regional hydrological conditions. Samples averaged over these ments carried in low-flow and flood conditions. For example, timescales would obscure such short-term shifts in source conthe average 137Cs concentration of flood sediments at Narran- tributions. dera (downstream of Wagga Wagga) is 2.1 +_ 0.1 (n = 116) Figure 4 can also be used to assess the effect of various while that from low flow is 3.0 _+ 0.3 (n = 47). This suggests assumptions in this analysis. For example, changing the aver-

7 WALLBRINK ET AL.: SOURCES AND TRANSIT TIMES OF SUSPENDED SEDIMENT 885 3O Cu " 430 I i i i i o. - '". _. - - _, Cc Pc C Pb O Cs Ps Deviations from mean in multiples of standard error Figure 4. Sensitivity of simple mixing model to changes in radionuclide concentration of input parameters. The solid line represents the contribution from uncultivated land, the dashed line that from cultivated land, and the dotted line the contribution from channel/gully sources. age channel/gully height from -3 m to -2 m is equal to about three standard errors of Pt, and Ct,, equivalent to a change in the relative contributions from the three sources of <5%. 3O O 21øPbex concentrations of --18 and 120 Bq kg -1, respectively. However, if erosion occurred from depths much greater than this (say, to 100 mm) the concentrations of Cc and Pc would decrease to values as low as Bq kg - for 137Cs and about Bq kg - for 2 øpbex; the greatereduction occurs for the latter because of its preferential distribution near the surface. Substituting these values into the sensitivity analysis, and allowing for the appropriate decay of 2 øpbex, alters the relative contributions amounts by less than 5%. The largest contribution (>80%) still occurs from gully and channel sources. The steady state addition of 21øPbex to Ps by direct fallout was calculated previously to be Bq kg-. Three times the standard error of the Ps value (i.e., _+4; Table 1) is equivalent to a change in this value of _+ 12 Bq kg -. This is at least a factor of 8 greater than the amount calculated from direct fallout. It is also significantly greater than any estimated uncertainties associated with the derivation of that value Specific Sediment Yield From Channel Banks and Cultivated and Uncultivated Lands The tracer-based estimates of relative contribution from the various land uses and landforms of the tributary catchments can be combined with the known annual sediment yield at Wagga Wagga to provide a quantitative estimate of specific yield from each source (assuming there is no gross increase or decrease in storage within the channels). These are summarized in Table 2 against years of residence and converted to specific sediment yields from the known surface area of each land use type. For example, the annual yield of sediment from uncultivated sources (-27,100 t yr- ), at 5 years residence, provides a specific sediment yield of t ha- yr, which is about a factor of 6 less than that from cultivated lands of t ha -1 yr - at the same time. These loss rates represent material actually delivered to the stream network. Presumably, a Conversely, if all the channel banks on the floodplains con- larger amount of soil is mobilized, but remains in storage on tained elevated concentrations of 137Cs and 21øPbex, say, dou- the slope [Walling, 1983]. ble the upland amounts (1.2 and 5.4 Bq kg -1, respectively), The estimated loss from channel/gully sources of 0.37 t ha- then this would increase the estimated contribution from chan- yr- at 5 years residence is consistent with values of 0.2 to 0.5 nel banks and uncultivated sources by 4% and 1%, respectively, and decrease that from cultivated lands by 5%. It was also assumed that surface erosion from the cultivated lands occurs to an average depth of <5 mm, yielding 137Cs and (m 3 ha- yr-l) from banks/gullies in some European loam and loess soils, although they are about an order of magnitude lower than that ( m 3 ha - yr - ) from ephemeral gullying in some Mediterranean environments [Poesen et al., 1996]. Table 2. Estimated Contribution Amounts and Sediment Yields From Subsoils (Channel Banks/Gullies), and Cultivated and Uncultivated Lands for Different Residence Times in Tributary Channels of the Mid-Murrumbidgee Uncultivated (A) Cultivated (B) Channel/Gully Banks (C) Residence Specific Specific Specific Time, Contribution, Yield, Yield, Contribution, Yield, Yield, Contribution, Yield, Yield, years % t x 103 t ha- yr- % t x 103 t ha- yr- % t x 103 t ha- yr Annual sediment load is -580,000 t yr -, total catchment area is -13,500 km 2, total channelength km. Specific yield from channel/gully sources is calculated using total mid-murrumbidgee catchment area of 13,500 km 2.

