Core extraction and analyses

A series o f seven cores were extracted from the reservoir at locations along the original main channel and in the vicinity of the dam wall. Studies have shown (Price, 1999) that there is minimal dispersive m ixing of the sedim ent laden storm w ater entering the reservoir, suggesting the main zones for coarser sedim ent deposition rem ain along the original stream channel (before reservoir construction) and behind the reservoir dam. Cores were extracted in the reservoir from a platform of inflatable boats, as shown in Plate 5.2a, following a technique developed by M ackereth (1969)

w hich enables an in situ, undisturbed record o f sedim ent to be collected. The

apparatus com prises of an open ended plastic cylindrical core that is forced into the basal sedim ent by com pressed air. O f the seven sedim ent cores extracted, cores 1 and 5 were considered to show significant signs of disturbance and were excluded from further analysis.

P late 5.2 (a) Platform o f boats and coring equipment at Wyresdale reservoir (left) and (b) the extracted sediment cores (right).

When selecting a single core (Plate 5.2b) for detailed particle size and radioceasium analysis, it is important to ensure that it is representative o f the sediment depositional characteristics o f the reservoir. M agnetic susceptibility assesses the degree to which a material can be magnetised (Foster et al., 1986), expressed per unit volume o f the sample (k) , which is directly related to the concentration o f ferromagnetic grains within

the materials composition. In the laboratory, the in situ volumetric magnetic

susceptibility o f each sediment core was determined using a Barlington MS2C

magnetic susceptibility loop and the Multius software. The magnetic susceptibility

profiles for four o f the sediment cores are presented in Figure 5.11. The similarity between the shape o f the profiles indicate that representative cores have been extruded

from the reservoir. Core 2 was selected for further analysis because on close

examination, the colour and size o f the sediment at the base o f the core suggested that the pre-inundation surface had been captured, indicating that a com plete record o f depositional sediment had been acquired.

core 7 core 3

•s-air butble breal

-10 -10

magnetics

core 4 magnetic score 2

CL <D

T 3

cay bubble bre;

>ble break

-10 -10

magnetics magnetics

F ig u re 5.11 M agnetic susceptibility of four sedim ent cores extracted from W yresdale reservoir.

The particle size distribution of 0.6 cm increm ents of the core were m easured using the C oulter LS-230 laser particle sizer. The equipm ent consists o f three com ponents: a fluid unit used to introduce the sedim ent into the equipm ent for analysis; the laser

particle sizer; and a PC which runs the software controlling the operation. The

particle size distribution of the sedim ent sample is determ ined by m easuring the laser diffraction patterns of sediment particles over a size range of 0.04-2,000 pm. Prior to analysis, 10cm3 o f 0.4% Calgon was added to the sedim ent to induce dispersal w ithout peroxide digestion of organic material (Duck, 1996) and each sample was placed on a shaker overnight.

The profiles presented in Figure 5.12 show the relative proportions o f sand, silt and clay throughout the length of the core. Figure 5.13 presents the vertical profiles of the core mean wet and dry bulk density. The clay fraction rem ains approxim ately constant at a low percentage throughout the length of the core, w hilst the sand fraction

increases and silt fraction decreases towards the base. Sharp increases in the

percentage sand fraction and the dry bulk density can be noted at a num ber of discrete intervals along the core length. As the sand com ponent o f the SSL at W yresdale has been found to be flow dependent (Goodwill, 1998), these isolated, denser horizons are indicative o f high m agnitude storm events. At a depth o f below 54 cm, there is a m arked increase in the sand content. M icroscopic exam ination of this coarse sandy m aterial showed it to be well-sorted, consistent with sedim ent deposited in a fluvial

environm ent, i.e. a pre-inundation environm ent (Dr. J. Row an pers. com.).

40

Silt Clay Sand

Percentage o f Core

F ig u re 5.12 Percentage content of clay (left profile), sand (middle profile) and silt (right profile) in Core 2.

C o re B u lk D en sity 0 10 20 30 S o •5 o. u a 4 0 5 0 60 7 0 0 500 1000 1500 D en sity kg/m 3

F ig u re 5.13 W et bulk density (right profile) and dry bulk density (left profile) of Core

2

5.5.4 C o re v alid atio n

In order to assist a com parison of both profiles, the core was independently analysed for 137Cs using gam m a spectrom etry to generate an absolute dating chronology. A sharp peak in 137Cs level is observed at a depth o f 8-10 cm, which is considered to be attributable to the 1986 Chernobyl disaster. A broader peak is located at a depth of 18-20cm w hich is caused by above ground atomic weapon tests which occurred

betw een 1959-1963 (c.f. Rowan et al., 1995). No further 137Cs was detected below

this depth. The resultant chronology from this analysis is shown in Figure 5.14, at the corresponding depth of sediment.

- 1986 Chernobyl

-1963

1954

5 0 -

10 15 20 25 30 35 40

Percentage sand fraction

F ig u re 5.14 Percentage sand content of Core 2 with 137Cs dating chronology.

It is not possible to quantitatively com pare the synthetic sedim ent accretion sequence, which is expressed in daily deposition load, with the sand fraction profile of the extracted sedim ent core. However, it has been dem onstrated at W yresdale that the sand com ponent of the SSL is strongly flow dependent. As a result, a qualitative com parison o f the sedim ent core and synthetic profile can be made by plotting the two sequences adjacently and matching peaks and troughs assisted by the 137Cs dates. As shown in Figure 5.15 the two profiles show good visual agreement.

