Stability of output fluxes of large rivers in South and East Asia during the last 2 million years: implications on floodplain processes

We compare the present-day sediment discharge (solid phase) of some of the largest rivers in Asia to the average discharge deduced from the mass accumulated in several sedimentary basins during the Quaternary. There is a very good correlation, especially for the largest rivers: the Ganges–Brahmaputra, the Changjiang, the Huanghe and, to a lesser extent, the Indus and the Zhujiang. This suggests that present-day average discharge at the outlet has remained constant throughout the Quaternary at least for very large rivers (drainage area of the order of 105–106 km2). This, in turn, suggests either that continental denudation of large Asian catchments has remained on average constant, implying a strong tectonic control on erosion during the Quaternary, or that the river network has the ability to buffer changes in hillslope erosion or in sea-level in order to conserve the total discharge at the outlet. We show how this buffering capacity relies on the characteristic reaction time-scale of Asian alluvial plains (of the order of 105–6 years), that is, much higher than the time-scales of the Quaternary climate oscillations (of the order of 104 years). A short-term perturbation originating in hillslopes will be diluted by the floodplain. At the outlet the signal should have a longer time span and a smaller amplitude. In the same manner, an alluvial plain should not instantaneously react to a 104-year sea-level drop because of its inertia. Along with long-term tectonic control we infer this buffering to be the main cause for the average constancy of sediment yield of large Asian rivers during the Quaternary. does this bear on upstream processes and transport INTRODUCTION capacity of the rivers? To address the relevance of long-term denudation rates The purpose of this paper is to look at the correlation possibly existing between long-term and short-term mass deduced from sediment accumulation, Granger et al. (1996) dated sands in rivers using cosmogenic isotopes. fluxes carried to the sea by large Asian rivers and to discuss the possible implications of such a correlation. They then deduced average denudation rates on the scale of 104 years and compared them to sediment accumuBecause rivers are the main carriers of erosion products from the continents to the oceans, measuring their lation. Their results show a good fit, indicating the relevance of using average denudation rates on that time sedimentary load provides invaluable constraints on the rate of continental denudation. There is, however, a scale. Lavé (1997) derived fluvial incision rates in the Himalayas from the study of strath terraces. He extrapoproblem in that these measurements cover a very short time-span, at most a few decades, while erosion processes lated his results to the entire range and compared the results to sediment volumes accumulated in the Ganga operating in a basin have very different time constants. Some are almost instantaneous, like avalanches (velocity plain and the Bengal fan (Métivier, 1996; Métivier et al., 1999). He showed a good average correlation between of the order of 100 to 102 m s−1) while others, like soil creep (velocity of the order of 10−10 to 10−12 m s−1), the two results, suggesting that short-term fluvial incision rates are comparable to long-term rates deduced from may be too slow to be observed (Goudie, 1995). This raises the following question: how representative of longsedimentary volumes. Recently, Burbank et al. (1998), conducting a similar study in the San Gabriel mountains, term average denudation rates are present-day measurements of solid loads in rivers? And what information found also good agreement between rates of denudation © 1999 Blackwell Science Ltd 293 F. Métivier and Y. Gaudemer at different time-scales. This correlation, however, does entire river bed assuming homogeneous turbulent mixing of the suspended load (Fournier, 1960). The data given not appear to be unique. Kirchner et al. (1998), comparing are generally in good agreement from one publication to sediment yield and denudation rates at the 104-year scale, the other. When they differ markedly, we selected the found that, for small to median size catchments (less most recently published value. than 5×104 km2), the suspended load was systematically These reports are incomplete however. There is, for lower than average denudation rates. They attribute it to instance, no value for the Salween river that drains part the fact that their record is too short to incorporate of SE Tibet and Burmese mountains before flowing into extreme erosion events like exceptional landslides. the Andaman sea. Similarly, there is no value for several To tackle this important problem of present-day and albeit small (drainage area A≤105 km2) rivers that drain long-term average fluxes of sediment, we have compared the Makran range in southern Iran and Pakistan and the present-day solid loads, Q p of several large rivers of bring sediments to