8 886 WALLBRINK ET AL.: SOURCES AND TRANSIT TIMES OF SUSPENDED SEDIMENT It is also equivalent to a minimum annual linear lateral retreat of - 18 mm (depending on clay content of the banks) over the Murrumbidgee. Improvements to the manuscript were suggested by J. Field and A. Scott, as well as three anonymous referees. period of suspended sediment measurement. However, this does not take into account the significant crenulations observed on gully walls within the tributaries. A crenulation ratio References of 2 for these features would decrease the estimated linear Basher, L. R., K. M. Matthews, and L. Zhi, Surface erosion assessment in the South Canterbury Downlands, New Zealand, using 137Cs lateral retreat by half. distribution, Aust. J. Soil Res., 33, , Bradford, J. M., and R. F. Piest, Erosional development of valley bottom gullies in the upper midwestern United States, in Thresholds 5. Conclusions in Geomorphology, edited by D. R. Coates and J. D. Vitek, pp , Allen and Unwin, Winchester, Mass., By using the fallout tracers 37Cs and 2 øpbex we have dem- Burch, G. J., C. J. Barnes, I. D. Moore, R. D. Barling, D. J. Mackenzie, onstrated that the dominant source of material (>80%) to and J. M. Olley, Detection and prediction of sediment sources in suspended sediments in channels of the mid-murrumbidgee is catchments: Use of 7Be and 137Cs, paper presented at Hydrology and Water Resources Symposium, Inst. of Eng., Aust. Natl. Univ., channel banks and gully walls. This is consistent with Neil and Canberra, Fogarty's [1991] conclusion that channel sediments dominate flux at smaller scales in the catchments' upper region, and with Coman, R. N.J., and D. E. Hockley, Kinetics of caesium sorption on illite, Geochim. Cosmochim. Acta, 56, , Osborn and Simanton [1989], who describe the substantial in- Dietrich, W. E., and T. Dunne, Sediment budget for a small catchment in mountainous terrain, Z. Geomorphol., 29, , fluence on yield by subsoils in U.S. catchments where channel- Dietrich, W. E., T. Dunne, N. F. Humphrey, and L. M. Reid, Coning occurs. Our data also suggested that the mean residence struction of sediment budgets for drainage basins, U.S. For. Serv. times of fine grain suspended sediments within the mid- Gen. Tech. Rep., PNW-141, 5-23, Murrumbidgee system is 10 _ 5 years. We presume this ma- Edwards, K., R. J. Blong, and O. P. Graham, Identification of sediment sources: Implications for erosion control, Aust. J. Soil Water Conterial to be stored within small transient bars commonly observ., 2(1), 18-27, served in the channel bed. Some fine-grained material remains Evans, D. W., J. Alberts, and R. A. Clarke, Reversible ion-exchange trapped within vegetated channel bars. Evidence that resi- fixation of caesium-137 leading to mobilisation from reservoir seddence times for some of the fines in this river could be of the iments, Geochim. Cosmochim. Acta, 47, , order of weeks to months, rather than years to decades, come Fisenne, I. M., Distribution of 21øpb and 226Ra in soil, Rep. UCRL , pp , U.S. Dep. of Energy, Washington, D.C., from (1) changes in the mean 37Cs concentrations between Foster, I. D., P. N. Owens, and D. E. Walling, Sediment yields and suspended sediments from flood and low-flow waters, and (2) sediment delivery in the catchments of Slapton Lower Ley, South the presence of the short-lived tracer 7Be in some flood water Devon, U. K., FieM Stud., 8, , sediments. Given some simplifying assumptions, we have also Froelich, W., D. L. Higgitt, and D. E. Walling, The use of caesium-137 to investigate soil erosion and sediment delivery from cultivated shown that sediment yield can be expressed as a linear function slopes in the Polish Carpathians, in Farm Land Erosion: In Temperof cumulative channel length in the mid-murrumbidgee. As a ate Plains Environment and Hills, edited by S. Wicherek, pp first approximation of sediment sources, this appears to be an 283, Elsevier Sci., New York, important finding; however its veracity will depend upon Glymph, L. M., Importance of sheet erosion as a source of sediment, Eos Trans. AGU, 38, , more formal analysis of the drainage network of this catch- He, O., and P. Owens, Determination of suspended sediment provement. Further development of this work should also include nance using caesium-137, unsupported lead-210 and radium-226: A the possible influence on tracer properties of longer-term stor- numerical mixing model