6.1. Riverine POC Transfer
 We have combined published estimates of suspended sediment transport for some rivers in the study area with our observations to estimate POC export (Table 4). Most of our suspended sediment sediment samples were collected at times when SSC was below the flux-weighted average SSC (Table 1) [Hicks et al., 2004b], which may lead to an overestimation of POC transfer if an inverse relation between SSC and Corg is assumed [c.f. Ludwig et al., 1996]. However, these small mountain rivers show a weak correlation between SSC and Corg and a nearly constant Corg at SSC above ∼100 mg L−1 (Table 1). We therefore assume that the Corg values presented in this study are representative of average SSC conditions. We estimate that the mean POC yield from the western Southern Alps is 57 tC km−2 a−1 (Table 4). This is >30 times the global average [Ludwig et al., 1996] and similar to estimates from other small mountainous catchments [Gomez et al., 2003]. Our estimate of the specific yield of POC in the catchment of the Hokitika river, 47 tC km−2 a−1, is similar to previous estimates of 43 and 44 tC km−2 a−1 [Lyons et al., 2002; Carey et al., 2005]. However, our estimate for the Haast catchment of 15 tC km−2 a−1 is much lower than the previous estimate of 71 tC km−2 a−1 [Lyons et al., 2002], due in part to a downward adjustment of the sediment yield from the catchment from 12,700 t km−2 a−1 to 4500 t km−2 a−1 [Lyons et al., 2005]. If the exhumation rate across the western Southern Alps is ∼5 mm a−1 [Bull and Cooper, 1986; Tippett and Kamp, 1993], then the long-term sediment yields from this domain are ∼12,500 t km−2 a−1. Therefore the POC yield presented here for the Haast is a minimum. To prepare a further discussion of the riverine transfer of POC from the western Southern Alps we must first consider the origin of the POC.
Table 4. Mean Measured Organic Carbon Concentration (Corg)a
|River||Mean Corg, %||Annual Sediment Yield, t km−2 a−1||POC Yield, tC km−2 a−1||Fraction Nonfossil POC||Nonfossil POC Yield, tC km−2 a−1||Silicate Weathering CO2 Consumption, tC km−2 a−1||Drainage Area, km2|
|Hokitika (n = 3)||0.75||6313b||47||0.80||38||3.6b||352|
|Wanganui (n = 2)||0.45||12,500c||56||0.66||37||1.7d||344|
|Poerua (n = 3)||0.35||26,200e||91||0.57||52||1.7d||136|
|Whataroa (n = 2)||0.99||10,325b||102||0.85||87||10.1b||453|
|Waitangitaona (n = 2)||0.66||12,500f||83||0.77||64||1.7d||72|
|Waiho (n = 2)||0.26||10,325b||27||0.43||12||1.7d||164|
|Fox (n = 2)||0.29||12,500c||37||0.49||18||1.7d||92|
|Haast (n = 3)||0.34||4500b||15||0.56||9||6.0b||1020|
 The δ13Corg and Corg measured in suspended sediment define a binary mixing of two distinct end-members (Figure 3). One end-member is isotopically light and enriched in Corg, the other is isotopically heavy and depleted in Corg. The mean of carbon values measured in bedrock samples from the study area lies on the mixing line (Figure 3), and we propose that fossil POC in bedrock best defines the depleted end-member. The enriched end-member sits within the field of isotopic compositions characteristic of C3 vegetation growing at <2000m elevation [Körner et al., 1988] and of soils in montane catchments [Bird et al., 1994]. Therefore the range of concentrations and isotopic compositions of the POC in suspended sediment and bed materials of rivers in the western Southern Alps can be explained by variable contributions of organic carbon from a biomass and soil source added to organic carbon contained as fossil POC in mobilized bedrock.
