The primary goal of this study is to provide background rates (over millennia) of landscape change in the George River basin using the in situ cosmogenic isotope beryllium-10 (10Bei) in fluvial sediment (Bierman and Steig, 1996; Brown et al., 1995; Granger et al., 1996). 10Bei production decreases exponentially with depth in rock and sediment near Earth's surface such that 10Bei concentrations at depths > 2 m are much lower compared to those measured closer to Earth's surface (Gosse and Phillips, 2001; Lal, 1991). 10Bei produced by muons dominates at depths > 2 m (Braucher et al., 2003; Gosse and Phillips, 2001; Heisinger et al., 1997), but muogenic 10Bei production is negligible when compared to near-surface spallogenic 10Bei production, except in rapidly eroding landscapes or landscapes with steep terrain (e.g., Dethier et al., 2014; Fellin et al., 2017; Rosenkranz et al., 2018; Scherler et al., 2014; Siame et al., 2011). Bioturbation homogenizes 10Bei concentrations in soils, in many places to depths of at least ∼ 1 m (Brown et al., 1995; Schaller et al., 2018), and thus 10Bei erosion rates are largely insensitive to widespread shallow erosion. This insensitivity allows 10Bei erosion rates to be a useful gauge of pre-disturbance rates of landscape change (Ferrier et al., 2005; Portenga et al., 2019; Schmidt et al., 2018; Vanacker et al., 2007), except where human land use is intensive (i.e., Schmidt et al., 2016) or the effects of human land use are exacerbated by climate extremes (i.e., Rosenkranz et al., 2018). Pre-disturbance 10Bei erosion data can thus inform approaches to reducing sediment delivery from the George River and support efforts to improve the ecological health of the Georges Bay estuary and possibly other watersheds in northeast Tasmania that share similar bedrock and topographic characteristics by providing a benchmark against which to compare modern sediment loads.

In addition to 10Bei, which is produced in rock and sediment, 10Be is also produced via spallation of oxygen and nitrogen in the atmosphere; this 10Be rains out or falls to Earth's surface (meteoric 10Be; 10Bem; Heikkilä and von Blanckenburg, 2015; Monaghan et al., 1986; Reusser et al., 2010) where it is readily adsorbed into sediment grain coatings. 10Bem has traditionally been used to trace sediment through landscapes (Brown et al., 1988; Helz et al., 1992; Portenga et al., 2017; Reusser and Bierman, 2010; Valette-Silver et al., 1986), but recently derived equations (along with a series of assumptions) now allow denudation rates to be calculated from measurements of 10Bem that are normalized to non-cosmogenic, stable 9Be, which weathers out of mineral grains (9Bereac; von Blanckenburg et al., 2012). 10Bem/9Bereac denudation rates have been used to quantify landscape evolution over a variety of spatial scales for different river basins (Dannhaus et al., 2018; Deng et al., 2020; Portenga et al., 2019; Rahaman et al., 2017; Wittmann et al., 2012, 2015; in some cases 10Bem is referred to as the reactive phase of 10Bem [10Bereac], and denudation rates may be referred to as 10Bereac/9Bereac denudation rates) and have shown promise in quantifying landscape dynamics in quartz-poor landscapes (Deng et al., 2020; Rahaman et al., 2017).

In this study, we use both 10Bei and 10Bem/9Bereac to measure the rates at which mass is lost from the George River basin's slopes. Over timescales sufficiently long that the assumption of steady state is approached, all of this mass will transported to the Georges Bay. Such mass loss from the George River basin is both chemical (dissolved load) and physical (sediment transport). The partitioning between these phases differs dramatically around the world depending on rock type, topography, and weathering regime and likely differs within the study basin. The assumptions underlying these two methods (10Bei and 10Bem/9Bereac) differ; thus, results from each method may not be the same. The concentration of 10Bei is biased towards mass loss within the upper meters of Earth's surface where rates of neutron spallation are high. Both chemical and physical mass losses within this surface layer of regolith are reflected by 10Bei concentrations. 10Bem/9Bereac, if the assumptions of the analytical model are met, reflects both physical and chemical mass loss throughout the regolith, regardless of depth.

The terms “erosion” and “denudation” have been used without precision in the literature, often as a replacement for one another. Erosion is applied more often to rates calculated using the concentration of 10Bei; rates calculated using 10Bem/9Bereac are more frequently referred to as denudation. We follow that convention in this paper. Because we have dissolved and suspended load data as well as river flow over time from the mouth of the George River, we attempt to provide a full discussion of what the rates we measure mean for landscape dynamics within the George River basin.

