Linking land use, erosion and sediment yields in river basins

Transcription

1 Hydrobiologia 410: , J. Garnier & J.-M. Mouchel (eds), Man and River Systems Kluwer Academic Publishers. Printed in the Netherlands. 223 Linking land use, erosion and sediment yields in river basins D. E. Walling Department of Geography, University of Exeter, Exeter, EX4 4RJ, U.K. Key words: erosion, sediment yield, sediment storage, sediment budget, floodplain sedimentation. Abstract Results obtained from erosion plots and catchment experiments provide clear evidence of the sensitivity of erosion rates to land use change and related human activity. Evidence for the impact of land use on the sediment yields of world rivers is less clear, although examples of rivers where sediment yields have both increased and decreased in recent decades can be identified. The apparent lack of sensitivity of river sediment loads to land use change reflects, at least in part, the buffering capacity associated with many river basins. This buffering capacity is closely related to the sediment delivery ratio of a river basin, in that basins with high sediment delivery ratios are likely to exhibit a reduced buffering capacity. Investigations of the impact of land use and related human activity on sediment yields should consider the overall sediment budget of a catchment rather than simply the sediment output. Information on the sediment budget of a drainage basin is difficult to assemble using traditional techniques, but recent developments in the application of fingerprinting techniques to establish sediment sources and in the use of environmental radionuclides, such as caesium-137 and lead-210, to document sediment storage offer considerable potential for providing such information. Sediment storage within a river basin can give rise to environmental problems where sediment-associated pollutants accumulate in sediment sinks. The accumulation of phosphorus on river floodplains as a result of overbank sedimentation can, for example, represent an important phosphorus sink. The context The results obtained from erosion plot experiments and experimental catchment studies in many different areas of the world provide clear evidence of the sensitivity of erosion rates to land use and related human activity. Table 1, for example, shows that the change from natural vegetation to cultivation can increase soil erosion rates by an order of magnitude or more. Increased rates of soil erosion can be expected to give rise to increased sediment transport by local streams and the literature contains results from many experimental catchment studies which show that vegetation clearance or land use change can again cause increases in sediment yield of an order of magnitude or more (cf. Table 2). Abernethy (1990) has similarly reported results from longer-term investigations of several small reservoir catchments in Southeast Asia which showed that during the present century annual sediment yields have progressively increased by between 2.5 and 6% per year as a result of land clearance and subsequent intensification of land use (Figure 1). Abernethy linked these increases to the rate of population growth in the areas concerned and found that the ratio of the increase in sediment yield to that for population was greater than unity and averaged ca. 1.6:1. Based on these findings, he suggested that annual suspended sediment yields from river basins in many developing countries could be expected to double in about 20 years. If the above generalisations are coupled with information on the global increase in population and the global extent of land clearance and intensification of land use, which indicates, for example, that the area of the earth s surface given over to crop production and livestock grazing has increased by more than five-fold over the past 200 years (cf. Buringh & Dudal, 1987), major increases in the sediment loads of the world s rivers could be inferred (cf. Douglas, 1967). The available evidence regarding the impact of human activity and land use change on the sediment loads of the world s rivers, is however, less clearcut. Existing estimates of the global transfer of sediment from the land surface of the globe to the oceans sug-

2 224 Table 1. A comparison of soil erosion rates under natural undisturbed conditions and under cultivation in selected areas of the world Country Natural Cultivated (kg m 2 year 1 ) (kg m 2 year 1 ) China < U.S.A Ivory Coast Nigeria India Belgium U.K Based on Morgan (1986). Figure 1. Trends of increasing sediment yield in several reservoir catchments in Southeast Asia (based on Abernethy, 1990). gest that this flux is currently about tonnes per year (cf. Milliman & Syvitski, 1992; Walling & Webb, 1996), but the extent to which this value has increased over and above the natural value remains uncertain. In the absence of long-term records of sediment transport for rivers in most areas of the world, it is necessary to look for other sources of evidence of the magnitude of this increase. In their classic study Erosion and Sediment Yield on the Earth, whichwas based on sediment yield data assembled from more than 3600 measuring stations on the world s rivers, Dedkov & Mozzherin (1984) used space time substitution to estimate the extent to which the sediment yields of the world s rivers had increased due to catchment disturbance by land use activity. They grouped the available data from both the mountainous and nonmountainous (plains) regions of the world into large (>5000 km 2 )andsmall(<5000 km 2 )basinsandfurther classified these according to the degree of disturbance by agricultural activity. Category I basins were largely undisturbed and characterized by either a forest cover >70% or a cultivated area <30%. Category II basins exhibited an average degree of disturbance (i.e. area under forest or cultivation of between 30 and 70%), whereas category III basins were characterized by the greatest degree of disturbance, with an area under forest of <30% or a cultivated area >70%. By comparing the average sediment yields of category I basins, with the equivalent values for category III basins, it was possible to derive an approximate measure of the magnitude of the increase in sediment yield associated with land disturbance by human activity (cf. Table 3). Although the results of this analysis are necessarily limited by the fact that they represent a simple comparison of mean values of sediment yield for drainage basins in different categories, and are therefore heavily dependent on the representativeness of the sample of river basins involved, they suggest that the sediment yields of many of the world s rivers may have increased ca. 5-fold as a result of disturbance by human activity. Milliman et al. (1987) provide provide a similar, although somewhat greater, estimate for the magnitude of the change in the sediment load of the Yellow River in China since the expansion of agriculture into the loess plateau, which began about 200 BC. In this case, an analysis of sediment deposits in the Yellow Sea was used to estimate that the sediment load of the Yellow river during the early and middle Holocene was an order of magnitude less than that associated with the more recent post-agriculture period. Other lines of evidence are, however, less conclusive in terms of confirming a substantial increase in the sediment loads of the world s rivers as a result of human activity and land use change. The work of Tardy et al. (1989), which has attempted to reconstruct global sediment yields over the geological past (cf. Walling, 1995), indicates that present-day sediment loads are ca. 1.6 times greater than those in the recent geological past (i.e. the Cenozoic, ca. 33 million years ago) (Figure 2A). However, although the impact of human activity is invoked as one explanation of this increase, these authors indicate that it may