 The measured concentration and isotopic composition of POC in sediment can be used to determine the relative contributions from these sources. This can be done in two ways. One considers only the Corg of the sediment and assumes that any enrichment of organic carbon in the suspended load, above the Corg of the fossil end-member, is due to an addition of nonfossil POC. For this purpose, the fossil POC end-member is assigned the mean measured bedrock value of Corg = 0.15%. The other method makes use of the isotopic composition of the carbon. The fossil POC, end-member is assigned a value of δ13Corg = −21.1‰, the mean value measured in bedrock, and the nonfossil end-member is attributed a value of δ13Corg = −28.3‰, obtained by extrapolation of the best fit to the data (Figure 3). These two approaches produce compatible estimates of the fraction of fossil POC, with an average discrepancy of 0.03. There are more measurements of Corg than of δ13Corg (Table 1), and on the basis of this characteristic we estimate that the fossil fraction of POC in river suspended load is on average 0.37 ± 0.07, with a range of 0.13 to 0.75 (Table 4), comparable to estimates from other small mountain rivers [Kao and Liu, 1996; Masiello and Druffel, 2001; Blair et al., 2003; Komada et al., 2005; Leithold et al., 2006]. Thus the fossil contribution to the POC load is large despite the dense hillslope organic carbon store, minimal upland anthropogenic disturbance and low Corg in metasedimentary bedrock in the western Southern Alps.
 Using the mean suspended load Corg and the mean fraction of nonfossil Corg for each catchment (Tables 1 and 4), the mean nonfossil biogenic POC transfer for the western Southern Alps is 39 tC km−2 a−1, ranging from 9 to 87 tC km−2 a−1 for individual catchments (Table 4). These POC yields do not account for bedload transport which can represent as much as 50% of the total riverine transport of sediment in an active orogen [Galy and France-Lanord, 2001; Dadson et al., 2003], and may contain a significant amount of biogenic POC when coarse woody debris is rigorously considered. The suspended sediment transfer of nonfossil POC represents ∼0.1% of the hillslope carbon store of ∼28,000 tC km−2 [Coomes et al., 2003] and ∼4% of the annual net primary productivity of ∼1100 tC km−2 a−1 [Whitehead et al., 2002]. From this we conclude that the high measured yields of nonfossil POC are sustainable on long timescales in a densely vegetated biome undergoing rapid uplift and erosion, such as the western Southern Alps.
 Burdige  has estimated the burial efficiency of terrestrial organic matter (relative to the riverine input) for nondeltaic continental margin sediments at ∼17%, and at sediment accumulation rates greater than 0.1 g cm−2 a−1 measured burial efficiency is consistently above 10%. Galy et al.  have reported even higher burial rates in the Bay of Bengal. On margins receiving sediment from small mountain rivers, sedimentation rates may approach 5 g cm−2 a−1 [Gomez et al., 2004b]. Moreover, rivers draining the western Southern Alps may on occasion discharge large amounts of sediment to the ocean at hyperpycnal concentrations [Mulder and Syvitski, 1995]. Hyperpycnal river plumes can turn into turbidity currents and result in thick sediment beds. Thus they provide a mechanism for delivering terrestrial POC rapidly to deep sea depocentres [Walsh and Nittrouer, 2003; Nakajima, 2006] and for shielding deposited POC from oxygenated ocean water. Therefore it seems reasonable to assume that at least 10% of the POC output from the western Southern Alps is preserved in long-lived sedimentary deposits. At this conservative preservation rate, the transfer of eroded biogenic POC to sediment is ∼4 tC km−2 a−1. Considering that the rate CO2 drawdown by silicate weathering in the same area is 3.5 tC km−2 a−1 [Jacobson and Blum, 2003; Lyons et al., 2005], the transfer and burial of nonfossil biogenic POC is likely to be the more significant way in which the Southern Alps consume atmospheric CO2 on geological timescales. By extrapolation, the harvesting and burial of modern biogenic POC in basins fed by steep river catchments in active mountain belts could be a globally significant mechanism of atmospheric CO2 drawdown on long timescales.