The small size and relatively uniform bedrock geology of the George River basin provide an ideal location to compare 10Bei erosion rates with 10Bem/9Bereac denudation rates (von Blanckenburg et al., 2012). 10Bem can be desorbed from sediment grain coatings under low-pH conditions (Aldahan et al., 1999; You et al., 1989), but 10Bem loss from soil profiles in solution is likely minimal in the George River basin, because measured soil pH values in the catchment range from 4.0–5.5 (Kidd et al., 2015) and long-term monitoring of stream water pH at two gauging stations – one in Ransom Creek and the other at the George River in St. Helens – shows that stream pH is consistently > 5 and mostly > 6 (DPIPWE, 2021a, b). The George River basin is a landscape of relative geological homogeneity in comparison to more geologically diverse landscapes with similar datasets (i.e., Deng et al., 2020; Portenga et al., 2019; Rahaman et al., 2017). Although the George River has a simple bedrock geology, it also has a long history of forestry and lode and placer tin mining that has, in the past, disturbed the hillslopes and fluvial systems (Knighton, 1991; Preston, 2012). Given that land use is speculated to affect the results of 10Bem/9Bereac-derived denudation rate calculations elsewhere (Portenga et al., 2019), we also explore how land use in the George River affects our interpretations of 10Be-based erosion and denudation calculations in this study.

Sediment samples for this study were collected in 2008 from several locations along the trunk (n = 2) and tributaries (n = 7) of the George River (samples TG-1 through TG-9; Fig. 2; Table 2), upstream of which channels are generally concave-up and therefore in geomorphic steady state (Fig. 5). At each site, sediment was collected from the streambed and/or in-channel bars to ensure active fluvial transport and mixing. Samples were sieved in the field to the 250–850 µm grain-size fraction. Although this grain size is finer than the mean natural grain size (30–50 mm; Knighton, 1991), previous studies show that 10Bei grain-size bias is minimal or not present in small, low-elevation, low-relief, temperate landscapes where landslides are uncommon (van Dongen et al., 2019); thus, 10Bei measured from the 250–850 µm grain-size fraction at George River can be interpreted as a geological erosion rate.

10Bem and the weathered reactive and silicate-bound in situ phases of 9Be (9Bereac and 9Bemin, respectively) were measured only from the 250–850 µm grain-size fraction from all seven tributary sites (TG-2 through TG-8) and one of the trunk channel sites (TG-9). When 10Bem is normalized to 9Bereac following von Blanckenburg et al.'s (2012) denudation rate equation, grain-size biases in resulting 10Bem/9Bereac-based denudation rates are diminished (Wittmann et al., 2012). Singleton et al. (2016) also showed the diminishment of grain-size bias in stream sand for 10Bem measurements when normalized to 9Bereac. Although it is possible to calculate erosion rates from 10Bem alone (Brown et al., 1988; Harrison et al., 2021; Willenbring and von Blanckenburg, 2010), this method does not include any normalization to 9Bereac, and 10Bem erosion rates are thus susceptible to grain-size bias, especially if the full grain-size distribution is not known and/or has not been analyzed. As our samples are of one grain-size fraction and were collected and sieved in the field prior to 10Bem erosion rate derivations, we only present 10Bem/9Bereac-based denudation rates in this study.

10Bei was extracted at the University of Vermont from quartz from each sample following standard methods, during which a known amount of a 9Be carrier (9Becarr) was added to each sample (Kohl and Nishiizumi, 1992; Corbett et al., 2016); relative to the amount of 9Becarr, no significant native Be was found in quartz concentrates from any sample, which can otherwise lead to significant overestimates of 10Bei-based erosion rates (Portenga et al., 2015). 10Bei/9Becarr ratios were measured by accelerator mass spectrometry at the Lawrence Livermore National Laboratory CAMS facility (Table 3); 10Bei measurements were blank-corrected (the average ratio of three blanks was subtracted from the ratio of each unknown sample) and normalized to the 07KNSTD3110 accelerator mass spectrometry 10Be standard material, which has a nominal 10Be/9Be ratio of 2.85 × 10-12 (Nishiizumi et al., 2007). 10Bei production was averaged across all sampled basins to a single point following Portenga and Bierman (2011), and the online erosion rate calculator described by Balco et al. (2008), which has been subsequently updated, was used to derive 10Bei erosion rates following the Lal (1991) and Stone (2000) scaling schemes (ε, Tables 4 and 5). Here, ε is presented in units of Mgkm-2kyr-1 (Table 5) allowing us to compare measurements of ε directly with 10Bem/9Bereac-based denudation rates (Dm; see below). Muogenic production of 10Bei is incorporated into ε; however, muogenic 10Bei is negligible relative to spallogenic 10Bei production given the George River's post-orogenic, low-elevation, low-relief setting.