3 225 Table 2. Some results from experimental basin studies of the impact of land use change on sediment yield Region Land use change Increase in sediment yield Reference Westland, New Zealand Clearfelling 8 O Loughlin et al. (1980) Oregon, U.S.A. Clearfelling 39 Fredriksen (1970) Northern England Afforestation (ditching and ploughing) 100 Painter et al. (1974) Texas, U.S.A. Forest clearance and cultivation 310 Chang et al. (1982) Maryland, U.S.A. Building construction Wolman & Schick (1967) Figure 2. Variation of sediment yields over geological time. Based (A) on Tardy et al. (1989) and (B) on Degens et al. (1991). also partly reflect the progressive increase in tectonic activity since Jurassic times. Studies of long-term sedimentation in water bodies receiving drainage from large areas, such as the Black Sea (cf. Degens et al., 1976, 1991), further suggest that recent increases in sediment input, although clearly evident, are of limited magnitude when compared with climatically induced changes occurring during the late Pleistocene and the Holocene(Figure2B). The apparent contradictions in the information presented above and the lack of clear evidence of substantial increases in sediment transport by the world s rivers as a result of human activity and land use change could reflect the averaging associated with larger river basins, whereby increased erosion rates may be limited to only parts of the basin. Equally, global variability in the magnitude of land use impacts will inevitably mean that significant increases are only found in some areas of the world. There is, however, also evidence to suggest that the marked increases in rates of soil loss cited above may not be reflected by increases of an equivalent magnitude in the sediment loads of world rivers. Alford (1992), for example, reports a study of the km 2 Chao Phraya river basin draining the highlands of northern Thailand which showed no evidence of a significant increase in sediment yield during the period extending from the late 1950s to the mid 1980s, despite substantial deforestation and extensive swidden agriculture within the basin. This apparent discrepancy may reflect the importance of sediment delivery processes in attenuating upstream increases in sediment mobilisation (cf. Walling, 1983), such that much of the additional sediment mobilised within a river basin may be stored within the system, either on the lower slopes and in the small tributary streams, or on the floodplains bordering the main channel system, and will not reach the basin outlet. There are also situations where human activity has caused a reduction, rather than an increase, in sediment yield. More particularly, the construction of reservoirs will commonly trap the majority of the sediment load formerly transported by the river. The closure of the Aswan Dam on the River Nile has, for example, reduced the annual sediment yield of that river from ca t year 1 to almost zero, and Meade & Parker (1985) cite the case of the Mississippi River, U.S.A., where the construction of five major dams on the Missouri River, one of its major tributaries, reduced the sediment yield at its mouth by more than 50% between the 1950s and the 1980s. In an attempt to scale up such findings to the global scale, Vorosmarty et al. (1997) estimated that more than 40% of the global river discharge is currently intercepted by large impoundments and that the global sediment retention by reservoirs may exceed

