Escarpment erosion and landscape evolution in southeastern Australia
Published:January 01, 2006
- PDF LinkChapter PDF
Arjun M. Heimsath, John Chappell, Robert C. Finkel, Keith Fifield, Abaz Alimanovic, 2006. "Escarpment erosion and landscape evolution in southeastern Australia", Tectonics, Climate, and Landscape Evolution, Sean D. Willett, Niels Hovius, Mark T. Brandon, Donald M. Fisher
Download citation file:
Passive margin escarpments are extensively studied around the world, and understanding their evolution continues to present one of the more compelling interdisciplinary challenges tackled by earth scientists today. Escarpments reflect the morphotectonic development of passive margins and can separate regions with different climatic histories, but the inferred rapid rates of escarpment retreat have been at odds with actual measurements of land surface denudation. In this paper we present results from extensive cosmogenic 10Be and 26Al analyses across the escarpment of southeastern Australia to quantify the erosional processes evolving the highland, lowland, and scarp face landscapes. We document new relationships between soil production rates and soil thicknesses for the highland and lowland landscapes and compare these soil production functions with those published in our earlier studies from the highlands and at the base of the escarpment. Both new functions define exponential declines of soil production rates with increasing soil depths, with inferred intercepts of 65 and 42 m/m.y. for the highland and lowland sites, respectively, and slopes of –0.02. Exposed bedrock at both of the new sites erodes more slowly than the maximum soil production rates, at 22 ± 3 and 9 ± 2 m/m.y., respectively, thus suggesting a “humped” soil production function. We suggest that instead of a humped function, lithologic variations set the emergence of bedrock, which evolves into the tors that are found extensively across the highlands and at the crest of the escarpment by eroding more slowly than the surrounding soil-mantled landscape. Compared to soil production rates from previous work using 10Be and 26Al measurements from two different sites, these results show remarkable agreement and specifically quantify a soil production function for the region where soil production rates decline exponentially with increasing soil thickness, with an intercept of 53 m/m.y. and a slope of –0.02. Erosion rates determined from 10Be concentrations from outcropping tors, bedrock, and saprolite from a main spur ridge perpendicular to the escarpment, and sediments from first- and zero-order catchments draining the main ridges, show a clear linear decline with elevation, from ∼35 m/m.y. near the escarpment base to ∼3 m/m.y. at the escarpment crest. This order of magnitude difference in erosion rates may be due to increases in stream incision with distance downslope on the escarpment, or to decreases in precipitation with elevation, neither of which we quantify here. The rates do agree, in general, with our soil production functions, suggesting that the biogenic processes actively eroding soil-mantled landscapes are shaping the evolution of the escarpment despite our observations of block fall and debris-flow processes across the steep regions near the scarp crest. Our results support recent results from studies using low-temperature thermochronology, which suggest that the escarpment is relatively stable after having retreated rapidly immediately following rifting. Differences between our rates of surface erosion caused by processes active today and the scarp retreat rates needed to place the escarpment in its present position need to be explained by future work to untangle the mysteries of escarpment evolution.
Passive continental margins and the escarpments that are typically associated with them have captured the attention of much geomorphic study, helping to build the application of numerical modeling as well as develop applications of new analytical techniques in the field of quantitative geomorphology (e.g., Bierman and Caffee, 2001; Cockburn et al., 2000; King, 1962; Ollier, 1982; Summerfield, 1999; Tucker and Slingerland, 1994; van der Beek and Braun, 1998). Interest in escarpments is well founded, as understanding landscape evolution of a passive margin escarpment forces an integration of climate, tectonic, and surface process studies. Traditional views of escarpment evolution have suggested high rates of escarpment retreat, parallel to the continental margin, in comparison with low rates of denudation in the low-relief highlands as well as across the more highly dissected lowlands.
Despite such long-standing interest in understanding the evolution of passive margin escarpments, relatively few data actually quantify the rates of surface erosion and long-term denudation across these margins to confirm qualitative models of rapid escarpment retreat. Recent work using low-temperature thermochronology place critical long-term constraints on escarpment denudation (e.g., Brown et al., 2002; Cockburn et al., 2000; Persano et al., 2002), while cosmogenic nuclides have defined shorter-term surface erosion rates above (Heimsath et al., 2001), across (Bierman and Caffee, 2001; Cockburn et al., 2000), and below (Heimsath et al., 2000) escarpments in southeastern Australia and Namibia. These data refute the traditional models of escarpment evolution, and instead suggest that relatively rapid rates of postrifting denudation are followed by long periods of lower erosion rates and the development of a long-term steady-state in the topographic form of the escarpment, where relatively slow landscape change is occurring.
We tackle the problem of escarpment evolution by applying the well-developed mass-balance approach initially articulated by Gilbert (1877), quantitatively laid out by Culling (1960, 1965), and eloquently made accessible to a broad audience by Carson and Kirkby (1972). This conceptual framework has been extensively applied recently to quantify soil production (e.g., Heimsath et al., 1997, 1999), landscape evolution (e.g., Dietrich et al., 1995), as well as model the dynamic responses of the land surface to changes in climate and tectonic forcing (e.g., Tucker and Slingerland, 1997; van der Beek and Braun, 1999). Specifically, we are interested in the vertical lowering rate of the land surface. For a bedrock surface, this rate is simply the erosion rate of the surface. For a soil-mantled landscape, this is the rate of conversion of the underlying weathered bedrock to mobile material: the soil production rate. Soil production rates have been hypothesized and recently documented to decline exponentially with increasing soil thickness in a relationship termed the soil production function (Heimsath et al., 1997). Note that soil production rates only equal total erosion rates (including solute loss), i.e., landscape lowering rates relative to a local datum, if local soil thickness is roughly constant over time. This conceptual framework applies to upland, soil-mantled landscapes with no recent history of glaciation, without significant eolian deposition, and only across divergent (convex-up) noses where there is no net soil deposition (Dietrich et al., 1995; Heimsath et al., 1997, 1999). Field verification of this function helps quantify recent landscape evolution models (Dietrich et al., 2003), and the local steady-state soil thickness assumption was verified at one of the field areas used here (Heimsath et al., 2000).
