Organic carbon in floodplain alluvium: Signature of historic variations in erosion processes associated with deforestation, Waipaoa River basin, New Zealand

Authors


Abstract

[1] We use C-org and δ13C to trace the origin of the alluvium deposited on the Waipaoa River floodplain at McPhail's Bend between 1853 and 2002. The overbank deposits exhibit a more positive range of δ13C values (−26.3 ± 0.6) than contemporary suspended sediment in transport at intermediate flows (−27.8 ± 0.2) when gully erosion releases most material to stream channels, reflecting the greater contribution made by shallow landsliding in the headwaters during large precipitation events. Overbank sediment associated with recent floods contains a higher (∼50%) amount of organic carbon than alluvium deposited by pre-1927 floods, which was derived from outcrops of weathered bedrock on steep, riparian hillslopes that were the primary sediment source before the soil-mantled hillslopes elsewhere in the headwaters of the basin were destabilized by deforestation. The recent alluvium resembles the Bw horizon of soils present on these hillslopes. Sediment deposited during floods generated by localized storms and an overbank event that did not feature widespread shallow landsliding also has a distinctive signature that provides an indication of its provenance. The organic carbon associated with the alluvium appears to be old (4031 ± 40 B.P., in the case of the carbon associated with sediment deposited in 2002) and to be derived from weathered Cretaceous and Tertiary sedimentary rocks.

1. Introduction

[2] Floodplain alluvium is an easily accessible, often high-resolution, sediment archive that contains chronologically ordered evidence of changing conditions in the catchment environment, and may provide proxy information about the past behavior of the upstream portion of a river basin. This information is contained in the physical and chemical properties of the alluvium and has, for example, been used to reconstruct changes in river basin sediment sources that accrued from land use changes in the United Kingdom during the 19th and 20th centuries [Collins et al., 1997; Owens et al., 1999; Owens and Walling, 2002]. These changes represent the most recent phase of a long (∼5000 year) history of anthropogenically induced land use change. By contrast, in New Zealand the succession of human impacts began with the disturbance of lowland forests by Polynesian settlers (Maori) only ∼700 years ago [McGlone, 1989; McFadgen, 2003], and culminated with the wholesale destruction of the indigenous forests by European colonists in the 19th and early 20th centuries. Here and elsewhere in the New World, once the land clearances extended beyond the lowlands and impacted the headwaters, erosion processes accelerated as the destruction of the natural vegetation and the resultant changes to the hillslope hydrological cycle and loss of root strength encouraged mass wasting and gully development [Wells and Andriamihaja, 1993; Page and Trustrum, 1997; Clark and Wilcock, 2000].

[3] Deforestation not only accelerates upland erosion, it also affects the cycling of carbon and other soil-derived biogenic elements [Lal, 1995; Ver et al., 1999], and since 1850 there has been a net release of carbon from terrestrial ecosystems as a result of the conversion of native forest to agricultural land [Houghton, 1995]. Some of the carbon that is exported to the aquatic environment is sequestered on floodplains as alluvial sediment accumulates [Mulholland, 1981; Stallard, 1998], and the majority is presumed to be derived from soils or regolith/saprolite removed from hillslopes by a variety of unconcentrated and concentrated erosion processes [Meybeck, 1982; Ludwig et al., 1996]. Small steepland rivers, many of which were deforested in the 19th and 20th centuries, have been recognized as a significant source of terrestrial sediment and carbon to the ocean [Milliman and Syvitski, 1992; Kao and Liu, 1996; Lyons et al., 2002]. Carbon isotopic properties have been used to trace the flow of organic matter from steepland river systems in the western United States across the continental shelf [Leithold and Blair, 2001; Masiello and Druffel, 2001]. The soils, colluvium and sedimentary rocks within these drainage basins reveal a range of δ13C values, however, organic particles in transport at progressively higher discharges appear to have more positive δ13C values [Leithold and Blair, 2001]; a trend that possibly reflects the greater contribution made by mass wasting during large precipitation events.

[4] In this paper we examine the cause of variations in the organic carbon content (as indicated by C-org and δ13C) of a suite of floodplain sediment, with a history that has been reconstructed stratigraphically, from the Waipaoa River drainage, New Zealand. First, using our previous work as a starting point we reiterate that, although mixing effaces many of the compositional differences that exist between the alluvium and the source materials from which it is derived, there is a strong signal preserved in the organic content of the sediment sequestered on the Waipaoa River floodplain at McPhail's Bend deposited between 1850 and 2002 [Gomez and Trustrum, 2005]. The signal is diagnostic of historical variations in sediment source(s) and the contribution made to the overbank deposits by shallow landsliding. Second, we show how C-org and δ13C can be used to trace the origin of this alluvium.