4 226 Table 3. Increases in the sediment yields of world rivers resulting from catchment disturbance by land use activities, (based on data presented by Dedkov & Mozzherin (1984)) Group Small Large All basins basins basins Lowland Rivers (n=1854) Mountain Rivers (n=1811) 25% of the global flux from the land to the oceans. It therefore seems likely that in some areas of the globe increases in sediment loads caused by land clearance and land use change may be balanced by reductions associated with reservoir development. Changing sediment yields The problem of assessing the magnitude of the impact of land use change and related human activity on river sediment loads is exacerbated by the general lack of long-term records of sediment transport by the world s rivers. Only a few records extend back to the 1920s and 1930s, and most cover only a few decades. Furthermore, some sediment load records may be of limited reliability (cf. Walling & Webb, 1981). Analysis of the available records can, however, provide more specific evidence regarding the sensitivity of river sediment loads to land use change, although in many cases there will be a need to disentangle the effects of both land use change and climate change, which may be closely linked. Figure 3 presents information concerning long-term trends in the sediment load of two rivers in the former Soviet Union. The record for the km 2 basin of the Dema River at Bochkarevo in Russia presented in Figure 3A shows a clear trend of increasing suspended sediment yield during the period of record which extends from 1949 to This upward trend of sediment yield is statistically significant (>95% level), but the minor increase in annual runoff over the period of record is not statistically significant. The double mass plot of sediment yield versus runoff suggests that the main increase in sediment yields dates from ca and that sediment loads have increased by about 1.4 times since that time. Bobrovitskaya (pers. com.) has indicated that the increased sediment loads evidenced by this river reflect the expansion of cultivation within the drainage basin. The second example, which is presented in Figure 3B, relates to the Dnestr River at Sambur, which drains an 850-km 2 catchment in the Ukraine. Here the trend line fitted to the annual sediment yields is again highly significant statistically and suggests that these have increased by as much as 5-fold since the early 1950s. This increase undoubtedly reflects the influence of forest clearance within the headwaters of the basin (Bobrovitskaya, pers. com.), but it is also a response to the general increase in runoff amounts that has occurred over the period and more particularly since the late 1960s and which also reflects climate change. The double mass plot suggests that the impact of forest clearance was particularly felt after 1968 and that this itself accounts for a 1.8-fold increase in the sediment load of the river. Figure 4 presents examples of two other rivers, which, in contrast to those shown in Figure 3, show evidence of declining sediment yields. The record for the River Isar at Munchen, Germany, which drains a catchment area of 2855 km 2, spans the period and shows a significant reduction in sediment yield. This reduction reflects the development of hydropower stations and associated storage reservoirs on this river, and, more particularly, the commissioning of the Sylvenstein Dam in 1959 (cf. Weiss, 1996). There is no significant trend in the annual runoff totals during the period of record, but sediment yields have decreased to only about 20% of their former level over this period and the trend of the double mass plot suggests that the reduction has intensified in recent years. The other river represented on Figure 4 is the Yellow River at Longmen, China, which drains a catchment area of km 2. This river is well known for its very high sediment load and for the very high erosion rates associated with the extensive area of thick loess deposits which occupies much of its middle reaches. The available record extends over 55 years and in this case shows evidence of a significant decline in both runoff and sediment yield during this period. Sediment yields can be seen to reduce by about 50% during the period of record and much of this reduction can be attributed to the extensive soil and water conservation works undertaken in the middle reaches of the Yellow River basin since the 1970s, which have reduced both runoff and sediment mobilisation. The reduction in runoff over this period also reflects a shift to a drier climate and reduced annual rainfall totals. Attempts by Chinese workers to assess the relative importance of soil and water conservation works and climate change

5 227 Figure 3. Trends in suspended sediment yield and runoff for the Dema River at Bochkarevo, Russia, (A) and the Dnestr River at Sambur, Ukraine, (B). (Based on data supplied by Professor Bobrovitskaya, State Hydrological Institute, St Petersburg, Russia.) in accounting for the reduced runoff and sediment loads in the main Yellow River have suggested that the former accounts for ca. 60% of the reduction in runoff and ca. 50% of the reduction in sediment yield (cf. Gu, 1994; Mou, 1996). The double mass plot further emphasises that the reduced sediment load is not simply the result of reduced runoff, since, if the relationship between annual sediment load and annual runoff can be assumed to be linear, no change in the slope of the double mass plot would be expected if the reduction in sediment yield was solely a reflection of reduced runoff. In this case, the double mass plot suggests that a reduction in sediment yield of approximately 40% has occurred since the late 1970s, which cannot be directly linked to the reduction in runoff during this period. The potential significance of soil and water conservation measures in reducing the sediment loads of large river basins is further emphasized by the records for the tributaries of the Yellow River which drain solely areas within the loess region. In the case of the Wuding River ( km 2 ), sediment yields have declined by almost 90% between the 1950s and the early 1990s (Figure 5), although part of this decline is again due to climate change and the associated reduction in precipitation. The double mass plot presented in Figure 5 suggests that a reduction in sediment yield of ca. 60% has occurred since 1971, which cannot be directly linked to the reduced runoff. As a final example of the sensitivity of river sediment loads to land use change, Figure 6 presents data for two rivers which show little evidence of changing sediment loads over a periods of more than 40 years. The record of annual sediment loads for the River Lech above Fussen in Bavaria (1422 km 2 )extends over more than 60 years and, since this river has been essentially unaffected by impoundment and regulation, it affords a valuable example of the response of an Alpine river to changing land use. The annual runoff record shows evidence of a small reduction over the period of record and this trend is statistically significant at the 95% level, but the sediment record shows little evidence of change, despite the likelihood of significant changes in both land use and land management practices within the catchment over the past