In this paper we present new results from a comprehensive study done across the passive margin escarpment (cf. Great Escarpment; Ollier, 1982) of southeastern Australia. We use two in situ–produced cosmogenic nuclides, 10Be and 26Al, to quantify erosion rates and processes from the low-relief coastal lowlands to the highlands above the escarpment. These erosion rates are compared with two new, well-defined, soil production functions that we present from the highlands at the escarpment crest and from the coastal lowlands. These soil production functions illustrate differences in the processes eroding the highlands in comparison with the coastal lowlands and escarpment base. These 44 new cosmogenic nuclide data are used with two independent data sets published using cosmogenic nuclides and landscape morphology from the base of the escarpment and from the highlands above the escarpment to show that erosion rates decrease with elevation from escarpment base to crest and that soil production rates decrease exponentially with increasing soil thickness. Our findings are consistent with recent studies (e.g., Cockburn et al., 2000; Matmon et al., 2002; Persano et al., 2002; van der Beek et al., 2002) that concluded that passive margin escarpments have been relatively stable and are not retreating as rapidly as suggested by earlier studies of escarpment evolution. To place our results in the context of the extensive work done to understand landscape evolution across the southeastern Australia rift margin, we begin with a brief review.
ERODING AND EVOLVING THE ESCARPMENT
The passive continental margin of southeastern Australia provides an excellent example of a landscape where the processes and rates of evolution have been studied over multiple time scales (e.g., Bishop, 1988; Dumitru et al., 1995; Lambeck and Stephenson, 1985; Nott, 1992; O'Sullivan et al., 1996; Ollier, 1995; Persano et al., 2002; Seidl et al., 1996; Stephenson and Lambeck, 1985; van der Beek and Braun, 1998; Wellman, 1987). This margin is thought to have begun with the rifting between the Australian continent and the Lord Howe Rise to the east ca. 85–100 Ma (Hayes and Ringis, 1973; Weissel and Hayes, 1977). While there has been much debate about the subsequent uplift, erosion, and evolution of the landscape associated with this rift, the large-scale morphology is roughly similar to other passive margins around the world (Matmon et al., 2002), although an important difference exists in that the drainage divide occurs inland of the scarp crest. The highland region of gentle topography, low-relief, and relatively slow erosion rates (e.g., Bishop, 1986; Bishop and Brown, 1992; Bishop et al., 1985; Nott, 1992; Ollier, 1978; Pain, 1985; Wellman, 1987) is separated from the more deeply incised coastal belt by what was once commonly referred to as the Great Escarpment (Ollier, 1982). Modeling of the evolution of the escarpment is extensive (e.g., Lambeck and Stephenson, 1985; van der Beek and Braun, 1998) and depends on field determination of erosion rates, but much modeling of this and other escarpments has been done without empirical constraints on relevant time scales.
Slow rates of Tertiary erosion were deduced for the southeastern highlands, west of the escarpment, in several studies by dividing the age of plateau-forming basalt flows into the elevation difference between the plateau and the channels that had incised through the flows (Bishop, 1985; Bishop et al., 1985; Wellman, 1979, 1987; Wellman and McDougall, 1974). Because of the likely posteruption erosion of the basalt surface, the rates calculated were the minimum rates of incision. All of the studies deduced rates less than 10 m/m.y., with significant uncertainty, for the southeastern highlands, and slight differences depended on the time since basalt emplacement. Young (1983) and Young and McDougall (1982, 1993) extended the examination of river incision through dated basalts to infer slow rates of scarp retreat both west and east of the Great Escarpment, suggesting rates as low as 18 m/m.y.
The approach of determining long-term denudation rates for the region by using low-temperature thermochronology measured across the southeastern highlands, as well as along the present coastline, was pursued relatively early, though often with few data (Dumitru et al., 1991, 1995; Foster and Gleadow, 1991; O'Sullivan et al., 1995, 1996). These studies were motivated, at least in part, by debate over the origin, not the erosion rates, of the southeastern highlands. O'Sullivan et al. (1995, 1996) used apatite fission-track thermochronology (AFTT) to deduce two separate periods of rapid denudation based on distinct periods of rapid cooling recorded in the apatite. They suggested that the initial continental extension associated with the splitting of the Tasman Sea in the mid-Cretaceous resulted in kilometer-scale (>2 km) denudation across much of the present-day southeastern highlands, possibly due to rock uplift following underplating inland from the rift. The second period of cooling recorded in samples from the coastal regions was several tens of millions of years later, associated with the continental breakup during the Late Cretaceous (ca. 80 Ma) through the Paleocene (ca. 60 Ma). These results implied that the highlands that were uplifted and highly eroded 90–100 Ma and the more recently eroded coastal belt arrived at a position near their present morphologies ca. 60 Ma. As Bishop (1986) pointed out, most of the models of the history of the southeast Australian highlands (as opposed to models of the evolution of the continental margin) implied a stable continental divide, which roughly coincided with the Great Escarpment. The inferred tectonic processes leading to the stability of this divide and the southeastern highlands were not stated, but the conclusions regarding very slow Tertiary denudation rates on the highlands were similar to those drawn from a passive isostatic rebound model as presented by Stephenson and Lambeck (1985). Closer examination of O'Sullivan et al.'s research (1995, 1996) shows few data from potentially anomalous areas, and the supposed peak in denudation rates on the coastal strip is not supported by numerous data from other regions (Gleadow et al., 2002; van der Beek et al., 2001). Recent AFTT work (Persano et al., 2002) concludes that the highlands have been tectonically stable with relatively constant denudation rates throughout continental breakup, and suggests erosion rates <10 m/m.y. since the late Paleozoic–early Mesozoic, similar to findings by Bishop and Goldrick (2000) and Moore et al. (1986).