[5] Our approach differs, in two fundamental respects, from other source provenance studies that have used the geochemical record preserved in floodplain sediment to provide information about changing conditions in the catchment environment during the historic period [cf. Collins et al., 1997; Owens et al., 1999; Owens and Walling, 2002]. First, we focus on the contribution made to the sediment deposited during high-magnitude events by an erosion process (shallow landsliding) whose signature varies through time (reflecting the shift from localized to more widespread slope instability following deforestation), rather than on the contributions made from geographically distinct sources. Second, we utilize total organic carbon concentration and its isotopic value that ofttimes may be nonconservative (i.e., affected by the amount and size distribution of the clastic sediment input, as well as by biological processes after the associated sediment has been deposited).

[6] Our results are of interest, not only because they demonstrate the potential for using total organic carbon concentration and its isotopic signature to reconstruct the historical record of sediment provenance on lowland river floodplains, but also because they provide information about the way in which carbon contents respond to changes in sediment source area and different erosion processes. The former approach has relevance to our understanding of the consequences of anthropogenic activity [Owens and Walling, 2002], and is a logical extension of the successful use of carbon, nitrogen and other stable isotopes for hydrological purposes [Kendall and McDonnell, 1998]. The latter effect influences the amount and composition of carbon buried on continental margins and is, therefore, of relevance to the study of the river-ocean carbon exchange system [Cole and Caraco, 2001; Leithold and Blair, 2001].

2. Study Area

[7] Situated on the East Coast Continental Margin, North Island, New Zealand (Figure 1), the 2200 km2 Waipaoa River basin drains the eastern flanks of the Raukumara Range. The river basin lies within the zone of active deformation associated with the Hikurangi subduction margin [Lewis and Pettinga, 1993; Reyners et al., 1999], and experiences a maritime climate that periodically is perturbed by cyclonic storms of tropical origin [Hessel, 1980]. Miocene-Pliocene mudstone and sandstone underlie the low hills surrounding the Poverty Bay Flats and the hills in the eastern section of the headwaters. Overthrusted Cretaceous to Tertiary mudstone and argillite, and upper Cretaceous sandstone and siltstone underlay the remainder of the headwaters [Mazengarb and Speden, 2000]. The first anthropogenic disturbances to the lowland forests occurred in association with Polynesian (Maori) settlement in the lower reaches of the river basin, which dates from the 13th century A.D. [Jones, 1988; Wilmshurst et al., 1999], but the wholesale clearance of the indigenous forest did not begin until European settlers arrived in the late 1820s. Most of the lowlands had been deforested by 1875. Clearances in the headwaters continued until 1920, and today only 2.5% of the basin remains under native forest. Reforestation began in 1960, and 13% of the basin is now covered by exotic forest.

Figure 1.

(a) Location map showing areas in the headwaters of the Waipaoa River basin underlain by overthrusted Cretaceous rocks that are susceptible to gully erosion (dark shading) and those areas of the headwaters impacted by deforestation after 1875 (light shading). The locations of Te Karaka (cross), Kanakanaia (solid dot), and McPhail's Bend also are shown. Isohyets (mm of precipitation) delimit the center of the August 2002 storm. (b) Configuration of McPhail's Bend (1868–1988) and location of core sites (cross).

[8] Deforestation destabilized the landscape and extensive, shallow landsliding was first observed in the hill country during the winters of 1893 and 1894, soon after the start of the most intensive period (1890–1910) of land use change in the headwaters of the Waipaoa River basin. The incidence of shallow landsliding increased during the first decade of the 20th century, and by the end of the second decade it had become a pervasive erosional process throughout the headwaters [Henderson and Ongley, 1920]. At the present time, 67% of the total basin area is susceptible to shallow landsliding during high-intensity rainstorms [Reid and Page, 2002]. Shallow landslides in the Waipaoa River basin characteristically involve small, <1 m deep, planar failures that originate at the surface of the weathered bedrock or within the soil profile [Reid and Page, 2002]. Soil released from the Bw horizon and regolith/saprolite account for the bulk of the material mobilized and exposed by landsliding, and the availability of this material increases dramatically once the threshold rainfall (∼200 mm in 72 hours) for landsliding is exceeded [Dymond et al., 1999; Reid and Page, 2002].