6 228 Figure 4. Trends in suspended sediment yield and runoff for the River Isar at Munich, Bavaria, (A) and the Yellow River at Longmen, China, (B). (Based on data supplied by Dr F.H. Weiss, Bayer Landesamt fur Wasserwirtschaft, Munich, and Dr Fang Xuemin, IWHR, Beijing.) 60 years (cf. Summer et al., 1996). A similar picture is provided by the annual runoff and sediment load data for the Upper Yangtze River at Yichang, China (Figure 5B). This river drains an area of km 2, which supports a population of ca. 140 million. The population has increased rapidly from ca. 60 million in 1953 and there are many reports of increased soil loss (cf. Dai & Tan, 1996). Despite such evidence of increased population pressure and intensification of land use, the time series of annual sediment loads is essentially stationary and there is no evidence of decreasing runoff, which might offset the impact of increased erosion. Some evidence of increased sediment loads is provided by the shorter records for some of the tributary rivers within the Upper Yangtze basin (cf. Lu & Higgitt, 1998), but again this suggests that sediment loads have exhibited only limited change. The results presented above afford coverage of only a small number of rivers and a limited range of physiographic conditions. Nevertheless, they provide evidence of the complexity of the causal link between land use change and associated increases in rates of erosion and river sediment loads. The records from the two rivers in the former Soviet Union depicted in Figure 3 show significant increases in sediment yields consequent upon land use change and land disturbance, and the records from the Yellow River presented in Figure 4B and Figure 5 similarly demonstrate the

7 229 Figure 5. Trends in suspended sediment yield and runoff for the Wuding River, China, (Based on data supplied by Mr Mou Jinze, YRCC, Zhengzhou.) sensitivity of the sediment yields from this river basin to reduction by means of soil and water conservation measures. Conversely, however, the records from the River Lech and the Upper Yangtze presented in Figure 6 provide examples of situations where sediment loads appear to be essentially insensitive to changes occurring within the catchment. These apparently conflicting findings reflect, at least in part, variations in what might be termed the buffering capacity of the catchments involved. The work of Meade & Trimble (1974) and Phillips (1992) in assessing the impact of post-colonial agricultural activity on the sediment loads of rivers draining to the coastal seaboard of the eastern United States has usefully demonstrated how much of the sediment eroded from their watersheds was deposited or stored within the river basin and failed to reach the basin outlet. Subsequent reductions in soil erosion associated with the decline of plantation agriculture and the introduction of soil conservation measures were also not reflected by downstream sediment loads, since reductions in sediment transport were balanced by remobilisation of stored sediment. Major changes in land use which would have been accompanied by changing erosion rates were therefore not reflected by major changes in sediment transport within the lower reaches of these rivers. The effectiveness of such buffering can be related to the classic sediment delivery ratio concept (cf. Walling, 1983). Rivers with a low sediment delivery ratio will be characterized by a high degree of buffering, since much of the sediment mobilized within the catchment will be deposited and these deposits can in turn be remobilized if sediment supply to the river system declines. Where, however, sediment delivery ratios are high, little sediment will be deposited and there will be only limited sediment available to be remobilized during times of reduced supply. Such river basins can therefore be expected to be poorly buffered and the sediment output will be sensitive to changing erosion rates associated with changing land use. Estimates of sediment delivery ratios available for the Middle Yellow River basin (cf. Gong & Xiong, 1980; Mou & Meng, 1980) indicate that much of this region is characterised by values close to 1.0 or 100%, due in part to the high drainage density and the frequent occurrence of hyperconcentrated flows, which reduce the potential for sediment deposition. These values are consistent with the low degree of buffering found in this river basin. Values of the sediment delivery ratio cited for the Upper Yangtze River basin by Dai & Tan (1996) and Liu & Zhang (1996) are much lower than those for the Middle Yellow River basin and averaged ca or 34%. Such values are likewise consistent with the much higher degree of buffering suggested by the record of sediment loads for the Upper Yangtze (cf. Figure 6B). No corresponding information on sedi-