Slow denudation rates are part of the assumed conditions used by Thomas (1989a, 1989b, 1995), Twidale (1971, 1985, 1993), and Twidale and Vidal Romani (1994) to suggest the differential weathering processes leading to the formation of relief in some landscapes, as well as features such as inselbergs and tors. The etching process reported by these authors involves the development of a saprolite (chemically weathered rock that retains its relict structure but has not been physically transported) resulting from the prolonged interaction of groundwater with the underlying bedrock. The weathering front progresses into the bedrock, penetrating into fractures and joints, and weathers the bedrock preferentially into saprolite and corestones. Saprolite may then be stripped (for reasons and by processes not specified by these authors, but potential mechanisms could include climate change to a wetter, more humid period likely to cause higher rates of erosion) to expose the differentially weathered bedrock, resulting in the formation of relief in the landscape. Depending on the spacing of the bedrock fractures and joints, the exposure of the corestones may result in the formation of landscape features as diverse as tors, inselbergs, angular turrets, or granite domes.
Seidl et al. (1996) and Weissel and Seidl (1997) focused on the mechanisms for propagation of the escarpment inland following the splitting of the Tasman Sea. They used a mass-balance approach to estimate denudation rates ranging from 0.04 m/m.y. to 12 m/m.y. east (downstream, or below) of the escarpment since the time of rifting. These rates were consistent with the minimum estimates of Dumitru et al. (1991), despite the rough constraints on their measurements and their assumption of initial topography, which was not substantiated. They also inferred escarpment retreat rates (Seidl et al., 1996) by dividing the average distance of many gorge heads from the rift margin (200 km) by the period since rifting (100 m.y.) to get a long-term average rate of 2 km/ m.y. This rate agreed roughly with retreat rates cited from other passive continental margins (Ollier and Marker, 1985), though the same simple assumptions were used, and the reasoning was not mechanistic.
Weissel and Seidl's (1998) analyses of cosmogenic nuclide measurements from bedrock rivers determined bedrock incision rates of ∼7 m/m.y. from upstream of the edge of the escarpment (which agrees with highland denudation rates of 1–10 m/m.y. for the highlands), and greater than 100 m/m.y. inferred from strath terraces downstream of the escarpment knickpoint. As they point out, their rates were consistent with their models of escarpment propagation across the coastal region with very low rates of erosion from the highland region (∼4 m/m.y.), higher rates of bedrock channel incision near the upper edge of the escarpment than from upstream (∼20 m/m.y. versus ∼7 m/m.y.), and the highest erosion rates across the gorge head (∼100 m/m.y.). Importantly, both studies (Seidl et al., 1996, and Weissel and Seidl, 1998) forward a process-based argument for the propagation of the escarpment. They suggested that as long as there is enough fluvial transport power in the streams, the propagation rate of the escarpment is primarily determined by the slope failure mechanisms and the mass wasting from the steep rock slopes at the escarpment head (also note modeling results of Tucker and Slingerland , which suggested that weathering-limited conditions across the escarpment knickpoint set the retreat rate of the escarpment).
In contrast with the extensive research on the evolution and denudation of the southeastern highlands, there is a paucity of data addressing the erosional history of the coastal plain, and work that has been done appears fraught with equivocation. The notable exceptions were the low-temperature thermochronology studies that found a discrete period of rock cooling from rock samples along the coast (O'Sullivan et al., 1995, 1996; Persano et al., 2002). Perhaps one reason there have been few studies similar to those carried out on the highlands is that the retreat of the escarpment has eroded most of the material used in the earlier studies, so they must now use datable basalts to infer the evolution of the highlands from the coastal region. While the highly dissected region adjacent to and below the escarpment suggests high erosion rates, the actual rates have not been well quantified beyond a few recent studies (Heimsath et al., 2000; Persano et al., 2002; Weissel and Seidl, 1998).
Young (1977) did, however, report mantles of Tertiary sediment, duricrust, and basalt on the coastal lowlands at the foot of the escarpment. Young and McDougall (1982) reported further on this work and used datable rock to report K-Ar ages averaging ca. 30 Ma from basalt flows that postdate the silcrete formation on the lowland sites. They used the basalt ages in relation with the silcrete development to conclude that the coastal lowland resulted from erosion following an early Tertiary uplift of the highland. Similarly, based on terrestrial sedimentary deposits examined along the southeastern coast, Nott et al. (1991) concluded that the uplift of the highland region and the development of the coastal lowland occurred by the early Tertiary. They proposed that a late Mesozoic period of deep weathering was followed by an early Tertiary period of uplift and erosion. Subsequently, by the early Miocene (ca. 22 Ma), global sea-level rise may have led to a period of deposition of sediments on the present coastal plain. These aggraded sediments were exposed when the drop in sea level in the late Miocene to Pliocene (ca. 8 Ma) allowed the incision of streams into the coastal landscape. There is considerable dispute surrounding these lowland studies.