[9] Numerous gullies also developed in the headwaters after the native beech, podocarp and mixed hardwood forest was replaced by pasture [Marden et al., 2005]. Gully erosion is associated with ∼10% of the terrain in the Waipaoa River basin, most of which is underlain by the overthrusted Cretaceous rocks of the East Coast Allochthon (Figure 1) [Gage and Black, 1979]. Many of the larger gullies and gully complexes appear to have developed on the sites of mass movements that were present under the native forest cover, and their initiation has been linked to the unusually wet winters of 1916, 1917, and 1918 [Hamilton and Kelman, 1952]. Most initial attempts to control erosion in the headwaters using fascines and check dams failed, but many smaller gullies eventually were stabilized after a program of exotic reforestation was implemented in 1960 [Allsop, 1973; Gomez et al., 2003b; Marden et al., 2005]. Currently, 420 gullies remain active. They vary in size from <0.01 to ∼0.2 km2, and the drainage basins that support them range from a few thousand square meters to gully complexes that are ∼0.5 km2 in area [cf. DeRose et al., 1998].

[10] Gully erosion, supplemented by shallow landsliding during intense rainstorms [Trustrum et al., 1999], generates large quantities of fine sediment, and the Waipaoa River annually delivers ∼15 Mt of suspended sediment to the Pacific Ocean (the standard deviation of the annual yield is 6.7 Mt) [Hicks et al., 2000]. Shallow landslides generate ∼15% of the Waipaoa River's annual suspended sediment yield at Kanakanaia [Reid and Page, 2002], the main stem gauging station which is located 48 km from the coast (Figure 1a). For comparison, the six gully complexes in the headwaters each generate between 1% and 3% of the Waipaoa River's annual suspended sediment load [cf. DeRose et al., 1998; Marden et al., 2005]. Together, the 420 active gullies generate sediment during events of all scales and show no threshold effects [Hicks et al., 2000]. Gully erosion not only sustains the catchment-averaged denudation rate of 1 to 2 mm yr−1 that is implied by the Waipaoa River's suspended sediment yield, it also regulates suspended sediment and particulate organic carbon yields across the entire range of flows. The gully complexes contribute sediment during baseflow conditions, and most established gullies are activated by small, frequent rainstorms. Consequently, frequent runoff events are relatively more important than large floods, both to the long-term suspended sediment and particulate organic carbon yields of the Waipaoa River [Hicks et al., 2000; Gomez et al., 2003a]. At Kanakanaia, 50% of the suspended sediment load is transported by flows <500 m3 s−1, and 83% by flows between the mean flow (34.7 m3 s−1) and the mean annual flood (1346 m3 s−1). The most effective flow (360 m3 s−1) is 0.23 times the mean annual flood and 11 times the mean flow.

[11] Most floods only last for a few tens of hours, and flows associated with overbank events transport only ∼24% of the mean annual suspended sediment load [Gomez et al., 1999]. However, suspended sediment concentrations in flood flows are very high (>30,000 mg L−1), and sediment accumulates rapidly on the floodplain during overbank events. During overbank events of 6–20 hours duration, contemporary rates of vertical accretion on the floodplain are 14 to 18 mm h−1 (which helps inhibit oxidation by reducing the residence time of organic matter near the surface of a deposit), and the average (post-1850) rate of vertical accretion is ∼60 mm yr−1 [Gomez et al., 1998, 1999]. The ‘recent’ alluvium also has a low organic matter content (1–2%), and a C/N ratio (9.4 ± 1.9) that suggests its breakdown was well advanced prior to deposition [Pullar, 1962; Gomez et al., 2003a].

[12] The high rate of deposition on the Waipaoa River floodplain in the historic period is a product of deforestation [Pullar, 1962; Pullar and Penhale, 1970]. At McPhail's Bend the floodplain is under pasture and has never been tilled, consequently the stratigraphic record for the historic period at this site is exceptionally detailed (Figure 2a). Most depositional units have been tied to specific events, because the high rates of fine-sediment deposition make it easy to distinguish the contribution of individual events identified from historical records (dating from 1820) to floodplain stratigraphy. Under the present hydrologic regime, overbank sedimentation occurs on McPhail's Bend at discharges ≥1500 m3 s−1 [Sy, 2003], and floods in excess of this magnitude often are associated with storms that exceed the threshold for landsliding [Kelliher et al., 1995; Gomez et al., 1999].