10Bem was extracted following Stone's (1998) fusion method, and a 9Be carrier solution was added to each sample. Through this process, some amount of 10Bei from bulk sediment is incorporated into the 10Bem sample; however, the amount of 10Bei is negligible, consistently 2 orders of magnitude less than 10Bem measurements (Table 3). 10Bem/9Becarr ratios of these fusion extracts were measured at the Lawrence Livermore National Laboratory CAMS facility, blank-corrected (ratio of one blank was subtracted from ratio of each unknown sample; Table 3) and normalized to the 07KNSTD3110 standard material (Nishiizumi et al., 2007). Sample material used to calculate 9Bereac was first subject to 6 N HCl acid leaching to remove sediment grain coatings (Greene, 2016; Portenga et al., 2019, their supplement); it was then fully digested in HF, and 9Bemin was measured in that solution. Both 9Bereac from sediment grain coatings and 9Bemin from the remaining mineral material were measured by inductively coupled plasma-optical emission spectrometry (ICP-OES) at the University of Vermont. Together, these data were used to derive denudation rates following von Blanckenburg et al. (2012; Table 4); two variables required to calculate denudation rates that we did not or were not able to measure are the deposition rate of meteoric 10Be (10BeFmet) and the amount of 9Be that is naturally occurring in bedrock (9Beparent). We use estimated values of 10BeFmet based on global deposition models presented in Graly et al. (2011), because it provides an estimate of 10BeFmet that is specific for each sampled basin (8.55 × 105 atomscm-2yr-1 to 1.46 × 106 atomscm-2yr-1). At the time of sample collection (2008), the equations to calculate Dm had not been published, and bedrock samples from the field area were not collected. We therefore use a value of 4.1 ppm for the amount of 9Beparent in our samples because the George River basin is underlain by biotite granites, and the average 9Beparent value of biotite granites comprising a subset from over 200 felsic intrusions measured across China and the Soviet Union in the mid-1900s was reported to be 4.1 ppm (Beus, 1962; additionally reported in Sainsbury, 1964). We discuss the use of Graly et al.'s (2011) 10BeFmet estimates and Beus's (1962) average 9Beparent for biotite granites in the Discussion section. In this study, Dm is presented in units of Mgkm-2yr-1 (Table 5).

We compare ε and Dm to various topographic and land-use factors to assess possible processes driving or related to background landscape evolution in the George River (Tables 1 and 2). Topographic data are derived from the Shuttle Radar Topography Mission (SRTM) 90 m resolution global dataset (Gallant et al., 2011). Mean local relief was calculated over a moving 10-cell (∼ 900 m) circular window. We do not compare ε or Dm to climate data from global gridded datasets for mean annual temperature and mean annual precipitation, although such data are available. This is because the gridded datasets are all models based on limited measurements and include a strong elevation component in their interpolation scheme (e.g., WorldClim, Fick and Hijmans, 2017) or have spatial resolutions that do not provide sufficient detail for the small size of the George River basin (e.g., TRMM, Huffman, 2021). These characteristics of gridded climate datasets make it difficult to attribute erosion to climatic drivers independent of their self-correlation with elevation. Thus, we rely on observed relationships between elevation and precipitation and temperature data from precipitation gauges (n = 10, each with > 4 years of daily data; Table 1, Figs. 2 and 3) and temperature loggers (n = 5, each with > 2 years of hourly data from at least 30 % of days reporting, average = 70 % of days reporting; Table 1, Figs. 2 and 3). Although the spatial coverage of rainfall gauges and temperature loggers is small relative to the coverage of interpolated, modeled, gridded data, they provide us an opportunity to work with measured, basin-specific data.

Proper interpretation of 10Bem-derived denudation rates requires an understanding of the potential for beryllium weathering and desorption from sediment grain coatings and mobility through regolith (von Blanckenburg et al., 2012). To this end, information on (1) the depth of regolith and (2) chemical weathering data across the George River basin is needed. A potentially relevant dataset available for Tasmania is an interpolated gridded map of the depth of regolith (Wilford et al., 2016). However, like the WorldClim precipitation and temperature datasets, the gridded regolith dataset was created by interpolating measured data from around Australia using a model and has an implicit dependence on elevation that does not reflect measured depths to bedrock in the George River basin. Only three boreholes exist in the George River basin that clearly go through regolith to bedrock, from which we extracted regolith depth (BoM, 2015; Fig. 2a; Table 1). They do not match the model results. These three boreholes, and others in the study area, have some units that could be alluvium or regolith; this differentiation is not clear, and therefore the depth of regolith could be overestimated if alluvium is marked as regolith. Thus, we do not know with certainty the depth of regolith across our field area, and we therefore cannot draw any clear conclusions about beryllium mobility in deep, weathered soils from the borehole data alone and do not explore it further.

Qualitative ratings of soil erosivity have previously been determined for Tasmania (Kidd et al., 2014, 2015) based on modeled soil loss should substantial vegetation and ground cover be removed; these ratings are strongly tied to hillslope angle within the George River basin (Fig. 6). Additionally, slope and erosion are strongly linked across the Great Dividing Range on the Australian mainland (Codilean et al., 2021). Thus, we compare erosion and denudation metrics against basin slope metrics, which enables us to compare our measurements of ε and Dm to basin slope to assess how Kidd et al.'s (2014, 2015) metrics for hillslope erodibility and erosion in the George River are related and to compare these new 10Bei erosion rates to those presented by Codilean et al. (2021) for the Australian mainland.