To quantify erosion rates and processes across the escarpment, we extended the field area that was used to quantify soil production rates with in situ–produced cosmogenic nuclides (10Be and 26Al; Heimsath et al., 2000; Figs. 1 and 2). Results from the Heimsath et al.'s (2000) study were later compared to soil production and erosion rates determined at a site on the highlands (Heimsath et al., 2001), but the highland site was roughly 60 km northwest of the escarpment base site and ∼12 km west of the escarpment crest. For the study presented here, we directly connect the coastal lowland with the escarpment by sampling along ridge and river transects (Fig. 3), as well as extensive new sampling on the highlands at the escarpment lip and also on the coastal lowlands. Sampling at the escarpment lip and on the highlands near Brown Mountain was also motivated by the desire to couple our quantification of physical erosion rates with recent and ongoing work into chemical weathering in the same field region (Banfield and Eggleton, 1989; Taunton et al., 2000a, 2000b).
The Nunnock River is a tributary to the Bemboka-Bega River, which drains into the Pacific Ocean at Bega (Fig. 1). The granite cliffs of the escarpment, downslope from the low-gradient headwaters region, and a relatively steep drop off the escarpment define the drainage area of Nunnock River. Evidence of debrisflow deposits are found in each of the second-order drainages downslope from the escarpment, though scour to bedrock on the escarpment face is relatively limited. The exposed cliff region across the top of the escarpment is not as dramatic or extensive as those from the Weissel and Seidl (1998) study area, but they are eroding by block failure across their face and thin exfoliation sheet erosion across their tops. There is little sediment accumulation beneath the cliffs, and it appears that the upper region of the escarpment is weathering-limited, as are regions of the channels (the channels are bedrock with the exception of patchy debrisflow cover). Spur ridges are soil-mantled and perpendicular to the escarpment. Slopes are sharply convex across the tops of the regions and steepen downslope to the channel (Heimsath et al., 2000). Between Nunnock River and the coastline, the landscape is composed of rolling, granodiorite hills with very few outcropping tors in comparison with the highland region. The granodiorites are Late Silurian to Early Devonian and are part of the Bega Batholith Igneous Suite (Lewis and Glen, 1995). Relief is <200 m across the coastal lowland and increases from the coast inland to the escarpment, which rises abruptly for ∼800 m directly west of the scarp base field site.
Rainfall on the coastal lowlands (∼900 mm/yr), and especially near the escarpment, is considerably higher than on the tablelands, though it does fall sporadically throughout the year (Richardson, 1976). Rainfall intensities are reportedly higher near the escarpment (personal commun. with landowners), though this has not been quantified beyond the past 10 yr. Using stations of the Australian Bureau of Meteorology (www.bom.gov.au/climate/averages) closest to the field sites, the average rainfall ranges from 870 mm/yr for the coastal lowland site (Snug, at ∼200 m elevation), to over 1200 mm/yr at the escarpment base (Nunnock River, at 400 m), to 690 mm/yr at the escarpment crest (Brown Mountain, at 1000 m), to 720 mm/yr at the highland site (Frogs Hollow, at 900 m). Observed vegetation at the escarpment base reflects this higher rainfall compared with the relatively dry and cold tablelands region, and forest cover is intermediate to wet schlerophyll forest with patches of dense understory, especially in the valley bottoms. Ferns grow widely on the more humid slopes around the Nunnock River field area, but are not observed on the slopes nearer the coast at the Snug site.
The field area of Heimsath et al. (2001) is representative of the low-relief plateau of the highlands above the escarpment, but is northwest of the Nunnock River portion of the escarpment leading to the coastal lowlands focused on here, motivating sampling another highland area around Brown Mountain. This highland site is the same region studied by Banfield and Eggleton (1989) and Taunton et al. (2000a, 2000b). The highlands, in general, are between ∼800 and 1500 m above sea level. Bedrock varies between Ordovician metasediments and Devonian granites across the region (Richardson, 1976). Morphology of the highlands directly above the Nunnock River site (i.e., Brown Mountain) is similar and is characterized by low-relief hillslopes (<100 m) that are gently sloping (<25°) and are soil-mantled, but punctuated frequently by tors. These tors are typical examples of the more coherent bedrock that outcrops through a saprolitic granite landscape (e.g., Linton, 1955; Twidale, 1985; Williams et al., 1986). The weathered nature of the soil-mantled saprolite is exposed in numerous road cuts around the field areas. These road cuts also reveal the prevalence of core-stones, some fraction of which are likely to emerge sometime in the future as tors.
The climate of the southeastern highlands is relatively cool, averaging ∼18–22 °C in the summer and 3–5 °C in the winter. There are periods of freezing temperatures and snow in the winter, but these rarely last long. Rainfall is not distinctly seasonal and, with the exception of drought periods, does not vary much from ∼550–750 mm/yr (McAlpine and Yapp, 1969; Richardson, 1976). Forest cover, where it occurs, is typically dry schlerophyll, with very little understory. Grassland and rolling savanna woodland vegetation are also widespread across this region of the tablelands. Climate changes during the Pleistocene and during the Pleistocene–Holocene transition (Bryant et al., 1994; Chappell, 1991; Dodson, 1977, 1987; Dodson and Wright, 1989; Harrison, 1993; Kershaw et al., 1991) probably affected the biotic communities and therefore the dominant soil production and transport processes. These processes were likely similar to the Nunnock River site since the biota are similar, but the drier and colder climate may reduce the frequency of the bioturbation at the highland sites. Potential impacts of historic agriculture were avoided by selecting protected areas for our work.
Soil production and transport is observed to be primarily due to biogenic processes across all field areas, although there is some evidence of overland flow, especially during storms following fires that remove any vegetative understory. Burrowing wombats were the most obviously disruptive biota and were observed to occupy all parts of the landscape except where soil thickness was less than ∼30 cm, and their burrows extend into the weathered bedrock in places. Other observed biogenic processes included burrowing echidnas and shallow burrowing lyrebirds, worms (worm trails extended into the saprolite and worm castings in the soil included quartz crystals that could have been conveyed up from the underlying saprolite), and tree throw (Fig. 4). Ant or termite mounds revealed significant quartz grains and pieces of unweathered granodiorite, but we did not observe ant burrows extending into the saprolite; therefore, we infer that such micro-fauna contribute to transport rather than production of the soil.