8 230 Table 4. The suite of sediment properties used to fingerprint suspended sediment sources and the results of the Multivariate Discriminant Function analysis used to select this composite fingerprint Tracer property N 51.9 Total P 74.3 Sr 77.9 Ni 82.4 Zn Ra Cs 88.2 unsupported 210 Pb 88.2 Fe 92.7 Al 94.1 Based on Walling et al. (1999a). Cumulative % samples classified correctly ment delivery ratios are available for the other rivers documented in Figures 3 6, although existing evidence would suggest that SDR values for the River Lech would be relatively low and thus indicative of a wellbuffered system, as suggested by the stability of the long-term record of sediment loads. Sediment budgets The above discussion emphasizes that any attempt to consider the impact of land use on erosion and sediment transfer through a river basin should direct attention not only to changes in sediment load at the catchment outlet, but also to the entire delivery or conveyance system and to the sinks or stores involved. This broader perspective will permit improved understanding of the link between land use, erosion and sediment yield as well as focusing attention on the fate of mobilised sediment. Although the detrimental effects of increased sediment loads in rivers are well documented (cf. Clark et al., 1985), accumulation of sediment in sinks and stores within a drainage basin is known to give rise to other environmental problems. The sediment budget concept (cf. Dietrich & Dunne, 1978; Walling, 1988) affords a convenient means of representing and interpreting this link and Figure 7, based on the classic work of Trimble (1981), provides a useful example of its application. In this case, data from a range of field and documentary sources were used to reconstruct the sediment budget of Coon Creek, a 360-km 2 basin in the Driftless Area of Wisconsin, U.S.A., for two periods in the past, namely and The first period represented one of poor land management which resulted in severe soil erosion, whereas the second was characterised by the introduction of conservation measures. During the first period, large volumes of soil were eroded from the slopes of the basin, but only a small proportion (ca. 5%) of this was transported out of the basin. Most of this material was stored within the catchment. During the latter period when widespread soil conservation measures were introduced, rates of soil loss from upland sheet and rill erosion were reduced by about 25%, but sediment yields at the basin outlet remained essentially the same, because sediment stored in the tributary valleys and upper main valley was remobilized. In this case, therefore, substantial changes in land use and land management within the upstream basin were not reflected in the downstream sediment yield because of the low SDR of the basin and thus its relatively high buffering capacity. The precise form of the sediment budget of a catchment, and thus its SDR and buffering capacity, will vary according to the local physiographic conditions, but considerable variation may also occur even within a relatively small area. Figure 8, for example, illustrates the very considerable diversity of sediment budget structure reported by Golosov et al. (1992) for small- and medium-sized drainage basins on the Russian Plain, a region which is heavily impacted by land use activities and soil erosion. In these examples, the sediment delivery ratios ranged from zero to 89%. In the former case, changes in land use and erosion rates would not be reflected in changes in sediment yield, whereas in the latter case sediment yields could be expected to be highly sensitive to such changes. Scaling up sediment budgets such as those shown in Figures 7 and 8 to much larger heterogeneous river basins clearly involves many problems in terms of assembling the information necessary to establish the key elements of the sediment budget and such studies have hitherto been constrained by the limitations of traditional monitoring techniques for quantifying sediment budgets. As a result much of our existing understanding of the sediment budgets of larger river basins is based on theory and inference rather than direct empirical evidence. However, recent work has demonstrated the potential of newly developed techniques for establishing the relative importance of different sediment sources within a drainage basin and for assessing the importance of overbank flood-

9 231 Figure 6. Trends in suspended sediment yield and ruonff for the River Lech at Fussen, Bavaria, (A) and the Yangtze River at Yichang, China, (B). (Based on data supplied by Dr F. H. Weiss, Bayer Landesamt fur Wasserwirtschaft, Munich and Dr Fang Xuemin, IWHR, Beijing.) plain deposition and associated floodplain storage as a component of the sediment budget. This potential can usefully be demonstrated by referring to the results of some recent work undertaken by the author and his co-workers in the catchment of the River Ouse in Yorkshire within the framework of the U.K. Land Ocean Interaction Study (LOIS). This work was aimed at elucidating the key features of the sediment budget of this 4000-km 2 catchment and involved establishing the relative importance of the primary sediment sources and quantifying the role of floodplain sedimentation as a sediment sink (cf. Walling et al., 1998, 1999a,b). The fingerprinting approach, involving multicomponent fingerprints (cf. Walling et al., 1993; Walling & Woodward, 1995), was used in the Ouse catchment to determine the relative importance of four potential sediment sources within the catchment, namely, the surface of cultivated areas, areas of permanent pasture and upland moorland, and areas of woodland, and channel banks and other sources of subsurface material. In essence, the fingerprinting approach in-

10 232 Table 5. Load-weighted contribution of the major source types to suspended sediment samples collected from the Rivers Swale, Ure, Nidd, Ouse and Wharfe during the period November 1994 to February 1997 River Number of Woodland Uncultivated Cultivated Channel suspended topsoil topsoil topsoil bank sediment (%) (%) (%) material samples (%) Swale Ure Nidd Ouse Wharfe Based on Walling et al. (1999a). Figure 8. The sediment budgets of four drainage basins on the Russian Plain, as documented by Golosov et al. (1992). Figure 7. The sediment budgets of Coon Creek, Wisconsin, U.S.A., reconstructed by Trimble (1983) for the periods and volves comparison of the geochemical properties of suspended sediment transported by the river with those of potential sources, in order to establish their relative importance. Use of statistically verified composite fingerprinting signatures involving several different sediment properties (cf. Table 4), enabled the four potential sources to be clearly discriminated and a multivariate mixing model was used to determine the relative contribution of the four sources to suspended sediment samples collected from the River Ouse and its major upstream tributaries and from the River Wharfe, which joins the Ouse below its tidal limit (cf. Figure 9). The load-weighted estimate values for the relative contributions from the four different sediment sources are listed in Table 5. These results confirm that a major proportion of the suspended sediment load of the River Ouse and its tributaries is derived from surface sources and is thus likely to be directly influenced by land use activities, including cultivation and grazing pressure. In upscaling from smaller to larger catchments, it is important to recognize that the river floodplains which border the main channel systems in larger river basins can represent an increasingly important sediment sink, which will attenuate changes in upstream sediment inputs. Walling & Quine (1993), for example, estimated that floodplain storage accounted for ca. 23% of