6.2. Routing of POC
 The occurrence of a single nonfossil end-member is intriguing, given that at the scale of individual plant species, litters and soils there can be a large range in both δ13Corg [Guehl et al., 1998] and Corg [Hart et al., 2003]. The δ13Corg of soils and vegetation tend to overlap in montane catchments [Bird et al., 1994; Townsend-Small et al., 2005]. The variable proportion of nonfossil POC in the riverine sediment (Table 4) could therefore represent a fairly stable amount of POC derived from rock and soil but a variable contribution from vegetation. The C/N of POC can be used to distinguish between nonfossil carbon from soil and vegetation [Townsend-Small et al., 2005; Holtvoeth et al., 2005]. C/N is 16 to 25 in surface soil horizons [Basher, 1986]. By contrast, in C3 vegetation in an indigenous montane forest ecosystem similar to that studied here, C/N is around 40 in the leaf component, 78 to 157 for twigs and small branches, through ∼250 for bark, to >600 in stem wood [Hart et al., 2003].
 Unlike the Corg and δ13Corg data, the C/N and Corg data for river sediment in the western Southern Alps shows significant scatter. Hence a simple binary mixing model described by the linear best fit with C/N = (a × Corg + b)/(c × Corg + d) does not describe the data well, R2 = 0.29 (Figure 5a). Some of the suspended sediment and most of the bed material have C/N and Corg close to the average composition of the Alpine Schist bedrock (Figure 5a), reemphasising the significant contribution of fossil POC from bedrock inferred from δ13Corg. Therefore scatter in the C and N composition of the suspended material is more likely to reflect a highly variable contribution of soil and vegetation. For example, the Hokitika River is characterized by the mixing of bedrock and soil material, the Poerua River has nonfossil POC derived mainly from vegetation, and nonfossil POC in the Whataroa and Waitangitanoa Rivers appears to be a mixture of materials derived from soils and vegetation (Figure 5a and Table 1). Notwithstanding this complexity, our data suggest that at the catchment scale, mobilization of POC occurs by processes that reduce the heterogeneity of the nonfossil POC and mix it with significant amounts of fossil POC from bedrock.
Figure 5a. Organic carbon concentration (Corg) versus organic carbon to nitrogen ratio (C/N) for suspended load (black diamonds) and bed material (gray diamonds). Black circle corresponds to the mean of the measured bedrock (Table 2). Solid and dashed curves represent the mixing of POC from bedrock with POC from vegetation and soil, respectively. Dotted lines represent the extent of the mixing envelope considering the range of measured values of bedrock, soil, and vegetation (Table 2) [Basher, 1986; Hart et al., 2003]. Suspended load data lie within the range of values expected from a mixture of vegetation, soil, and bedrock.
Download figure to PowerPoint
 The processes by which this mixing occurs can be examined by looking at landslide deposits and the nature of POC within individual deposits. Hillslopes in the western Southern Alps have significant stores of carbon in standing biomass, partially degraded litter and coarse woody debris, and surface soils [Coomes et al., 2003; P. J. Bellingham, unpublished data, 2003]. The measured compositions of POC in landslide debris occupy a section of the ternary mixing space of bedrock, soil and vegetation end-members larger than the river sediment discussed above (Figure 5b) However, their range is restricted when compared with Corg and C/N measured in soil and vegetation, and values for POC in landslide debris plot very near the bedrock end-member. This shows that POC from vegetation, soil and bedrock have been mixed in landslide debris.
Figure 5b. Landslide sediment samples together with the same mixing relationships (Figure 5a). Solid and open triangles show clay silt (<63 μm) and sand (<500 μm, >63 μm) fractions, respectively. Numbers correspond to each landslide site (Table 3).