There is one long-term water quality and stream gauging station in the George River basin at the inlet to the local water treatment plant drawing water from the trunk channel of the George River in the town of St. Helens (Fig. 2). Thus, we can only estimate the dissolved load for the entire George River basin, not individual tributaries. Chemical weathering rates for the George River at St. Helens were calculated using these water quality data (i.e., dissolved major and trace element data) and discharge data (John Fawcett, TasWater, personal communication, 2021). Discharge measurements were taken at intervals ranging from 4 to 96 times per day from 1968 to 2021; 26 complete years of discharge data were available. Water quality measurements have been conducted since 2015 and we used the data from July 2015 to September 2021 in our derivation of the dissolved load for the George River basin.

We matched water quality measurements with the nearest discharge measurement in time; when times did not line up exactly, we used the average of the nearest two discharge measurements. We then explored the relationship between discharge and each water quality parameter. For parameters that are invariant with discharge (iron, potassium, sulfate, silica), we calculated the mean concentration of the parameter (Table 6). For parameters that scale with discharge (calcium, magnesium), we used a rating curve to determine how discharge relates to each water quality parameter; we then applied the mean measured values and rating curves, as appropriate, to every discharge measurement for years with complete discharge records (Table 6). Sodium and chlorine were balanced (suggesting a sea salt contribution) and were thus omitted from the calculation. Carbonate that balanced the calcium and magnesium present was included; the rest was assumed to be from atmospheric sources (Table 6). Silica concentrations were measured independently, once annually from 1974 to 1981 (John Fawcett, TasWater, personal communication, 2021), and we used all eight of those measurements; measurements in individual years were taken in March, June, August, October, and November (Table 6). We report total dissolved solids (TDS) measurements that are the sum of potassium, sulfate, silica, calcium, magnesium, and carbonate concentrations following West et al.'s (2005) chemical weathering rate calculation. We used a similar method to calculate the total suspended sediment (TSS) for each year of complete discharge data; TSS scales with discharge and so we applied a rating curve (Table 6).

Erosion rates, ε, based on measured concentrations of 10Bei, range from 13.1 to 66.2 Mgkm-2yr-1. They have integrated landscape dynamics in the George River basin since ∼ 24–122 ka (Table 4). The average ε from tributaries (36.8±1.3 Mgkm-2yr-1) is greater than from either of the trunk channel samples (TG-1 =25.9±2.2 Mgkm-2yr-1; TG-9 =22.4±1.9 Mgkm-2yr-1). Tributary values for ε are greater in the high-elevation, western headwaters of the George River basin and decrease systematically eastwards towards the lower-elevation coast (Fig. 6; R2=0.91, p<0.001). Relationships between ε in tributary catchments and mean local relief, mean basin slope, and the percent of each basin that is categorized as being greater than or equal to “High” erosivity are weak and not significant (R2=0.28, R2=0.17, R2=0.05, respectively, p≥0.13). Taking the product of ε and basin area provides us with the average annual mass loss for each catchment. Making the assumption of steady state and no change in storage over time, we can then compare mass export rates across the catchment. Following this approach, we find that a similar mass exited sampled tributaries (10511±390 Mgyr-1) as the mass that passes through the trunk channel sites (TG-1 = 10 286 ± 860 Mgyr-1; TG-9 =9555±820 Mgyr-1). This comparison suggests little to no contribution of mass from the lowland, mainstem George River valley below the tributaries and above the basin outlet sampling sites.

Based on an assumed 9Beparent value of 4.1 ppm (Beus, 1962), 10Bem/9Bereac-based denudation rates, Dm, range from 16.7 to 36.9 Mgkm-2yr-1. Most values for Dm in tributaries do not replicate well the 10Bei-derived erosion rates, ε, with the exception of TG-4 (Fig. 8). The 10Bem/9Bereac-based denudation rate at the trunk channel site, TG-9 (23.4±0.3 Mgkm-2yr-1), replicates the 10Bei erosion rate (22.4±1.9 Mgkm-2yr-1). In general, 10Bem-based measures Dm of tributaries are not significantly related to any topographic or basin metric such as mean basin elevation, mean local relief, or mean basin slope (R2=0.12, R2=0.06, R2=0.11, respectively; p>0.44). 10Bem-based measures Dm of tributaries appear to be moderately related to the percentage of each basin that Kidd et al. (2014, 2015) categorizes with a land use of “High” to “Extreme” erosivity, though we note this relationship is not significant at high-confidence levels (R2=0.42; p=0.18; Fig. 9).