In situ–produced cosmogenic nuclides (e.g., 10Be and 26Al) are used increasingly to quantify bedrock erosion (e.g., Lal, 1991; Nishiizumi et al., 1991), soil production (e.g., Heimsath et al., 1997; Small et al., 1999), and spatially averaged rates of erosion (e.g., Bierman and Steig, 1996; Brown et al., 1995; Granger et al., 1996). Concentrations of these nuclides accumulate in materials at or near Earth's surface as cosmic rays bombard atoms, such as Si and O in quartz and other minerals in rock and sediments (Lal and Arnold, 1985; Lal and Peters, 1967). Cosmicray production of nuclides is balanced by radioactive decay of the nuclides in surfaces that are not eroding and by removal of the target material in eroding surfaces. Application of cosmogenic nuclide analysis to understanding landscapes is extensively and well-reviewed (Bierman, 1994; Cockburn and Summerfield, 2004; Gosse and Phillips, 2001; Lal, 1988; Nishiizumi et al., 1993) and will only be summarized here.
Nuclide concentrations measured from quartz have been shown to reflect the exposure history of the sample and are dependent on the production and decay rates of the nuclide, as well as the erosion rate of the sample (Lal, 1988; Lal and Arnold, 1985; Nishiizumi et al., 1986, 1991). As shown by Lal (1991), and neglecting muon effects, the nuclide concentration, C, depends on the production of the nuclide at an exponential depth, zx, function of the surface production rate, P0, and its radioactive decayconstant, λ,equation 1 can be solved analytically (Lal, 1991), yielding, Dunne et al., 1999). Under steady-state conditions, where local soil thickness is relatively constant with time, soil production, ϵ s, equals erosion, ϵ, at the surface, which is also equal to the rate of lowering of the soil-saprolite interface (Heimsath et al., 1997) and can therefore be equated to the local denudation rate. Equation 2 can therefore be solved to yield steady-state soil production or erosion rates. Nuclide concentration accumulated in river or catchment sediments can be interpreted by equation 2 to infer catchment-averaged steady-state erosion rates, assuming relatively short transport times and relatively homogeneous catchment lithology (e.g., Granger et al., 1996; Bierman and Steig, 1996).
We apply this methodology to determine erosion rates in three different ways. Most simply, we sample the outermost centimeter or two of exposed bedrock (tor tops as well as the top of the cliff region of the escarpment) that shows field evidence of eroding by grain-grain spallation or by exfoliation of thin sheets. Concentrations of 10Be and 26Al in these samples yield the inferred steady-state erosion rate of the exposed bedrock surfaces, assuming that samples have had simple exposure histories (e.g., no significant burial or shielding that is now not evident) and have been eroding at nearly constant rates (i.e., not by large block failure or landsliding). For soil production rate determination, we followed the typical geomorphic transect from bedrock exposed at convex ridge crests onto the soil-mantled part of the landscape and dug soil pits to sample the top centimeter or two of saprolite or weathered bedrock directly beneath different depths of the relatively thin (<1 m) soil cover. Concentrations of nuclides from these samples were used with a corrected nuclide production rate that accounts for the shielding of the soil mantle above the sampled layer to yield a soil production rate, which is equivalent to a landscape-lowering rate relative to a local datum (Heimsath et al., 1999). Finally, we collected stream and creek sediments or sediment from zero-order catchments, measured the concentrations of the nuclides, and used the nuclide concentrations to infer a catchment-averaged erosion rate (e.g., Granger et al., 1996). Each of these applications depends on careful geomorphic site selection, such that the numerous assumptions and simplifications detailed in the above studies are met.
Roughly 500 g of bedrock, saprolite, or sediment from each sample were crushed, sieved to separate particles <1 mm size, and chemically purified following the procedure of Kohl and Nishiizumi (1992) to yield ∼50 g of quartz, from which Be and Al were extracted. For results reported here we only measure concentrations of 10Be, because sample lithology and exposure history were similar to those reported in Heimsath et al. (2000, 2001) and did not warrant the test for steady-state exposure history enabled by measuring 26Al (e.g., Nishiizumi et al., 1993). Samples for 10Be measurement were spiked with a 9Be carrier solution calibrated by a Be atomic absorption standard that differed by <2% from the Be carrier used for the Nishiizumi et al. (1989) analyses. Preparation and chemistry on these samples were done at Dartmouth College as well as at the Research School of Earth Sciences at the Australian National University (ANU). We measured concentrations of 10Be at the Lawrence Livermore National Laboratory, Center for Accelerator Mass Spectrometry (LLNL-CAMS) facility (Davis et al., 1990) as well as at the ANU Nuclear Physics facility and Purdue University's PRIME Lab, and standardized the measurements to the ICN (ICN Biomedical, Inc.) 10Be standards. By doing the chemistry on these samples in two very different labs and measuring the nuclide concentrations on three very different accelerator mass spectrometers, we also have an excellent test for the potential analytical uncertainties in the cosmogenic nuclide methodology used for geomorphic studies.
Production rates for 10Be and 26Al in quartz were based on the sea-level and high-latitude production rates of 6 and 36.8 atoms g−1 yr−1 (Nishiizumi et al., 1989), rescaled to 5.1 and 31.1 atoms g−1 yr−1, respectively (Stone, 2000), and were corrected for latitude and altitude effects (Lal, 1991), as well as for the slope and shielding of the sample (Dunne et al., 1999; Masarik and Vanya, 2000; Nishiizumi et al., 1989). These production rates were used for consistency with previous reports of erosion rates or exposure ages, despite the growing debate over production rates (Clark et al., 1995; Dunai, 2000; Nishiizumi et al., 1996; Stone et al., 1998a) and the potential contribution of muons to nuclide concentrations under moderate and high erosion rates (e.g., Granger and Smith, 2000; Stone et al., 1998b).
