Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 Small mountainous rivers discharge disproportionate amounts of sediment and carbon to the Earth's oceans. Our New Zealand data demonstrates that localized erosion plays a greater role in C budgets than has been recognized in national and global studies. We estimate that New Zealand's rivers export 4 ± 1 Mg C km−2 yr−1 of dissolved organic carbon (DOC) and 10 ± 3 Mg C km−2 yr−1 of particulate organic carbon (POC) (2 and 6 times the global average), which is equivalent to 40% of New Zealand's fossil fuel emissions. Under intact native vegetation in mountain-belt hot spots, POC export greatly exceeds CO2 consumption from mineral weathering. Moreover, deforestation of fertile steepland greatly accelerates POC loss, evidenced by 1.7% of New Zealand's land area which generates 20% of exported POC. Thus, localized erosion deserves increased attention in C budgets and accounting.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Rivers play an important role in the transport of C between land and the atmosphere, oceans and sediments [Ludwig et al., 1996; Mayorga et al., 2005; Richey et al., 2002; Stallard, 1998]. The Marrakech Accords do not require that human-induced riverine C fluxes be accounted for under Articles 3.3 or 3.4 of the Kyoto Protocol during 2008–2012, yet more complete C accounting regimes could include riverine C transport where changes in land-use and management influence net C balance. The possibility of high C mobilization through erosion has particular relevance for countries around the tectonically active Pacific Rim, which contribute as much as 35% of the global sediment and POC flux to the oceans from only 3% of Earth's land area [Kao and Liu, 1997; Lyons et al., 2002; Milliman and Syvitski, 1992]. The greatest POC fluxes occur in areas where rapid uplift exposes soft or highly fractured rocks to frequent large storms [Gomez et al., 2003; Hovius et al., 2000; Kao et al., 2003]. Over geologic timescales (∼106 yr), POC burial around the Pacific Rim may also influence paleo-atmospheric CO2 levels [Lyons et al., 2002], by helping to determine the mass of CO2 contained in the biosphere/atmosphere/hydrosphere system. Tectonic uplift drives CO2 consumption by exposing new crust to mineral weathering and facilitating subsequent oceanic carbonate burial [Berner et al., 1983], as well as by contributing nutrients that influence the rates of oceanic production and burial of phytoplankton-derived organic C [Berner, 2003]. A strong relationship exists between tectonism and paleo-atmospheric CO2 [Berner, 2003; Berner et al., 1983]; however, high rates of POC delivery from the Pacific Rim [Gomez et al., 2003; Kao and Liu, 1997; Lyons et al., 2002] suggest there is a need to evaluate the effect preservation of buried organic C associated with largely unweathered terrigenous sediment derived from young mountain belts [France-Lanord and Derry, 1997] has on the global C-budget.
 To address these questions about the role of rivers in C budgets, we quantified riverine C exports from New Zealand (267,304 km2), which lies on the active boundary of two major tectonic plates. We estimated long-term riverine POC and dissolved organic C (DOC) yields from 47 and 44 catchments, respectively, representing ∼45% of New Zealand's land area. The hydrophysical characteristics of these catchments were used as a basis for extrapolating C yields across the entire country (Table 1).
Values of POC and DOC from 47 river stations were extrapolated to the national scale. YPOC and YDOC represent the POC and DOC yield (Mg C km−2 y−1), MAP is mean annual precipitation, MAS is mean annual solar radiation, MAT is mean annual temperature, slope is obtained from a 25-m digital elevation model (DEM), αı and β are regression parameters, and ET731 and ET732 represent the proportion of area in landslide and earth flow/gully susceptible terrains, respectively. Models 1a and 1b differ in using %C in TSS and TON/Turbidity, respectively, as surrogate variables for estimating the yield at the station.
 Samples were collected for 1 year at monthly intervals for DOC, POC, and total suspended sediment (TSS) concentrations from 44 National Institute of Water and Atmospheric Research (NIWA) water quality stations within New Zealand [Smith et al., 1996]. These stations have a 12-year record of monthly water-quality measurements, and provide good spatial coverage of the entire country. High-flow sampling was also performed within the Manawatu, Waipaoa, and Motueka catchments, and the Waipaoa dataset was augmented by previous research [Gomez et al., 2003].