approach, in Sediment and Water Quality in age sites in the system and of other sources not directly con- River Catchments, edited by I. D. Foster, A.M. Gurnell, and B. W. nected to the drainage network, such as discontinuous gullies Webb, pp , John Wiley, New York, Hickey, J. J., Variations in low water streambed elevations at selected and roads. In summary, we have shown how analysis of fallout gaging stations in northwestern California, U.S. Geol. Surv. Water nuclides in a large catchment has allowed us to (1) quantify the Supply Pap., 1879-E, 33 pp., proportional contributions from various sediment sources, (2) Horton, R. E., Erosional developments of streams and their drainage calculate their respective specific yields, and (3) estimate basins; hydrophysical approach to quantitative morphology, Geol. boundaries on sediment transit times in channels. These vari- Soc. Am. Bull., 56, , Hutchinson, S. M., Use of magnetic and radiometric measurements to ables are essential in the construction of any sediment budget investigat erosion and sedimentation in a British upland catchment, that aims to describe the fate of eroded sediments and their Earth Surf. Processes Landforms, 20, , downstream impacts in large catchment such as the Murrumbidgee. Longmore, M. E., B. M. O'Leary, C. W. Rose, and A. L. Chandica, Mapping soil erosion and accumulation with the fallout isotope caesium-137, Aust. J. Soil Res., 21, , Loughran, R. J., B. L. Campbell, and G. L. Elliott, The identification and quantification of sediment sources using Cs-137, in Recent De- Acknowledgments. This project was primarily funded through CSIRO's Land and Water Care program. Additional funding was provided by the Murray Darling Basin Commission and the New South Wales Environment Protection Authority. In-kind support was provided by New South Wales Department of Land and Water Conservation. The authors would also like to thank G. Hancock, G. Caitcheon, C. Smith, M. Rake, and C. Turley for their assistance in collecting suspended sediment samples during flood events in this river system. C. Saunders, C. Smith, and R. Rake processed the suspended sediment and catchment soils for radionuclide analyses. I. Prosser, Cooperative Research Centre for Catchment Hydrology, provided the area and bifurcation ratio used in the stream net analysis. R. J. Wasson, Australian National University, and Peter Fogarty, Department of Conservation and Land Management, helped provide a description of the likely fate and storage sites of fine-grained sediment in the midvelopments in the Explanation and Prediction of Erosion and Sediment YieM, IAHS Publ., 137, , Lucas, S., Proposals for integrated management of soil erosion and related land degradation; Mid Murrumbidgee catchment, vol. 2, Statistical information for selected subcatchments, internal report, Dep. of Conserv. and Land Manage., Wagga, Madej, M. A., Changes in channel-stored sediment, Redwood Creek, northwestern California, , U.S. Geol. Surv. Prof. Pap., , Matthews, K. M., and K. Potipin, Extraction of fallout 21øpb from soils and its distribution in soil profiles, J. Environ. Radioact., 2, , Moore, W. S., and S. E. Poet, Pb210 fluxes determined from 2 øpb and 226Ra soil profiles, J. Geophys. Res., 81, , Murray, A. S., R. Marten, A. Johnston, and P. Martin, Analysis for

9 WALLBRINK ET AL.: SOURCES AND TRANSIT TIMES OF SUSPENDED SEDIMENT 887 naturally occurring radionuclides at environmentalevels by gamma of 137Cs with suspended sediments in estuarine and coastal environspectrometry, J. Radiat. Nucl. Chem., 115, , ments, paper presented at International Symposium on the Impacts Neil, D., and P. Fogarty, Land use and sediment yield on the Southern of Radionuclide Releases Into the Marine Environment, Int. At. Tablelands of NSW, Aust. J. Soil Water Conserr., 4(2), 33-39, Energy Agency, Vienna, Oldfield, F., T. A. Rummery, R. Thompson, and D. E. Walling, Iden- Strahler, A. N., Quantitative analysis of erosionalandforms, in Phystification of suspended sediment sources by means of magnetic mea- ical Geography, pp , John Wiley, New York, surements, Water Resour. Res., 15, , Swanson, F. J., R. J. Janda, T. Dunne, and D. N. Swanston, Sediment Olive, L. J., and W. A. Reiger, Low Australian sediment yields--a budgets and routing in forested drainage basins, U.S. For. Serv. Gen. question of inefficient delivery, in Drainage Basin Sediment Delivery, Tech. Rep., PNW-141, 165 