11 Table 6. A comparison of the estimates of total storage of sediment on the floodplains bordering the main channel system of the River Ouse and its primary tributaries, with the estimated suspended sediment loads for the study rivers River Floodplain Mean annual Total sediment Floodplain storage suspended derived to storage as % (t year 1 ) sediment load channel of sediment (t year 1 ) (t year 1 ) input Swale Nidd Ure a Ouse a Total to Ouse gauging station Total to tidal limit Wharfe a Ure is the River Ure to its confluence with the River Swale, while Ouse refers to River Ure/Ouse from below this point to the tidal limit. Based on Walling et al. (1998). the sediment delivered to the main channel system of the 6850-km 2 catchment of the River Severn in the U.K. Attempts to document rates of overbank sedimentation on river floodplains, in order to establish the importance of such conveyance losses, face many practical problems. However, recent developments in the application of the environmental radionuclide Cs- 137 to estimate rates of overbank sedimentation offer considerable potential in this context (cf. Walling & He, 1997a,b). This approach was used by Walling et al. (1998) to establish the significance of floodplain storage in the suspended sediment budget of the main channel systems of the Rivers Ouse (3315 km 2 ) and Wharfe (818 km 2 ). In this case, more than 250 sediment cores were collected from 26 representative transects located along the main channel systems of the Yorkshire Ouse and its major tributaries and the River Wharfe (cf. Figure 9) for Cs-137 analysis. The estimates of mean sedimentation rates for the individual transects obtained from these cores, which averaged ca. 0.2 g cm 2 year 1, were extrapolated to the individual reaches between adjacent transects and the mean annual conveyance loss associated with overbank sedimentation on the floodplain bordering the main channel system was calculated. By relating these losses to the mean annual sediment loads of the rivers (Table 6), it was possible to establish the relative importance of floodplain storage in the sediment budget of the main channel system (Table 6, Figure 10). In the case of the main River Ouse system, floodplain deposition accounted for ca. 40% of the total amount of suspended sediment delivered to the main channel system over the past 40 years and for the River Wharfe the equivalent value was ca. 49%. These conveyance loss values of 40% or more, associated with the floodplains bordering the main channel system of a 4000-km 2 river basin, serve to emphasize further the potential importance of storage in attenuating the link between upstream erosion and downstream sediment loads and thus the impact of land use change on sediment yield. The application of environmental radionuclides in quantifying the role of river floodplains as sediment sinks also offers the possibility of investigating changes in sedimentation rates through time. In this way, studies of changing sediment output from a drainage basin could be paralleled by studies of changing sediment storage. By measuring both the Cs-137 and unsupported Pb-210, content of floodplain cores, it is possible to derive estimates of sedimentation rates over the past years (Cs-137) and years (unsupported Pb-210) which can be compared (cf. Walling & He, 1994, 1999). The basis of these estimates precludes separation of the record into consecutive periods, but, by comparing the longer-term with the shorter-term estimate, it is possible to establish whether sedimentation rates have increased or decreased in recent years. Table 7, which is based on the work of Walling & He (1999), identifies the trends evidenced by single cores collected from the floodplains of a representative selection of 21 rivers

12 234 Figure 9. The catchments of the River Ouse and Wharfe, Yorkshire, U.K. in the U.K. The results show no evidence of major changes in sedimentation rates over the past 100 years, but the individual rivers provide examples of situations where sedimentation rates have both increased and decreased and remained effectively stable. Increases and decreases have been defined as instances where the sedimentation rate for the past 33 years has increased or decreased by more than 10% relative to the sedimentation rate for the past 100 years. This period was marked by significant changes in land use and land management practices, but it appears that such changes had relatively little impact on rates of floodplain sedimentation and storage. In these circumstances, it is likely that sediment outputs from the river basins represented also changed relatively little. Thus, although floodplain storage could represent an important buffer for changing sediment fluxes, such effects would appear to have been of limited import-

