CORRELATING EROSION THRESHOLD AND PHYSICOCHEMICAL PROPERTIES OF NATURAL COHESIVE SEDIMENTS

Abstract: The erosion threshold of cohesive sediments is an important parameter, because in water systems fine-grained particles are the major transport agent of many contaminants. However, the erosion behaviour of cohesive sediments, is not yet well understood. To gain some insight into the processes governing the stabilisation of cohesive sediments, various sediment properties were analysed for their correlation with the erosion threshold of natural fine-grained bed sediments, which was determined as a function of sediment depth. Sediment consolidation, grain-size, and biological slimes (EPS) were found to be significant factors influencing erosion resistance. The combined effect of these mechanisms could either be considered with the consistency index as one overall parameter or by combining the water content, the percentage of grains smaller than 20 µm, and the content of colloidal carbohydrates in a multiple linear regression model. Hence, the results highlight that the erosion threshold of fine-grained cohesive sediments is not governed by one single mechanism, but rather by a combination of physical, chemical, and biological processes which are active at the same time. For a better understanding of the erosion behaviour of cohesive sediments it seems therefore advisable to conduct further experiments with both, well defined artificial and undisturbed natural sediments.

1 INTRODUCTION

Most water contaminants are primarily transported in association with fine-grained cohesive sediments (FCS). From an environmental management perspective it is therefore of utmost importance to be able to predict the erosion threshold of FCS (Haag et al., 2000) . However, whilst the erosion behaviour of non-cohesive coarse-grained sediments is well understood and can be predicted since the landmark work of Shields (1936) , up to date no universally valid theory for the erosion behaviour of FCS exists. The objective of the present study is to shed some light on the mechanisms that control the erosion threshold of natural FCS, by correlating experimentally determined critical erosion shear stresses (τ_c,e) with various biological and physicochemical sediment properties.

2 BACKGROUND

Erosion resistance of FCS is mainly controlled by inter-particle forces. The strength of these forces is governed by biological sediment parameters and electrochemical properties of sediment and pore water. Depending on species assemblage and organism density bioturbation by macro-biota (e.g. worms) may cause both, increasing or decreasing erosion resistance. Micro-organisms (algae, bacteria, fungi) however, generally increase sediment stability. The stabilising effect of surface films (mats) of algae or bacteria has been investigated quite frequently for both, artificial and natural sediments. In these investigations it was found that biological slimes, the so called extracellular polymeric substances (EPS) play a major role in biological sediment stabilisation, by making the surface smoother and interconnecting particles. EPS or as a surrogate colloidal carbohydrate concentration has been reported to correlate with τ_c,e (e.g. Paterson, 1997) .

The strength of electrochemical bonding is primarily governed by the surface charge density of the sediment particles and the valencies and concentration of the ions present in the pore water solution (Sposito, 1984) . Even the effect of EPS, which are macromolecular polyelectrolytes, can partially be interpreted as electrochemical. Charge density tends to increase with increasing organic carbon content and decreasing grain size. Also mechanical sediment consolidation and the decrease of porosity (increase of bulk density) does enhance the strength of electrochemical inter-particle bonding. Consequently, a variety of chemical and physical parameters, including the sodium adsorption ratio, organic carbon content, grain-size distribution, water content, and bulk density have been proposed as master variables for the erosion behaviour of FCS (for an overview see Haag et al., 1999b) .

Most of the erosion investigations correlating erosion thresholds to sediment properties have been performed with artificial or disturbed sediments. Although Lau and Droppo (2000) showed that the conditions under which FCS were deposited play a major role for their erosion behaviour, cohesive sediments with natural layering and state of consolidation have rarely been investigated. Up to date most investigations, in particular those dealing with biological stabilisation, have focused on the sediment surface. Less attention has been paid to deeper sediment layers which have been subject to long term consolidation. However, since EPS are excreted by almost all micro-organisms (including anaerobic bacteria) they can also be expected to control the bio-stabilisation of deeper sediment layers. Also most studies concentrated on one single sediment property or stabilising mechanism, not considering the combined effect of several mechanisms which is likely to be important for natural sediments.