pp., IAHS Publ., 159, , Trimble, S. W., A sediment budget for Coon Creek Basin in the Olive, L. J., J. M. Olley, P. J. Wallbrink, and A. S. Murray, Spatial driftless area, Wisconsin, Am. J. Sci., 283, , variation in suspended sediment transport in the Murrumbidgee Wallbrink, P. J., and A. S. Murray, Use of fallout radionuclides as River, New South Wales, Australia, in Variability in Erosion and indicators of erosion processes, J. Hydrol. Processes, 7, , Stream Transport, edited by L. J. Olive, R. Loughran, and M. Kesby, IAHS Publ., 224, , Wallbrink, P. J., and A. S. Murray, Distribution and variability of 7Be Olive, L. J., J. M. Olley, P. J. Wallbrink, and A. S. Murray, Down- in soils under different surface cover conditions and its potential for stream patterns of sedimentransport during floods in the Murrum- describing soil redistribution processes, Water Resour. Res., 32, 467- bidgee River, NSW, Australia, Z. Geomorphol. N. F., Suppl. Bd., 476, , , Wallbrink, P. J., J. M. Olley, and A. S. Murray, Mechanisms controlling Olley, J. M., A. S. Murray, D. M. McKenzie, and K. Edwards, Identi- radionuclide concentration variations found within transported sedfication of sediment sources in a gullied catchment using natural and iments, paper presented at Environmental Radioactivity Workshop, artificial radioactivity, Water Resour. Res., 29, , LESE, Papeete, Tahiti, Sept. 2-6, Osborn, H. B., and J. R. Simanton, Gullies and sediment yield, Range- Walling, D. E., The sediment delivery problem, J. Hydrol., 65, , lands, (2), 51-56, Owens, P. N., D. E. Walling, and Q. He, The behaviour of bomb- Walling, D. E., and S. B. Bradley, The use of caesium-137 measurederived caesium-137 fallout in catchment soils, J. Environ. Radioact., ments to investigate sediment delivery from cultivated areas in 32, , Devon, U. K., in Sediment Budgets, IAHS Publ., 174, , Peart, M. R., and D. E. Walling, Fingerprinting sediment source: The Walling, D. E., and J. C. Woodward, Tracing sources of suspended example of a drainage basin in Devon, U. K., in Drainage Basin sediment in river basins: A case study of the River Culm, Devon, Sediment Delivery, IAHS Publ., 159, 41-54, UK, Mar. Freshwater Res., 46, , Poesen, J. W., K. Vandaele, and B. Van Wesemael, Contribution of Wasson, R. J., Annual and decadal variance in sediment yield, in gully erosion to sediment production on cultivated lands and range- Variability in Erosion and Stream Transport, edited by L. J. Olive, R. lands, in Erosion and Sediment Yield: Global and Regional Perspec- Loughran, and M. Kesby, IAHS Publ., 224, , tives, edited by D. E. Walling, and B. W. Webb, IAHS Publ., 236, Wasson, R. J., R. L. Clark, P.M. Nanninga, and J. Waters, 2 øpb as a , chronometer and tracer, Burrinjuck Reservoir, Australia, Earth Surf. Ritter, J. R., Changes in the channel morphology of Trinity River and Processes Landforms, 12, , eight tributaries, California, , U.S. Geol. Surv. Open File Zucker, C. L., C. R. Olsen, I. L. Larsen, and N.H. Cutshall, Invento- Rep., 60 pp., ries and sorption-desorption trends of radiocesium and radiocobalt Roberts, J., and G. Sainty, Listening to the Lachlan, pp. 1-98, Sainty in James River Estuary sediments, Environ. Geol. Water Sci., 6(3), and Assoc., Griffith, N. S. W., Australia, , Santschi, P. H., Y. Li, D. M. Adler, M. Amdurer, J. Bell, and U. P. Nyffeler, The relative mobility of natural (Th, Pb and Po) and fallout A. S. Murray, Nordic Laboratory for Luminescence Dating, Riso (Pu, Am, Cs) radionuclides in the coastal marine environment: Re- National Laboratory, P.O. Box 49, DK-4000 Roskilde, Denmark. sults from model ecosystems (MERL) and Narragansett Bay, L. J. Olive, Australian Defence Force Academy, Northcott Drive, Geochim. Cosmochim. Acta, 47, , Campbell, ACT, 2600, Australia. Sawhney, B. L., Potassium and caesium ion selectivity in relation to J. M. Olley and P. J. Wallbrink, CSIRO Land and Water, P.O. Box clay mineral structure, Clays Clay Miner., 18, 47-52, , Canberra 2601, A. C. T., Australia. ( peter.wallbrink Sebire, A., Protecting Lake Burley Griffin water quality through sion control, Aust. J. Soil Water Conserv., 4(3), 19-26, Singh, B., and R. J. Gilkes, Sorption-desorption behaviour of caesium in some Western Australian soils, Aust. J. Soil Res., 28, , (Received November 28, 1996; revised November 5, 1997; Stanners, D. A., and S. R. Aston, Factors controlling the interactions accepted November 28, 1997.)