13 235 Table 7. A comparison of mean annual sedimentation rates for the past 33 and 100 years estimated for a selection of sites on the floodplains of British rivers River/Location Sedimentation rate (g cm 2 year 1 ) Trend a past 33 years past 100 years 1. River Ouse near York Stable 2. River Vyrnwy near Llanymynech Decrease 3. River Severn near Atcham Decrease 4. River Wye near Preston on Wye Decrease 5. River Severn near Tewkesbury Stable 6. Warwickshire Avon near Pershore Decrease 7. River Usk near Usk Decrease 8. Bristol Avon near Langley Burrell Increase 9. River Thames near Dorchester Decrease 10. River Torridge near Great Torrington Decrease 11. River Taw near Barnstaple Stable 12. River Tone near Bradford on Tone Increase 13. River Exe near Stoke Canon Stable 14. River Culm near Silverton Stable 15. River Axe near Colyton Increase 16. Dorset Stour near Spetisbury Stable 17. River Rother near Fittleworth Decrease 18. River Arun near Billingshurst Decrease 19. River Adur near Partridge Green Decrease 20. River Medway near Penshurst Decrease 21. River Start near Slapton Increase a The trend indicates the change in sedimentation rate when comparing the past 33 years with the past 100 years. A change of >±10% is taken to represent a significant increase or decrease. Based on Walling & He (1999). ance in these river basins. It is, nevertheless, possible that significant changes in erosion rates were buffered by loose coupling of the slope and channel systems and sediment storage in headwater areas, since such sinks are likely to be of similar importance to floodplain storage within the lower reaches of the main river system. Further work is clearly required to elucidate the behaviour of this other potentially important component of the catchment sediment budget. Sediment storage Figure 10. The role of overbank floodplain sedimentation in the sediment budget of the main channel systems of the River Ouse and its major tributaries and the River Wharfe, Yorkshire, U.K. Recognition of the potential importance of sediment storage within a drainage basin and its role in attenuating the impact of land use change on downstream sediment loads also directs attention to the wider environmental significance of such storage. Deposition of sediment and, more importantly, sediment-associated nutrients and contaminants in sediment sinks could constitute a significant environmental problem in terms of both their accumulation and storage and the po-

14 236 tential for subsequent remobilisation, such that the storage response of a river basin to land use change could be of equal importance to changes in sediment output. Marron (1987), for example, cites the case of the Belle Fourche River in South Dakota, U.S.A., where gold mining within the upstream catchment during the period extending from the 1880s to 1977 caused the accumulation of large quantities of arseniccontaminated sediment within the river floodplain and it was estimated that, even with no further inputs from mining activity, arsenic contamination would remain a problem within this river system for many centuries to come due to remobilisation of contaminated sediment. Recent concern for phosphorus cycling within terrestrial and freshwater ecosystems has also directed attention to the potential role of river floodplains in storing sediment associated-p mobilized from the upstream catchment. Since particulate-p can account for a major proportion of the total-p load of a river, overbank sedimentation on floodplains can result in substantial conveyance losses of P and appreciable quantities of P can accumulate in such floodplain sinks. Figure 11 provides information on the P content of a sediment core collected from the floodplain of the River Culm in Devon U.K. In this case, total-p concentrations in the upper horizons of the floodplain sediment, which average ca mg kg 1 are more than twice those associated with soils in adjacent areas above the level of the floodplain.. These higher concentrations reflect the deposition of sediment-associated P mobilised from the upstream watershed P in association with fine sediment. Selective erosion, involving preferential mobilisation of fines and organic matter, causes enrichment of the P content of suspended sediment relative to the bulk soil, and this enrichment is in turn reflected in the high concentrations found in the floodplain sediments. In addition to being characterized by higher total-p concentrations, the floodplain sediments also possess a much greater P stock and this stock will increase through time as deposition continues. Figure 11 also presents information on the downcore variation of Cs-137 concentrations in the floodplain core and this information can be used to estimate the rate of sediment deposition, because the level associated with the peak Cs-137 concentration can be used to establish the floodplain surface at the time of peak fallout rates which occurred in 1963 (cf. He & Walling, 1996). In this case the deposition rate is estimated to be 0.31 g cm 2 year 1 and, if this value is combined with the average total P content of sediment in the upper part of the core (i.e. post 1963), an annual input of sediment-associated P to the floodplain surface of 8.2 gm 2 year 1 can be estimated. Equivalent values for the total P concentration of post 1963 floodplain sediment deposits and the P accumulation rate have been obtained for sediment cores collected from 19 further sites on the floodplains of U.K. rivers and these values are presented in Table 8. The range of P concentrations associated with floodplain sediments listed in Table 8 closely reflects the intensity of agricultural activity in the upstream catchment, although point-source inputs of P to the river system from sewage works will clearly also exert a significant influence. The values of P accumulation presented in Table 8 are of a similar order of magnitude to that of 9 g m 2 year 1 reported by Fustec et al. (1995) for a floodplain site on the River Seine at Maizières, France, and further underscore the potential importance of river floodplains as a phosphorus sink. Remobilisation of these sediment stores could release large amounts of P back into the rivers, although rates of bank erosion associated with U.K. rivers are such that this sediment is likely to have a residence time of 10 2 to 10 4 years and can be viewed as being in long-term storage. Major changes in channel mobility and floodplain erosion, such as might accompany climate change, must, however, be seen as a potential cause of more rapid remobilisation. Consideration of downcore changes in the P content of floodplain sediment also affords a means of deriving some information on past changes in the P content of suspended sediment transported by the river concerned. In this case it is necessary to recognize the potential for post-depositional changes in P content, due to mobilisation into solution, plant uptake etc. but, since most of the sediment-associated P is likely to be firmly fixed to the sediment and not readily mobilised, downcore changes in P concentration will provide some tentative evidence concerning past changes in the P content of deposited sediment. By coupling this information with information on average sedimentation rates derived from the Cs-137 profile, it is possible to establish the likely trend of the P content of deposited sediment over the past 40 years or more. Such information is presented in Figure 12 for a representative selection of five U.K. rivers. In this case there is close relationship between both the P concentrations involved and the trend of these concentrations over the period since 1950, and the intensity of agricultural activity in the upstream catchment. The catchments of both the Usk and the Upper Severn are characterized by low intensity agriculture with substantial areas of upland sheep pasture and the sediment collected from