Figure 2.

(a) Generalized stratigraphy and probable dates associated with major flood units identified in the sediment core from McPhail's Bend [after Gomez et al., 1999]. The radiocarbon date was obtained from wood fragments incorporated in sediment near the base of the core. (b) Percentage of total organic carbon. Shading shows the time period when the headwaters were undisturbed, the horizontal dotted line indicates the appearance of a deforestation-related signal, and vertical dashed lines denote mean values for the periods before and after deforestation impacted the depositional record [after Gomez and Trustrum, 2005]. (c–e) Component scores (PC1, PC2, and PC3) for metals and related elements (Co, Cr, Cs, Fe, Rb, Sc, Sr, Zn, Al, Ba, Ca, K, Mn, Na, Ti, and V) and rare earth and related elements (As, Ce, Dy, Eu, Hf, La, Lu, Nd, Sb, Sm, Tb, Th, Yb, and Zr). The analysis was performed on normalized (to upper continental crust) concentrations. Together, the first three components explain 61% and 77% of the cumulative variance, respectively, and the absence of any downcore trend in the composite variables is replicated in the data for individual elements. (f) Percentage of principal clay minerals.

[13] The organic carbon content of the fine-grained alluvium (D50 12.6 ± 4.8 μm) sequestered on the Waipaoa River floodplain at McPhail's Bend retains the impact of deforestation [Gomez and Trustrum, 2005]. The accumulated signal is manifest as a ∼50% increase in the amount of total organic carbon associated with the overbank sediment deposited on McPhail's Bend from the flood of 1927 onward (Figure 2b). The lag between the clearances in the headwaters and the appearance of a deforestation-related signal in the alluvium is a product of two factors. First, the occurrence of landslides on hillslopes that have been converted to pasture is conditional upon the (gradual decay and eventual) loss of the root strength afforded by the former forest cover. Second, shallow landsliding is a threshold phenomenon, triggered by large magnitude low frequency events; ∼50% of landslides occur during storms with a recurrence interval of <7 yr and 75% during storms with a recurrence interval of <25 yr [Reid and Page, 2002].

3. Sampling and Analysis

[14] Bulk samples were taken from the bottom or center of compositionally similar (fine-medium silt, D50 12.6 ± 4.8 μm) units in a single, 9.7-m-long sediment core with an established stratigraphy that had been stored in a cold room (maintained at ∼4°C) [Gomez et al., 1998, 1999]. This core was obtained near the locus of McPhail's Bend as it appeared in 1868 (Figure 1b). On the basis of the core stratigraphy, which was constrained by a radiocarbon date of 1820 ± 88 B.P. obtained on wood fragments recovered at 9.3 m, all the samples considered here were associated with post-1850 flood events (Figure 2a), the most recent of which occurred in March 1988. A shallow (200-mm-long) core was also obtained from approximately the same location, in the week following a flood that occurred on 6 and 7 August 2002. This core was frozen within hours of being collected and, in addition to the 2002 flood deposit, contained sediment deposited during a flood that occurred in March 1996 (D50 17.9 and 12.0 μm, respectively). We surveyed and sampled the August 2002 flood deposit throughout the 44-km-long reach downstream from Kanakanaia [cf. Gomez et al., 1999]. High water marks indicate that ∼5 km2 of the floodplain between Kanakanaia and the confluence of the Te Arai River was inundated during the 2002 event, and the duration of overbank flow at McPhail's Bend was ∼7 hours. On the basis of the average depth of sediment measured along nine transects and a measured bulk density of 1.08 mg m3, ∼0.22 Mt of sediment were deposited within the 44-km-long reach. This amounts to ∼4% of the event suspended sediment yield (∼5 Mt) measured at Kanakanaia, and 14% of the suspended sediment discharged at stages in excess of bankfull.

[15] We also sampled a representative range of the earth materials (weathered bedrock, and soil and regolith/saprolite on hillslopes prone to landsliding) present in the headwaters of the Waipaoa River basin, as well as the suspended sediment discharged from two active gullies and the suspended sediment in transport at the mainstem gauging station (Kanakanaia) during low and intermediate flows [Gomez et al., 2003a]. All these samples were air dried and stored in a cold room prior to analysis, which included identification of the common clay groups, the determination of element concentrations, stable carbon and nitrogen isotope analyses, and particle size analysis. Details of the sampling program were provided in Gomez et al. [2003a] and are not reiterated here. Summary data are presented in Table 1.