 Annual POC yields were estimated using 2 approaches. The 1st method applied the relationship between %C and TSS (model 1a and 2) in conjunction with 20-year daily TSS records generated from sediment rating curves [Hicks et al., 2004]. The 2nd approach applied the relationship between %C and total organic nitrogen (TON)/Turbidity for each region (model 1b) in conjunction with flow and sediment 20-year records. For DOC, DOC concentrations for each station from the 1-year sampling campaign were extrapolated across the 12-year dataset using filtrate absorption coefficients at 340 nm and a regional coefficient. DOC yields were then estimated at each station using the relationship between DOC concentration and discharge across the 20-year flow record.
 The estimated DOC yields from each station were then used to build a log-linear regression model to extrapolate DOC yields across the landscape. Several variables (e.g., mean annual temperature (MAT), mean annual precipitation (MAP), mean annual solar radiation (MAS), soil C content, P-retention, catchment slope) were then tested to develop a spatial prediction model. Significant variables included MAP, MAT, MAS and slope.
 For POC, 3 models (2 linear, 1 power-law) (Table 1) were applied to the dataset to obtain a statistical relationship suitable for scaling the variability to produce a national POC estimate. MAP was the primary variable in all 3 models, and a geomorphological variable was required to capture the high POC fluxes within the East Cape region of New Zealand, where large gully systems provide sediment and organic C to the river network.
 The spatial DOC & POC models were then applied across the landscape at 1-km grid resolution to estimate New Zealand's DOC & POC yields. Predicted and surrogate yields were largely in agreement for each model (Figure 1). The predicted POC yields between models 1a, 1b, and 2 provided a range in POC yield, and were used to generate a standard deviation.
 We estimate that New Zealand's rivers discharge 2.7 ± 1.0 Tg C yr−1 of POC (10 ± 3 Mg C km−2 yr−1) and 1.1 ± 0.3 Tg C yr−1 (4 ± 1 Mg C km−2 yr−1) of DOC to the ocean. These yields are 2 and 6 times the global average POC [Ludwig et al., 1996; Meybeck, 1993; Stallard, 1998] and DOC [Aitkenhead and McDowell, 2000; Ludwig et al., 1996] fluxes, respectively, distinguishing New Zealand's tectonically active setting from typical continental environments. The total riverine POC yield is a C flux equivalent to 30% of the country's total 2002 fossil fuel CO2 emissions; globally this ratio is less than 5%. These results demonstrate that, compared with most other land masses, New Zealand's riverine POC flux is relatively more important to both the global C cycle and aspects of C accounting that relate land-use change and forestry to industrial activities.
 Maps of POC and DOC (Figure 2) yield indicate that, just as there are hot spots of riverine C export on a global scale [Lyons et al., 2002], local hot spots of POC and DOC yield exist within New Zealand, characterized by a 10,000-fold range in POC yield and a 1000-fold range in DOC yield. We define a threshold of 20 Mg C km−2 y−1 as a change in soil organic C stock of similar magnitude to other land-use changes relevant to the global C budget [Tate et al., 2003]. About 3% of New Zealand's land area has a DOC yield exceeding this threshold, and this area accounts for 25% of New Zealand's total DOC yield to the ocean. Similarly, 9% of New Zealand has a POC yield in excess of the threshold, accounting for 65% of total POC export to the ocean.
 These POC and DOC ‘hot spot’ terrains are primarily blanketed in natural vegetation cover – mainly primary forests, secondary forests, subalpine tussock grasslands and alpine herbfields. POC and DOC fluxes from these areas can largely be considered representative of conditions prior to human disturbance. The principal exception is steep (typically >26°) terrain representing 1.7% of New Zealand's land area, typically underlain by soft mudstones and crushed sedimentary rocks [Jessen et al., 1999; Page et al., 2004]. These soft-rock steeplands have a specific yield of 110 Mg C km−2 y−1, in agreement with previous estimates [Page et al., 2004], and deliver 20% of New Zealand's POC to the ocean. Over half this terrain is in pastoral agriculture, and a large portion of the remainder once cleared for pasture has been recently reforested with indigenous or exotic species. Evidence that erosion rates in this terrain are elevated in pastoral agriculture relative to forest vegetation [Blaschke et al., 1992] suggests the associated C fluxes can be considered relevant to Articles 3.3 and 3.4 of the Kyoto Protocol because they can be considered ‘human-induced’. Determining the fate of this large human-induced POC flux is therefore highly relevant to New Zealand's Kyoto commitments and policies.