15 237 Figure 11. Down-core variations in total phosphorus (A) and caesium-137 (B) for a sediment core collected from the floodplain of the River Culm near Columbjohn, Devon, U.K. Figure 12. Variations in the total phosphorus content of sediment deposited on the floodplains of several British rivers during the period the floodplains of these two rivers exhibits relatively low P concentrations with only limited evidence of an upward trend. This is consistent with the relatively low levels of fertiliser application in these two catchments. In contrast, the data obtained for the Rivers Axe, Arun and Torridge show both higher P concentrations and a more well-defined upward trend, which are consistent with the more intensive agriculture in their catchments. The catchment of the River Torridge is occupied primarily by permanent pasture, whereas the catchments of the Axe and Arun are characterized by areas of both both pasture and arable cultivation. In the case of the River Torridge, the P content of sediment deposited on the floodplain has increased by nearly 50% over the past 40 years, whereas the increases for the Arun and Axe are ca. 94 and 170%, respectively. Equivalent values for 15 other sites, for which information on the P content of floodplain sediment and P accumulation rates were provided in Table 8, are also are presented in Table 8. This larger data set shows further evidence of substantial increases in the P content of sediment deposited in the floodplain store of British rivers over the past 40 years, which are likely to reflect land use change and increased fertilizer application within the upstream catchments. Recognition of the significance of sediment as a carrier for nutrients and contaminants clearly introduces a need to take account of both the amount and the properties of sediment

16 238 Table 8. Phosphorus storage on the floodplains of British Rivers River/Location Mean Total P Total P Increase in concentration of storage Total P content post 1963 sediment since 1963 of deposited (mg kg 1 ) (g m 2 year 1 ) sediment (%) 1. River Vyrnwy near Llanymynech River Severn near Atcham River Wye near Preston on Wye River Usk near Usk River Teme near Broadwas River Torridge near Great Torrington River Taw near Barnstaple River Exe near Stoke Canon River Culm near Silverton River Start near Slapton River Tone near Bradford on Tone Bristol Avon near Langley Burrell River Thames near Dorchester Warwickshire Avon near Pershore River Axe near Colyton Dorset Stour near Shillingstone River Rother near Fittleworth River Arun near Billingshurst River Adur near Partridge Green River Medway near Penshurst stored within the delivery system and the influence of land use on both aspects. Perspective This contribution has endeavoured to review existing understanding of the relationship between land use change, erosion and sediment yield. Although the impact of land use change on rates of soil loss, and more particularly the impact of land clearance and cultivation in increasing erosion rates, has been extensively documented, the evidence for major changes in the sediment loads of larger rivers is less clear. Existing long-term records provide evidence of cases where sediment loads have both increased and decreased as a result of land use change and other human activity, but it is clear that river basins and more particularly river systems possess considerable capacity to buffer such changes. The extent of this buffering capacity can be expected to be closely related to the sediment delivery ratio of the basin, such that river basins with low sediment delivery ratios will be characterized by a high buffering capacity and vice versa. Recognition of the importance of sediment storage within a drainage basin necessarily means that attempts to understand the linkages between land use, erosion and sediment yield should consider the overall sediment budget and the associated sources and sinks, rather than only the sediment outputs. Changes in the functioning of these stores could be of greater significance than changes in sediment output from a drainage basin. Recent work has demonstrated an improved capacity for assembling the information needed to construct a sediment budget and investigations of the role of floodplain sedimentation as a sediment sink have increasingly emphasised the potential importance of this store. By directing attention to the overall sediment budget of a river basin and to the associated sediment stores and sinks, as well as the sediment output, the wider environmental significance of these components of the sediment budget becomes increasingly apparent. They can represent important sinks for sediment-associated contaminants and any attempt to fully understand the