RESULTS AND DISCUSSION
Results from the new nuclide analyses are reported here for 10Be concentrations, with the erosion rates or soil production rates determined from them (Table 1). Table 1 does not distinguish between samples prepared at different laboratories or measurements made at different accelerators. The results presented below incorporate these differences and agree so well that we believe the soundness of our methodology is maintained. Seven samples from directly beneath the actively eroding soil mantle define a soil production function for the highland landscape at Brown Mountain (Fig. 5A). Two samples from bedrock exposed at the soil surface yield steady-state erosion rates and suggest that maximum soil production rates occur under some finite soil mantle (Ahnert, 1987; Carson and Kirkby, 1972; Dietrich et al., 1995; Gilbert, 1877), such that the soil production function could have a “humped” form (Cox, 1980). Measured nuclide concentrations were corrected for observed soil depths (Heimsath et al., 1999), and steady-state erosion rates, calculated with equation 2, define an exponential decline of apparent soil production rates with increasing soil depths for the saprolite samples. The variance-weighted best fit to these points, excluding the bedrock points, isTable 1 in m/m.y., and H is in cm. Comparison of these data with the soil production and exposed bedrock erosion rates reported for the highland site of Heimsath et al. (2001) shows remarkable similarity in rates and form despite the sites being nearly 60 km apart and the samples being collected and analyzed about four years later in a different laboratory (Heimsath et al., 2001, analyses were done at the University of California, Berkeley; Fig. 5B). The apparent soil production function reported by Heimsath et al. (2001) was fit to five samples using both 10Be and 26Al from beneath soil thicknesses from 25 to 65 cm and had a dramatically steeper slope to it, such that the function defined an intercept that was unreasonably high for any rates measured across the highlands: (Fig. 5B) would define a humped soil production function, as first articulated by Carson and Kirkby (1972) and also discussed by Cox (1980). This is discussed below, after presentation of all the soil production data.
Similarities in morphology between the two highlands sites suggest that the conclusions reached by Heimsath et al. (2001) are applicable to the escarpment crest region. Specifically, that extensive outcropping tors are found at both sites and that the weathering profiles visible in the road cuts, and analyzed at a site very close to our Brown Mountain site by Banfield and Eggleton (1989) and Taunton et al. (2000a, 2000b), support the notion of a propagating weathering front into the relatively stable highlands. Their measurements suggest considerable erosion by solute processes, while our cosmogenic nuclide–determined lowering rates are recording a total lowering rate with no differentiation between chemical and physical weathering processes. It is clear from the weathering profiles that solute processes are significant, and it is also clear from the active bioturbation across the landscape that physical soil production and transport are significant. At this point, we cannot resolve the relative roles that both play, though it is clear that they are coupled as Riebe et al. (2001, 2003) found elsewhere, and we are examining the problem with a detailed look at how the extent and rate of weathering varies across the landscape (Burke and Heimsath, 2003).
Average erosion rates from 10Be concentrations from stream sediments taken from small catchments draining to Nunnock River above the escarpment range from 13 ± 1.7 to 20 ± 2.5 m/m.y. (Table 1; Fig. 3B). These rates are at the low end of the escarpment erosion rates discussed below, but agree well the highland soil production rates shown in Figures 5A and 5B. This is encouraging support for the use of small catchment sediments to quantify average erosion rates for a landscape, as the processes of soil production and transport are the dominant geomorphic forces shaping the highlands. Landslides and extensive gullying are only observed where there is a clear link between overgrazing by livestock and local degradation of the land surface following increases in rainfall. Tor erosion rates are slower, as discussed in Heimsath et al. (2000, 2001), and by quantifying the sediment budget for the catchment, the interpretation of the nuclide concentrations in the sediment samples reflects well the proportional contributions of tor erosion and soil production from the saprolite (Granger et al., 2001; Heimsath et al., 2001). The overlap between average and point specific rates supports the suggestion that the highlands may have reached a steady-state since emerging from the last glacial cycle, when periglacial processes were likely to be dominant, a suggestion that was raised in the “Departures from Steady State: (II) Soil Production” section of Heimsath et al. (2001, p. 180–181).
Ten samples from the saprolite-soil boundary at the coastal lowland site (Snug) define soil production rates for soil depths between 27 and 90 cm (Fig. 6A). No intermediate soil depths were observed, and the sample from bedrock exposed at the soil surface led to a lower soil production rate than the sample from beneath the shallowest soil, a finding similar to the highland sites. Analyses of the nuclide concentrations followed procedures outlined above, correcting for overlying soil thickness to infer steady-state soil production rates. Using the saprolite soil production rates to determine a variance-weighted best fit for the soil production function, we determine that,Heimsath et al. (2000). The notable difference between the Snug and Nunnock River sites was the absence of intermediate soil thicknesses, and further investigation is required to decided whether this is true across the coastal lowland region, as we observe it to be for the highlands, or if there is the potential for human impact that may have left a plow layer. A plow layer is a surficial layer of soils with near-uniform thickness that would be present if any plowing of the site had homogenized soils down to a finite depth, typically ∼30 cm. The soil production function defined from Nunnock River, at the base of the escarpment, was from both 10Be and 26Al concentrations measured in 15 samples of saprolite and bedrock and showed the following very well-defined exponential decline of soil production rates with increasing soil thicknesses (Heimsath et al., 2000): (Fig. 6B). This overlap is surprising, as the coastal lowlands have lower relief and are more gently sloping than the soil-mantled slopes adjacent to the escarpment base. Conventional wisdom would suggest that the lowlands and highlands would be eroding more slowly than the escarpment base, yet our cosmogenic nuclide data suggest otherwise.