The annual dissolved load in the George River at St. Helens for the 26 years between 1969 and 2020 is between 1820 and 10 770 Mgyr-1 (mean: 4400±2230 Mgyr-1, 1σ) and the total suspended sediment load ranges from 280 to 10 560 Mgyr-1 (mean: 1830±2180 Mgyr-1, 1σ). The water treatment plant from which the dissolved load data were obtained is close to site TG-9, and data from this site allow us to place 10Be-inferred erosion and denudation rates for the whole George River basin in context. These data show that the dissolved load export rate averages to about 10.3 Mgkm-1yr-1, which is < 50 % of ε (22.4 Mgkm-2yr-1), based on decades of flow records and 5 years of discontinuous water sampling at the same sampling location. The suspended sediment export rate out of the George River basin is lower, at 4.3 Mgkm-2yr-1.

Erosion rates in the George River basin are strongly related to mean basin elevation, which varies greatly across the catchment as the study area extends east from the Rattler Range and Mt. Victoria (1213 m) to the coast at sea level (Fig. 2). In contrast, we find no evidence to suggest that ε is related to slope in the George River over millennial timescales. This result differs from many studies, which show strong correlations between ε and mean basin slope at a global scale (Portenga and Bierman, 2011) and at regional scales across the Great Dividing Range on Australia's mainland (Fig. 10; Codilean et al., 2021; Nichols et al., 2014). Our results also differ from prior assessments of the George River basin using measured climate data, bedrock structure, topographic analysis, water quality models, and geographical landscape characterization that suggest slope imparts a large control over erosion and sediment generation in the catchment on human timescales (Jerie et al., 2003; Kragt and Newham, 2009).

Any process-based explanation for the correlation of erosion rates with elevation requires that we consider how relevant geomorphic and geochemical processes vary across the George River basin. Climatic data collected from stations in and near the George River basin indicate that both mean annual temperature and mean annual precipitation are strongly correlated with elevation (Fig. 3). At higher elevations, rocks are experiencing lower temperatures more frequently and receive more precipitation than those at lower elevations, increasing the potential for both mechanical (frost cracking) and chemical weathering (dissolution). Frost cracking rates are greatest in rocks where mean annual temperature is above freezing (which is the case for all of the George River basin), but temperatures go below freezing both long and frequently enough to crack rocks, which is also the case across much of the basin (Delunel et al., 2010; Hales and Roering, 2007). In the George River basin, the only temperature-related metric that correlates with elevation is mean annual temperature, in contrast to, for example, the time spent below freezing, likely because temperature inversions, with cold air drainage to lower elevation valleys, are common (Webb et al., 2018, 2020). Additionally, the underlying mechanics that lead to rock fracturing in the first place have been demonstrated to be strongly linked to climate and the availability of water (Eppes and Keanini, 2017; Eppes et al., 2018). While water is plentiful across the George River basin, we see that ε is greater at higher elevations where rainfall is also greater, facilitating faster breakdown of rock.

Mean annual precipitation at meteorology stations in the George River basin varies less (2.7-fold) from low to high elevations (681–1836 mmyr-1; Fig. 3) than ε (4.8–24.5 mmkyr-1; a 5.1-fold difference; Table 4). The elevation-induced precipitation and erosion rate gradients we observe are consistent with suggestions made at regional and global scales that the relationship between slope and erosion becomes secondary to precipitation in low-slope, low-elevation, post-tectonic settings (Henck et al., 2011; Mishra et al., 2019). We note that Mishra et al. (2019) also suggest that at the global scale, the erosional effects of increased precipitation may be balanced by increased vegetation cover, which serves to stymie erosion. However, the George River basin is densely vegetated throughout, and forests are no more prevalent at higher than lower elevations in our field area. We propose that in the George River basin, ε is related to elevation in large part because precipitation is strongly correlated with elevation. This interpretation seems to hold true for bedrock outcrops, the erosion rates of which are most closely correlated to mean annual rainfall in aseismic landscapes, globally; however, basin-wide erosion rates in aseismic areas globally remain more strongly correlated to mean basin slope and subsequently to elevation and climate-related processes (Portenga and Bierman, 2011), which stands in contrast to the relationship we observe here between elevation and ε.