the Indus fan in the Arabian Sea. For Asia (catchment area greater than 105 km2) with the the Huang He (Yellow River), we take the value of 100 filling rate, Q f of the sedimentary basins fed by the same million tons per year (hereafter t year−1) given by rivers averaged over the last two million years (Fig. 1). Milliman et al. (1987), which is corrected for the anthroThis time span approximately corresponds to the pogenic activity, namely agriculture on the loess plateau Quaternary Era for which data are available from drill in the last 2000 years (Wang et al., 1998). logging in almost every Cenozoic basin of Asia (Métivier Another source of uncertainty is the lack of reliable et al., 1999). We show that the correlation we obtain estimates of bed load transport as no precise measurement suggests that rates of mass transport to the sea by large technique is available at present. Clearly, bed load can Asian rivers has remained, on average, a constant. We be important in the sediment budget as has been shown furthermore propose this constancy to be explained either by more or less empirical relationships between bed load by sustained erosion rates in the catchments, that would transport and discharge or stream power (Ashmore, mainly be controlled by high tectonic uplift rates, or by 1988). In the absence of any applicable relationship or buffering of variations of the sediment influx from the measurements we will assume a large uncertainty similar catchment by the river floodplain. We illustrate this latter to the Q f uncertainty (i.e. 40%). possibility by considering a 104-year sea-level drop and We did not include solute load transport for two by looking at the possible effects on the catchments and reasons. First, the sediments deposited in the basins of on erosion fluxes. Asia are almost exclusively composed of clastics transported as solids by the river systems (Métivier, 1996). DATA ANALYSIS Second, apart from the Changjiang, the contribution of solute transport appears to be largely negligible and in The average filling rates (Q f in Table 1) for the last two any case remains within the 40% error bars we assume million years were derived from a study of mass accumu(Summerfield & Hulton, 1994). lation in the sedimentary basins of Asia (Métivier et al., Finally, a very important point is that the Quaternary 1999). Local depths and thickness of sediments measured infill of the basins we study is not only composed of the on isopach maps, cross-sections and drill-logs, either fine portion of sediments that leave a catchment, the published or provided by TOTAL-CFP company, were so-called washload made of particles smaller than 0.2 mm interpolated, then integrated over the entire depositional in size. One observes the presence of clays, sands, gravels areas in order to provide volumes of sediments deposited and even conglomerates in large amounts in all the basins in the basins of Asia for several periods covering the studied (Métivier, 1996; Wang et al., 1998; Métivier Cenozoic. The volume estimates are fraught with an et al., 1999), and especially in their proximal and thickest uncertainty of the order of 35–40%. A complete descripparts where coarse material can account for most of the tion of the method and associated errors is given in sediment pile. Therefore in the rivers we study the coarse Métivier & Gaudemer (1997) and Métivier et al. (1998, fraction of erosion, that is the size fraction produced at 1999). It is at present impossible to achieve a time the outcrops and stored in the floodplains and alluvial resolution smaller than about 2 million years because of fans, reaches, in some proportion, the sedimentary basins the lack of reliable dating of recent coarse-grained clastic through the fluvial system. Analysis of these river basins sediments for all Asian basins. should therefore allow us to discuss floodplain processes Suspended load values (Q p in Table 1) were gathered that involve both the suspended and bedload components from several published sources (Milliman & Meade, 1983; of sediment transport. Milliman & Syvitski, 1992; Summerfield & Hulton, 1994; Meybeck & Ragu, 1995). Sampling rates are usually of the order of one sample each day over a period on order RESULTS AND DISCUSSION of 1–10 years. Sampling techniques vary between precise Equality of the present-day and Quaternary sampling – at different depths on several vertical profiles average fluxes across the river section as, for example, in the case for the Indus river – and one point sampling (Fournier, Comparing the average discharge deduced from basin filling, Q f with the corresponding present-day discharge, 1960). In the last case the results are extrapolated to the © 1999 Blackwell Science Ltd, Basin Research, 11, 293–303 294 R ier stab iity, S E A ia Fig. 1. Average accumulation rates qf (x, y) (t km−2 yr−1) in Asian basins during the last 2 million years (Métivier, 1996; Métivier et al., 1999). Regional rates (over area S) were derived from this database following Q f=∆∆Sqf(x, y)dxdy. © 1999 B lckw ll Sence L t,B sin R esrch, 1, 29–303