Download figure to PowerPoint
 All sampled landslide deposits show a lack of vertical trends in profiles of both Corg and C/N (Figure 4). This homogeneity most likely originates from destruction, and effective mixing of hillslope biomass and soil with bedrock during mass wasting. In addition, the ratio of Corg in the clay silt/sand fractions is always >1 (Table 3). This may not be expected for landslide sediment where fresh bedrock may have been broken into a large range of grain sizes. It suggests that POC is dominantly fine grained and that a large proportion of the POC is probably associated or bound to mineral surfaces. As mean C/N values are very similar for the clay silt and sand fractions in landslide debris (Table 3), the POC at these grain sizes has essentially identical sources. Its concentration is set by mineral dilution. Taken together, these observations suggest that the composition, concentration and distribution of POC in landslide debris are due to a combination of two processes. Landsliding harvests bedrock, soil and standing biomass. During downslope transport the grain size of this material is mechanically reduced, and contributions from different sources are mixed. This results in homogenization of POC in landslide debris, and a dilution of high concentrations of POC in soil and vegetation materials through addition of fragmented bedrock material with much lower POC concentrations. However, this mechanism cannot fully explain the dominance of fine POC in landslide deposits and its association to mineral surfaces. These attributes are likely to be due to pedogenic processes [Sollins et al., 1996] acting on hillslope materials prior to slope failure, and on landslide debris after deposition.
 Given that the POC in landslide debris is a mixture of materials from fossil and nonfossil sources, the bedrock end-member (Figure 3) can be used to estimate that the nonfossil component of the POC ranges from 0% to 95% of the total. Considerable intersite variability, seen both in Corg and in C/N (Figure 5b), can best be explained with reference to the characteristics of the sampled landslides. The depth of a landslide sets the relative bedrock contribution, and the state of hillslope vegetation prior to landsliding determines the biogenic input to landslide debris. In agreement with this notion, in the small, shallow landslide of site 4 the mean Corg is 4 to 6 times higher than at sites 2 and 3 in debris sourced from a large, deep failure with a greater contribution from the bedrock. Moreover, the differences in Corg and C/N between these sites (Figure 5b) may reflect differences in the source of the nonfossil carbon–standing biomass for sites 2 and 3, and soil and biomass for site 4. Site 5 was fed by a recent landslide located within an older (>60 years), larger landslide scar on which vegetation had not reached full maturity when failure reoccurred. In contrast, site 4 was sourced from a slide that disrupted mature forest. Lower values of Corg and C/N at site 5 relative to site 4 (Figure 5b) may reflect depleted biomass and thin soil in the source area of site 5 prior to failure. Together, the five sites at which we have sampled landslide deposits span a range of source areas from ∼0.01 km2 to ∼0.3 km2. This covers a substantial segment of the total size range of landslides mapped in the western Southern Alps [Hovius et al., 1997]. Therefore we are confident that our sample of landslide deposits should provide a reasonable constraint on the composition and abundance of POC contained in landslide debris throughout the mountain belt.
 At the mountain belt scale a large number of bedrock landslides contribute sediment to rivers [Hovius et al., 1997]. For example, analysis of a time series of aerial photographs has revealed that in the Whataroa catchment alone (Figure 2) at least 315 landslides occurred between 1965 and 1985. If rivers source their sediment and POC mainly from the numerous landslide deposits in their catchment area, then the POC load of a river at the mountain front represents the integral of contributions from all active landslide sources. Mixing of the biogenic POC from different hillslope stores has already happened in the landslides (Figure 5b), and this is why the composition of POC in river suspended load can simply be described as a binary mix of material from fossil and nonfossil sources (Figures 3 and 5a). Surface runoff, which can deliver organic rich particulate material from standing biomass and litter, cannot explain the large fossil POC component measured in the rivers of the western Southern Alps. Landslides explain the low biogenic component in the riverine POC (of 63 ± 7%) and should be important in the routing of POC from other rapidly uplifting mountain belts where metamorphic bedrock builds steep topography.