The concave-up geometry of sampled streams (Fig. 5) demonstrates that values of ε presented here come from streams that are in steady state. Thus, the very strong relationship between elevation, climate (both mean annual rainfall and temperature), and ε would likely not have emerged had our 10Bei samples been affected by clast attrition (Carretier et al., 2009); deep-seated landslides (Aguilar et al., 2014; Gonzalez et al., 2016; Puchol et al., 2014); or intensive erosion associated with mining, forestry, or agriculture (Barreto et al., 2014; Neilson et al., 2017). Even intensive tin mining, which supplied > 106 m3 to the George River over the last two centuries (Knighton, 1991) seems not to have had a long-lasting diluting effect on 10Bei in sampled stream sediment. It is possible that mining efforts, especially sluice mining, did not lead to 10Bei dilution because of the homogenizing effect of 10Bei in bioturbated soils (Brown et al., 1995; Schaller et al., 2018) or because the size of the George River basin is large enough to buffer the effects of mining efforts in a similar way that large catchments may buffer the effects of landslide material (Niemi et al., 2005; Yanites et al., 2009). It is also possible that mining activity did lead to 10Bei dilution, but concentrations have returned to pre-disturbance levels in the same way that bedload characteristics returned to pre-disturbance levels (Knighton, 1991) and similar to the rapid, two-year recovery of 10Bei concentrations following storm-triggered landslides in Puerto Rico (Grande et al., 2021).

Overall, the close relationship between 10Bei erosion rates and climate across the George River basin demonstrates that 10Bei erosion rates reflect background, geologically meaningful rates of landscape evolution on millennial timescales, even in areas with long histories of intensive human land use (e.g., Barreto et al., 2014; Rosenkranz et al., 2018; Vanacker et al., 2007). Secondarily, that higher values of ε are observed where there is more rainfall and there are colder temperatures suggests that more sediment is being generated per unit area in the western portion of the catchment. There, larger volumes of rainfall and colder temperatures facilitate the generation, erosion, entrainment, and delivery of more sediment to trunk channels than in the eastern portion of the catchment.

Since pre-disturbance stream flow and bedload conditions were re-established by the 1990s (Knighton, 1991), it appears the greatest risk of enhanced sediment flux from the George River to Georges Bay in the future comes from land-use changes involving the widespread disturbance of surficial soils, such as through forestry (Wilson, 1999). The percentage of land used for production forestry in native environments has been decreasing throughout the 21st century (Fig. 4). Although some land previously used for production forestry in native environments is being supplanted by conservation and protected native land cover, which could buffer the effects of widespread erosion, much is being replaced by grazing and agriculture, which would likely increase erosion, particularly in the headwater catchments where geological erosion rates are naturally higher (Fig. 4). Given recent land-use trends, the 10Bei erosion rates presented here provide a useful benchmark level of sediment delivery to the George River, Georges Bay, and other fluvial systems in northeast Tasmania that share topographic and geologic characteristics similar to those of the George River basin.

Taking the product of ε and basin area calculates the annual mass exported from sampled basins. The mass leaving the tributaries (mean: 10510±390 Mgyr-1) is about the same as the mass passing through TG-1 (10290±860 Mgyr-1) and the mass of sediment leaving TG-9 (9560±820 Mgyr-1). We infer from these data that the 10Bei measured at TG-1 and TG-9 trunk channel locations is dominated by mass produced in the higher-elevation tributary basins with minimal sediment input from the George River valley bottoms. Similar interpretations have been made elsewhere, albeit in much larger river basins (i.e., Wittmann et al., 2009, 2011, 2016). Given these similarities, we average ε from the two trunk channel sites to produce a nominal average erosion rate for the George River basin as a whole (24.1±1.4 Mgkm-2yr-1; or 8.9±0.5 mmkyr-1 when dividing ε by rock density, ρ=2.7 gcm-3), which is of the same magnitude as the average erosion rate of catchments draining the eastern flanks of the Great Dividing Range along the southeastern passive margin of mainland Australia (11.6 mmkyr-1; Fig. 10; Codilean et al., 2021). Average ε from the George River basin is most consistent with erosion of basins across the Bass Strait, which share similar topographic characteristics and geological histories to the George River basin (Codilean et al., 2021). The similarity between the geology, topography, and climate of newly sampled basins and derived 10Bei erosion rates in Tasmania from this study and those from southeast mainland Australia supports the notion that evolution of landscapes that share similar climatic, topographic, and geologic characteristics is similar.

Once delivered to Earth's surface in temperate regions, 10Bem concentrates in the uppermost soil horizons (Graly et al., 2010; Willenbring and von Blanckenburg, 2010). This behavior differs from that of 10Bei, the concentration of which remains homogenous in well-mixed, bioturbated soils for millennia (Jungers et al., 2009). Thus, any disturbance of large volumes of topsoil (i.e., agriculture, forestry, wildfire erosion, or mining activities) strips material with the highest concentrations of 10Bem and introduces that material into streams, a process similar to that identified following early land-use changes and deforestation in the Chesapeake Bay and San Francisco Bay (Portenga et al., 2019; Valette-Silver et al., 1986; van Geen et al., 1999). In contrast, the strong relationship between 10Bei erosion rates and elevation, and thus both precipitation and temperature, across the George River basin (Fig. 7) suggests that 10Bei erosion rates, ε, are unaffected by land use.