 Understanding the specific greenhouse gas implications of POC and DOC losses can be difficult because traditional land-atmosphere C exchange measurements only record losses of CO2 from the ecosystem, while inventory-based methods will not record C losses if the soil C pool recovers before the subsequent measurement. The net impact of POC and DOC on land-atmosphere C exchange depends on the proportion of the C subsequently oxidized and returned to the atmosphere as CO2.
 Carbon lost from terrestrial ecosystems to rivers as DOC and DIC is likely to be returned to the atmosphere as CO2 through a combination of photo-oxidation, heterotrophic decomposition [Amon and Benner, 1996] and evasion [Richey et al., 2002], causing potential errors in terrestrial land-atmosphere CO2 exchange measurements. However, POC delivered to the ocean by high-yielding rivers is more likely to be protected from oxidation by rapid burial [Sommerfield et al., 1999], which also hinders biological degradation associated with bioturbation [France-Lanord and Derry, 1997]. Thus, high POC export causes errors in both land-atmosphere CO2 exchange and C inventory, which can be further confounded by recovery of soil C stocks on eroded land [Harden et al., 1999; Stallard, 1998]. When combined, POC export and burial therefore paradoxically represent a C sink.
 The significance of POC export and burial goes beyond present-day C budgets. If burial protects a significant fraction of POC from oxidation, the relative importance of POC burial versus CO2 consumption by weathering must be re-examined vis-à-vis paleo-atmospheric CO2 levels. Figure 3 compares recently observed rates of CO2 consumption by weathering [Jacobson and Blum, 2003] with map-based POC yields derived using approaches 1a, 1b and 2 (Table 1). The weathering rates plotted in Figure 3 describe CO2 delivery to rivers from both silicate and carbonate weathering rather than the rate of CaCO3 burial. Thus, the protection of even a small fraction of POC by burial is likely to represent a greater sink for CO2 than weathering consumption. These results therefore lend strong independent support to previous suggestions that mountain building primarily influences atmospheric CO2 through POC burial, rather than silicate weathering [France-Lanord and Derry, 1997]. The data in Figure 3 can be assumed representative of the geologic past because POC delivered from high-yielding areas of South Island does not result from human land-use change.
5. Concluding Statements
 We conclude that mountain building on New Zealand's high-standing islands has strong implications for C cycling on the timescales of land-use change and C accounting under the UNFCCC and Kyoto Protocol, as well as on geologic timescales. In the case of geological erosion in nearly pristine mountain belts such as the Southern Alps, our estimates and those made previously [Lyons et al., 2002] are based on scarce data, suggesting more measurements are needed for high-yielding catchments. On timescales where ‘human-induced’ changes to the C cycle are a concern, further research should reveal the degree to which reforestation reduces the erosion and subsequent oxidation of POC–a critical aspect of confirming or refuting the suggestion that erosive steeplands should be a global priority for tree-planting because of the double benefit of carbon sequestration and erosion reduction. Critical uncertainties in the net greenhouse gas implications of erosion remain, including the need to better quantify the fraction of POC buried in sediments that is oxidized, the contribution of sediments to methane and nitrous oxide production, as well as the rate of soil C recovery on eroded land. Nations with hot spots of erosion will need to be aware of the uncertainties erosion introduces into the fair and equitable trading of C credits for afforestation, reforestation, and land management under the Kyoto Protocol, as these nations finalize policies to devolve C credits to individual land owners and projects.
 We thank H. Brackley, M. D'Ath and B. Rosser for technical assistance; R.H. Wilde for providing soil data; M. Jessen, I. Lynn and L Basher for assistance with terrain classification; J. Leathwick for climate data; Regional Councils for support of the stream gauging; and K. Walter for assistance with river flow data. Landcare Research provided primary funding. Additional funding from NZ NSOF to RAW, DMH, and WTB. BG acknowledges support from National Science Foundation Grant BCS-0317570.