If we plot soil production rates from all four of the field sites from lowland to highland, the overlap becomes especially clear (Fig. 7). Indeed, using all these data, again with the exception of the exposed bedrock samples that are eroding more slowly (discussed below), to determine a variance-weighted soil production function tightens the relationship of equation 6 such that
That exposed bedrock at the highland and lowland sites appears to be eroding more slowly than the maximum soil production rates determined under relatively thin depths of soil raises an important distinction between soil production functions (Carson and Kirkby, 1972; Cox, 1980; Dietrich et al., 1995). Specifically, if lithology is uniform beneath the actively eroding soil layer, then lower rates of erosion for exposed bedrock document the hypothesis articulated by Gilbert (1877) from his observations that weathering processes attributed to frost and solute processes would occur most rapidly beneath a thin soil cover. A thin soil cover would enable moisture retention at the soil-rock interface, such that chemical and physical weathering processes could occur. Dietrich et al.'s (1995) modeling yielded an unreasonable modeling prediction of the spatial variation of soil thickness using such a humped soil production function. Specifically, they used both humped and exponentially declining soil production functions in a model predicting spatial patterns of soil thickness at the same soil-mantled field site in northern California, which was used subsequently by Heimsath et al. (1997, 1999). Using the humped function predicted widespread bedrock exposure not observed in the field, while using the exponential decline predicted soil depths consistent with observations. Carson and Kirkby (1972) explained that having maximum soil production rates beneath a finite soil thickness would lead to landscapes with either exposed bedrock, or soil thicknesses greater than the finite depth, with no intermediate soil depths. This suggestion was documented for the highland site of Heimsath et al. (2001), where modeling spatial variation of soil depth led to a better prediction of soil thickness when the exponential decline function was used, and Small et al. (1999) presented similar findings for an alpine landscape. Ahnert's (1987) modeling presented a different perspective, which was that variations in the underlying bedrock could lead to soil depth variations across a hypothetical landscape eroding at the same rate everywhere. Importantly, variations in lithology documented at the highland site may help to explain why the soil production functions from the highlands and lowlands are different from those presented from the escarpment base and northern California (Heimsath et al., 1997, 1999), emphasizing the importance of knowing the chemical weathering story by quantifying both the extents and rates of chemical weathering across all sites following, for example, methods of Riebe et al. (2001, 2003).
Across the Escarpment
Eight samples define catchment-averaged erosion rates from small, first- and zero-ordered streams draining the escarpment to Nunnock River, while ten samples from a main spur ridge define point-specific erosion rates for an altitudinal transect along a main spur ridge from coastal lowland to the Brown Mountain site (BM; Table 1). We include tops of three granite tors from the escarpment crest, near Brown Mountain, as well as three samples from the tops of outcropping granite faces along the steep escarpment face, where the only cliff exposures are found for this region of the escarpment (plotted as Tor Top). Together these 24 samples, along with average soil production rates from the scarp base site and the lowland site (plotted as separate ridge crest points), yield a striking trend of erosion rates as a function of elevation (Fig. 8), from 3.26 ± 0.28 m/m.y. at the scarp crest to an apparent peak of 36.3 ± 4.8 m/m.y. near the scarp base and declining to 24.5 ± 2.5 m/m.y. across the coastal lowlands. Local agreement between the rates determined for different processes suggests local steady-state between soil production, bedrock erosion, and small channel incision.
There are at least two mechanistic reasons to predict this order of magnitude decline in erosion rates, which are not entirely intuitive, because conventional wisdom might suggest increases in erosion rate with elevation. Firstly, observations that precipitation decreases with elevation across this escarpment would help explain the decline in erosion rates with elevation if there were a clear process link. At this stage we can simply suggest that higher precipitation at the escarpment base may support higher densities of biota, which are, in turn, likely to be more active across the landscape, potentially leading to higher erosion rates. While this suggestion is extremely speculative and needs field testing, it is critical that a process-based explanation be determined. Interestingly, the order of magnitude decrease of denudation with a <1000 m rise in elevation reported here is nearly the opposite of the less than a factor of 1.4 difference in rates reported by Riebe et al. (2004) across a <700 m altitudinal transect used as a steep climatic gradient in the Santa Rosa Mountains of Nevada. Their observations suggest a sharp decline in chemical weathering with altitude by coupling precipitation with total denudation, which may play a role here. We are currently testing this hypothesis. Secondly, both increasing stream power with distance downslope and increasing relief at the escarpment base would predict higher erosion rates with lower elevations across our field areas. Morphometrically, the correlation shown in Figure 8 fits with the low relief and gently convex-up ridges near and upslope of the scarp crest, as well as across the coastal lowlands, in comparison with the much higher relief and sharply convex ridge crests downslope on the scarp and onto the escarpment base site of Heimsath et al. (2000; see inset topographic map of Fig. 1). That is, lower erosion rates would be expected at the top of the escarpment and out on the coastal lowland based on the morphology, which is what the data show (note Fig. 6A, the soil production and bedrock erosion rates from Snug at an elevation of just over 200 m). The highest rates shown by Figure 8 are from the sharply convex-up ridge crest across and just upslope of the Heimsath et al. (2000) study area and from small catchments draining that ridge. Local relief, i.e., elevation difference between main spur ridge and the second- and third-order tributary draining the escarpment to Nunnock River, is greater at these sample sites than it is anywhere else on the escarpment, on the order of 50–60 m, and we observed debrisflow deposits in the main tributaries.