Assuming a 9Beparent value of 4.1 ppm (Beus, 1962), calculated values of Dm do not consistently replicate ε (Fig. 8), nor does Dm replicate the spatial patterns or yield the same relationships with topographic parameters that we observe with ε in the small, geologically homogeneous landscape of the George River basin (e.g., Fig. 7). In fact, we calculate similar Dm values at TG-2 and TG-7, which have the lowest and highest calculated values for ε (Fig. 8). We know that decades-old historical mining activities and historical bushfires in the George River were restricted to lower catchment areas and tributaries where measurements of Dm are highest (Figs. 4 and 9). Additionally, we infer from the moderate correlation observed between Dm and the percent of tributary basins classified as “High” to “Extreme” erosivity (R2 = 0.42; Fig. 9; Kidd et al., 2014, 2015) that 10Bem/9Bereac-derived denudation rates appear to be sensitive to recent land-use practices that disturb soils. The highest denudation rates, Dm, we measured are those from basins with past histories of intense surface disruption through mining and forestry (i.e., TG-4, TG-5).

The similarity of Dm and ε for the entire George River basin, however – within a factor of 3 (Fig. 8) – provides general support for the hypothesis that 10Bem/9Bereac-based denudation rates more-closely resemble 10Bei-based erosion rates in small river basins where geological heterogeneity is minimized. This observed similarity between Dm and ε supports the continued exploration and application of 10Bem/9Bereac denudation rates in geomorphological studies (i.e., Dannhaus et al., 2017; Deng et al., 2020; Portenga et al., 2019; Rahaman et al., 2017; Wittmann et al., 2012, 2015). However, data presented here suggest that this method should be used with caution in landscapes with recent soil disturbance.

The values of Dm we present in this study are calculated with assumed values for the amount of 9Be naturally occurring in bedrock in the field area (9Beparent) and the rate at which meteoric 10Be is delivered from the atmosphere to Earth's surface (10BeFmet) because we did not measure these values specifically for the field area. Thus, we carry out a sensitivity analysis of both variables to assess how much Dm responds to changes in these values.

Grew (2002) suggests that Earth's crustal average concentration of 9Beparent is 3 ppm, though it is not unheard of for 9Beparent to be < 1 ppm in (ultra)mafic lithologies and that 9Beparent can vary 10-fold within the same igneous complex. Von Blanckenburg et al. (2012), who first present calculations for Dm, cite a slightly lower crustal average for 9Beparent of 2.5 ppm. London and Evensen (2002) present 9Beparent concentrations measured from felsic granites, which range from 1.6–160 ppm; for S-type granites or those that are tin-bearing – the same as the Blue Tier batholith in our field area (Higgins, 1985) – 9Beparent ranges from 2.3–130 ppm (n = 11, mean: 18 ppm). Additionally, Sainsbury (1964) presents data from a tin-bearing biotite granite in Alaska, showing that 9Beparent concentrations range from 2–26 ppm (n = 5, mean: 16.6 ppm). Thus, it seems a reasonable range of values for 9Beparent that might apply to bedrock in this study are as low as crustal averages (2.5 ppm) or as high as tin-bearing biotite granites elsewhere (> 100 ppm). We choose to calculate and analyze Dm from a more modest estimate of 4.1 ppm (Beus, 1962), because single-digit concentrations of Be are most common for felsic igneous intrusions (London and Evensen, 2002). At lower 9Beparent concentrations (2.5 ppm), Dm values across the George River basin increase such that Dm values replicate ε within a factor of two. However, when conservative but higher 9Beparent concentrations are used (18 ppm; the average of values presented for S-type and tin-bearing granites presented by London and Evensen, 2002), Dm values decrease across the field area such that all Dm values are lower than ε by at least a factor of 3 (Fig. 8). The results of this sensitivity analysis highlight the importance of collecting representative bedrock samples throughout a field area to ascertain appropriate measures of 9Beparent when using von Blanckenburg et al.'s (2012) 10Bem/9Bereac denudation rate method because of the highly sensitive dependency of Dm on 9Beparent.