Table 1. Summary Data of Carbon and Nitrogen Contents and Stable Isotope Ratios of Alluvial Sediment and Source Materials
 Number of SamplesCarbon, %δ13C, ‰Nitrogen, %δ15N, ‰C/N
  • a

    Sandstone, siltstone, and argillite lithologies.

  • b

    Four soil types on steep (30°–40°), pastoral, landslide-prone hillslopes.

  • c

    After Gomez et al. [2003a].

  • d

    Excluding 1944, 1988, 1996, and 2002 (see text for discussion).

  • e

    No data available.

Weathered rock (W)a110.4 ± 0.2−26.0 ± 1.20.04 ± 0.013.4 ± 1.09.6 ± 4.8
Bw horizonb110.9 ± 0.3−26.1 ± 0.70.07 ± 0.035.7 ± 1.112.68 ± 3.0
AB/BA horizonb102.5 ± 0.9−26.1 ± 0.70.2 ± 0.077.7 ± 0.512.9 ± 2.1
Ap horizonb134.4 ± 1.0−26.8 ± 0.60.5 ± 0.16.4 ± 1.09.8 ± 0.8
Gulliesc20.42−28.10.052.28.5
Pre-1927 floodplain sediment120.5 ± 0.05−26.5 ± 0.10.04 ± 0.010.8 ± 1.09.6 ± 1.3
Post-1927 floodplain sedimentd60.7 ± 0.05−26.1 ± 0.70.08 ± 0.021.9 ± 0.710.3 ± 2.3
Suspended sedimentc240.6 ± 0.07−27.8 ± 0.2NAeNAeNAe

[16] The mineralogy of the clay size fraction in the 9.7 m core was determined by X-ray diffraction (XRD) at Massey University. Air dried subsamples (10 g) were treated with hydrochloric acid, hydrogen peroxide, a citrate reagent, sodium bicarbonate, and sodium dithonite to remove calcium, organic matter and iron and aluminum oxides. The clay particles were then separated by centrifuge, and Mg++ and K+ saturated slides prepared and analyzed using the procedures outlined by Whitton and Churchman [1987]. Estimated uncertainties for each identified mineral are of the order of ±5 to 15%.

[17] All of the following analyses were undertaken on bulk samples and did not focus on any particular size fraction. Concentrations of 32 major and minor elements in 200 mg, oven dried (at 30°C) subsamples were determined by neutron activation analysis at Missouri University Research Reactor. The procedure involves two irradiations and three gamma counts that yielded determinations for Al, As, Ba, Ca, Ce, Co, Cr, Cs, Dy, Eu, Fe, Hf, K, La, Lu, Mn, Na, Nd, Rb, Sb, Sc, Sm, Sr, Ta, Tb, Th, Ti, U, V, Yb, Zn, and Zr. Reference standards (SRM-1633a Flyash and SRM-688 Basalt) were irradiated and measured with each batch of samples for calibration, and the analytical error is between 1 and 2% [Glascock, 1992].

[18] The carbon and nitrogen contents and isotopic composition of 20 mg subsamples were analyzed in duplicate at the Institute of Geological and Nuclear Sciences, New Zealand, using a Europa Geo 20/20 isotope ratio mass spectrometer, interfaced to an ANCA SL elemental analyzer in continuous flow mode (EA IRMS). Each subsample was demineralized with 1M HCl to remove inorganic carbon, rinsed with deionized water until neutral, then dried in an oven at 30°C, prior to being placed in a tin capsule for automated combustion in dual isotope mode. The carbon dioxide gas was resolved from nitrogen gas using gas chromatographic separation on a column at 85°C, and analyzed simultaneously for isotopic abundance as well as total organic carbon and nitrogen. Standards and blanks were included during each run for calibration. The δ13C value (given in ‰) is the ratio of 13C to 12C in a sample relative to the international Pee Dee Belemnite (PDB) standard, where: δ13C = {[(13C/12C Sample)/(13C/12C Standard)] − 1} × 1000. Similarly, the δ15N value (‰) is the ratio of 15N to 14N in a sample relative to atmospheric N2, where: δ15N = {[(15N/14NSample)/(15N/14NAir)] − 1} × 1000. The analytical precision of the measurements is ±0.1‰, and the reproducibility of the results is within ±0.1‰ for carbon and ±0.3‰ for nitrogen. A sample from the top of the 200 mm core taken from McPhail's Bend immediately after the 2002 event was also analyzed for 14C concentration at the Institute of Geological and Nuclear Sciences, Rafter Radiocarbon Laboratory.