3 METHODS AND MATERIALS

During the years 1997 and 1998 several undisturbed sediment cores (13.5 cm in diameter and 60 to 130 cm in length) were taken in the backwater region of a reservoir of the River Neckar in south-west Germany. At least two parallel cores were sampled in close proximity to each other. Vertical profiles of bulk densities of all cores were measured non-intrusively by using a γ-ray-densitometer. Comparing density profiles of parallel cores allows to check whether they have similar vertical profiles of sediment properties. For details about sampling procedures and density measurements see Haag et al. (1999a) . If, on the basis of the density profiles, parallel cores were considered to be similar, one of them served to experimentally determine the critical shear stress of mass erosion (τ_c,e) as a function of sediment depth. Erosion experiments were carried out in the SETEG-system (Kern et al., 1999) . The second one of the parallel cores was cut into sections. In an attempt to gain almost uniformly textured sub-samples, the core was cut at depths of significant bulk density changes. The resulting 147 sub-samples of eight cores were homogenised and the following sediment properties were determined for all of them: grain size fraction smaller than 20 µm (d<20 [%]), water content relative to dry mass (WC [%]), total organic carbon content (TOC [%]), bulk density (ρ [g/cm³]), and cation exchange capacity (CEC [mmol_c/kg]). For the sub-samples of two cores, liquid and plastic limits, plasticity and consistency indexes, and the content of colloidal carbohydrates (as an indicator for the EPS content) were determined. Detailed record of the procedures used to gain these sediment properties is given by Haag et al. (1999b) .

4 RESULTS AND DISCUSSION

With a mean TOC, clay and silt content of 5, 22 and 61%, respectively, the sediments were found to be rather fine-grained and rich in organic carbon on average. However, few layers with coarse particles, low TOC and consequently increased bulk densities could be detected. They are most probably the result of sedimentation during the receding limbs of flood events (Haag et al., 2000) . Often these layers were also characterised by sudden decreases of τ_c,e in the corresponding parallel core, indicating the predominance of non-cohesive particles (Fig. 1). Because of this clearly non-cohesive character the corresponding data of these layers were excluded from the data pool prior to further analysis. After excluding clearly non-cohesive layers, the remaining values of τ_c,e were ascribed to the corresponding sediment property values of the parallel cores.

Fig. 1 Vertical profiles of bulk density and of τ_c,e of a sediment core from the River Neckar.

Correlation analyses of

and the eleven sediment properties given in Table 1 were performed. The most commonly used correlation coefficient after Pearson (r) assumes a linear relation. Kendall’s (tau) and Spearman’s (rho) correlation coefficients are based on rank analyses and are therefore also applicable to non-linear relationships. The absolute values of the three correlation measures are not directly comparable, but their significance (p, probability of error) is (Helsel and Hirsch, 1992).

Table 1 shows the computed correlation coefficients and their significance. Sediment depth, bulk density, the fraction of fines, and the consistency index are positively correlated to

with high significance. Also the EPS content (measured as colloidal carbohydrates) seems to be positively correlated with

. However, the probability of error is about 10%, which might be due to the small number of values (32) that were available for this analysis. The water content is negatively correlated with

. The negative correlation of

with TOC and CEC is very uncertain, and cannot be explained by any of the mechanisms causing cohesion.

Table 1 Correlation coefficients of

with sediment properties and their significance.