Values for the rate at which 10Bem is delivered from the atmosphere to Earth's surface (10BeFmet) have been measured and modeled in various ways at both local and global scales, each with its own strengths. In the South Pacific region, for instance, Reusser et al. (2010a) directly measured 10BeFmet in a dated New Zealand paleosol (1.68 to 1.72 × 106 atomscm-2yr-1) and Graham et al. (2003) report 10BeFmet values measured from rainfall across New Zealand, finding a wider range of 10Bem deposition rates (1.7 to 2.9 × 106 atomscm-2yr-1). In the absence of direct measurement, 10BeFmet must be estimated or modeled. Heikkilä and von Blanckenburg (2015) integrate 10BeFmet through the Holocene while others integrate 10BeFmet for total atmospheric thickness, all at a global scale (Masarik and Beer, 2009; Willenbring and von Blanckenburg, 2010), but the resolution of these models is coarse, relative to the small spatial scale of this study, and 10BeFmet would be the same for each sampled basin (1.0–1.5 × 106 atomscm-2yr-1 for Holocene integrated or ∼ 7 × 105 atomscm-2yr-1 for atmospheric depth-integrated 10BeFmet). Graly et al. (2011), however, present an equation that estimates 10BeFmet from a location's mean annual precipitation and latitude, which provides a more specific value for 10BeFmet for a given study site. We choose to use 10BeFmet modeled from Graly et al.'s (2011) equation because of its ability to provide basin-specific values of 10BeFmet, but we present Dm calculations for all basins using other 10BeFmet values to assess the sensitivity of Dm to 10BeFmet (Fig. 11). In doing so, we find that Dm calculated from Reusser et al.'s (2010a) and Graham et al.'s (2003) values of 10BeFmet are consistently higher than using the Graly et al.'s (2011) model, likely owing to precipitation rate differences between northeast Tasmania and New Zealand, thousands of kilometers away. Dm calculated using 10BeFmet values integrated through total atmospheric thickness (Masarik and Beer, 2009; Willenbring and von Blanckenburg, 2010) are consistently lower than those calculated using Graly et al.'s (2011) model, but those using 10BeFmet values averaged through the Holocene (Heikkilä and von Blanckenburg, 2015) are remarkably consistent with results from the Graly et al. (2011) model. We suggest that the consistency of Dm modeled using the Graly et al. (2011) 10BeFmet values and Dm calculated using Heikkilä and von Blanckenburg's (2015) 10BeFmet values provides support for our decision to use Graly et al.'s model. Additionally, we suggest our use of Graly et al.'s (2011) estimates of 10BeFmet is reasonable because Dm values using Graly et al.'s 10BeFmet values plot between Dm values calculated using 10BeFmet values from both global climate models (Heikkilä and von Blanckenburg, 2015; Masarik and Beer, 2009; Willenbring and von Blanckenburg, 2010), at least for northeast Tasmania.

If chemical weathering occurs primarily in the uppermost meters of the landscape, where most 10Bei is produced, then the erosion rate, ε, we calculate represents total landscape mass loss over time – a combination of physical and chemical mass loss. We could then partition ε along the trunk channel at the mouth of the George River basin (TG-9 = 22.4 Mgkm-2yr-1) into mass flux removed in the measured dissolved load (10.3 Mgkm-2yr-1) and the remainder, mass flux removed as solid sediment (12.1 Mgkm-2yr-1). Of the sediment mass flux, it appears that 4.3 Mgkm-2yr-1 is transported as suspended load (measured from water quality data), and the difference of 7.8 Mgkm-2yr-1 is bedload (i.e., 12.1 Mgkm-2yr-1 minus 4.3 Mgkm-2yr-1). Our measure of ε at TG-9 (22.4 Mgkm-2yr-1) is similar to the 10Bem/9Bereac measure of denudation at this site (23.4 Mgkm-2yr-1), which, if the assumptions of the method are met, represents total physical and chemical mass loss. Taken at face value, Dm assuming 9Beparent = 4.1 ppm is an accurate measure of total mass loss from the George basin at TG-9. However, given the wide range of Dm possible using other reasonable values for 9Beparent, it is difficult to know how well the two measures of landscape change, ε and Dm, truly compare.

If the majority of chemical weathering occurs below the penetration depth of most cosmic rays (< 2 m), then the chemical denudation and physical mass loss are at least in part and perhaps wholly disconnected. In this case, ε (TG-9 = 22.4 Mgkm-2yr-1) would need to be summed with the measured chemical mass flux (10.3 Mgkm-2yr-1), and together (30.7 Mgkm-2yr-1) they would estimate the total mass loss from the landscape. In this case, the summed total is ∼ 24 % greater than Dm at TG-9. The presence of bedrock outcrops in some of the George River basin channels suggests that regolith thickness is limited in places, and in that case 10Bei measurements incorporate much of the chemical mass loss from the basin. However, the few boreholes that extend to unweathered bedrock (n = 3; Fig. 2; Table 1) clearly indicate that regolith is deeper in some parts of the catchment. With the paucity of available data, we cannot determine how much of the dissolved load is coming from below the penetration depth of cosmic-ray neutrons, but it could be significant.

Thus, despite the fact that ε and Dm at TG-9 are similar (22.4 and 23.4 Mgkm-2yr-1, respectively), 10Bemet/9Bereac-based denudation rates appear to have little correlation with landscape-scale metrics suggesting that they do not reflect the rate of geomorphic processes controlling mass loss over time. Dm is highest in basins with known histories of intensive land-use disturbance and high erosivity (Figs. 2 and 8), a relationship that exists regardless of what value is used for 9Beparent. In contrast, ε is well-correlated to elevation and thus temperature and precipitation across the George River basin.