[19] The size distribution of 200 mg, oven dried (at 30°C) subsamples, which were dispersed by (NaPO3)6 and ultrasound, was determined at Indiana State University, using a CILAS l064 laser granulometer. The error and reproducibility of mean values are ±0.6 μm and ±0.5%, respectively.

4. Sediment Sources and Variations in Sediment Properties

[20] Once sediment is in transport, mixing rapidly effaces the slight compositional differences that exist between different source materials in the headwaters [cf. D'Ath, 2002], thus, it has not previously proved possible to differentiate between the contributions made by sediment from different geographic sources in the Waipaoa River basin to the floodplain alluvium. Indeed, there are no obvious downcore trends either in the chemical characteristics or in the clay mineralogy of the overbank deposits (Figures 2c–2f), nor in their particle size [Gomez et al., 2003a]. The identification of sediment sources also has been confounded by the realization that, through sediment supply and erosion threshold effects, the different hillslope erosion processes generate characteristic signatures in the relationships between suspended sediment concentration and water discharge, and in the magnitude-frequency characteristics of event sediment yields [Hicks et al., 2000]. Consequently, the composition of the floodplain sediment cannot be assumed to be generally representative of the bulk of the suspended sediment transported by the river at low to intermediate discharges. Fortuitously, however, this information permits us to differentiate between sediment delivery mechanisms and the fundamental processes (shallow landsliding and gully erosion) responsible for generating sediment during events of different magnitude [Hicks et al., 2000, 2004]. Additionally, the temporal change in organic carbon content observed in Figure 2b is consistent with the observation that prior to deforestation failures frequently occurred in weathered bedrock, on steep riparian hillslopes that only possessed a thin, skeletal soil cover and were strongly coupled to river channels [cf. Henderson and Ongley, 1920; Hamilton and Kelman, 1952]. Whereas, following their conversion to pasture (1880–1920), landslides were initiated on hillslopes with better-developed soils elsewhere in the headwaters.

[21] Following the methodology developed for stable isotopes in surface water [cf. Cravotta, 1997], we employed bivariate plots of δ13C, C-org, and δ15N to explore the relationship between the source materials and the floodplain sediment further. Our results are summarized in Figure 3.

Figure 3.

(a) Bivariate plot of δ13C versus C-org in floodplain, suspended and gully-derived sediment (unlabeled open squares represent alluvium deposited by pre-1927 floods). Bars indicate the range of values obtained for 11 weathered bedrock (W) sources and four representative soils (Bw horizon) on landslide-prone hillslopes in the headwaters of the Waipaoa River basin. (b) Data ranges for each component (Ap, AB/BA, and Bw horizons) of the soil profiles and weathered bedrock and gully sources (shaded area delimits the region covered by Figure 3a, and the system of horizon notation is that of Clayden and Hewitt [1989]). (c) Concentration-weighted mixing triangle for floodplain sediment. The isotopic ratios at the vertices of the triangle are mean values for the specified sediment source (compare Table 1).

[22] Figure 3a is a plot of δ13C versus percent organic carbon for the dated floodplain sediment and suspended sediment samples taken at intermediate flows (discharge 11–23% bankfull). The more positive range of δ13C values for the floodplain sediment, by comparison with the range of values for suspended sediment in transport at intermediate flows (Table 1), reflects the processes that are responsible for delivering sediment to the channel system during events of different magnitude [cf. Gomez et al., 2003a]. Gullies contribute most sediment during the small, frequent events that are characterized by more negative δ13C values, whereas large amounts of sediment are generated by shallow landsliding during the large magnitude events that are associated with the more positive range of δ13C values. The sediment associated with the large floods (peak discharge 1520–4000 m3 s−1) that occurred from 1927 onward also plots in a different field to the alluvium that was deposited by the 12 pre-1927 floods (Figure 3a). Our data do not permit us to distinguish between different types of organic matter in the sediment (e.g., organic matter sorbed onto the surface of mineral particles or organic matter incorporated into mineral structures). However, the signature of the pre-1927 floods is similar to that of weathered bedrock in the headwaters, whereas the alluvium deposited from 1927 onward more closely resembles the Bw horizon in soils present on steep (30°–40°) landslide-prone hillslopes. As noted above, the separation between the two groups of flood deposits reflects the transition from predeforestation failures on steep, riparian hillslopes that only possessed a thin, skeletal soil cover, to postdeforestation landsliding on hillslopes with better-developed soils elsewhere in the headwaters.