Increases in both slope and drainage area (a proxy for discharge) to the main tributaries draining the escarpment, of which Nunnock River is one example, with distance from the headwaters would predict the highest stream incision rates near the base of the escarpment. While the higher relief and over-steepened slopes adjacent to the streams at lower elevations at the escarpment base would suggest that stream incision rates are greater than ridge crest lowering rates, the lower elevation samples plotted in Figure 8 show similarity between the rates, within error. Importantly, the open boxes are average rates of first- and zero-order catchments draining from the spur ridges into the main tributaries, and are therefore not measures of the incision of Nunnock River or of the other main incision dissecting tributaries. We specifically chose the small catchments to avoid input of sediments or bedrock eroded by debris flows or landslides, such that the nuclide concentrations in the sediments would reflect average denudation rates of the upslope contributing area (Granger et al., 1996; Bierman and Steig, 1996). That ridge crest and basin averaged rates are similar suggests that the biogenic processes driving hillslope erosion are keeping up with the increased fluvial incision at the escarpment base, such that there is the suggestion for local steady-state where these rates are nearly equal.
Recent studies using low-temperature thermochronology and cosmogenic nuclides inferred denudational histories for passive margin escarpments, including a transect across the same field area that we are using here, and concluded that rift escarpments experience most of their erosion and retreat soon after rift development (Cockburn et al., 2000; Matmon et al., 2002; Persano et al., 2002). These studies provide important new constraints for the denudational histories of escarpments, which are used in modeling efforts that had previously predicted the same finding (Gilchrist et al., 1994; van der Beek and Braun, 1998). The mechanism suggested by Cockburn et al. (2000) and predicted by the above modeling studies, is that a major inland drainage divide fed rivers that rapidly eroded the newly formed escarpment shortly after formation. Studies from the southeastern Australia escarpment further north of our study area focused on the mechanisms of escarpment propagation, but could not resolve the greater than two orders of magnitude difference between inferred retreat rates of ∼3 km/m.y. and denudation rates of 0.04–12 m/m.y. estimated since the time of rifting (Seidl et al., 1996; Weissel and Seidl, 1997, 1998). Rates of surface processes reported here and in Heimsath et al. (2000) are more than an order of magnitude lower than the inferred local escarpment retreat rate of ∼500 m/m.y. (dividing the roughly 50 km distance between the coast and scarp lip by 100 m.y., the rough time since rifting), highlighting the need to resolve the discrepancy between inferred and observed rates of escarpment retreat. If, instead of using the gross estimate of horizontal retreat obtained by dividing distance from the coast by time since rifting, we use the geometric relationship between lowering rates and retreat rates (where retreat rate equals lowering rate divided by the tangent of the mean escarpment angle), then a contemporary retreat can be estimated. Specifically, lowering rates range from ∼12 to ∼36 m/m.y. (excluding tors) across the scarp with an average gradient of 0.25, suggesting an average retreat rate between ∼40 and 140 m/m.y., which is an order of magnitude slower than the rate required to retreat the escarpment from the coastline at a continuous pace since rifting. If escarpments such as the one studied here did indeed retreat rapidly soon after rifting and then settle into a period of relative erosional stability, as our measurements and the above thermochronology studies suggest, then a geomorphic mechanism needs to be discovered that can accomplish the initial rapid retreat.
Our study used 44 new measurements of the in situ–produced cosmogenic nuclide, 10Be, combined with two previously published studies using 46 measurements of both 10Be and 26Al, all measured from bedrock, saprolite, and stream sediments, to enable a comprehensive quantification of the erosion rates and processes across the passive margin escarpment of southeastern Australia. These rates further our understanding of the surface processes shaping landscapes with different climatic and tectonic histories. Specifically, we have documented two new relationships between soil production and the overlying soil thickness, termed the soil production function. These two new functions, from the coastal lowlands and from the highlands near the study areas used for previous studies, agree remarkably well with the earlier findings. We then connect the lowlands with the highlands by reporting erosion rates from samples taken from two different transects across the escarpment face. The first transect samples the convex-up ridge crest leading upslope from the study area at the escarpment base to the highlands above the escarpment lip. The second transect begins near the headwaters of the major escarpment-dissecting river and samples small catchments draining the spur ridges from the highlands to just upslope of the scarp base study area. Rates from these transects define a linear decrease in erosion rate with elevation across the face of the escarpment, with an apparent peak in denudation rate at the scarp base, and agree with the rates determined by our soil production studies. We conclude that biogenic processes of soil production and removal are keeping up with other observed processes of debris flows and bedrock landsliding (that we did not quantify here) across the escarpment to set the overall rate of escarpment retreat. This rate is at least an order of magnitude less than previously inferred, long-term scarp retreat rates long suggested by workers focusing on the problem and therefore calls for the need to resolve the discrepancy between observed rates of landscape denudation and inferred rates of scarp retreat.
We thank Rishi Satyadharma (Steve Bateman) and his family for continued access to and unbounded hospitality at the Nunnock River field area. Bill Dietrich, Ian Prosser, and Kuni Nishiizumi helped initiate this work across the escarpment, and continued discussions have helped our thinking immensely. Paul Rustomji sampled and helped to process the nuclide samples from the coastal lowland site, and we thank him. Comments by Mike Summerfield and an anonymous reviewer helped improve the original manuscript. Cosmogenic nuclide measurements and field support were funded by Dartmouth College and National Science Foundation grants DEB-0128995 and EAR-0239655 to Heimsath. Thirteen nuclide measurements for this study and all the previous analyses were partially performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory (LLNL) under contract W-7405-Eng-48, while the research schools of Earth Sciences and Physical Sciences and Engineering at the Australian National University supported the rest.
Figures & Tables
Tectonics, Climate, and Landscape Evolution
- absolute age
- alkaline earth metals
- cosmogenic elements
- erosion rates
- landform evolution
- New South Wales Australia
- passive margins
- radioactive isotopes
- Bega River basin