Recapitulación

The main challenges encountered in this research relate to the capacity of the sediment fingerprinting approach to account for natural geochemical variability within catchment scale processes. This is both within the sediment sources as well as within river transport systems. Sediment sources can display a large geochemical range within short spatial scales as well as within geological and pedological groups (Chapter 4 & Chapter 6). This inherent intra-source geochemical variability presents a challenge for characterization of sediment sources since geochemically unique sediment sources can in some instances arise from similar processes and conversely, geochemically similar sediment sources can be produced from different erosion and deposition processes. This continues during sediment transport where disparate mixing of suspended sediment at, and downstream of, river confluences (Chapter 3) as well as the high temporal fluctuations of suspended sediment (Chapter 5) adds uncertainty to the retention of

sediment source geochemical character. This is compounded within larger catchments providing a challenge for clear characterization of sources.

One approach to negate this issue is to avoid sampling the sediment sources directly, but instead sample the channel sediment which represents each tributary because sediment from within the channel is less-variable than catchment soil and sediment sources (Wilkinson et al., 2013). In Chapter 3, a confluence approach was employed to target the main sub-catchments flowing into the main stem of the Manawatu River in order to establish whether there was enough geochemical variability to discriminate between sediment exiting the main sub- catchments; Upper Manawatu, Tiraumea, Mangatainoka, Mangahao, Pohangina and Oroua. Quantification of sediment source contributions was not possible due to the low sample replication, but sub-catchment sediment sources evidenced clear discrimination between respective upstream sources. This suggested potential for the technique to simplify larger complex catchment systems into manageable sections and be integrated into a traditional sediment sampling programme for step-wise discrimination. The discriminant function analysis between each of the confluences was able to effectively differentiate each of the upstream sediment source components. Each confluence used a different array of elements to differentiate sources, with only CaO, P2O5 and Cu being utilized in each confluence analysis. The C3 and C5 confluence (Fig. 15) had over twice as many significant elements available for discrimination mostly from REEs (Table 9). In the case of C5, this is attributed to sediment recycling occurring in the dune environment supported by the extremely high Zr concentrations which are associated with REE’s. In C3, the importance of the REE possibly reflects a ‘cleaner’ geological signal coming from the greywacke sandstone in the Tararua Range which contrasts to the mudstone dominated sediment from the Upstream Manawatu sediment source. Despite being able to discriminate the upstream sources, there was still a significant amount of geochemical variation displayed within the channel sediment, particularly closer to the confluence point and between point bars. This suggests that homogeneous mixing may not occur as quickly as assumed, with implications for the sample position within a channel. If the channel sediment is sampled, sampling density needs to be significantly higher even through channel sediment is considered less geochemically variable. Further source sediment sampling was undertaken throughout the catchment (Chapter 4) and involved targeting the sediment sources within the Manawatu Catchment with the intention that this would delineate erosion processes. However, exclusive and clear relationships between geological sources and processes made this impractical highlighting methodological limitations facing sediment fingerprinting in this catchment, which involve a balance between

an appropriate sampling density to cater for the geochemical variability and the total spatial extent, all ultimately limited by resources available. This is compounded when vertical stratification produces multiple geochemical sources contained within a single erosion mechanism (e.g. landslide). Conversely, a unique geochemical signature can occur in multiple erosion process settings throughout the catchment and not reside exclusively within a particular process or spatial location. This is well displayed in the Manawatu catchment through the Mudstone sediment source which is abundant throughout the catchment, and although the Mudstone provides a geochemically unique signature relative to other sediment sources (Chapter 4), introduction of mudstone derived sediment into the active channel can occur through a variety of pathways including cliff collapse, channel incision, gullying, and landslides, and in spatially assorted patterns. Because of this issue, source groups were defined based on geological and geomorphological contexts in order to produce geochemically unique sources. This provided eight potential source groups including; Mudstone, Hill Subsurface, Hill Surface, Channel Bank, Mountain Range, Loess, Gravel Terrace and Limestone.

Limitations for the geochemical characterization within a large catchment are the challenge to adequately sample all appropriate erosion sources and account for the unknown issues occurring throughout transport, which require longer time frames for sediment to move from source to catchment outflow. This relates to issues raised by several authors, notably Koiter et al. (2013b) identifying the black-box approach to sediment fingerprinting and the limited knowledge of what happens to the sediment properties between input and output. This particularly relates to the storage component and in the indirect relationship between source and suspended sediment sample and the changes to the sediment properties throughout this time. Geochemical Source Characterization

All the analysed geochemical variables displayed significantly different concentrations between at least two sediment sources attributed to the high number of sediment sources being characterized and the vastly different geochemistry from Limestone. Discriminant Function Analysis (DFA) produced a 16 variable solution consisting of (in order of significance) CaO, Lu, Cs, Sr, Tm, Na2O, P2O5, Fe2O3, Pb, U, Hf, MnO, Zn, MgO, Nb, and Y. The most readily differentiated sediment sources were Limestone, Mountain Range, Hill Surface and Loess due to their distinctive geochemistry, while Channel Bank and Hill Subsurface provided the most difficulty for geochemical classification, possessing closer geochemical signatures and similar origins. This is displayed in the DFA plot (Fig. 26) where Channel Bank sediment is plotted relatively centrally to all other source groups reflecting the mixed origin of the floodplain deposits, and raises challenges pertaining to differentiation of primary and secondary sources

due to sediment recycling. Retaining a similar geochemical signature to the primary sediment source is important for understanding historical sediment sources and requires conservative tracers to retain the geochemical signature. However, for contemporary sediment fingerprinting those same sediments need to have chemically altered enough to be able to differentiate between the channel bank sediment as a distinct source. It is possible that some of the channel bank sediment samples retained enough geochemical character from their primary source to limit full distinction from other sources, but alludes to the need to incorporate an understanding of weathering pathways and individual geochemical behaviour into the sediment fingerprinting research in a catchment of this nature.

Individual geochemical concentrations can be explained for some of the source groups, but full geochemical interpretation requires additional information. Limestone characterization is clearly provided by the CaO and Sr concentrations associated with mineral calcite, while Mountain Range sediment, Loess and Gravel Terrace tend to display relatively higher concentrations of trace elements and REEs. In the example of Mountain Range, the geochemical values reflect the rock fragments and muddy matrix of greywacke material as well as the depleted Si, Ca and Na concentrations of the weathering profile on a greywacke range (Anderson et al., 2002). In contrast, Hill Surface and Hill Subsurface display lower trace and REE concentrations than most other sources and display similar geochemical signatures due to the same mineral origins. The most significant differentiation between these two sources is due to higher P2O5 concentrations in the Hill Surface source.