[23] It is estimated that shallow landslides generated >50% of the suspended sediment transported by the Waipaoa River during the 1988 flood [Page et al., 1999], and the more negative δ13C value associated with sediment deposited during this event (peak discharge ∼4000 m3 s−1) probably reflects the different time periods over which sediment generated by gully erosion and by shallow landsliding is delivered to stream channels during very large magnitude events. Gully erosion intensifies as rainfall increases but shallow landsliding usually is initiated toward the end of storm events, once a rainfall threshold has been surpassed. Thus the signal of shallow landsliding is likely to be strongest in sediment deposited during the latter stages of such events. The sample used to characterize the 1988 flood was obtained from the center of a fining-upward, sand-silt-clay unit that is presumed to have been deposited on the falling limb of the flood hydrograph, but elevated suspended sediment concentrations were observed long (53 hours) after the flood peak and active overbank flow had ceased [cf. Gomez et al., 1998]. The implication is that the sample may not fully represent the latter stages of the 1988 event, when shallow landsliding likely was a more important source of sediment than gully erosion.

[24] Sediment deposited during the floods of 1944, 1996, and 2002 also plot in apparently uncharacteristic positions (Figure 3a). Field observations indicate that widespread shallow landsliding was not a feature of the rainstorm that generated the 1996 flood (peak discharge, 2030 m3 s−1). Gully and sheet erosion generated most of the sediment, and the elevated organic carbon content likely is related to the contribution made by surface wash on riparian hillslopes. During the rainstorms that generated the floods of 1944 and 2002 (peak discharge 2240 and 1872 m3 s−1, respectively) the heaviest rainfalls were recorded in the lower reaches of the basin. In March 1944, 167 mm of rain fell at Te Karaka in 24 hours [Soil Conservation and Rivers Control Council, 1957], and in August 2002 the southeastern portion of the Waipaoa River basin received >200 mm of rain in a 36 hour period (Figure 1). Most landsliding occurred on steep and precipitous riparian hillslopes in the upper reaches of the Waihora, Wharekopae and Waikohu rivers, as opposed to hillslopes with better-developed soils elsewhere in the headwaters. Consequently the signature of these two events is similar to that of the 12 pre-1927 floods (Figure 3a).

[25] We employed a mixing model in an attempt to partition the sources for the sediment deposited during the 1996 and 2002 floods (Figure 3c). The model, developed by Phillips and Koch [2002], assumes that for each element, a source's contribution is proportional to the contributed mass times the elemental concentration in that source. We consider the case of two isotopic signatures (δ13C and δ15N) and three sources (sediment from gullies, soil released from the Bw horizon and weathered bedrock). Average values for δ13C, C, δ15N, and N were used to characterize each source (Table 1). The lines connecting the vertices of the mixing triangle are curved because elemental concentrations in the sources differ (Figure 3c), and represent two-source mixing lines (e.g., a point halfway between two vertices corresponds to mixture that contains 50% of each source). For the 2002 flood, the C and N proportions of the weathered bedrock source present in the alluvium both are >98%, and in the case of the 1996 event, the C and N proportions of the gully source present in the alluvium are 74% and 77%, respectively, with weathered bedrock accounting for the remainder of the source contribution. Data points for other events plot outside the mixing triangle (Figure 3c). We suggest that this is because biological processes affect nitrogen in soils over time, hence the δ15N value of soil organic matter commonly decreases several or more ‰ near the soil surface [cf. Mariotti et al., 1980; Brenner et al., 2001]. Specifically, nitrogen-fixing plants such as clover and grass contribute nitrogen (δ15N near 0‰) to the alluvium on the floodplain which, in time, draws the nitrogen isotope signature of the system away from the higher δ15N values that are characteristic of the sources (gullies and weathered bedrock), and which are preserved only in the most recent flood deposits (the lower values being retained during burial).