Sediment property	depth	ρ	d<20	WC	TOC	CEC	W_L	W_P	I_P	I_C	EPS*
number of pairs n	119	118	118	119	118	119	34	34	34	36	32
Kendall’s tau	0.34	0.18	0.19	-0.23	-0.01	-0.06	0.17	0.12	0.18	0.50	0.20
p [%]	<0.1	0.4	0.2	<0.1	90.9	30.6	15.0	32.8	14.2	<0.1	10.8
Spearman’s rho	0.49	0.26	0.30	-0.33	-0.02	-0.09	0.25	0.18	0.26	0.66	0.28
p [%]	<0.1	0.5	0.1	<0.1	83.4	32.1	16.0	30.2	13.8	<0.1	11.9
Pearson’s r	0.47	0.22	0.29	-0.34	-0.02	-0.13	0.20	0.20	0.20	0.64	0.31
p [%]	<0.1	1.7	0.1	<0.1	86.8	17.0	25.3	26.8	26.3	<0.1	8.3

The correlation with depth, density, and water content can be attributed to the effect of sediment consolidation, which increases with depth and causes a higher bulk density and lower water content. On the other hand also a grain size effect is still apparent: a higher fraction of d < 20 µm does have a positive effect on sediment stability. Since the bulk density is negatively correlated with d<20, the grain size effect may obscure part of the consolidation effect and vice versa.

Despite the significance of several correlations the magnitude of Pearson’s r is rather low, indicating that the percentage of variance of

being explained by linear regression against one single parameter is well below 25% except for the consistency index. The strong correlation between erosion threshold and consistency index is not surprising, because I_C is an overall measure of the cohesion within the sediment, comprising the effects of various mechanisms. The relationship however is clearly not linear. As shown in Figure 2, rather the logarithms of

are linearly correlated to the consistency index. The regression which could be fitted best to the data is also given in Figure 2 along with the 95%-prediction interval. Note that the regression is based on two sediment cores (36 samples) only and that

is plotted in a logarithmic scale. As indicated by the 95%-prediction interval the scatter of the data around the regression line is rather high, with a standard error of the estimate of 1.6 N/m².

The residuals of the regression given in Figure 2 were not significantly correlated with any of the remaining ten sediment properties, indicating that I_C already comprises most of the influences of the other sediment parameters. Consequently the remaining uncertainty cannot be ascribed to the influence of any single sediment property.

Since the consistency index comprises the effects of various mechanisms that sum up to the total cohesion and erosion resistance of the sediment, it is not possible from the given relationship to gain further insight into the mechanisms underlying sediment stabilisation. Therefore, on the basis of the correlation analyses, as an alternative a multiple regression approach was considered. As pointed out above sediment depth, bulk density, and water content are measures of sediment consolidation; d<20 is a measure of the influence of grain size on

(influencing the chemical bonding); EPS is a measure of the biological sediment stabilisation. Therefore, a regression model combining one of the three consolidation parameters with d<20 and EPS should be applicable to predict the erosion threshold. Plots of

versus the above mentioned sediment properties indicated linear relationships. Hence a multiple linear regression model was chosen. The model combining WC with d<20 and EPS yielded the highest r² and the lowest standard error, and was therefore chosen as most appropriate. In Figure 3 the performance of the multiple linear regression model is compared with that of the single parameter model only including I_C. Even though the multiple regression model includes three parameters, the standard error is only slightly reduced. For both statistical models the data scatter around the line of perfect agreement increases with

. Thus, higher erosion thresholds are associated with higher prediction uncertainties. The remaining uncertainty is likely owing to inaccuracies in determining

and sediment properties, the influence of additional parameters not considered here, and probably most importantly to errors arising from considering two parallel cores instead of one single core.

Considering the scatter and the resulting standard errors neither of the equations in Figure 3 can be used to accurately predict the erosion threshold of Neckar River sediments or of cohesive sediments in general, but they may well serve as a rough estimate. The results also highlight the importance of considering the influence of several mechanisms which contribute to erosion resistance of FCS. Sediment consistency and stability both arise from the combined effects of biological, chemical, and physical processes. Hence, the integral parameter I_C is the only single parameter considered in this study which can be used to explain an appreciable part of the variance of

. Similarly high degrees of correlation might be gained with rheometric parameters (Otsubo and Muraoka, 1988) . In contrast, none of the single properties which can clearly be interpreted as biological, chemical, or physical stabilisation parameter explains more than ca. 20% of the total variance of

(Table 1). Only a combination of parameters taking all three mechanisms into account, performs similarly well in explaining the variance of

as does I_C alone.