1996 1995 1986 1980/81 1976 1963 1966 On <D -a 1956 1951/54 1946 "O_<D CO 1936 1926 -191607. 1916 1.0 0.5

Simulated annual sediment accretion (g/cm2) Percentage core slice >63 microns

F ig u re 5.15 Sim ulated and observed accretion sequence

O f particular interest are the high rates of sedim ent deposition sim ulated by the model during the winters of 1916/1917, 1927, 1954, 1963 and 1980/81. There appears to be corresponding evidence of significant deposition horizons in the lake at these tim e

points, reflected by the increased percentage sand fractions in the core. The

significance of the larger storms predicted from the m odel are additionally confirm ed from independent evidence corroborated from local new spaper records. For exam ple, on 28th O ctober 1980 the model estim ated that 6.5 tonnes of sedim ent was deposited in the reservoir. On 31st O ctober 1980 the Garstang Guardian headlined with ‘W ater up to 4ft in places’; ’M6 closed for several hours’, including m aking specific references to Scorton village. The model estim ated that 6.45 tonnes o f sedim ent were deposited on 17th Septem ber 1957 and subsequently on 20th Septem ber 1957 the

Lancaster Guardian reported high flood w ater levels with the headline: ‘Wye overflow ’; ‘Scorton road to A6 closed’. The analogous trends identified between the two profiles in Figure 5.15 indicate that the D BM m odels utilised in this analysis have to some degree captured the dom inant sedim ent dynam ics o f the catchm ent-reservoir system and an encouraging reconstruction of the historical sedim ent accretion in the reservoir has been possible.

5 .6 C

This chapter has discussed the application of the D BM m ethodology to model historical sedim entation sequences at W yresdale Park catchm ent. A novel nonlinear sedim ent m odel has been used to simulate suspended sedim ent loads from rainfall records alone, providing inform ation on sedim ent transport dynam ics over much longer periods than are currently available in m ost m onitoring programm es.

Initially, an investigation o f the rainfall-flow dynam ics at the W yresdale catchm ent was m ade to help in the development of a rainfall-sedim ent model. Rainfall-flow m odelling at the seasonal scale showed evidence o f some tim e varying behaviour w ithin the m odel param eters such that a single model with seasonally adjusting param eters could potentially be estimated. However, as these relationships were not clearly defined, a model of this com plexity was deem ed to add unnecessary

uncertainty. As a result, the rainfall-flow and rainfall-sedim ent processes at

W yresdale, were m odelled based upon the com plete two years of tim e series data. The identified rainfall-sedim ent model is sim ilar in principle to the rainfall-flow model. It is com prised of a nonlinear and linear element: the nonlinear com ponent provides a

m easure of the rainfall and flow that contributes to the transport of the sediment through the catchm ent system; and a second linear TF w hich relates this transform ed rainfall to suspended sedim ent load.

This prelim inary study has dem onstrated the potential of the D B M m ethodology for producing historical tim e series. This is increasingly im portant w here SSL data is

unavailable in sufficient quantities and where it is expensive to collect. These

historical data can be used to explore m agnitude-frequency relationships and their im plications for event horizon preservation within sedim ent profiles. The TF model used to generate the synthetic sedim ent sequence assumes that the dynam ics of the contem porary catchm ent are stationary and can be extended over historical time- scales. However, the sand fraction of the sedim ent core and the average dry bulk density increase below a depth of 20 cm as shown in Figures 5.13 and 5.15, which may suggest the system is actually time variant. The prelim inary results of this study indicate that the m odelling scheme is able to reproduce the calibration data in the sense that the model has been able to predict the location o f a num ber of discrete flood horizons. However, the increasing trend in the sand fraction and bulk density of the sedim ents can not be explained by the model. The deviation o f the observed data from the stationary model predictions, particularly in the low er portion of the profile, suggests that the behavioural response o f the catchm ent system may have changed, possibly due to land use or land m anagem ent modifications. Although the results obtained in this study are not conclusive, it represents the first such exercise of its type, and in this regard, the results are very encouraging.

The original prem ise o f the experimental program m e at W yresdale was not intended to produce data for rainfall-flow or rainfall-sedim ent m odelling. N evertheless, the DBM approach has dem onstrated for the first time that even with lim ited data it is possible

to m odel the nonlinear rainfall-sedim ent relationship. It is hoped that this D BM

approach could be developed further for other better suited catchm ents for use as an additional design tool for water resources management. Ideally, to explore the full potential of this D BM approach, it w ould require a catchm ent with extended time series which exhibit greater dynamics, i.e. sam pled m ore frequently e.g. every 15 m inutes, or choosing a larger catchment where daily sam pling w ould be adequate.

The w ork sum m arised in this chapter highlights how the D B M m odelling m ethodology can be applied to the identification and estim ation of nonlinear

hydrological models. Future w ork aims to look m ore closely at the nonlinear

transform ed rainfall term which lumps the sedim ent production processes through FIS/SD PM analysis. The study also provides a good platform for other research initiatives, such as evaluating the effects of model uncertainty and the im plem entation of K alm an filter-based forecasting m odels for both discharge and sedim ent transm ission.

Ch a p t e r 6