[26] Finally, the results presented in this paper, and in our previous work (which demonstrated that during high flows the particulate organic carbon content of suspended sediment in the Waipaoa River attains a low stable value [Gomez et al., 2003a]), suggest that much of the associated organic material likely is fossil carbon derived from weathered sedimentary rocks. The carbon in the sample from the top of the shallow core taken from McPhail's Bend immediately after the 2002 event yielded a radiocarbon age of 4031 ± 40 B.P. and contained 60 ± 0.3% modern carbon, as defined by Stuiver and Polach [1977], indicating that this is indeed a tenable assumption. The age of the carbon is considerably greater than the residence time of soil organic carbon, which has a global mean value of ∼850 yr [Harrison et al., 1993], and the percentage of modern carbon present is consistent with the assumption that the particulate organic carbon associated with fluvial sediment results from the mixing of ancient organic matter derived from sedimentary rocks with younger material. However, by comparison with data from other turbid steepland rivers [cf. Kao and Liu, 1996], the percentage of modern carbon is relatively high. Commensurate with the localized impact of the 2002 event (on tributary basins largely unaffected by gully erosion), this likely reflects the contribution of organic matter supplied by sheet erosion, which it is estimated may have contributed as much as 70% of the sediment supplied by the Waikohu River alone (A. Sidorchuk, personal communication, 2003).

5. Conclusion

[27] The organic carbon content of floodplain sediment depends not only on the supply and preservation of organic matter, but also on the supply of inorganic clastic sediment [Morozova and Smith, 2003]. On floodplains bordering turbid steepland rivers, such as the Waipaoa River, the high frequency of flooding and rate of overbank deposition limit the time available for alluvial soils to develop, and consequently they exhibit a low total organic carbon content. Much of the organic carbon is a product of transfers effected by geomorphological processes, and is associated with fluvial (suspended) sediment generated by gully erosion, supplemented by shallow landsliding during high-magnitude events [Gomez et al., 2003a].

[28] Sediment deposited on the Waipaoa River floodplain, at McPhail's Bend, during 22 overbank events that occurred in the past ∼150 yr retains the signature of landsliding, and exhibits more positive δ13C values than suspended sediment transported during flood events of low to intermediate magnitude (Figure 3a). Alluvium deposited in the decades before and during the time deforestation impacted the headwaters (1880–1920) also generally contained a lower amount of total organic carbon than sediment deposited during flood events dating from 1927 onward. This reflects its derivation from failures in weathered bedrock initiated on steep, riparian hillslopes mantled with thin, skeletal soils that were the primary sediment source before hillslopes with better-developed soils elsewhere in the headwaters were destabilized by deforestation. Sediment deposited during floods generated by localized storms, and an overbank event that did not feature widespread shallow landsliding, also have distinctive signatures that provide an indication of their provenance. An attempt to apportion the contributions made by different sediment sources using a mixing model indicated that biological processes influence the nitrogen isotope signature (Figure 3c). Specifically, contributions from clover and grass growing on the floodplain draw the δ15N value of the alluvium away from the higher values characteristic of the gully and weathered bedrock sources. However, because this is a time-dependent process the isotopic signature of the source material is preserved in contemporary alluvium (i.e., sediment deposited in 1996 and 2002). By contrast, the carbon appears relatively recalcitrant, and its age (4031 ± 40 B.P., in the case of the 2002 flood deposit) reinforces our opinion that, as in steepland basins elsewhere [cf. Kao and Liu, 1996], a large proportion is derived from weathered sedimentary rock. Our results thus provide support for the suggestion that the more positive δ13C values of organic particles transported by steepland rivers at higher discharges are a product of the contribution made by mass wasting during large precipitation events [Leithold and Blair, 2001].

[29] Our study was facilitated by the dominance of two erosion processes, one of which (shallow landsliding) had a signature that varied through time. By contrast, other studies that have used the record preserved in floodplain deposits to reconstruct recent historical changes in river basin sediment sources have highlighted the importance of temporal changes in distinct geographic sources that encompass a variety of erosion processes [cf. Collins et al., 1997]. In either case, however, it is necessary to differentiate potential sources in an unequivocal manner, and our results demonstrate the potential for using concentrations and isotopic signatures of carbon and, perhaps, nitrogen to elucidate sediment provenance and the effects of anthropogenic activity in the catchment environment.

Acknowledgments

[30] This paper is a contribution to Manaaki Whenua−Landcare Research's Waipaoa Catchment Study. The research was supported by the National Science Foundation (grants SBR-9807195 and BCS-0136375); the New Zealand Foundation for Research, Science and Technology (contract C09X0013); and Indiana State University. We thank Michel D'Ath, Dennis Eden, Mike Marden, Dave Peacock, Ted Pinkney, Brenda Rosser, Yuko Siguta, and Joe Whitton for their assistance with the coring and analyses and Troy Baisden and Brendan Hicks for their commentary on our results.

Ancillary