5 CONCLUSIONS

The erosion investigations conducted in this study clearly demonstrate, that the critical shear stress of erosion for natural cohesive sediments can only be explained as the combined effect of various mechanisms. The erosion threshold is highly correlated to the consistency index of the sediments. However, the consistency itself is the product of other parameters or processes, and therefore this correlation does not give further insight into the actual cause of erosion resistance. Considering three parameters which can be ascribed to biological, chemical, and physical (consolidation) stabilisation mechanisms yields similarly good predictions as the consistency index. This allows one to interpret the overall erosion resistance as an effect of a combination of all three processes.

Hence, this study shows, that investigating natural sediments (with depth) does very well serve to elucidate the mechanisms which govern the erosion resistance of cohesive sediments. On the other hand, experiments with artificial sediments allow to gradually change single sediment properties and to evaluate the effects of these changes on the erosion behaviour. However, it is difficult to account for all possible influences and to simulate long term processes such as consolidation with artificial systems. Therefore, in order to come closer to a unified theory of erosion resistance for fine-grained cohesive sediments, a combination of both approaches seems to be most promising.

The authors would like to thank the federal state of Baden-Württemberg/Germany for financing this work; grant PW 96 182. We are also indebted to Dr. Ulrich Kern for many invaluable discussions on the topic.

Haag, I., Kern, U. and Westrich, B., 1999a. Comprehensive assessment of contaminated fluvial sediments - erosion risk and ecological hazard. In: G.H. Jirka (Editor), Proceedings of XXVIII^th IAHR congress, Graz.

Haag, I., Kern, U. and Westrich, B., 1999b. Mobilität von Schadstoffen in den Sedimenten staugeregelter Flüsse - Dynamik und Bilanzierung von Schwebstoffen in einer Stauhaltungskette. 99/11 (VA 21), Universität Stuttgart, Institut für Wasserbau.

Haag, I., Kern, U. and Westrich, B., 2000. Erosion investigation and sediment quality measurements for a comprehensive risk assessment of contaminated aquatic sediments. Science of the Total Environment, (in press).

Helsel, D.R. and Hirsch, R.M., 1992. Statistical Methods in Water Resources. Studies in Environmental Science. Elsevier, Amsterdam, 522 pp.

Kern, U., Haag, I., Holzwarth, M., Schürlein, V. and Westrich, B., 1999. Ein Strömungskanal zur Ermittlung der tiefenabhängigen Erosionsstabilität von Gewässersedimenten: das SETEG-System. Wasserwirtschaft, 89(2): 72-77.

Lau, Y.L. and Droppo, I.G., 2000. Influence of antecedent conditions on critical shear stress of bed sediments. Water Research, 34(2): 663-667.

Otsubo, K. and Muraoka, K., 1988. Critical shear stress of cohesive bottom sediments. Journal of Hydraulic Engineering, 114(10): 1241-1256.

Paterson, D.M., 1997. Biological mediation of sediment erodibility: Ecology and physical dynamics. In: N. Burt, Parker, R. and Watts, J. (Editors), Cohesive Sediments. J. Wiley and Sons, Chichester, pp. 215-230.

Shields, A., 1936. Anwendung der Ähnlichkeitsmechanik und der Turbulenzforschung auf die Geschiebebewegung, Preussische Versuchsanstalt für Wasser-, Erd- und Schiffbau.

Sposito, G., 1984. The Surface Chemistry of Soils. Oxford University Press, New York.